Download to read offline
![Programming Google App Engine
by Dan Sanderson
Copyright © 2010 Dan Sanderson. All rights reserved.
Printed in the United States of America.
Published by O’Reilly Media, Inc., 1005 Gravenstein Highway North, Sebastopol, CA 95472.
O’Reilly books may be purchased for educational, business, or sales promotional use. Online editions
are also available for most titles (http://my.safaribooksonline.com). For more information, contact our
corporate/institutional sales department: (800) 998-9938 or corporate@oreilly.com.
Editor: Mike Loukides Indexer: Ellen Troutman Zaig
Production Editor: Sumita Mukherji Cover Designer: Karen Montgomery
Proofreader: Sada Preisch Interior Designer: David Futato
Illustrator: Robert Romano
Printing History:
November 2009: First Edition.
Nutshell Handbook, the Nutshell Handbook logo, and the O’Reilly logo are registered trademarks of
O’Reilly Media, Inc. Programming Google App Engine, the image of a waterbuck, and related trade dress
are trademarks of O’Reilly Media, Inc.
Many of the designations used by manufacturers and sellers to distinguish their products are claimed as
trademarks. Where those designations appear in this book, and O’Reilly Media, Inc., was aware of a
trademark claim, the designations have been printed in caps or initial caps.
While every precaution has been taken in the preparation of this book, the publisher and author assume
no responsibility for errors or omissions, or for damages resulting from the use of the information con-
tained herein.
TM
This book uses RepKover™, a durable and flexible lay-flat binding.
ISBN: 978-0-596-52272-8
[M]
1257864694](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-5-320.jpg)
![Let’s upgrade the clock app to use the webapp framework. Replace the contents of
main.py with the version shown in Example 2-3. Reload the page in your browser to
see the new version in action. (You won’t notice a difference other than an updated
time. This example is equivalent to the previous version.)
Example 2-3. A simple request handler using the webapp framework
from google.appengine.ext import webapp
from google.appengine.ext.webapp.util import run_wsgi_app
import datetime
class MainPage(webapp.RequestHandler):
def get(self):
time = datetime.datetime.now()
self.response.headers['Content-Type'] = 'text/html'
self.response.out.write('<p>The time is: %s</p>' % str(time))
application = webapp.WSGIApplication([('/', MainPage)],
debug=True)
def main():
run_wsgi_app(application)
if __name__ == '__main__':
main()
Example 2-3 imports the module google.appengine.ext.webapp, then defines a request
handler class called MainPage, a subclass of webapp.RequestHandler. The class defines
methods for each HTTP method supported by the handler, in this case one method for
HTTP GET called get(). When the application handles a request, it instantiates the
handler class, sets self.request and self.response to values the handler method can
access and modify, then calls the appropriate handler method, in this case get(). When
the handler method exits, the application uses the value of self.response as the HTTP
response.
The application itself is represented by an instance of the class
webapp.WSGIApplication. The instance is initialized with a list of mappings of URLs to
handler classes. The debug parameter tells the application to print error messages to the
browser window when a handler returns an exception if the application is running
under the development web server. webapp detects whether it is running under the
development server or running as a live App Engine application, and will not print
errors to the browser when running live even if debug is True. You can set it to False to
have the development server emulate the live server when errors occur.
The script defines a main() function that runs the application using a utility method.
Lastly, the script calls main() using the Python idiom of if __name__ ==
'__main__': ..., a condition that is always true when the script is run by the web server.
This idiom allows you to import the script as a module for other code, including the
classes and functions defined in the script, without running the main() routine.
Developing the Application | 29](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-50-320.jpg)
![Defining a main() function this way allows App Engine to cache the
compiled handler script, making subsequent requests faster to execute.
For more information on app caching, see Chapter 3.
A single WSGIApplication instance can handle multiple URLs, routing the request to
different RequestHandler classes based on the URL pattern. But we’ve already seen that
the app.yaml file maps URL patterns to handler scripts. So which URL patterns should
appear in app.yaml, and which should appear in the WSGIApplication? Many web
frameworks include their own URL dispatcher logic, and it’s common to route all dy-
namic URLs to the framework’s dispatcher in app.yaml. With webapp, the answer
mostly depends on how you’d like to organize your code. For the clock application, we
will create a second request handler as a separate script to take advantage of a feature
of app.yaml for user authentication, but we could also put this logic in main.py and
route the URL with the WSGIApplication object.
Users and Google Accounts
So far, our clock shows the same display for every user. To allow each user to customize
the display and save her preferences for future sessions, we need a way to identify the
user making a request. An easy way to do this is with Google Accounts.
Let’s add something to the page that indicates whether the user is signed in, and pro-
vides links for signing in and signing out of the application. Edit main.py to resemble
Example 2-4.
Example 2-4. A version of main.py that displays Google Accounts information and links
from google.appengine.api import users
from google.appengine.ext import webapp
from google.appengine.ext.webapp.util import run_wsgi_app
import datetime
class MainPage(webapp.RequestHandler):
def get(self):
time = datetime.datetime.now()
user = users.get_current_user()
if not user:
navbar = ('<p>Welcome! <a href="%s">Sign in or register</a> to customize.</p>'
% (users.create_login_url(self.request.path)))
else:
navbar = ('<p>Welcome, %s! You can <a href="%s">sign out</a>.</p>'
% (user.email(), users.create_logout_url(self.request.path)))
self.response.headers['Content-Type'] = 'text/html'
self.response.out.write('''
<html>
<head>
<title>The Time Is...</title>
</head>
30 | Chapter 2: Creating an Application](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-51-320.jpg)
![<body>
%s
<p>The time is: %s</p>
</body>
</html>
''' % (navbar, str(time)))
application = webapp.WSGIApplication([('/', MainPage)],
debug=True)
def main():
run_wsgi_app(application)
if __name__ == '__main__':
main()
In a real application, you would use a templating system for the output, separating the
HTML and display logic from the application code. While many web application
frameworks include a templating system, webapp does not. Since the clock app only
has one page, we’ll put the HTML in the handler code, using Python string formatting
to keep things organized.
The Python runtime environment includes a version of Django, whose
templating system can be used with webapp. When Google released
version 1 of the Python runtime environment, the latest version of
Django was 0.96, so this is what the runtime includes. For more infor-
mation on using Django templates with webapp, see the App Engine
documentation.
Reload the page in your browser. The new page resembles Figure 2-5.
Figure 2-5. The clock app with a link to Google Accounts when the user is not signed in
Developing the Application | 31](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-52-320.jpg)
![if not user:
navbar = ('<p>Welcome! <a href="%s">Sign in or register</a> to customize.</p>'
% (users.create_login_url(self.request.path)))
tz_form = ''
else:
userprefs = models.get_userprefs()
navbar = ('<p>Welcome, %s! You can <a href="%s">sign out</a>.</p>'
% (user.nickname(), users.create_logout_url(self.request.path)))
tz_form = '''
<form action="/prefs" method="post">
<label for="tz_offset">
Timezone offset from UTC (can be negative):
</label>
<input name="tz_offset" id="tz_offset" type="text"
size="4" value="%d" />
<input type="submit" value="Set" />
</form>
''' % userprefs.tz_offset
time += datetime.timedelta(0, 0, 0, 0, 0, userprefs.tz_offset)
self.response.headers['Content-Type'] = 'text/html'
self.response.out.write('''
<html>
<head>
<title>The Time Is...</title>
</head>
<body>
%s
<p>The time is: %s</p>
%s
</body>
</html>
''' % (navbar, str(time), tz_form))
application = webapp.WSGIApplication([('/', MainPage)],
debug=True)
def main():
run_wsgi_app(application)
if __name__ == '__main__':
main()
To enable the preferences form, we need a request handler to parse the form data and
update the datastore. Let’s implement this as a new request handler script. Create a file
named prefs.py with the contents shown in Example 2-7.
Example 2-7. A new handler script, prefs.py, for the preferences form
from google.appengine.ext import webapp
from google.appengine.ext.webapp.util import run_wsgi_app
import models
class PrefsPage(webapp.RequestHandler):
def post(self):
Developing the Application | 35](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-56-320.jpg)
![userprefs = models.get_userprefs()
try:
tz_offset = int(self.request.get('tz_offset'))
userprefs.tz_offset = tz_offset
userprefs.put()
except ValueError:
# User entered a value that wasn't an integer. Ignore for now.
pass
self.redirect('/')
application = webapp.WSGIApplication([('/prefs', PrefsPage)],
debug=True)
def main():
run_wsgi_app(application)
if __name__ == '__main__':
main()
This request handler handles HTTP POST requests to the URL /prefs, which is the
URL (“action”) and HTTP method used by the form. The handler calls the
get_userprefs() function from models.py to get the UserPrefs object for the current
user, which is either a new unsaved object with default values, or the object for an
existing entity. The handler parses the tz_offset parameter from the form data as an
integer, sets the property of the UserPrefs object, then saves the object to the datastore
by calling its put() method. The put() method creates the object if it doesn’t exist, or
updates the existing object.
If the user enters a noninteger in the form field, we don’t do anything. It’d be appro-
priate to return an error message, but we’ll leave this as is to keep the example simple.
Finally, edit app.yaml to map the handler script to the URL /prefs in the handlers:
section, as shown in Example 2-8.
Example 2-8. A new version of app.yaml mapping the URL /prefs, with login required
application: clock
version: 1
runtime: python
api_version: 1
handlers:
- url: /prefs
script: prefs.py
login: required
- url: /.*
script: main.py
The login: required line says that the user must be signed in to Google Accounts to
access the /prefs URL. If the user accesses the URL while not signed in, App Engine
automatically directs the user to the Google Accounts sign-in page, then redirects her
36 | Chapter 2: Creating an Application](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-57-320.jpg)
![class MainHandler(webapp.RequestHandler):
def get(self):
host = self.request.headers['Host']
self.response.out.write('Host: %s' % host)
App IDs and Versions
Every App Engine application has an application ID that uniquely distinguishes the
app from all other applications. As described in Chapter 2, you can register an ID for
a new application using the Administration Console. Once you have an ID, you add it
to the app’s configuration so the developer tools know that the files in the app root
directory belong to the app with that ID.
The app’s configuration also includes a version identifier. Like the app ID, the version
identifier is associated with the app’s files when the app is uploaded. App Engine retains
one set of files and frontend configuration for each distinct version identifier used dur-
ing an upload. If you do not change the app version in the configuration when you
upload, the existing files for that version of the app are replaced.
Each distinct version of the app is accessible at its own domain name, of the following
form:
version-id.latest.app-id.appspot.com
When you have multiple versions of an app uploaded to App Engine, you can use the
Administration Console to select which version is the one you want the public to access.
The Console calls this the “default” version. When a user visits your Google Apps
domain (and configured subdomain), or the appspot.com domain without the version
ID, he sees the default version.
The appspot.com domain containing the version ID supports an additional domain part,
just like the default appspot.com domain:
anything.version-id.latest.app-id.appspot.com
Unless you explicitly prevent it, anyone who knows your application ID
and version identifiers can access any uploaded version of your appli-
cation using the appspot.com URLs. You can restrict access to nondefault
versions of the application using code that checks the domain of the
request and only allows authorized users to access the versioned do-
mains. You can’t restrict access to static files this way.
Another way to restrict access to nondefault versions is to use Google
Accounts authorization, described later in this chapter. You can restrict
access to app administrators while a version is in development, then
replace the configuration to remove the restriction just before making
that version the default version.
70 | Chapter 3: Handling Web Requests](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-91-320.jpg)
![If you’d like browsers to reload a static file resource automatically every
time you launch a new major version of the app, you can use the mul-
tiversion URL handler just discussed, then use the CURRENT_VERSION_ID
environment variable as the “version” in the static file URLs:
self.response.out('<script src="/js/' +
os.environ['CURRENT_VERSION_ID'] +
'/code.js" />')
Static files in Java
As we saw earlier, the WAR directory structure for a Java web application keeps all
application code, JARs, and configuration in a subdirectory named WEB-INF/. Typi-
cally, files outside of WEB-INF/ represent resources that the user can access directly,
including static files and JSPs. The URL paths to these resources are equivalent to the
paths to these files within the WAR.
Say an app’s WAR has the following files:
main.jsp
forum/home.jsp
images/logo.png
images/cancelbutton.png
images/okbutton.png
terms.html
WEB-INF/classes/com/example/Forum.class
WEB-INF/classes/com/example/MainServlet.class
WEB-INF/classes/com/example/Message.class
WEB-INF/classes/com/example/UserPrefs.class
WEB-INF/lib/appengine-api.jar
This app has four static files: three PNG images and an HTML file named terms.html.
When the app is uploaded, these four files are pushed to the static file servers. The
frontends know to route requests for URL paths equivalent to these file paths (such
as /images/logo.png) to the static file servers.
The two .jsp files are assumed to be JSPs, and are compiled to servlet classes and
mapped to the URL paths equivalent to their file paths. Since these are application
code, they are handled by the application servers. The JSP source files themselves are
not pushed to the static file servers.
By default, all files in the WAR are pushed to the application servers, and are accessible
by the application code via the filesystem. This includes the files that are identified as
static files and pushed to the static file servers. In other words, all files are considered
resource files, and all files except for JSPs and the WEB-INF/ directory are considered
static files.
Configuring the Frontend | 79](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-100-320.jpg)
![In practice, neither method of importing offers much advantage. Late importing may
actually be slower if the import is inside a loop or frequent code path, since importing
an already-imported module takes some time, even if it doesn’t load files or evaluate
module code. It’s best to import modules in the way that makes the most sense for code
clarity, typically at the top of each module (source file) that uses the module. Once
you’ve developed your app, if you notice you have a significant slowdown during im-
ports of the first request, you could consider importing selected modules in their code
paths if you can demonstrate that doing so would skip a lot of unneeded code in a
significant number of cases.
App Engine includes a feature similar to import caching to achieve similar savings for
the handler scripts themselves. In general, when the app server invokes a handler script,
it evaluates all of the script’s code, regardless of whether the script has been invoked
by the app instance in the past. But if the script defines a function named main() that
takes no arguments (or arguments with default values), App Engine treats the handler
script like a module. The first invocation evaluates the complete script as usual, but for
all subsequent invocations, App Engine just calls the main() function.
Consider an early version of the clock application we built in Chapter 2, repeated here
as Example 3-4. This handler script imports three modules, defines three global ele-
ments (a class, a global variable, and the main() function), then calls the main() function.
Example 3-4. A simple request handler illustrating the use of app caching
from google.appengine.ext import webapp
from google.appengine.ext.webapp.util import run_wsgi_app
import datetime
class MainPage(webapp.RequestHandler):
def get(self):
time = datetime.datetime.now()
self.response.headers['Content-Type'] = 'text/html'
self.response.out.write('<p>The time is: %s</p>' % str(time))
application = webapp.WSGIApplication([('/', MainPage)],
debug=True)
def main():
run_wsgi_app(application)
if __name__ == '__main__':
main()
The first time an app instance invokes this script, the full script is evaluated, thereby
importing the modules and defining the global elements in the script’s namespace
within the app instance. Evaluating the script also calls the main() function.
How the App Is Run | 91](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-112-320.jpg)
![A property with the null value is not the same as an unset property. Consider the fol-
lowing Python code:
class Entity(db.Expando):
pass
a = Entity()
a.prop1 = 'abc'
a.prop2 = None
a.put()
b = Entity()
b.prop1 = 'def'
b.put()
This creates two entities of the kind Entity. Both entities have a property named
prop1. The first entity has a property named prop2; the second does not.
Of course, an unset property can be set later:
b.prop2 = 123
b.put()
# b now has a property named "prop2."
Similarly, a set property can be made unset. In the Python API, you delete the property
by deleting the attribute from the object, using the del keyword:
del b.prop2
b.put()
# b no longer has a property named "prop2."
In Java, the low-level datastore API’s Entity class has methods to set properties
(setProperty()) and unset properties (removeProperty()).
Multivalued Properties
As we mentioned earlier, a property can have multiple values. We’ll discuss the more
substantial aspects of multivalued properties when we talk about queries and data
modeling. But for now, it’s worth a brief mention.
A property can have one or more values. A property cannot have zero values; a property
without a value is simply unset. Each value for a property can be of a different type,
and can be the null value.
The datastore preserves the order of values as they are assigned. The Python API returns
the values in the same order as they were set.
In Python, a property with multiple values is represented as a single Python list value:
e.prop = [1, 2, 'a', None, 'b']
Property Values | 113](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-134-320.jpg)
![Because a property must have at least one value, it is an error to assign
an empty list ([] in Python) to a property on an entity whose Python
class is based on the Expando class.
class Entity(db.Expando):
pass
e = Entity()
e.prop = [] # ERROR
In contrast, the Model base class includes a feature that automatically
translates between the empty list value and “no property set.” You’ll see
this feature in Chapter 7.
In the Java low-level datastore API, you can store multiple values for a property using
a Collection type. The low-level API returns the values as a java.util.List. The items
are stored in the order provided by the Collection type’s iterator. For many types, such
as SortedSet or TreeSet, this order is deterministic. For others, such as HashSet, it is
not. If the app needs the original data structure, it must convert the List returned by
the datastore to the appropriate type.
Keys and Key Objects
The key for an entity is a value that can be retrieved, passed around, and stored like
any other value. If you have the key for an entity, you can retrieve the entity from the
datastore quickly, much more quickly than with a datastore query. Keys can be stored
as property values, as an easy way for one entity to refer to another.
The Python API represents an entity key value as an instance of the Key class, in
the db package. To get the key for an entity, you call the entity object’s key() method.
The Key instance provides access to its several parts using accessor methods, including
the kind, key name (if any), and system-assigned ID (if the entity does not have a key
name).
The Java low-level API is similar: the getKey() method of the Entity class returns an
instance of the Key class.
When you construct a new entity object and do not provide a key name, the entity
object has a key, but the key does not yet have an ID. The ID is populated when the
entity object is saved to the datastore for the first time. You can get the key object prior
to saving the object, but it will be incomplete.
e = Entity()
e.prop = 123
k = e.key() # key is incomplete, has neither key name nor ID
kind = k.kind() # 'Entity'
e.put() # ID is assigned
114 | Chapter 4: Datastore Entities](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-135-320.jpg)
![k_str = str(k)
# ...
k = db.Key(k_str)
And in Java:
String k_str = KeyFactory.keyToString(k);
// ...
Key k = KeyFactory.stringToKey(k_str);
The Java Key class’s toString() method does not return the key’s string encoding. You
must use KeyFactory.keyToString() to get the string encoding of a key.
Using Entities
Let’s look briefly at how to retrieve entities from the datastore using keys, how to inspect
the contents of entities, and how to update and delete entities. The API methods for
these features are straightforward.
Getting Entities Using Keys
Given a complete key for an entity, you can retrieve the entity from the datastore.
In the Python API, you can call the get() function in the db package with the Key object
as an argument:
from google.appengine.ext import db
k = db.Key('Entity', 'alphabeta')
e = db.get(k)
If you know the kind of the entity you are fetching, you can also use the get() class
method on the appropriate entity class. This does a bit of type checking, ensuring that
the key you provide is of the appropriate kind:
class Entity(db.Expando):
pass
e = Entity.get(k)
To fetch multiple entities in a batch, you can pass the keys to get() as a list. Given a
list, the method returns a list containing entity objects, with None values for keys that
do not have a corresponding entity in the datastore.
entities = db.get([k1, k2, k3])
116 | Chapter 4: Datastore Entities](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-137-320.jpg)
![complete key (such as a key with a kind and a key name) and an entity already exists
with that key, the datastore replaces the existing entity with the new one.
If you want to test that an entity with a given key does not exist before
you create it, you can do so using the transaction API. You must use a
transaction to ensure that another process doesn’t create an entity with
that key after you test for it and before you create it. For more informa-
tion on transactions, see Chapter 6.
If you have several entity objects to save, you can save them all in one call using the
put() function in the db package. The put() function can also take a single entity object.
db.put(e)
db.put([e1, e2, e3])
As with a batch get, a batch put performs a single call to the service. The total size of
the call is subject to the API call limits for the datastore. The entity count is also subject
to a limit.
In Java, you can save entities using the put() method of a DatastoreService
instance. As with get(), the method takes a single Entity for a single put, or an
Iterable<Entity> for a batch put.
When the call to put() returns, the datastore is up to date, and all future queries in the
current request handler and other handlers will see the new data. The specifics of how
the datastore gets updated are discussed in detail in Chapter 6.
Deleting Entities
Deleting entities works similarly to putting entities. In Python, you can call the
delete() method on the entity object, or you can pass entity objects or Key objects to
the delete() function.
e = db.get('Entity', 'alphabeta')
e.delete()
db.delete(e)
db.delete([e1, e2, e3])
# Deleting without first fetching the entity:
k = db.Key('Entity', 'alphabeta')
db.delete(k)
In Java, you call the delete() method of the DatastoreService with either a single Key
or an Iterable<Key>.
As with gets and puts, a delete of multiple entities occurs in a single batch call to the
service, and is faster than making multiple service calls.
Using Entities | 119](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-140-320.jpg)
![impossible to get the list of players for a particular trophy. (A query filter can’t match
patterns within string values.)
Another option is to record each trophy win in a separate property named after the
trophy. To get the list of players with a trophy, you just query for the existence of the
corresponding property. However, getting the list of trophies for a given player would
require either coding the names of all of the trophies in the display logic, or iterating
over all of the Player entity’s properties looking for trophy names.
With multivalued properties, we can store each trophy name as a separate value for the
trophies property. To access a list of all trophies for a player, we simply access the
property of the entity. To get a list of all players with a trophy, we use a query with an
equality filter on the property.
Here’s what this example looks like in Python:
p = Player.get_by_key_name(user_id)
p.trophies = ['Lava Polo Champion',
'World Building 2008, Bronze',
'Glarcon Fighter, 2nd class']
p.put()
# List all trophies for a player.
for trophy in p.trophies:
# ...
# Query all players that have a trophy.
q = Player.gql("WHERE trophies = 'Lava Polo Champion'")
for p in q:
# ...
MVPs in Python
The Python API represents the values of a multivalued property as a Python list. Each
member of the list must be of one of the supported datastore types.
class Entity(db.Expando):
pass
e = Entity()
e.prop = [ 'value1', 123, users.get_current_user() ]
Remember that list is not a datastore type; it is only the mechanism for manipulating
multivalued properties. A list cannot contain another list.
A property must have at least one value, otherwise the property does not exist. To
enforce this, the Python API does not allow you to assign an empty list to a property.
Notice that the API can’t do otherwise: if a property doesn’t exist, then the API cannot
know to represent the missing property as an empty list when the entity is retrieved
from the datastore. (This being Python, the API could return the empty list whenever
a nonexistent property is accessed, but that might be more trouble than it’s worth.)
154 | Chapter 5: Datastore Queries](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-175-320.jpg)
![Because it is often useful for lists to behave like lists, including the ability to contain
zero items, the Python data modeling API provides a mechanism that supports assign-
ing the empty list to a property. We’ll look at this mechanism in Chapter 7.
MVPs and Equality Filters
As you’ve seen, when a multivalued property is the subject of an equality filter in a
query, the entity matches if any of the property’s values are equal to the filter value.
e1 = Entity()
e1.prop = [ 3.14, 'a', 'b' ]
e1.put()
e2 = Entity()
e2.prop = [ 'a', 1, 6 ]
e2.put()
# Returns e1 but not e2:
q = Entity.gql('WHERE prop = 3.14')
# Returns e2 but not e1:
q = Entity.gql('WHERE prop = 6')
# Returns both e1 and e2:
q = Entity.gql("WHERE prop = 'a'")
Recall that a query with a single equality filter uses an index that contains the keys of
every entity of the given kind with the given property and the property values. If an
entity has a single value for the property, the index contains one row that represents
the entity and the value. If an entity has multiple values for the property, the index
contains one row for each value. The index for this example is shown in Figure 5-14.
This brings us to the first of several odd-looking queries that nonetheless make sense
for multivalued properties. Since an equality filter is a membership test, it is possible
for multiple equality filters to use the same property with different values and still return
a result. An example in GQL:
SELECT * FROM Entity WHERE prop = 'a' AND prop = 'b'
App Engine uses the “merge join” algorithm described earlier for multiple equality
filters to satisfy this query using the prop single-property index. This query returns the
e1 entity because the entity key appears in two places in the index, once for each value
requested by the filters.
The way multivalued properties appear in an index gives us another way of thinking
about multivalued properties: an entity has one or more properties, each with a name
and a single value, and an entity can have multiple properties with the same name. The
API represents the values of multiple properties with the same name as a list of values
associated with that name.
Queries and Multivalued Properties | 155](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-176-320.jpg)
![Figure 5-14. An index of two entities with multiple values for the “prop” property, with results for
WHERE prop=‘a’
The datastore does not have a way to query for the exact set of values in a multivalued
property. You can use multiple equality filters to test that each of several values belongs
to the list, but there is no filter that ensures that those are the only values that belong
to the list, or that each value appears only once.
MVPs and Inequality Filters
Just as an equality filter tests that any value of the property is equal to the filter value,
an inequality filter tests that any value of the property meets the filter criterion.
e1 = Entity()
e1.prop = [ 1, 3, 5 ]
e1.put()
e2 = Entity()
e2.prop = [ 4, 6, 8 ]
e2.put()
# Returns e1 but not e2:
q = Entity.gql("WHERE prop < 2")
# Returns e2 but not e1:
q = Entity.gql("WHERE prop > 7")
# Returns both e1 and e2:
q = Entity.gql("WHERE prop > 3")
Figure 5-15 shows the index for this example, with the results of prop > 3 highlighted.
In the case of an inequality filter, it’s possible for the index scan to match rows for a
single entity multiple times. When this happens, the first occurrence of each key in the
156 | Chapter 5: Datastore Queries](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-177-320.jpg)
![Figure 5-15. An index of two entities with multiple values for the “prop” property, with results for
WHERE prop > 3
index determines the order of the results. If the index used for the query sorts the
property in ascending order, the first occurrence is the smallest matching value. For
descending, it’s the largest. In the example above, prop > 3 returns e2 before e1 because
4 appears before 5 in the index.
MVPs and Sort Orders
To summarize things we know about how multivalued properties are indexed:
• A multivalued property appears in an index with one row per value.
• All rows in an index are sorted by the values, possibly distributing property values
for a single entity across the index.
• The first occurrence of an entity in an index scan determines its place in the result
set for a query.
Together, these facts explain what happens when a query orders its results by a mul-
tivalued property. When results are sorted by a multivalued property in ascending or-
der, the smallest value for the property determines its location in the results. When
results are sorted in descending order, the largest value for the property determines its
location.
This has a counterintuitive—but consistent—consequence:
e1 = Entity()
e1.prop = [ 1, 3, 5 ]
e1.put()
e2 = Entity()
Queries and Multivalued Properties | 157](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-178-320.jpg)
![e2.prop = [ 2, 3, 4 ]
e2.put()
# Returns e1, e2:
q = Entity.gql("ORDER BY prop ASC")
# Also returns e1, e2:
q = Entity.gql("ORDER BY prop DESC")
Because e1 has both the smallest value and the largest value, it appears first in the result
set in ascending order and in descending order. See Figure 5-16.
Figure 5-16. Indexes of two entities with multiple values for the “prop” property, one ascending and
one descending
MVPs and the Query Planner
The query planner tries to be smart by ignoring aspects of the query that are redundant
or contradictory. For instance, a = 3 AND a = 4 would normally return no results, so
the query planner catches those cases and doesn’t bother doing work it doesn’t need
to do. However, most of these normalization techniques don’t apply to multivalued
properties. In this case, the query could be asking, “Does this MVP have a value that
is equal to 3 and another value equal to 4?” The datastore remembers which properties
are MVPs (even those that end up with one or zero values), and never takes a shortcut
that would produce incorrect results.
But there is one exception. A query that has both an equality filter and a sort order will
drop the sort order. If a query asks for a = 3 ORDER BY a DESC and a is a single-value
property, the sort order has no effect because all of the values in the result are identical.
For an MVP, however, a = 3 tests for membership, and two MVPs that meet that
condition are not necessarily identical.
158 | Chapter 5: Datastore Queries](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-179-320.jpg)
![SELECT * FROM Message
WHERE ANCESTOR IS KEY(MessageBoard, 'The_Archonville_Times')
ORDER BY post_date DESC
LIMIT 10
Most queries that use an ancestor filter need custom indexes. There is one unusual
exception: a query does not need a custom index if the query also contains equality
filters on properties (and no inequality filters or sort orders). In this exceptional case,
the “merge join” algorithm can use a built-in index of keys along with the built-in
property indexes. In cases where the query would need a custom index anyway, the
query can match the ancestor to the keys in the custom index.
As we mentioned in Chapter 5, the datastore supports queries over entities of all kinds.
Kindless queries are limited to key filters and ancestor filters. Since ancestors can have
children of disparate kinds, kindless queries are useful for getting every child of a given
ancestor, regardless of kind:
SELECT * WHERE ANCESTOR IS KEY('MessageBoard', 'The_Archonville_Times')
While ancestor queries can be useful, don’t get carried away building
large ancestor trees. Remember that every entity with the same root
belongs to the same entity group, and more simultaneous users that
need to write to a group mean a greater likelihood of concurrency
failures.
If you want to model hierarchical relationships between entities without
the consequences of entity groups, consider using multivalued proper-
ties to store paths. For example, if there’s an entity whose path in your
hierarchy can be represented as /A/B/C/D, you can store this path as:
e.parents = ['/A', '/A/B', '/A/B/C'] Then you can perform a query
similar to an ancestor query on this property: ... WHERE parents = '/
A/B'.
What Can Happen in a Transaction
Entity groups ensure that the operations performed within a transaction see a consistent
view of the entities in a group. For this to work, a single transaction must limit its
operations to entities in a single group. The entity group determines the scope of the
transaction.
Within a transaction, you can fetch, update, or delete an entity using the entity’s key.
You can also create a new entity that either is a root entity of a new group that becomes
the subject of the transaction, or that has a member of the transaction’s entity group
as its parent.
You can perform queries over the entities of a single entity group in a transaction. A
query in a transaction must have an ancestor filter that matches the transaction’s entity
group. The results of the query, including both the indexes that provide the results as
168 | Chapter 6: Datastore Transactions](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-189-320.jpg)
![p = Entity()
e = Entity(parent=p)
The parent key does not have to represent an entity that exists. You can construct a
fake key using db.Key.from_path() and use it as the parent:
p_key = db.Key.from_path('Entity', 'fake_parent')
e = Entity(parent=p_key)
You can use methods on the model object to get the key for the entity’s parent, or fetch
the entity’s parent from the datastore. These methods return None if the entity is a root
entity for its entity group.
p = e.parent()
p_key = e.parent_key()
Similarly, the Key object can return the Key of its parent, or None:
e_key = e.key()
p_key = e_key.parent()
The Python API uses function objects to handle transactions. To perform multiple
operations in a transaction, you define a function that executes the operations, then
you pass it to the run_in_transaction() function (in the ...ext.db module).
import datetime
from google.appengine.ext import db
class MessageBoard(db.Expando):
pass
class Message(db.Expando):
pass
def create_message_txn(board_name, message_name, message_title, message_text):
board = db.get(db.Key.from_path('MessageBoard', board_name))
if not board:
board = MessageBoard(key_name=board_name)
board.count = 0
message = Message(key_name=message_name, parent=board)
message.title = message_title
message.text = message_text
message.post_date = datetime.datetime.now()
board.count += 1
db.put([board, message])
# ...
try:
db.run_in_transaction(create_message_txn,
170 | Chapter 6: Datastore Transactions](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-191-320.jpg)
![Batch Updates
When you read, create, update, or delete an entity, the runtime environment makes a
service call to the datastore. Each service call has some overhead, including serializing
and deserializing parameters and transmitting them between machines in the data cen-
ter. If you need to update multiple entities, you can save time by performing the updates
together as a batch in one service call.
We introduced batch calls in Chapter 4. Here’s a quick example of the Python batch
API:
# Creating multiple entities:
e1 = Message(key_name='m1', text='...')
e2 = Message(key_name='m2', text='...')
e3 = Message(key_name='m3', text='...')
message_keys = db.put([e1, e2, e3])
# Getting multiple entities using keys:
message_keys = [db.Key('Message', 'm1'),
db.Key('Message', 'm2'),
db.Key('Message', 'm2')]
messages = db.get(message_keys)
for message in messages:
# ...
# Deleting multiple entities:
db.delete(message_keys)
When the datastore receives a batch call, it bundles the keys or entities by their entity
groups, which it can determine from the keys. Then it dispatches calls to the datastore
machines responsible for each entity group. The datastore returns results to the app
when it has received all results from all machines.
If the call includes changes for multiple entities in a single entity group, those changes
are performed in a single transaction. There is no way to control this behavior, but
there’s no reason to do it any other way. It’s faster to commit multiple changes to a
group at once than to commit them individually, and no less likely to result in concur-
rency failures.
Each entity group involved in a batch update may fail to commit due to a concurrency
failure. If a concurrency failure occurs for any update, the API raises the concurrency
failure exception—even if updates to other groups were committed successfully.
Batch updates in disparate entity groups are performed in separate threads, possibly
by separate datastore machines, executed in parallel to one another. This can make
batch updates especially fast compared to performing each update one at a time.
Remember that if you use the batch API during a transaction, every entity or key in the
batch must use the same entity group as the rest of the transaction. Also, be aware that
App Engine puts a maximum size on service calls and responses, and the total size of
the call or its responses must not exceed these limits. The number of items in a batch
Batch Updates | 179](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-200-320.jpg)
![s.song_key = 'C# min'
# The song_key attribute is stored as the
# datastore property named 'key'.
s.put()
You can declare that a property should contain only one of a fixed set of values by
providing a list of possible values as the choices argument. If None is not one of the
choices, this acts as a more restrictive form of required: the property must be set to one
of the valid choices using a keyword argument to the constructor.
_KEYS = ['C', 'C min', 'C 7',
'C#', 'C# min', 'C# 7',
# ...
]
class Song(db.Model):
song_key = db.StringProperty(choices=_KEYS)
s = Song(song_key='H min') # db.BadValueError
s = Song() # db.BadValueError, None is not an option
s = Song(song_key='C# min') # OK
All of these features validate the value assigned to a property, and raise a
db.BadValueError if the value does not meet the appropriate conditions. For even
greater control over value validation, you can define your own validation function and
assign it to a property declaration as the validator argument. The function should take
the value as an argument, and raise a db.BadValueError (or an exception of your choos-
ing) if the value should not be allowed.
def is_recent_year(val):
if val < 1923:
raise db.BadValueError
class Book(db.Model):
copyright_year = db.IntegerProperty(validator=is_recent_year)
b = Book(copyright_year=1922) # db.BadValueError
b = Book(copyright_year=1924) # OK
Nonindexed Properties
In Chapter 5, we mentioned that you can set properties of an entity in such a way that
they are available on the entity, but are considered unset for the purposes of indexes.
In the Python API, you establish a property as nonindexed using a property declaration.
If the property declaration is given an indexed argument of False, entities created with
that model class will set that property as nonindexed.
188 | Chapter 7: Data Modeling with Python](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-209-320.jpg)
![is saved for the first time, and left untouched thereafter. If the current user is not signed
in, the value is set to None.
class BookReview(db.Model):
created_by_user = db.UserProperty(auto_current_user_add=True)
last_edited_by_user = db.UserProperty(auto_current_user=True)
br = BookReview()
br.put() # created_by_user and last_edited_by_user set
# ...
br.put() # last_edited_by_user set again
At first glance, it might seem reasonable to set a default for a
db.UserProperty this way:
from google.appengine.api import users
class BookReview(db.Model):
created_by_user = db.UserProperty(
default=users.get_current_user())
# WRONG
This would set the default value to be the user who is signed in when
the class is imported. Requests that use a cached instance of the appli-
cation may end up using a previous user instead of the current user as
the default.
To guard against this mistake, db.UserProperty does not accept
the default argument. You can use only auto_current_user or
auto_current_user_add to set an automatic value.
List Properties
The data modeling API provides a property declaration class for multivalued properties,
called db.ListProperty. This class ensures that every value for the property is of the
same type. You pass this type to the property declaration, like so:
class Book(db.Model):
tags = db.ListProperty(basestring)
b = Book()
b.tags = ['python', 'app engine', 'data']
The type argument to the db.ListProperty constructor must be the Python represen-
tation of one of the native datastore types. Refer back to Table 4-1 for a complete list.
The datastore does not distinguish between a multivalued property with no elements
and no property at all. As such, an undeclared property on a db.Expando object can’t
store the empty list. If it did, when the entity is loaded back into an object, the property
simply wouldn’t be there, potentially confusing code that’s expecting to find an empty
190 | Chapter 7: Data Modeling with Python](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-211-320.jpg)
![list. To avoid confusion, db.Expando disallows assigning an empty list to an undeclared
property.
The db.ListProperty declaration makes it possible to keep an empty list value on a
multivalued property. The declaration interprets the state of an entity that doesn’t have
the declared property as the property being set to the empty list, and maintains that
distinction on the object. This also means that you cannot assign None to a declared list
property—but this isn’t of the expected type for the property anyway.
The datastore does distinguish between a property with a single value and a multivalued
property with a single value. An undeclared property on a db.Expando object can store
a list with one element, and represent it as a list value the next time the entity is loaded.
The example above declares a list of string values. (basestring is the Python base
type for str and unicode.) This case is so common that the API also provides
db.StringListProperty.
You can provide a default value to db.ListProperty using the default argument. If you
specify a nonempty list as the default, a shallow copy of the list value is made for each
new object that doesn’t have an initial value for the property.
db.ListProperty does not support the required validator, since every list property tech-
nically has a list value (possibly empty). If you wish to disallow the empty list, you can
provide your own validator function that does so:
def is_not_empty(lst):
if len(lst) == 0:
raise db.BadValueError
class Book(db.Model):
tags = db.ListProperty(basestring, validator=is_not_empty)
b = Book(tags=[]) # db.BadValueError
b = Book() # db.BadValueError, default "tags" is empty
b = Book(tags=['awesome']) # OK
db.ListProperty does not allow None as an element in the list because it doesn’t match
the required value type. It is possible to store None as an element in a list for an unde-
clared property.
Models and Schema Migration
Property declarations prevent the application from creating an invalid data object, or
assigning an invalid value to a property. If the application always uses the same model
classes to create and manipulate entities, then all entities in the datastore will be con-
sistent with the rules you establish using property declarations.
Property Declarations | 191](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-212-320.jpg)
![b.put()
br = BookReview()
br.book = b # sets br's 'book' property to b's key
br.book = b.key() # same thing
We’ll explain what collection_name does in a moment.
The referenced object must have a “complete” key before it can be assigned to a refer-
ence property. A key is complete when it has all of its parts, including the string name
or the system-assigned numeric ID. If you create a new object without a key name, the
key is not complete until you save the object. When you save the object, the system
completes the key with a numeric ID. If you create the object (or a db.Key) with a key
name, the key is already complete, and you can use it for a reference without saving it
first.
b = Book()
br = BookReview()
br.book = b # db.BadValueError, b's key is not complete
b.put()
br.book = b # OK, b's key has system ID
b = Book(key_name='The_Grapes_of_Wrath')
br = BookReview()
br.book = b # OK, b's key has a name
db.put([b, br])
A model class must be defined before it can be the subject of a db.ReferenceProperty.
To declare a reference property that can refer to another instance of the same class, you
use a different declaration, db.SelfReferenceProperty.
class Book(db.Model):
previous_edition = db.SelfReferenceProperty()
b1 = Book()
b2 = Book()
b2.previous_edition = b1
Reference properties have a powerful and intuitive syntax for accessing referenced ob-
jects. When you access the value of a reference property, the property fetches the entity
from the datastore using the stored key, then returns it as an instance of its model class.
A referenced entity is loaded “lazily”: it is not fetched from the datastore until the
property is dereferenced.
br = db.get(book_review_key)
# br is a BookReview instance
title = br.book.title # fetches book, gets its title property
Modeling Relationships | 193](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-214-320.jpg)
![The key list method is best suited for situations where there are fewer objects on one
side of the relationship than on the other, and the short list is small enough to store
directly on an entity. In this example, many players each belong to a few guilds; each
player has a short list of guilds, while each guild may have a long list of players. We
put the list property on the Player side of the relationship to keep the entity small, and
use queries to produce the long list when it is needed.
The link model method
The link model method represents each relationship as an entity. The relationship entity
has reference properties pointing to the related classes. You traverse the relationship
by going through the relationship entity via the back-references.
class Player(db.Model):
name = db.StringProperty()
class Guild(db.Model):
name = db.StringProperty()
class GuildMembership(db.Model):
player = db.ReferenceProperty(Player, collection_name='guild_memberships')
guild = db.ReferenceProperty(Guild, collection_name='player_memberships')
p = Player()
g = Guild()
db.put([p, g])
gm = GuildMembership(player=p, guild=g)
db.put(gm)
# Guilds to which a player belongs:
for gm in p.guild_memberships:
guild_name = gm.guild.name
# ...
# Players that belong to a guild:
for gm in g.player_memberships:
player_name = gm.player.name
# ...
This technique is similar to how you’d use “join tables” in a SQL database. It’s a good
choice if either side of the relationship may get too large to store on the entity itself.
You can also use the relationship entity to store metadata about the relationship (such
as when the player joined the guild), or model more complex relationships between
multiple classes.
The link model method is more expensive than the key list method. It requires fetching
the relationship entity to access the related object.
Modeling Relationships | 197](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-218-320.jpg)
![self.first_name = first_name
self.surname = surname
def is_valid(self):
return (isinstance(self.first_name, unicode)
and isinstance(self.surname, unicode)
and len(self.surname) >= 6)
class PlayerNameProperty(db.Property):
data_type = basestring
def validate(self, value):
value = super(PlayerNameProperty, self).validate(value)
if value is not None:
if not isinstance(value, PlayerName):
raise db.BadValueError('Property %s must be a PlayerName.' %
(self.name))
# Let the data class have a say in validity.
if not value.is_valid():
raise db.BadValueError('Property %s must be a valid PlayerName.' %
self.name)
# Disallow the serialization delimiter in the first field.
if value.surname.find('|') != -1:
raise db.BadValueError(('PlayerName surname in property %s cannot ' +
'contain a "|".') % self.name)
return value
def get_value_for_datastore(self, model_instance):
# Convert the data object's PlayerName to a unicode.
return (getattr(model_instance, self.name).surname + u'|'
+ getattr(model_instance, self.name).first_name)
def make_value_for_datastore(self, value):
# Convert a unicode to a PlayerName.
i = value.find(u'|')
return PlayerName(first_name=value[i+1:],
surname=value[:i])
And here’s how you’d use it:
p = Player()
p.player_name = PlayerName('Ned', 'Nederlander')
p.player_name = PlayerName('Ned', 'Neder|lander')
# db.BadValueError, surname contains serialization delimiter
p.player_name = PlayerName('Ned', 'Neder')
# db.BadValueError, PlayerName.is_valid() == False
Here, the application value type is a PlayerName instance, and the datastore value type
is that value encoded as a Unicode string. The encoding format is the surname field,
followed by a delimiter, followed by the first_name field. We disallow the delimiter
character in the surname using the validate() method. (Instead of disallowing
Creating Your Own Property Classes | 203](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-224-320.jpg)
![selects the requested fields and assembles the results. This is only true if one of the
requested fields is a datastore property, and is not true if the only field is a key field (@Id).
If the query is for one field, each result is a value of the type of that field. If the query
is for multiple fields, each result is an Object[] whose elements are the field values in
the order specified in the query.
Query query = em.createQuery("SELECT isbn, title, author FROM Book");
// Fetch complete Book objects, then
// produce result objects from 3 fields
// of each result
@SuppressWarnings("unchecked")
List<Object[]> results = (List<Object[]>) query.getResultList();
for (Object[] result : results) {
String isbn = result[0];
String title = result[1];
String author = result[2];
// ...
}
You specify filters on fields using a WHERE clause and one or more conditions:
SELECT b FROM Book b WHERE author = "John Steinbeck"
AND copyrightYear >= 1940
To filter on the entity key, refer to the field that represents the key in the data class (the
@Id field):
SELECT b FROM Book b WHERE author = "John Steinbeck"
AND isbn > :firstKeyToFetch
You can perform an ancestor filter by establishing a parent key field (as we did in the
previous section) and referring to that field in the query:
SELECT br FROM BookReview br WHERE bookKey = :pk
As with find(), you can use any of the four key types with parameterized queries, but
the most portable way is to use the type used in the class.
You specify sort orders using an ORDER BY clause. Multiple sort orders are comma-
delimited. Each sort order can have a direction of ASC (the default) or DESC.
SELECT b FROM Book b ORDER BY rating DESC title
The App Engine implementation of JPQL includes a couple of additional tricks that
the datastore can support natively. One such trick is the string prefix trick:
SELECT b FROM Book b WHERE title LIKE 'The Grape%'
The implementation translates this to WHERE title >= 'The Grape', which does the
same thing: it returns all books with a title that begins with the string The Grape, in-
cluding "The Grape", "The Grapefruit", and "The Grapes of Wrath".
Queries and JPQL | 219](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-240-320.jpg)
![Keys are strings. The API also accepts a memcached-style tuple of a hash code and a string
as a key for compatibility, but it ignores the hash code. The string can be any size;
strings larger than 250 bytes are hashed to 250 bytes.
Values can be any Python value that can be serialized using the pickle module. This
includes most simple values, and data structures and objects whose members are
pickleable values. Pretty much any db.Model instance falls into this category. You can
check the documentation for the pickle module for a more complete description.
Setting and Getting Values in Python
To store a value in memcache, you call the set() function (or method, if you’re using
the Client class) with the key and the value:
from google.appengine.api import memcache
# headlines is a pickle-able object
headlines = ['...', '...', '...']
success = memcache.set("headlines", headlines)
if not success:
# Problem accessing memcache...
set() returns True on success, or False if the value could not be stored.
If there already exists a value with the given key, set() overwrites it. If there is no value
for the key, set() sets it.
To set a value only if there is no value for the key, you call the add() function with
similar arguments:
success = memcache.add("headlines", headlines)
if not success:
# Key already set, or another problem setting...
add() works like set(), but does not update the value and returns False if a value is
currently set for the given key.
To update a value only if it’s already set, you call the replace() function, also with
similar arguments:
success = memcache.replace("headlines", headlines)
if not success:
# Key not set, or another problem setting...
To get a value, you call the get() method with the key:
headlines = memcache.get("headlines")
if headlines is None:
# Key not set, or another problem getting...
get() returns the value, or None if there is no value set for the given key.
The Python Memcache API | 229](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-250-320.jpg)
![Setting and Getting Multiple Values
You can set and get multiple values at a time using the batch interface. A batch call to
memcache is performed as a single call to the service, which is faster than multiple
individual calls for many items. Like the datastore batch operations, the total amount
of data sent to or received from memcache in a single call must not exceed one
megabyte.
set_multi(), add_multi(), and replace_multi() are similar to their singular counter-
parts, except instead of taking a key and a value as arguments, they take a Python
mapping (such as a dict) of keys to values.
article_summaries = {'article00174': '...',
'article05234': '...',
'article15820': '...',
}
failed_keys = memcache.set_multi(article_summaries)
if failed_keys:
# One or more keys failed to update...
These methods return a list of keys that failed to be set. A completely successful batch
update returns the empty list.
To get multiple values in a single batch, you call the get_multi() method with a list of
keys:
article_summary_keys = ['article00174',
'article05234',
'article15820',
]
article_summaries = memcache.get_multi(article_summary_keys)
get_multi() returns a dict of keys and values that were present and retrieved
successfully.
All of the batch operations accept a key_prefix keyword argument that prepends every
key with the given prefix string. When used with get_multi(), the keys on the returned
values do not include the prefix.
article_summaries = {'00174': '...',
'05234': '...',
'15820': '...',
}
failed_keys = memcache.set_multi(article_summaries,
key_prefix='article')
if failed_keys:
# ...
# ...
article_summary_keys = ['00174',
230 | Chapter 9: The Memory Cache](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-251-320.jpg)
!['05234',
'15820',
]
article_summaries = memcache.get_multi(article_summary_keys,
key_prefix='article')
Batch calls to the memcache service are not atomic actions.
set_multi() may be successful for some items and not others.
get_multi() will return the data seen at the time of each individual get,
and does not guarantee that all retrieved data is read at the same time.
Memcache Namespaces
You can specify an optional namespace for a memcache key with any setter or getter
call. Namespaces let you segment the key space by the kind of thing you are storing,
so you don’t have to worry about key collisions between disparate uses of the cache.
All of the setter and getter functions accept a namespace argument, a string that declares
the namespace for all keys involved in the call. The batch functions use one namespace
for all keys in the batch.
article_summaries = {'article00174': '...',
'article05234': '...',
'article15820': '...',
}
failed_keys = memcache.set_multi(article_summaries,
namespace='News')
if failed_keys:
# ...
# ...
summary = memcache.get('article05234',
namespace='News')
Namespaces are different from key prefixes in batch calls. Key prefixes merely prepend
keys in a batch with a common string, as a convenience for batch processing. The
namespace is a separate part of the key, and must be specified with an explicit
namespace argument in all calls.
Cache Expiration
You can tell the memcache to delete (or evict) a value automatically after a period of
time when you set or update the value.
All of the setter methods accept a time argument. Its value is either a number of seconds
from the current time up to one month (up to 2,678,400 seconds), or it is a Unix epoch
time (seconds since January 1, 1970, such as returned by time.time()) in the future.
The Python Memcache API | 231](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-252-320.jpg)
![The callback function is called without arguments. But a common use for a callback
function is to process the results of the fetch, so the function needs access to the RPC
object. There are several ways to give the callback function access to the object.
One way is to use a bound method, a feature of Python that lets you refer to a method
of an instance of a class as a callable object. Define a class with a method that processes
the results of the fetch, using an RPC object stored as a member of the class. Create an
instance of the class, then create the RPC object, assigning the bound method as the
callback by passing it as the callback argument to the create_rpc() function. Exam-
ple 10-6 demonstrates this technique.
Example 10-6. Using an object method as a callback to access the RPC object
from google.appengine.api import urlfetch
# ...
class CatalogUpdater(object):
def prepare_urlfetch_rpc(self):
self.rpc = urlfetch.create_rpc(callback=self.process_results)
urlfetch.make_fetch_call(self.rpc, 'http://api.example.com/catalog_feed')
return self.rpc
def process_results(self):
try:
results = self.rpc.get_result()
# Process results.content...
except urlfetch.Error, e:
# Handle urlfetch errors...
class MainHandler(webapp.RequestHandler):
def get(self):
rpcs = []
catalog_updater = CatalogUpdater(self.response)
rpcs.append(catalog_updater.prepare_urlfetch_rpc())
# ...
for rpc in rpcs:
rpc.wait()
Another way to give the callback access to the RPC object is to use a nested function
(sometimes called a closure). If the callback function is defined in the same scope as a
variable whose value is the RPC object, the function can access the variable when it is
called. To allow for this style of calling, the URL Fetch API lets you assign the callback
function to the RPC object’s callback property after the object has been created.
Example 10-7 demonstrates the use of a nested function as a callback. The
create_callback() function creates a function object, a lambda expression, that calls
another function with the RPC object as an argument. This function object is assigned
to the callback property of the RPC object.
248 | Chapter 10: Fetching URLs and Web Resources](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-269-320.jpg)
![Example 11-1. An example of sending an email message in Python, using several features
from google.appengine.api import mail
def send_registration_key(user_addr, software_key_data):
message_body = '''
Thank you for purchasing The Example App, the best
example on the market! Your registration key is attached
to this email.
To install your key, download the attachment, then select
"Register..." from the Help menu. Select the key file, then
click "Register".
You can download the app at any time from:
http://www.example.com/downloads/
Thanks again!
The Example Team
'''
html_message_body = '''
<p>Thank you for purchasing The Example App, the best
example on the market! Your registration key is attached
to this email.</p>
<p>To install your key, download the attachment, then select
<b>Register...</b> from the <b>Help</b> menu. Select the key file, then
click <b>Register</b>.</p>
<p>You can download the app at any time from:</p>
<p>
<a href="http://www.example.com/downloads/">
http://www.example.com/downloads/
</a>
</p>
<p>Thanks again!</p>
<p>The Example Team<br />
<img src="http://www.example.com/images/logo_email.gif" /></p>
'''
message = mail.EmailMessage(
sender='The Example Team <admin@example.com>',
to=user_addr,
subject='Your Example Registration Key',
body=message_body,
html=html_message_body,
attachments=[('example_key.txt', software_key_data)])
message.send()
260 | Chapter 11: Sending and Receiving Mail and Instant Messages](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-281-320.jpg)
![// There was an error contacting the Mail service.
// ...
}
As shown here, you call methods on the MimeMessage to set fields and to add recipients
and content. The simplest message has a sender (setFrom()), one “To” recipient
(addRecipient()), a subject (setSubject()), and a plain-text message body (setText()).
The setFrom() method takes an InternetAddress. You can create an InternetAddress
with just the email address (a String) or the address and a human-readable name as
arguments to the constructor. The email address of the sender must meet the require-
ments described earlier. You can use any string for the human-readable name.
The addRecipient() method takes a recipient type and an InternetAddress. The
allowed recipient types are Message.RecipientType.TO (“To,” a primary recipient),
Message.RecipientType.CC (“Cc” or “carbon-copy,” a secondary recipient), and
Message.RecipientType.BCC (“Bcc” or “blind carbon-copy,” where the recipient is sent
the message but the address does not appear in the message content). You can call
addRecipient() multiple times to add multiple recipients of any type.
The setText() method sets the plain-text body for the message. To include an
HTML version of the message body for mail readers that support HTML, you create a
MimeMultipart object, then create a MimeBodyPart for the plain-text body and another
for the HTML body and add them to the MimeMultipart. You then make the
MimeMultipart the content of the MimeMessage:
import javax.mail.Multipart;
import javax.mail.internet.MimeBodyPart;
import javax.mail.internet.MimeMultipart;
// ...
String textBody = "...text...";
String htmlBody = "...HTML...";
Multipart multipart = new MimeMultipart();
MimeBodyPart textPart = new MimeBodyPart();
textPart.setContent(textBody, "text/plain");
multipart.addBodyPart(htmlPart);
MimeBodyPart htmlPart = new MimeBodyPart();
htmlPart.setContent(htmlBody, "text/html");
multipart.addBodyPart(htmlPart);
message.setContent(multipart);
You attach files to the email message in a similar way:
Multipart multipart = new MimeMultipart();
// ...
byte[] fileData = getBrochureData();
String fileName = "brochure.pdf";
262 | Chapter 11: Sending and Receiving Mail and Instant Messages](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-283-320.jpg)
![The development server console (http://localhost:8080/_ah/admin/) includes a fea-
ture for simulating incoming email by submitting a web form. The development server
cannot receive actual email messages.
If the app has the incoming mail service enabled but does not have a
request handler for the appropriate URL, or if the request handler re-
turns an HTTP response code other than 200 for the request, the mes-
sage gets “bounced” and the sender receives an error email message.
Receiving Email in Python
To receive email in Python, you map the incoming email URL path to a script handler
in app.yaml file:
handlers:
- url: /_ah/mail/.+
script: handle_email.py
The app address used for the message is included in the URL path, so you can set up
separate handlers for different addresses directly in the configuration:
handlers:
- url: /_ah/mail/support%40.*app-id.appspotmail.com
script: support_contact.py
- url: /_ah/mail/.+
script: handle_email.py
Email addresses are URL-encoded in the final URL, so this pattern uses %40 to represent
an @ symbol. Also notice you must include a .* before the application ID when using
this technique, so the pattern works for messages sent to version-specific addresses
(such as support@dev.latest.app-id.appspotmail.com).
The Python SDK includes a class for parsing the POST content into a convenient object,
called InboundEmailMessage (in the google.appengine.api.mail package). It takes the
multipart MIME data (the POST body) as an argument to its constructor. Here’s an
example using the webapp framework:
from google.appengine.api import mail
from google.appengine.ext import webapp
from google.appengine.ext.webapp.util import run_wsgi_app
class IncomingMailHandler(webapp.RequestHandler):
def post(self):
message = mail.InboundEmailMessage(self.request.body)
# sender = message.sender
# recipients = message.to
# body = list(message.bodies(content_type='text/plain'))[0]
# ...
application = webapp.WSGIApplication([('/_ah/mail/.+', IncomingMailHandler)],
debug=True)
264 | Chapter 11: Sending and Receiving Mail and Instant Messages](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-285-320.jpg)
![def main():
run_wsgi_app(application)
if __name__ == '__main__':
main()
The InboundEmailMessage object includes attributes for the fields of the message, similar
to EmailMessage. sender is the sender’s email address, possibly with a displayable name
in the standard format (Mr. Sender <sender@example.com>). to is a list of primary
recipient addresses, and cc is a list of secondary recipients. (There is no bcc on an
incoming message, because blind-carbon-copied recipients are not included in the
message content.) subject is the message’s subject.
The InboundEmailMessage object may have more than one message body: an HTML
body and a plain-text body. You can iterate over the MIME multipart parts of the types
text/html and text/plain using the bodies() method. Without arguments, this method
returns an iterator that returns the HTML parts first, then the plain-text parts. You can
limit the parts returned to just the HTML or plain-text parts by setting the
content_type parameter. For example, to get just the plain-text bodies:
for text_body in message.bodies(content_type='text/plain'):
# ...
In the example earlier, we extracted the first plain-text body by passing the iterator to
the list() type, then indexing its first argument (which assumes one exists):
text = list(message.bodies(content_type='text/plain'))[0]
If the incoming message has file attachments, then these are accessible on the
attachments attribute. As with using EmailMessage for sending, this attribute is a list of
tuples whose first element is the filename and whose second element is the data byte
string. InboundEmailMessage allows all file types for incoming attachments, and does
not require that the filename accurately represent the file type. Be careful when using
files sent to the app by users, since they may not be what they say they are, and have
not been scanned for viruses.
The Python SDK includes a convenient webapp handler base class
for processing incoming email, called InboundMailHandler in the
google.appengine.ext.webapp.mail_handlers package. You use the handler by creating
a subclass that overrides the receive() method, then installing it like any other handler.
When the handler receives an email message, the receive() method is called with an
InboundEmailMessage object as its argument.
from google.appengine.ext.webapp import mail_handlers
class MyMailHandler(mail_handlers.InboundMailHandler):
def receive(self, message):
# ...
application = webapp.WSGIApplication([('/_ah/mail/.+', MyMailHandler)],
Receiving Email Messages | 265](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-286-320.jpg)
![The body content of the HTTP POST request is a MIME multipart message, with a
part for each field of the message:
from
The sender’s JID.
to
The app JID to which this message was sent.
body
The message body content (with characters as they were originally typed).
stanza
The full XML stanza of the message, including the previous fields (with XML spe-
cial characters escaped); useful for communicating with a custom client using
XML.
The Python and Java SDKs include classes for parsing the request data into objects.
When the service is enabled, all incoming messages are accepted and routed to the app.
All incoming chat invitations are accepted automatically. The app always appears as
connected and available when someone inquires for the app’s presence.
The development server console (http://localhost:8080/_ah/admin/) includes a fea-
ture for simulating incoming XMPP messages by submitting a web form. The devel-
opment server cannot receive actual XMPP messages.
Receiving XMPP Messages in Python
In Python, you map the URL path to a script handler in app.yaml file, as usual:
handlers:
- url: /_ah/xmpp/message/chat/
script: handle_xmpp.py
You can parse the incoming message into a Message object by passing a mapping of the
POST parameters to its constructor. With the webapp framework, this is a simple
matter of passing the parsed POST data (a mapping of the POST parameter names to
values) in directly:
from google.appengine.api import xmpp
from google.appengine.ext import webapp
from google.appengine.ext.webapp.util import run_wsgi_app
class IncomingXMPPHandler(webapp.RequestHandler):
def post(self):
message = xmpp.Message(self.request.POST)
message.reply('I got your message!')
application = webapp.WSGIApplication([('/_ah/xmpp/message/chat/',
IncomingXMPPHandler)],
debug=True)
Receiving XMPP Messages | 273](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-294-320.jpg)
![def main():
run_wsgi_app(application)
if __name__ == '__main__':
main()
The Message object has attributes for each message field: from, to, and body. It also
includes a convenience method for replying to the message: reply(), which takes the
body of the reply as its first argument.
The Python SDK includes an extension for webapp that makes it easy to implement
chat interfaces that perform user-issued commands. Here’s an example that responds
to the commands /stats and /score username, and ignores all other messages:
from google.appengine.api import xmpp
from google.appengine.ext import webapp
from google.appengine.ext.webapp import xmpp_handlers
from google.appengine.ext.webapp.util import run_wsgi_app
def get_stats():
# ...
def get_score_for_user(username):
# ...
class UnknownUserError(Exception):
pass
class ScoreBotHandler(xmpp_handlers.CommandHandler):
def stats_command(self, message):
stats = get_stats()
if stats:
message.reply('The latest stats: %s' % stats)
else:
message.reply('Stats are not available right now.')
def score_command(self, message):
try:
score = get_score_for_user(message.arg)
message.reply('Score for user %s: %d' % (message.arg, score))
except UnknownUserError, e:
message.reply('Unknown user %s' % message.arg)
application = webapp.WSGIApplication([('/_ah/xmpp/message/chat/', ScoreBotHandler)],
debug=True)
def main():
run_wsgi_app(application)
if __name__ == '__main__':
main()
274 | Chapter 11: Sending and Receiving Mail and Instant Messages](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-295-320.jpg)
![public class XMPPReceiverServlet extends HttpServlet {
public void doPost(HttpServletRequest req,
HttpServletResponse resp)
throws IOException {
XMPPService xmpp = XMPPServiceFactory.getXMPPService();
Message message = xmpp.parseMessage(req);
// ...
}
}
You access the fields of a Message using methods:
getFromJid()
The sender JID.
getMessageType()
The MessageType of the message (MessageType.CHAT or MessageType.NORMAL).
getRecipientJids()
The app JID used, as a single-element JID[].
getBody()
The String content of the message.
getStanza()
The String raw XML of the message.
276 | Chapter 11: Sending and Receiving Mail and Instant Messages](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-297-320.jpg)
![the sqlite3 module installed by starting a Python shell and importing it. If nothing
happens, the module is installed:
% python
Python 2.6.1 (r261:67515, Jul 7 2009, 23:51:51)
[GCC 4.2.1 (Apple Inc. build 5646)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import sqlite3
>>>
If you do not have the module, you can get it from the SQLite website. Note that if you
compiled your Python installation from source, you will need to recompile it with
SQLite support.
http://www.sqlite.org/download.html
If you’d prefer not to install SQLite, you can disable the use of the progress datafile in
the bulk loader tool by providing this command-line argument when running the tool:
--db_filename=skip.
Backup and Restore
One of the easiest features of the bulk loader to use is the backup and restore feature.
The bulk loader can download every entity of a given kind to a datafile, and can upload
that datafile back to the app to restore those entities. It can also upload the datafile to
another app, effectively moving all of the entities of a kind from one app to another.
You can also use these features to download data from the live app, and “upload” it to
the development server for testing (using the --url=... argument mentioned earlier).
These features require no additional configuration or code beyond the installation of
the remote API handler.
To download every entity of a given kind, run the bulkloader.py tool with the --dump,
--kind=..., and --filename=... arguments:
bulkloader.py --dump
--app_id=app-id
--url=http://app-id.appspot.com/remote_api
--kind=kind
--filename=backup.dat
(This is a single command line. The characters are line continuation markers, sup-
ported by some command shells. You can remove these characters and type all of the
arguments on one line.)
To upload a datafile created by --dump, thereby re-creating every entity in the file, use
the --restore argument instead of --dump:
bulkloader.py --restore
--app_id=app-id
--url=http://app-id.appspot.com/remote_api
--kind=kind
--filename=backup.dat
Using the Bulk Loader Tool | 281](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-302-320.jpg)
![return datetime.datetime.strptime(s, '%m/%d/%Y').date()
else:
return None
class BookLoader(bulkloader.Loader):
def __init__(self):
bulkloader.Loader.__init__(self, 'Book',
[('title', str),
('author', str),
('copyright_year', int),
('author_birthdate', get_date_from_str),
])
loaders = [BookLoader]
The loader definition file must do several things:
1. It must import or define a data model class for each kind to import, using the
Python data modeling API. The loader file does not have the app directory in the
module load path, so the loader file must add it. The modules should be imported
directly into the local namespace.
2. It must define (or import) a loader class for each kind to import, a subclass of the
Loader class in the package google.appengine.tools.bulkloader. This class must
have a constructor (an __init__() method) that calls the parent class’s constructor
with arguments that describe the CSV file’s columns.
3. It must define a module-global variable named loaders, whose value is a list of
loader classes.
The loader class’s __init__() method calls its parent class’s __init__() method with
three arguments: self (the object being initialized), the kind of the entity to create as
a string, and a list of tuples. Each tuple represents a column from the CSV file, in order
from left to right. The first element of each tuple is the property name to use. The second
element is a Python function that takes the CSV file value as a string and returns the
desired datastore value in the appropriate Python value type. (Refer to Table 4-1 for a
list of the Python value types for the datastore.)
For simple data types, you can use the constructor for the type as the type conversion
function. In the example above, we use str to leave the value from the CSV file as a
string, and int to parse the CSV string value as an integer. Depending on the output of
your spreadsheet program, you can use this technique for long, float, db.Text, and
other types whose constructors accept strings as initial arguments. For instance,
users.User takes an email address as a string argument.
The bool type constructor returns False for the empty string, and True in all other cases.
If your CSV file contains Boolean values in other forms, you’ll need to write a converter
function, like this:
def bool_from_string(s):
return s == 'TRUE'
Using the Bulk Loader Tool | 283](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-304-320.jpg)
![Dates, times, and date-times usually require a custom converter function to make an
appropriate datetime.datetime value from the string. As shown previously, you can use
the datetime.datetime.strptime() function to parse the string in a particular format.
Tailor this to the output of your spreadsheet program.
To create a multivalued property, you can use a type conversion function that returns
a list value. However, the function must get all of its data from a single cell (though you
can work around this by tweaking the file reader routine, described below). It cannot
reference other cells in the row, or other rows in the table.
Storing a key value (a reference to another entity) is usually a straightforward matter
of defining a type conversion function that uses db.Key.from_path() with the desired
kind and a key name derived from a CSV file value.
By default, the bulk loader creates each entity using a key with no ancestors, the given
kind, and a system-assigned numeric ID. You can tell the bulk loader to use a custom
key name instead of a numeric ID by defining a method of your loader class named
generate_key(). This method takes as arguments a loop counter (i) and the row’s col-
umn values as a list of strings (values), and returns either a string key name or a
db.Key object. The db.Key object can include an entity group parent key. If the function
returns None, a system-assigned numeric ID will be used.
class BookLoader(bulkloader.Loader):
# ...
def generate_key(self, i, values):
# Use the first column as the key name.
return values[0]
If an entity is uploaded with the same key as an entity that already exists
in the datastore, the new entity will replace the existing entity. You can
use this effect to download, modify, then reupload data to update ex-
isting entities instead of creating new ones.
You can “skip” an unused column in the CSV file by giving it a property name that
begins with an underscore. The Python data modeling API does not save attributes
beginning with an underscore as properties, so the final entity will not have a property
with this name. You can reuse the same dummy property name for all columns you
wish to skip:
class BookLoader(bulkloader.Loader):
def __init__(self):
bulkloader.Loader.__init__(self, 'Book',
[('_UNUSED', lambda x: None),
('title', str),
('author', str),
('_UNUSED', lambda x: None),
('_UNUSED', lambda x: None),
('copyright_year', int),
284 | Chapter 12: Bulk Data Operations and Remote Access](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-305-320.jpg)
![# ...
])
If you have a column that is to be the entity key name (via generate_key()) and you do
not want this value to also become a property, you must “skip” the column in the
column list using this technique. The bulk loader will complain if it finds more columns
than are declared in the list.
The bulk loader uses the Python data modeling API to construct, validate, and save
entities. If your app is written in Python, you just use the model classes of your app for
this purpose, and get the validation for free.
If your app is written in Java, you have a choice to make. You can either implement
Python data model classes that resemble your application’s data model, or you can
define Python data models that allow arbitrary properties to be set by the loader without
validation. For information on defining data models in Python, see Chapter 7. To define
a model class that allows all properties and does no validation, simply define a class
like the following for each kind in the loader file, above the definition of the loader
classes:
from google.appengine.ext import db
class Book(db.Expando):
pass
The loader class can do postprocessing on an entity after it is created but before it is
saved to the datastore. This gives you the opportunity to inspect the entity, add or
remove properties, and even create additional entities based on each row of the datafile.
To perform postprocessing, implement a method named handle_entity() on the loader
class. This takes an instance of the model class as its only argument (entity). It should
return the instance to create, a list of model instances if more than one entity should
be created, or None if no entity should be created.
You can extend the loader class to read file formats other than CSV files. To do
this, you override the generate_records() method. This method takes a filename
(filename), and returns a Python generator. Each iteration of the generator should yield
a list of strings for each entity to create, where each string corresponds to a property
defined in the loader’s constructor. You can use a generate_records() routine to mas-
sage the columns processed by the loader so that all the information needed for a single
property appears in the column’s value, such as to assemble what will eventually be-
come a multivalued property from multiple columns in the datafile.
You can override the initialize() method of the loader to perform actions at the be-
ginning of an upload. This method takes the input file’s name (filename) and a string
given on the command line as the --loader_opts=... argument (loader_opts), which
you can use to pass options to this method. Similarly, you can override the
finalize() method to do something after the upload is complete.
Using the Bulk Loader Tool | 285](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-306-320.jpg)
![specifies which property goes in which column of the resulting CSV file, and can include
value transformation functions to get the entity data into a format accepted by your
spreadsheet program.
Here is a simple exporter definition file that is symmetric with the loader definition
example in the previous section. Data downloaded with this exporter can be loaded
back in with the loader. Let’s call this file exporters.py:
from google.appengine.tools import bulkloader
import datetime
# Import the app's data models directly into
# this namespace. We must add the app
# directory to the path explicitly.
import sys
import os.path
sys.path.append(
os.path.abspath(
os.path.dirname(
os.path.realpath(__file__))))
from models import *
def make_str_for_date(d):
if d:
return d.strftime('%m/%d/%Y')
else:
return ''
class BookExporter(bulkloader.Exporter):
def __init__(self):
bulkloader.Exporter.__init__(self, 'Book',
[('title', str, None),
('author', str, ''),
('copyright_year', int, ''),
('author_birthdate', make_str_for_date, ''),
])
exporters = [BookExporter]
The exporter definition file must do several things:
1. It must import or define a data model class for each kind to export, using the Python
data modeling API. The exporter file does not have the app directory in the module
load path, so the exporter file must add it. The modules should be imported directly
into the local namespace.
2. It must define (or import) an exporter class for each kind to export, a subclass of
the Exporter class in the package google.appengine.tools.bulkloader. This class
must have a constructor (an __init__() method) that calls the parent class’s con-
structor with arguments that describe the CSV file’s columns.
3. It must define a module-global variable named exporters, whose value is a list of
exporter classes.
Using the Bulk Loader Tool | 287](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-308-320.jpg)
![Python 2.5.1 (r251:54863, Feb 6 2009, 19:02:12)
[GCC 4.0.1 (Apple Inc. build 5465)]
The db, users, urlfetch, and memcache modules are imported.
clock> import os.path
clock> import sys
clock> sys.path.append(os.path.realpath('.'))
clock> import models
clock> books = models.Book.all().fetch(6)
clock> books
[<models.Book object at 0x7a2c30>, <models.Book object at 0x7a2bf0>,
<models.Book object at 0x7a2cd0>, <models.Book object at 0x7a2cb0>,
<models.Book object at 0x7a2d30>, <models.Book object at 0x7a2c90>]
clock> books[0].title
u'The Grapes of Wrath'
clock> from google.appengine.api import mail
clock> mail.send_mail('juliet@example.com', 'test@example.com',
'Test email', 'This is a test message.')
clock>
To exit the shell, press Ctrl-D.
Using the Remote API from a Script
You can call the remote API directly from your own Python scripts using a library from
the Python SDK. This configures the Python API to use the remote API handler for your
application for all service calls, so you can use the service APIs as you would from a
request handler directly in your scripts.
Here’s a simple example script that prompts for a developer account email address and
password, then accesses the datastore of a live application:
#!/usr/bin/python
import getpass
import sys
# Add the Python SDK to the package path.
# Adjust these paths accordingly.
sys.path.append('~/google_appengine')
sys.path.append('~/google_appengine/lib/yaml/lib')
from google.appengine.ext.remote_api import remote_api_stub
from google.appengine.ext import db
import models
# Your app ID and remote API URL path go here.
APP-ID = 'app-id'
REMOTE_API_PATH = '/remote_api'
def auth_func():
email_address = raw_input('Email address: ')
password = getpass.getpass('Password: ')
Using the Remote API from a Script | 291](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-312-320.jpg)
![return email_address, password
def initialize_remote_api(app-id=APP-ID,
path=REMOTE_API_PATH):
remote_api_stub.ConfigureRemoteApi(
app-id,
path,
auth_func)
remote_api_stub.MaybeInvokeAuthentication()
def main(args):
initialize_remote_api()
books = models.Book.all().fetch(10)
for book in books:
print book.title
return 0
if __name__ == '__main__':
sys.exit(main(sys.argv[1:]))
The ConfigureRemoteApi() function (yes, it has a TitleCase name) sets up the remote
API access. It takes as arguments the application ID, the remote API handler URL path,
and a callable that returns a tuple containing the email address and password to use
when connecting. In this example, we define a function that prompts for the email
address and password, and pass the function to ConfigureRemoteApi().
The function also accepts an optional fourth argument specifying an alternate domain
name for the connection. By default, it uses app-id.appspot.com, where app-id is the
application ID in the first argument.
The MaybeInvokeAuthentication() function sends an empty request to verify that the
email address and password are correct, and raises an exception if they are not. (With-
out this, the script would wait until the first remote call to verify the authentication.)
Remember that every call to an App Engine library that performs a service call does so
over the network via an HTTP request to the application. This is inevitably slower than
running within the live application. It also consumes application resources like web
requests do, including bandwidth and request counts, which are not normally con-
sumed by service calls in the live app.
On the plus side, since your code runs on your local computer, it is not constrained by
the App Engine runtime sandbox or the 30-second request deadline. You can run long
jobs and interactive applications on your computer without restriction, using any Py-
thon modules you like—at the expense of consuming app resources to marshal service
calls over HTTP.
292 | Chapter 12: Bulk Data Operations and Remote Access](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-313-320.jpg)
![And here’s the request handler script:
from google.appengine.ext import webapp
from google.appengine.ext.webapp.util import run_wsgi_app
from google.appengine.ext import deferred
import invitation
class SendInvitationHandler(webapp.RequestHandler):
def get(self):
# recipient = ...
deferred.defer(invitation.send_invitation, recipient)
# ...
application = webapp.WSGIApplication([
('/sendinvite', SendInvitationHandler),
], debug=True)
def main():
run_wsgi_app(application)
if __name__ == '__main__':
main()
You set up the request handler for deferred work using the following configuration in
your app.yaml file:
handlers:
- url: /_ah/queue/deferred
script: $PYTHON_LIB/google/appengine/ext/deferred/
login: admin
This maps the URL path /_ah/queue/deferred to the deferred module in the API.
($PYTHON_LIB is replaced with the path to the Python directory containing the SDK on
App Engine.) Tasks created with defer() use this URL path by default. login: admin
restricts this handler to task queues and application developers.
The defer() function enqueues a task on the default queue that calls the given callable
object with the given arguments. (Despite its use of the URL /_ah/queue/deferred for
deferred tasks, this utility uses the default queue, not a queue named deferred.) The
arguments are serialized and deserialized using Python’s pickle module; all argument
values must be pickle-able.
Most Python callable objects can be used with defer(), including functions and classes
defined at the top level of a module, methods of objects, class methods, instances of
classes that implement __call__(), and built-in functions and methods. defer() does
not work with lambda functions, nested functions, nested classes, or instances of nested
classes. The task handler must be able to access the callable object by name, possibly
via a serializable object, since it does not preserve the scope of the call to defer().
You also can’t use a function or class in the same module as the request handler class
from which you call defer(). This is because pickle believes the module of the request
handler class to be __main__ while it is running, and so it doesn’t save the correct
Task Queues | 303](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-324-320.jpg)
![The url() method sets a custom URL for the task. If no custom URL is set, the task
calls the default URL for the queue.
The param() method adds a request parameter to be included with the task’s HTTP
request. The method takes the name of the parameter, and either a String or a
byte[] as the value of the parameter. Parameters are added to the HTTP request body
URL-encoded, similar to a web form. The removeParam() method removes a parameter
from the TaskOptions object; it takes the parameter name as its argument.
You can also set the HTTP request body with the payload() method. This method takes
the request body as a String. You can also include a character set specification (such
as "utf-8") as the second argument. You can’t set both a payload and parameters for
the same task.
You can control other aspects of the HTTP request as well. The method() method sets
the HTTP method; the default is "POST". You can also add arbitrary HTTP headers by
calling the header() method (which takes a String key and String value as arguments)
for each header, or by calling the headers() method with a Map<String, String>.
You can set a delay for the task. The countdownMillis() method sets the delay as a
number of milliseconds (a long) from the time the task is enqueued. The etaMillis()
method sets the delay to be an absolute date and time, as a number of milliseconds
from midnight, January 1, 1970, UTC.
To set the name for a task, you call the name() method. As described earlier, the name
ensures that only one task with that name can be enqueued for a period of time, at least
seven days with the current implementation. If you do not specify a name, the system
assigns a unique name that has not been used in the past seven days.
An app can have up to 10 named queues, each with its own task execution rate (token
replenishment rate and bucket size). In a Java app, you configure queues using a con-
figuration file named queue.xml, in the WEB-INF/ directory, which looks like this:
<queue-entries>
<queue>
<name>fastest</name>
<rate>10/s</rate>
<bucket-size>10</bucket-size>
</queue>
<queue>
<name>limitedemail</name>
<rate>1000/d</rate>
<bucket-size>5</bucket-size>
</queue>
</queue-entries>
The file includes one <queue> element per queue to be configured. Each queue has a
<name>, a token replenishment <rate>, and a <bucket-size>.
306 | Chapter 13: Task Queues and Scheduled Tasks](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-327-320.jpg)
![Because App Engine promises not to make backward-incompatible changes to the
runtime environment, the default version of Django on App Engine (in version 1 of the
Python runtime) is still 0.96. However, App Engine includes a mechanism for explicitly
selecting a newer version, such that all subsequent attempts to import Django modules
will use the new version. You can use this mechanism in your main request handler
script to use Django 1.1 for your app.
Before you do that, you must download and install Django 1.1 on your local computer.
You won’t need to add it to your application directory (it’s available when running on
App Engine), but you will need it locally while building and testing your application.
Go to the Django project website and download Django 1.1:
http://www.djangoproject.com/download/1.1/tarball/
Follow the instructions to install the Python modules in your local Python environment
by running the following commands. (The sudo command will require your computer’s
administrator password.)
tar xzvf Django-1.1.tar.gz
cd Django-1.1
sudo python setup.py install
Django 1.1 was the latest release when this book was published. If you
see a newer version available on the Django website, be sure to check
the App Engine documentation to confirm that the new version has been
added to the Python runtime environment before using it locally.
If you’d prefer to use your own copy of Django instead of the one pro-
vided by the Python runtime environment (such as a newer version not
yet in the environment), you can add it directly to your application di-
rectory. Doing so requires an additional step to remove the version of
Django that’s part of the runtime environment from the Python module
search path. To do this, put this in your main request handler script
prior to importing django modules:
for k in [k for k in sys.modules if k.startswith('django')]:
del sys.modules[k]
Creating a Django Project
In Django, the complete set of code, configuration, and static files for a website is called
a project. You create a new project by running a command that was installed with
Django, called django-admin.py, like so:
django-admin.py startproject myproject
This command creates in your current working directory a new directory named after
the project, in this case myproject/, with several starter files:
Creating a Django Project | 315](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-336-320.jpg)
![__init__.py
A file that tells Python that code files in this directory can be imported as modules
(this directory is a Python package).
manage.py
A command-line utility you will use to build and manage this project, with many
features.
settings.py
Configuration for this project, in the form of a Python source file.
urls.py
Mappings of URL paths to Python code, as a Python source file.
If you’re following along with a Django tutorial or book, the next step is usually to start
the Django development server using the manage.py command. If you did so now, you
would be running the Django server, but it would know nothing of App Engine. To
run the project in the App Engine development server, we must connect the Django
project to the App Engine runtime environment using a request handler script and
configuration in the app.yaml file.
The Request Handler Script
All that’s left to do is to invoke the Django project from a handler script. Exam-
ple 14-1 is a minimal script that hands the request off to Django 1.1 using the WSGI
runner function from the webapp library. Create a file named main.py in the Django
project directory you created in the previous step (e.g., myproject/).
Example 14-1. A simple request handler for invoking the Django framework
from google.appengine.dist import use_library
use_library('django', '1.1')
import os
os.environ['DJANGO_SETTINGS_MODULE'] = 'settings'
import django.core.handlers.wsgi
from google.appengine.ext.webapp import util
def main():
# Run Django via WSGI.
application = django.core.handlers.wsgi.WSGIHandler()
util.run_wsgi_app(application)
if __name__ == '__main__':
main()
The use_library() function is an App Engine utility that tells the runtime environment
that version 1.1 of Django should be used when importing libraries from the django
package. Without this call, importing from django would use Django 0.96.
316 | Chapter 14: The Django Web Application Framework](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-337-320.jpg)
![InstallAppengineHelperForDjango(), and adding the version identifier as an argument
to that function call. For instance, to select Django 1.0:
InstallAppengineHelperForDjango('1.0')
Now that the Helper is installed, you can start the App Engine development server
using manage.py as you would with any other Django project:
python manage.py runserver
Thanks to the Helper, this is no different than running the dev_appserver.py command
or using the Launcher to start the server. If you’d prefer to use those methods, you may,
without losing any functionality of the Helper.
Test that the default project is working locally by visiting the server’s URL in a web
browser. By default, Django uses port 8000, not 8080, so this URL is:
http://localhost:8000/
The Helper adds several features to manage.py that work with the App Engine devel-
opment server. One especially useful feature is the shell command, which starts a
Python shell in the context of the Django project—and the development server. You
can use this shell to import App Engine modules and interact with the development
server, the simulated sandbox, and the simulated services.
python manage.py shell
Here is a simple shell session that demonstrates calling the development server’s
simulated Mail service (several harmless warning messages have been elided):
% python manage.py shell
Python 2.6.1 (r261:67515, Jul 7 2009, 23:51:51)
[GCC 4.2.1 (Apple Inc. build 5646)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
(InteractiveConsole)
>>> from google.appengine.api import mail
>>> mail.send_mail('sender@example.com', 'to@example.com', 'my subject',
'message body')
INFO:root:MailService.Send
INFO:root: From: sender@example.com
INFO:root: To: to@example.com
INFO:root: Subject: my subject
INFO:root: Body:
INFO:root: Content-type: text/plain
INFO:root: Data length: 12
INFO:root:You are not currently sending out real email. If you have sendmail
installed you can use it by using the server with --enable_sendmail
>>>
You can open a Python shell that connects to the live application via remote_api using
the console command. This requires that you set up remote_api as described in Chap-
ter 12 using a URL path of /remote_api.
python manage.py console
The Django App Engine Helper | 319](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-340-320.jpg)
![from appengine_django.models import BaseModel
from google.appengine.ext import db
class Book(BaseModel):
title = db.StringProperty()
author = db.StringProperty()
copyright_year = db.IntegerProperty()
author_birthdate = db.DateProperty()
class BookReview(BaseModel):
book = db.ReferenceProperty(Book, collection_name='reviews')
review_author = db.UserProperty()
review_text = db.TextProperty()
rating = db.StringProperty(choices=['Poor', 'OK', 'Good', 'Very Good', 'Great'],
default='Great')
create_date = db.DateTimeProperty(auto_now_add=True)
Edit views.py to query the database for books and pass the query object to the template,
like this:
from django.shortcuts import render_to_response
from google.appengine.ext import db
from bookstore import models
def home(request):
q = models.Book.all().order('title')
return render_to_response('bookstore/index.html',
{ 'books': q })
Finally, edit templates/bookstore/index.html to display the list of books:
<html>
<body>
<p>Welcome to The Book Store!</p>
<p>Books in our catalog:</p>
<ul>
{% for book in books %}
<li>{{ book.title }}, by {{ book.author }} ({{ book.copyright_year }})</li>
{% endfor %}
</ul>
</body>
</html>
Notice that we’re passing the db.Query object directly to the template. The template
uses the query object’s iterator interface to fetch the Book entities from the datastore.
If you load the page right now (and feel free to do so to test the code), you’ll see an
empty list of books. Let’s create some books from the Python shell. Stop the server (hit
Ctrl-C), then start the shell:
python manage.py shell
You can create objects in the development server’s datastore by typing Python code
directly into the shell. Here is a complete shell session, including output, that creates
several Book entities:
Using App Engine Models With Django | 323](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-344-320.jpg)
![% python manage.py shell
Python 2.6.1 (r261:67515, Jul 7 2009, 23:51:51)
[GCC 4.2.1 (Apple Inc. build 5646)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
(InteractiveConsole)
>>> from bookstore import models
>>> import datetime
>>> b = models.Book(title='Grapes of Wrath, The', author='Steinbeck, John',
copyright_year=1939, author_birthdate=datetime.date(1902, 2, 27))
>>> b.put()
datastore_types.Key.from_path(u'Book', 1,
_app_id_namespace=u'google-app-engine-django')
>>> b = models.Book(title='Cat in the Hat, The', author='Dr. Seuss',
copyright_year=1957, author_birthdate=datetime.date(1904, 3, 2))
>>> b.put()
datastore_types.Key.from_path(u'Book', 2,
_app_id_namespace=u'google-app-engine-django')
>>>
Exit the shell (press Ctrl-D), then restart the server and load the page.
Using Django Unit Tests and Fixtures
The Django App Engine Helper includes support for unit tests using Django’s unit test
runner and the App Engine development server. The unit test runner runs tests inside
the development server’s environment, so tests can access the simulated services, in-
cluding the simulated datastore. The test runner supports tests defined with the
standard Python libraries unittest or doctest, or other test frameworks with test runner
customizations. Django provides its own class for defining unittest test cases, called
django.test.TestCase, which adds several Django-specific features.
For unittest tests, the runner looks in files in the Django application directory (e.g.,
bookstore/) named models.py and tests.py for classes that inherit from
unittest.TestCase or django.test.TestCase. Each method of a test case class whose
name begins with test is a separate test. For each test, the runner initializes the test
environment, instantiates the TestCase class, and calls the method for the test. If the
class defines a setUp() method, it calls it before each test. If the class defines a tear
Down() method, it calls it after each test. You can use setUp() and tearDown() to create
and clean up test data and preparation for every test in a test case.
The TestCase class provides methods that cause the test to report failure if the condi-
tions tested are not met. self.assertEqual() takes two arguments, and causes the test
to fail if the arguments are not equal. self.assert_() takes one argument, and fails the
test if the argument is a false value (False, None, 0, the empty string, and so forth).
self.assertRaises() takes an exception class, a callable object, and arguments for the
callable object, calls the callable, then fails the test if the callable does not raise the given
exception. There are several other test methods, as well. Each of these methods takes
an optional message argument that is printed in the test report if the test fails.
324 | Chapter 14: The Django Web Application Framework](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-345-320.jpg)
![Enter Python statements to create test data in the datastore. Here is a shell session that
creates a few Book and BookReview entities:
% python manage.py shell
Python 2.6.1 (r261:67515, Jul 7 2009, 23:51:51)
[GCC 4.2.1 (Apple Inc. build 5646)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
(InteractiveConsole)
>>> from google.appengine.ext import db
>>> from bookstore import models
>>> b1 = models.Book(title="one", author="")
>>> b1.put()
datastore_types.Key.from_path(u'Book', 1,
_app_id_namespace=u'google-app-engine-django')
>>> br1 = models.BookReview(book=b1, review_text="review1")
>>> br2 = models.BookReview(book=b1, review_text="review2")
>>> br3 = models.BookReview(book=b1, review_text="review3")
>>> db.put(br1)
datastore_types.Key.from_path(u'BookReview', 2,
_app_id_namespace=u'google-app-engine-django')
>>> db.put(br2)
datastore_types.Key.from_path(u'BookReview', 3,
_app_id_namespace=u'google-app-engine-django')
>>> db.put(br3)
datastore_types.Key.from_path(u'BookReview', 4,
_app_id_namespace=u'google-app-engine-django')
>>> b2 = models.Book(title="two", author="")
>>> b2.put()
datastore_types.Key.from_path(u'Book', 5,
_app_id_namespace=u'google-app-engine-django')
>>> b2br1 = models.BookReview(book=b2, review_text="review1")
>>> b2br2 = models.BookReview(book=b2, review_text="review2")
>>> db.put(b2br1)
datastore_types.Key.from_path(u'BookReview', 6,
_app_id_namespace=u'google-app-engine-django')
>>> db.put(b2br2)
datastore_types.Key.from_path(u'BookReview', 7,
_app_id_namespace=u'google-app-engine-django')
>>>
Exit the shell (Ctrl-D).
Create a directory named fixtures/ in the bookstore/ app directory. Now run the
dumpdata command, and redirect its output to a file in this new directory:
python manage.py dumpdata bookstore >bookstore/fixtures/bookstore.json
The resulting file contains representations of all of the datastore entities for the kinds
whose models are defined in the bookstore app, in the JSON format.
To load this fixture for every test in a test case, you create a class attribute for the
django.test.TestCase named fixtures, whose value is a list of the names of the fixture
files to load from the fixtures/ directory. Here’s another (trivial) test case that uses
fixtures, which you can put in models.py:
326 | Chapter 14: The Django Web Application Framework](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-347-320.jpg)
![# ...
class TestBooksAndReviews(test.TestCase):
fixtures = ['bookstore.json']
def testBookReviewOrder(self):
d = Book.all().filter('title =', 'one').get()
self.assert_(d, 'Book "one" must be in test fixture')
q = BookReview.all().filter('book =', d)
q.order('-create_date')
reviews = q.fetch(3)
self.assertEqual('review3', reviews[0].review_text)
self.assertEqual('review2', reviews[1].review_text)
self.assertEqual('review1', reviews[2].review_text)
Run the new test using the test command, as before:
python manage.py test bookstore
For each test in the TestBooksAndReviews test case, Django flushes the simulated data-
store, loads the fixtures, calls the test case’s setUp() method (if any), then calls the test
method, and finally calls the tearDown() method (if any).
You can also load fixtures into the project’s development datastore manually using the
python manage.py loaddata command:
python manage.py loaddata bookstore/fixtures/bookstore.json
See the Django documentation for more information about unit testing and fixtures.
Using Django Forms
Django includes a powerful feature for building web forms based on data model defi-
nitions. The Django forms library can generate HTML for forms, validate that submit-
ted data meets the requirements of the model, and redisplay the form to the user with
error messages. The default appearance is useful, and you can customize the appearance
extensively.
Django’s version of the form library only works with the Django data modeling API,
and does not work directly with App Engine models. Thankfully, the App Engine Py-
thon runtime environment includes an App Engine–compatible implementation of
Django forms, in the package google.appengine.ext.db.djangoforms.
We won’t go into the details of how Django forms work—see the Django documen-
tation for a complete explanation—but let’s walk through a quick example to see how
the pieces fit together. Our example will use the following behavior for creating and
editing Book entities:
• An HTTP GET request to /books/book/ displays an empty form for creating a new
Book.
Using Django Forms | 327](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-348-320.jpg)
![To display the form, we simply pass the BookForm object to a template. There are several
methods on the object for rendering it in different ways; we’ll use the as_p() method
to display the form fields in <p> elements.
Create that template now, named templates/bookstore/bookform.html:
<html>
<body>
{% if book_id %}
<p>Edit book {{ book_id }}:</p>
<form action="/books/book/{{ book_id }}" method="POST">
{% else %}
<p>Create book:</p>
<form action="/books/book/" method="POST">
{% endif %}
{{ form.as_p }}
<input type="submit" />
</form>
</body>
</html>
The template is responsible for outputting the <form> tag and the Submit button. The
BookForm does the rest.
Before we test this, let’s make one more change to the book-listing template to add
some “create” and “edit” links. Edit templates/bookstore/index.html:
<html>
<body>
<p>Welcome to The Book Store!</p>
<p>Books in our catalog:</p>
<ul>
{% for book in books %}
<li>{{ book.title }}, by {{ book.author }} ({{ book.copyright_year }})
[<a href="/books/book/{{ book.key.id }}">edit</a>]</li>
{% endfor %}
</ul>
<p>[<a href="/books/book/">add a book</a>]</p>
</body>
</html>
Restart the development server, then load the book list URL (/books/) in the browser.
Click “add a book” to show the book creation form. Enter some data for the book, then
submit it to create it.
The default form widget for a date field is just a text field, and it’s finicky
about the format. In this case, the “author birthdate” field expects input
in the form YYYY-MM-DD, such as 1902-02-27.
330 | Chapter 14: The Django Web Application Framework](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-351-320.jpg)
![Figure 15-3. The Administration Console log browser
The following command downloads request logs for the app in the development di-
rectory clock, and saves them to a file named logs.txt:
appcfg.py request_logs clock logs.txt
Request data appears in the file in a common format known as the Apache Combined
(or “NCSA Combined”) logfile format, one request per line (shown here as two lines
to fit on the page):
127.0.0.1 - - [05/Jul/2009:06:46:30 -0700] "GET /blog/ HTTP/1.1" 200 14598 -
"Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10_5_8; en-us)...,gzip(gfe)"
From left to right, the fields are:
• The IP address of the client
• A - (an unused field retained for backward compatibility)
• The email address of the user who made the request, if the user is signed in using
Google Accounts; - otherwise
• The date and time of the request
• The HTTP command string in double quotes, including the method and URL path
• The HTTP response code returned by the server
• The size of the response, as a number of bytes
• The “Referrer” header provided by the client, usually the URL of the page that
linked to this URL
• The “User-Agent” header provided by the client, usually identifying the browser
and its capabilities
340 | Chapter 15: Deploying and Managing Applications](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/85/Programming-google-app-engine-361-320.jpg)
![Programming Google App Engine
by Dan Sanderson
Copyright © 2010 Dan Sanderson. All rights reserved.
Printed in the United States of America.
Published by O’Reilly Media, Inc., 1005 Gravenstein Highway North, Sebastopol, CA 95472.
O’Reilly books may be purchased for educational, business, or sales promotional use. Online editions
are also available for most titles (http://my.safaribooksonline.com). For more information, contact our
corporate/institutional sales department: (800) 998-9938 or corporate@oreilly.com.
Editor: Mike Loukides Indexer: Ellen Troutman Zaig
Production Editor: Sumita Mukherji Cover Designer: Karen Montgomery
Proofreader: Sada Preisch Interior Designer: David Futato
Illustrator: Robert Romano
Printing History:
November 2009: First Edition.
Nutshell Handbook, the Nutshell Handbook logo, and the O’Reilly logo are registered trademarks of
O’Reilly Media, Inc. Programming Google App Engine, the image of a waterbuck, and related trade dress
are trademarks of O’Reilly Media, Inc.
Many of the designations used by manufacturers and sellers to distinguish their products are claimed as
trademarks. Where those designations appear in this book, and O’Reilly Media, Inc., was aware of a
trademark claim, the designations have been printed in caps or initial caps.
While every precaution has been taken in the preparation of this book, the publisher and author assume
no responsibility for errors or omissions, or for damages resulting from the use of the information con-
tained herein.
TM
This book uses RepKover™, a durable and flexible lay-flat binding.
ISBN: 978-0-596-52272-8
[M]
1257864694](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-5-2048.jpg)
![Let’s upgrade the clock app to use the webapp framework. Replace the contents of
main.py with the version shown in Example 2-3. Reload the page in your browser to
see the new version in action. (You won’t notice a difference other than an updated
time. This example is equivalent to the previous version.)
Example 2-3. A simple request handler using the webapp framework
from google.appengine.ext import webapp
from google.appengine.ext.webapp.util import run_wsgi_app
import datetime
class MainPage(webapp.RequestHandler):
def get(self):
time = datetime.datetime.now()
self.response.headers['Content-Type'] = 'text/html'
self.response.out.write('<p>The time is: %s</p>' % str(time))
application = webapp.WSGIApplication([('/', MainPage)],
debug=True)
def main():
run_wsgi_app(application)
if __name__ == '__main__':
main()
Example 2-3 imports the module google.appengine.ext.webapp, then defines a request
handler class called MainPage, a subclass of webapp.RequestHandler. The class defines
methods for each HTTP method supported by the handler, in this case one method for
HTTP GET called get(). When the application handles a request, it instantiates the
handler class, sets self.request and self.response to values the handler method can
access and modify, then calls the appropriate handler method, in this case get(). When
the handler method exits, the application uses the value of self.response as the HTTP
response.
The application itself is represented by an instance of the class
webapp.WSGIApplication. The instance is initialized with a list of mappings of URLs to
handler classes. The debug parameter tells the application to print error messages to the
browser window when a handler returns an exception if the application is running
under the development web server. webapp detects whether it is running under the
development server or running as a live App Engine application, and will not print
errors to the browser when running live even if debug is True. You can set it to False to
have the development server emulate the live server when errors occur.
The script defines a main() function that runs the application using a utility method.
Lastly, the script calls main() using the Python idiom of if __name__ ==
'__main__': ..., a condition that is always true when the script is run by the web server.
This idiom allows you to import the script as a module for other code, including the
classes and functions defined in the script, without running the main() routine.
Developing the Application | 29](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-50-2048.jpg)
![Defining a main() function this way allows App Engine to cache the
compiled handler script, making subsequent requests faster to execute.
For more information on app caching, see Chapter 3.
A single WSGIApplication instance can handle multiple URLs, routing the request to
different RequestHandler classes based on the URL pattern. But we’ve already seen that
the app.yaml file maps URL patterns to handler scripts. So which URL patterns should
appear in app.yaml, and which should appear in the WSGIApplication? Many web
frameworks include their own URL dispatcher logic, and it’s common to route all dy-
namic URLs to the framework’s dispatcher in app.yaml. With webapp, the answer
mostly depends on how you’d like to organize your code. For the clock application, we
will create a second request handler as a separate script to take advantage of a feature
of app.yaml for user authentication, but we could also put this logic in main.py and
route the URL with the WSGIApplication object.
Users and Google Accounts
So far, our clock shows the same display for every user. To allow each user to customize
the display and save her preferences for future sessions, we need a way to identify the
user making a request. An easy way to do this is with Google Accounts.
Let’s add something to the page that indicates whether the user is signed in, and pro-
vides links for signing in and signing out of the application. Edit main.py to resemble
Example 2-4.
Example 2-4. A version of main.py that displays Google Accounts information and links
from google.appengine.api import users
from google.appengine.ext import webapp
from google.appengine.ext.webapp.util import run_wsgi_app
import datetime
class MainPage(webapp.RequestHandler):
def get(self):
time = datetime.datetime.now()
user = users.get_current_user()
if not user:
navbar = ('<p>Welcome! <a href="%s">Sign in or register</a> to customize.</p>'
% (users.create_login_url(self.request.path)))
else:
navbar = ('<p>Welcome, %s! You can <a href="%s">sign out</a>.</p>'
% (user.email(), users.create_logout_url(self.request.path)))
self.response.headers['Content-Type'] = 'text/html'
self.response.out.write('''
<html>
<head>
<title>The Time Is...</title>
</head>
30 | Chapter 2: Creating an Application](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-51-2048.jpg)
![<body>
%s
<p>The time is: %s</p>
</body>
</html>
''' % (navbar, str(time)))
application = webapp.WSGIApplication([('/', MainPage)],
debug=True)
def main():
run_wsgi_app(application)
if __name__ == '__main__':
main()
In a real application, you would use a templating system for the output, separating the
HTML and display logic from the application code. While many web application
frameworks include a templating system, webapp does not. Since the clock app only
has one page, we’ll put the HTML in the handler code, using Python string formatting
to keep things organized.
The Python runtime environment includes a version of Django, whose
templating system can be used with webapp. When Google released
version 1 of the Python runtime environment, the latest version of
Django was 0.96, so this is what the runtime includes. For more infor-
mation on using Django templates with webapp, see the App Engine
documentation.
Reload the page in your browser. The new page resembles Figure 2-5.
Figure 2-5. The clock app with a link to Google Accounts when the user is not signed in
Developing the Application | 31](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-52-2048.jpg)
![if not user:
navbar = ('<p>Welcome! <a href="%s">Sign in or register</a> to customize.</p>'
% (users.create_login_url(self.request.path)))
tz_form = ''
else:
userprefs = models.get_userprefs()
navbar = ('<p>Welcome, %s! You can <a href="%s">sign out</a>.</p>'
% (user.nickname(), users.create_logout_url(self.request.path)))
tz_form = '''
<form action="/prefs" method="post">
<label for="tz_offset">
Timezone offset from UTC (can be negative):
</label>
<input name="tz_offset" id="tz_offset" type="text"
size="4" value="%d" />
<input type="submit" value="Set" />
</form>
''' % userprefs.tz_offset
time += datetime.timedelta(0, 0, 0, 0, 0, userprefs.tz_offset)
self.response.headers['Content-Type'] = 'text/html'
self.response.out.write('''
<html>
<head>
<title>The Time Is...</title>
</head>
<body>
%s
<p>The time is: %s</p>
%s
</body>
</html>
''' % (navbar, str(time), tz_form))
application = webapp.WSGIApplication([('/', MainPage)],
debug=True)
def main():
run_wsgi_app(application)
if __name__ == '__main__':
main()
To enable the preferences form, we need a request handler to parse the form data and
update the datastore. Let’s implement this as a new request handler script. Create a file
named prefs.py with the contents shown in Example 2-7.
Example 2-7. A new handler script, prefs.py, for the preferences form
from google.appengine.ext import webapp
from google.appengine.ext.webapp.util import run_wsgi_app
import models
class PrefsPage(webapp.RequestHandler):
def post(self):
Developing the Application | 35](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-56-2048.jpg)
![userprefs = models.get_userprefs()
try:
tz_offset = int(self.request.get('tz_offset'))
userprefs.tz_offset = tz_offset
userprefs.put()
except ValueError:
# User entered a value that wasn't an integer. Ignore for now.
pass
self.redirect('/')
application = webapp.WSGIApplication([('/prefs', PrefsPage)],
debug=True)
def main():
run_wsgi_app(application)
if __name__ == '__main__':
main()
This request handler handles HTTP POST requests to the URL /prefs, which is the
URL (“action”) and HTTP method used by the form. The handler calls the
get_userprefs() function from models.py to get the UserPrefs object for the current
user, which is either a new unsaved object with default values, or the object for an
existing entity. The handler parses the tz_offset parameter from the form data as an
integer, sets the property of the UserPrefs object, then saves the object to the datastore
by calling its put() method. The put() method creates the object if it doesn’t exist, or
updates the existing object.
If the user enters a noninteger in the form field, we don’t do anything. It’d be appro-
priate to return an error message, but we’ll leave this as is to keep the example simple.
Finally, edit app.yaml to map the handler script to the URL /prefs in the handlers:
section, as shown in Example 2-8.
Example 2-8. A new version of app.yaml mapping the URL /prefs, with login required
application: clock
version: 1
runtime: python
api_version: 1
handlers:
- url: /prefs
script: prefs.py
login: required
- url: /.*
script: main.py
The login: required line says that the user must be signed in to Google Accounts to
access the /prefs URL. If the user accesses the URL while not signed in, App Engine
automatically directs the user to the Google Accounts sign-in page, then redirects her
36 | Chapter 2: Creating an Application](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-57-2048.jpg)
![class MainHandler(webapp.RequestHandler):
def get(self):
host = self.request.headers['Host']
self.response.out.write('Host: %s' % host)
App IDs and Versions
Every App Engine application has an application ID that uniquely distinguishes the
app from all other applications. As described in Chapter 2, you can register an ID for
a new application using the Administration Console. Once you have an ID, you add it
to the app’s configuration so the developer tools know that the files in the app root
directory belong to the app with that ID.
The app’s configuration also includes a version identifier. Like the app ID, the version
identifier is associated with the app’s files when the app is uploaded. App Engine retains
one set of files and frontend configuration for each distinct version identifier used dur-
ing an upload. If you do not change the app version in the configuration when you
upload, the existing files for that version of the app are replaced.
Each distinct version of the app is accessible at its own domain name, of the following
form:
version-id.latest.app-id.appspot.com
When you have multiple versions of an app uploaded to App Engine, you can use the
Administration Console to select which version is the one you want the public to access.
The Console calls this the “default” version. When a user visits your Google Apps
domain (and configured subdomain), or the appspot.com domain without the version
ID, he sees the default version.
The appspot.com domain containing the version ID supports an additional domain part,
just like the default appspot.com domain:
anything.version-id.latest.app-id.appspot.com
Unless you explicitly prevent it, anyone who knows your application ID
and version identifiers can access any uploaded version of your appli-
cation using the appspot.com URLs. You can restrict access to nondefault
versions of the application using code that checks the domain of the
request and only allows authorized users to access the versioned do-
mains. You can’t restrict access to static files this way.
Another way to restrict access to nondefault versions is to use Google
Accounts authorization, described later in this chapter. You can restrict
access to app administrators while a version is in development, then
replace the configuration to remove the restriction just before making
that version the default version.
70 | Chapter 3: Handling Web Requests](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-91-2048.jpg)
![If you’d like browsers to reload a static file resource automatically every
time you launch a new major version of the app, you can use the mul-
tiversion URL handler just discussed, then use the CURRENT_VERSION_ID
environment variable as the “version” in the static file URLs:
self.response.out('<script src="/js/' +
os.environ['CURRENT_VERSION_ID'] +
'/code.js" />')
Static files in Java
As we saw earlier, the WAR directory structure for a Java web application keeps all
application code, JARs, and configuration in a subdirectory named WEB-INF/. Typi-
cally, files outside of WEB-INF/ represent resources that the user can access directly,
including static files and JSPs. The URL paths to these resources are equivalent to the
paths to these files within the WAR.
Say an app’s WAR has the following files:
main.jsp
forum/home.jsp
images/logo.png
images/cancelbutton.png
images/okbutton.png
terms.html
WEB-INF/classes/com/example/Forum.class
WEB-INF/classes/com/example/MainServlet.class
WEB-INF/classes/com/example/Message.class
WEB-INF/classes/com/example/UserPrefs.class
WEB-INF/lib/appengine-api.jar
This app has four static files: three PNG images and an HTML file named terms.html.
When the app is uploaded, these four files are pushed to the static file servers. The
frontends know to route requests for URL paths equivalent to these file paths (such
as /images/logo.png) to the static file servers.
The two .jsp files are assumed to be JSPs, and are compiled to servlet classes and
mapped to the URL paths equivalent to their file paths. Since these are application
code, they are handled by the application servers. The JSP source files themselves are
not pushed to the static file servers.
By default, all files in the WAR are pushed to the application servers, and are accessible
by the application code via the filesystem. This includes the files that are identified as
static files and pushed to the static file servers. In other words, all files are considered
resource files, and all files except for JSPs and the WEB-INF/ directory are considered
static files.
Configuring the Frontend | 79](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-100-2048.jpg)
![In practice, neither method of importing offers much advantage. Late importing may
actually be slower if the import is inside a loop or frequent code path, since importing
an already-imported module takes some time, even if it doesn’t load files or evaluate
module code. It’s best to import modules in the way that makes the most sense for code
clarity, typically at the top of each module (source file) that uses the module. Once
you’ve developed your app, if you notice you have a significant slowdown during im-
ports of the first request, you could consider importing selected modules in their code
paths if you can demonstrate that doing so would skip a lot of unneeded code in a
significant number of cases.
App Engine includes a feature similar to import caching to achieve similar savings for
the handler scripts themselves. In general, when the app server invokes a handler script,
it evaluates all of the script’s code, regardless of whether the script has been invoked
by the app instance in the past. But if the script defines a function named main() that
takes no arguments (or arguments with default values), App Engine treats the handler
script like a module. The first invocation evaluates the complete script as usual, but for
all subsequent invocations, App Engine just calls the main() function.
Consider an early version of the clock application we built in Chapter 2, repeated here
as Example 3-4. This handler script imports three modules, defines three global ele-
ments (a class, a global variable, and the main() function), then calls the main() function.
Example 3-4. A simple request handler illustrating the use of app caching
from google.appengine.ext import webapp
from google.appengine.ext.webapp.util import run_wsgi_app
import datetime
class MainPage(webapp.RequestHandler):
def get(self):
time = datetime.datetime.now()
self.response.headers['Content-Type'] = 'text/html'
self.response.out.write('<p>The time is: %s</p>' % str(time))
application = webapp.WSGIApplication([('/', MainPage)],
debug=True)
def main():
run_wsgi_app(application)
if __name__ == '__main__':
main()
The first time an app instance invokes this script, the full script is evaluated, thereby
importing the modules and defining the global elements in the script’s namespace
within the app instance. Evaluating the script also calls the main() function.
How the App Is Run | 91](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-112-2048.jpg)
![A property with the null value is not the same as an unset property. Consider the fol-
lowing Python code:
class Entity(db.Expando):
pass
a = Entity()
a.prop1 = 'abc'
a.prop2 = None
a.put()
b = Entity()
b.prop1 = 'def'
b.put()
This creates two entities of the kind Entity. Both entities have a property named
prop1. The first entity has a property named prop2; the second does not.
Of course, an unset property can be set later:
b.prop2 = 123
b.put()
# b now has a property named "prop2."
Similarly, a set property can be made unset. In the Python API, you delete the property
by deleting the attribute from the object, using the del keyword:
del b.prop2
b.put()
# b no longer has a property named "prop2."
In Java, the low-level datastore API’s Entity class has methods to set properties
(setProperty()) and unset properties (removeProperty()).
Multivalued Properties
As we mentioned earlier, a property can have multiple values. We’ll discuss the more
substantial aspects of multivalued properties when we talk about queries and data
modeling. But for now, it’s worth a brief mention.
A property can have one or more values. A property cannot have zero values; a property
without a value is simply unset. Each value for a property can be of a different type,
and can be the null value.
The datastore preserves the order of values as they are assigned. The Python API returns
the values in the same order as they were set.
In Python, a property with multiple values is represented as a single Python list value:
e.prop = [1, 2, 'a', None, 'b']
Property Values | 113](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-134-2048.jpg)
![Because a property must have at least one value, it is an error to assign
an empty list ([] in Python) to a property on an entity whose Python
class is based on the Expando class.
class Entity(db.Expando):
pass
e = Entity()
e.prop = [] # ERROR
In contrast, the Model base class includes a feature that automatically
translates between the empty list value and “no property set.” You’ll see
this feature in Chapter 7.
In the Java low-level datastore API, you can store multiple values for a property using
a Collection type. The low-level API returns the values as a java.util.List. The items
are stored in the order provided by the Collection type’s iterator. For many types, such
as SortedSet or TreeSet, this order is deterministic. For others, such as HashSet, it is
not. If the app needs the original data structure, it must convert the List returned by
the datastore to the appropriate type.
Keys and Key Objects
The key for an entity is a value that can be retrieved, passed around, and stored like
any other value. If you have the key for an entity, you can retrieve the entity from the
datastore quickly, much more quickly than with a datastore query. Keys can be stored
as property values, as an easy way for one entity to refer to another.
The Python API represents an entity key value as an instance of the Key class, in
the db package. To get the key for an entity, you call the entity object’s key() method.
The Key instance provides access to its several parts using accessor methods, including
the kind, key name (if any), and system-assigned ID (if the entity does not have a key
name).
The Java low-level API is similar: the getKey() method of the Entity class returns an
instance of the Key class.
When you construct a new entity object and do not provide a key name, the entity
object has a key, but the key does not yet have an ID. The ID is populated when the
entity object is saved to the datastore for the first time. You can get the key object prior
to saving the object, but it will be incomplete.
e = Entity()
e.prop = 123
k = e.key() # key is incomplete, has neither key name nor ID
kind = k.kind() # 'Entity'
e.put() # ID is assigned
114 | Chapter 4: Datastore Entities](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-135-2048.jpg)
![k_str = str(k)
# ...
k = db.Key(k_str)
And in Java:
String k_str = KeyFactory.keyToString(k);
// ...
Key k = KeyFactory.stringToKey(k_str);
The Java Key class’s toString() method does not return the key’s string encoding. You
must use KeyFactory.keyToString() to get the string encoding of a key.
Using Entities
Let’s look briefly at how to retrieve entities from the datastore using keys, how to inspect
the contents of entities, and how to update and delete entities. The API methods for
these features are straightforward.
Getting Entities Using Keys
Given a complete key for an entity, you can retrieve the entity from the datastore.
In the Python API, you can call the get() function in the db package with the Key object
as an argument:
from google.appengine.ext import db
k = db.Key('Entity', 'alphabeta')
e = db.get(k)
If you know the kind of the entity you are fetching, you can also use the get() class
method on the appropriate entity class. This does a bit of type checking, ensuring that
the key you provide is of the appropriate kind:
class Entity(db.Expando):
pass
e = Entity.get(k)
To fetch multiple entities in a batch, you can pass the keys to get() as a list. Given a
list, the method returns a list containing entity objects, with None values for keys that
do not have a corresponding entity in the datastore.
entities = db.get([k1, k2, k3])
116 | Chapter 4: Datastore Entities](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-137-2048.jpg)
![complete key (such as a key with a kind and a key name) and an entity already exists
with that key, the datastore replaces the existing entity with the new one.
If you want to test that an entity with a given key does not exist before
you create it, you can do so using the transaction API. You must use a
transaction to ensure that another process doesn’t create an entity with
that key after you test for it and before you create it. For more informa-
tion on transactions, see Chapter 6.
If you have several entity objects to save, you can save them all in one call using the
put() function in the db package. The put() function can also take a single entity object.
db.put(e)
db.put([e1, e2, e3])
As with a batch get, a batch put performs a single call to the service. The total size of
the call is subject to the API call limits for the datastore. The entity count is also subject
to a limit.
In Java, you can save entities using the put() method of a DatastoreService
instance. As with get(), the method takes a single Entity for a single put, or an
Iterable<Entity> for a batch put.
When the call to put() returns, the datastore is up to date, and all future queries in the
current request handler and other handlers will see the new data. The specifics of how
the datastore gets updated are discussed in detail in Chapter 6.
Deleting Entities
Deleting entities works similarly to putting entities. In Python, you can call the
delete() method on the entity object, or you can pass entity objects or Key objects to
the delete() function.
e = db.get('Entity', 'alphabeta')
e.delete()
db.delete(e)
db.delete([e1, e2, e3])
# Deleting without first fetching the entity:
k = db.Key('Entity', 'alphabeta')
db.delete(k)
In Java, you call the delete() method of the DatastoreService with either a single Key
or an Iterable<Key>.
As with gets and puts, a delete of multiple entities occurs in a single batch call to the
service, and is faster than making multiple service calls.
Using Entities | 119](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-140-2048.jpg)
![impossible to get the list of players for a particular trophy. (A query filter can’t match
patterns within string values.)
Another option is to record each trophy win in a separate property named after the
trophy. To get the list of players with a trophy, you just query for the existence of the
corresponding property. However, getting the list of trophies for a given player would
require either coding the names of all of the trophies in the display logic, or iterating
over all of the Player entity’s properties looking for trophy names.
With multivalued properties, we can store each trophy name as a separate value for the
trophies property. To access a list of all trophies for a player, we simply access the
property of the entity. To get a list of all players with a trophy, we use a query with an
equality filter on the property.
Here’s what this example looks like in Python:
p = Player.get_by_key_name(user_id)
p.trophies = ['Lava Polo Champion',
'World Building 2008, Bronze',
'Glarcon Fighter, 2nd class']
p.put()
# List all trophies for a player.
for trophy in p.trophies:
# ...
# Query all players that have a trophy.
q = Player.gql("WHERE trophies = 'Lava Polo Champion'")
for p in q:
# ...
MVPs in Python
The Python API represents the values of a multivalued property as a Python list. Each
member of the list must be of one of the supported datastore types.
class Entity(db.Expando):
pass
e = Entity()
e.prop = [ 'value1', 123, users.get_current_user() ]
Remember that list is not a datastore type; it is only the mechanism for manipulating
multivalued properties. A list cannot contain another list.
A property must have at least one value, otherwise the property does not exist. To
enforce this, the Python API does not allow you to assign an empty list to a property.
Notice that the API can’t do otherwise: if a property doesn’t exist, then the API cannot
know to represent the missing property as an empty list when the entity is retrieved
from the datastore. (This being Python, the API could return the empty list whenever
a nonexistent property is accessed, but that might be more trouble than it’s worth.)
154 | Chapter 5: Datastore Queries](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-175-2048.jpg)
![Because it is often useful for lists to behave like lists, including the ability to contain
zero items, the Python data modeling API provides a mechanism that supports assign-
ing the empty list to a property. We’ll look at this mechanism in Chapter 7.
MVPs and Equality Filters
As you’ve seen, when a multivalued property is the subject of an equality filter in a
query, the entity matches if any of the property’s values are equal to the filter value.
e1 = Entity()
e1.prop = [ 3.14, 'a', 'b' ]
e1.put()
e2 = Entity()
e2.prop = [ 'a', 1, 6 ]
e2.put()
# Returns e1 but not e2:
q = Entity.gql('WHERE prop = 3.14')
# Returns e2 but not e1:
q = Entity.gql('WHERE prop = 6')
# Returns both e1 and e2:
q = Entity.gql("WHERE prop = 'a'")
Recall that a query with a single equality filter uses an index that contains the keys of
every entity of the given kind with the given property and the property values. If an
entity has a single value for the property, the index contains one row that represents
the entity and the value. If an entity has multiple values for the property, the index
contains one row for each value. The index for this example is shown in Figure 5-14.
This brings us to the first of several odd-looking queries that nonetheless make sense
for multivalued properties. Since an equality filter is a membership test, it is possible
for multiple equality filters to use the same property with different values and still return
a result. An example in GQL:
SELECT * FROM Entity WHERE prop = 'a' AND prop = 'b'
App Engine uses the “merge join” algorithm described earlier for multiple equality
filters to satisfy this query using the prop single-property index. This query returns the
e1 entity because the entity key appears in two places in the index, once for each value
requested by the filters.
The way multivalued properties appear in an index gives us another way of thinking
about multivalued properties: an entity has one or more properties, each with a name
and a single value, and an entity can have multiple properties with the same name. The
API represents the values of multiple properties with the same name as a list of values
associated with that name.
Queries and Multivalued Properties | 155](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-176-2048.jpg)
![Figure 5-14. An index of two entities with multiple values for the “prop” property, with results for
WHERE prop=‘a’
The datastore does not have a way to query for the exact set of values in a multivalued
property. You can use multiple equality filters to test that each of several values belongs
to the list, but there is no filter that ensures that those are the only values that belong
to the list, or that each value appears only once.
MVPs and Inequality Filters
Just as an equality filter tests that any value of the property is equal to the filter value,
an inequality filter tests that any value of the property meets the filter criterion.
e1 = Entity()
e1.prop = [ 1, 3, 5 ]
e1.put()
e2 = Entity()
e2.prop = [ 4, 6, 8 ]
e2.put()
# Returns e1 but not e2:
q = Entity.gql("WHERE prop < 2")
# Returns e2 but not e1:
q = Entity.gql("WHERE prop > 7")
# Returns both e1 and e2:
q = Entity.gql("WHERE prop > 3")
Figure 5-15 shows the index for this example, with the results of prop > 3 highlighted.
In the case of an inequality filter, it’s possible for the index scan to match rows for a
single entity multiple times. When this happens, the first occurrence of each key in the
156 | Chapter 5: Datastore Queries](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-177-2048.jpg)
![Figure 5-15. An index of two entities with multiple values for the “prop” property, with results for
WHERE prop > 3
index determines the order of the results. If the index used for the query sorts the
property in ascending order, the first occurrence is the smallest matching value. For
descending, it’s the largest. In the example above, prop > 3 returns e2 before e1 because
4 appears before 5 in the index.
MVPs and Sort Orders
To summarize things we know about how multivalued properties are indexed:
• A multivalued property appears in an index with one row per value.
• All rows in an index are sorted by the values, possibly distributing property values
for a single entity across the index.
• The first occurrence of an entity in an index scan determines its place in the result
set for a query.
Together, these facts explain what happens when a query orders its results by a mul-
tivalued property. When results are sorted by a multivalued property in ascending or-
der, the smallest value for the property determines its location in the results. When
results are sorted in descending order, the largest value for the property determines its
location.
This has a counterintuitive—but consistent—consequence:
e1 = Entity()
e1.prop = [ 1, 3, 5 ]
e1.put()
e2 = Entity()
Queries and Multivalued Properties | 157](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-178-2048.jpg)
![e2.prop = [ 2, 3, 4 ]
e2.put()
# Returns e1, e2:
q = Entity.gql("ORDER BY prop ASC")
# Also returns e1, e2:
q = Entity.gql("ORDER BY prop DESC")
Because e1 has both the smallest value and the largest value, it appears first in the result
set in ascending order and in descending order. See Figure 5-16.
Figure 5-16. Indexes of two entities with multiple values for the “prop” property, one ascending and
one descending
MVPs and the Query Planner
The query planner tries to be smart by ignoring aspects of the query that are redundant
or contradictory. For instance, a = 3 AND a = 4 would normally return no results, so
the query planner catches those cases and doesn’t bother doing work it doesn’t need
to do. However, most of these normalization techniques don’t apply to multivalued
properties. In this case, the query could be asking, “Does this MVP have a value that
is equal to 3 and another value equal to 4?” The datastore remembers which properties
are MVPs (even those that end up with one or zero values), and never takes a shortcut
that would produce incorrect results.
But there is one exception. A query that has both an equality filter and a sort order will
drop the sort order. If a query asks for a = 3 ORDER BY a DESC and a is a single-value
property, the sort order has no effect because all of the values in the result are identical.
For an MVP, however, a = 3 tests for membership, and two MVPs that meet that
condition are not necessarily identical.
158 | Chapter 5: Datastore Queries](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-179-2048.jpg)
![SELECT * FROM Message
WHERE ANCESTOR IS KEY(MessageBoard, 'The_Archonville_Times')
ORDER BY post_date DESC
LIMIT 10
Most queries that use an ancestor filter need custom indexes. There is one unusual
exception: a query does not need a custom index if the query also contains equality
filters on properties (and no inequality filters or sort orders). In this exceptional case,
the “merge join” algorithm can use a built-in index of keys along with the built-in
property indexes. In cases where the query would need a custom index anyway, the
query can match the ancestor to the keys in the custom index.
As we mentioned in Chapter 5, the datastore supports queries over entities of all kinds.
Kindless queries are limited to key filters and ancestor filters. Since ancestors can have
children of disparate kinds, kindless queries are useful for getting every child of a given
ancestor, regardless of kind:
SELECT * WHERE ANCESTOR IS KEY('MessageBoard', 'The_Archonville_Times')
While ancestor queries can be useful, don’t get carried away building
large ancestor trees. Remember that every entity with the same root
belongs to the same entity group, and more simultaneous users that
need to write to a group mean a greater likelihood of concurrency
failures.
If you want to model hierarchical relationships between entities without
the consequences of entity groups, consider using multivalued proper-
ties to store paths. For example, if there’s an entity whose path in your
hierarchy can be represented as /A/B/C/D, you can store this path as:
e.parents = ['/A', '/A/B', '/A/B/C'] Then you can perform a query
similar to an ancestor query on this property: ... WHERE parents = '/
A/B'.
What Can Happen in a Transaction
Entity groups ensure that the operations performed within a transaction see a consistent
view of the entities in a group. For this to work, a single transaction must limit its
operations to entities in a single group. The entity group determines the scope of the
transaction.
Within a transaction, you can fetch, update, or delete an entity using the entity’s key.
You can also create a new entity that either is a root entity of a new group that becomes
the subject of the transaction, or that has a member of the transaction’s entity group
as its parent.
You can perform queries over the entities of a single entity group in a transaction. A
query in a transaction must have an ancestor filter that matches the transaction’s entity
group. The results of the query, including both the indexes that provide the results as
168 | Chapter 6: Datastore Transactions](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-189-2048.jpg)
![p = Entity()
e = Entity(parent=p)
The parent key does not have to represent an entity that exists. You can construct a
fake key using db.Key.from_path() and use it as the parent:
p_key = db.Key.from_path('Entity', 'fake_parent')
e = Entity(parent=p_key)
You can use methods on the model object to get the key for the entity’s parent, or fetch
the entity’s parent from the datastore. These methods return None if the entity is a root
entity for its entity group.
p = e.parent()
p_key = e.parent_key()
Similarly, the Key object can return the Key of its parent, or None:
e_key = e.key()
p_key = e_key.parent()
The Python API uses function objects to handle transactions. To perform multiple
operations in a transaction, you define a function that executes the operations, then
you pass it to the run_in_transaction() function (in the ...ext.db module).
import datetime
from google.appengine.ext import db
class MessageBoard(db.Expando):
pass
class Message(db.Expando):
pass
def create_message_txn(board_name, message_name, message_title, message_text):
board = db.get(db.Key.from_path('MessageBoard', board_name))
if not board:
board = MessageBoard(key_name=board_name)
board.count = 0
message = Message(key_name=message_name, parent=board)
message.title = message_title
message.text = message_text
message.post_date = datetime.datetime.now()
board.count += 1
db.put([board, message])
# ...
try:
db.run_in_transaction(create_message_txn,
170 | Chapter 6: Datastore Transactions](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-191-2048.jpg)
![Batch Updates
When you read, create, update, or delete an entity, the runtime environment makes a
service call to the datastore. Each service call has some overhead, including serializing
and deserializing parameters and transmitting them between machines in the data cen-
ter. If you need to update multiple entities, you can save time by performing the updates
together as a batch in one service call.
We introduced batch calls in Chapter 4. Here’s a quick example of the Python batch
API:
# Creating multiple entities:
e1 = Message(key_name='m1', text='...')
e2 = Message(key_name='m2', text='...')
e3 = Message(key_name='m3', text='...')
message_keys = db.put([e1, e2, e3])
# Getting multiple entities using keys:
message_keys = [db.Key('Message', 'm1'),
db.Key('Message', 'm2'),
db.Key('Message', 'm2')]
messages = db.get(message_keys)
for message in messages:
# ...
# Deleting multiple entities:
db.delete(message_keys)
When the datastore receives a batch call, it bundles the keys or entities by their entity
groups, which it can determine from the keys. Then it dispatches calls to the datastore
machines responsible for each entity group. The datastore returns results to the app
when it has received all results from all machines.
If the call includes changes for multiple entities in a single entity group, those changes
are performed in a single transaction. There is no way to control this behavior, but
there’s no reason to do it any other way. It’s faster to commit multiple changes to a
group at once than to commit them individually, and no less likely to result in concur-
rency failures.
Each entity group involved in a batch update may fail to commit due to a concurrency
failure. If a concurrency failure occurs for any update, the API raises the concurrency
failure exception—even if updates to other groups were committed successfully.
Batch updates in disparate entity groups are performed in separate threads, possibly
by separate datastore machines, executed in parallel to one another. This can make
batch updates especially fast compared to performing each update one at a time.
Remember that if you use the batch API during a transaction, every entity or key in the
batch must use the same entity group as the rest of the transaction. Also, be aware that
App Engine puts a maximum size on service calls and responses, and the total size of
the call or its responses must not exceed these limits. The number of items in a batch
Batch Updates | 179](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-200-2048.jpg)
![s.song_key = 'C# min'
# The song_key attribute is stored as the
# datastore property named 'key'.
s.put()
You can declare that a property should contain only one of a fixed set of values by
providing a list of possible values as the choices argument. If None is not one of the
choices, this acts as a more restrictive form of required: the property must be set to one
of the valid choices using a keyword argument to the constructor.
_KEYS = ['C', 'C min', 'C 7',
'C#', 'C# min', 'C# 7',
# ...
]
class Song(db.Model):
song_key = db.StringProperty(choices=_KEYS)
s = Song(song_key='H min') # db.BadValueError
s = Song() # db.BadValueError, None is not an option
s = Song(song_key='C# min') # OK
All of these features validate the value assigned to a property, and raise a
db.BadValueError if the value does not meet the appropriate conditions. For even
greater control over value validation, you can define your own validation function and
assign it to a property declaration as the validator argument. The function should take
the value as an argument, and raise a db.BadValueError (or an exception of your choos-
ing) if the value should not be allowed.
def is_recent_year(val):
if val < 1923:
raise db.BadValueError
class Book(db.Model):
copyright_year = db.IntegerProperty(validator=is_recent_year)
b = Book(copyright_year=1922) # db.BadValueError
b = Book(copyright_year=1924) # OK
Nonindexed Properties
In Chapter 5, we mentioned that you can set properties of an entity in such a way that
they are available on the entity, but are considered unset for the purposes of indexes.
In the Python API, you establish a property as nonindexed using a property declaration.
If the property declaration is given an indexed argument of False, entities created with
that model class will set that property as nonindexed.
188 | Chapter 7: Data Modeling with Python](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-209-2048.jpg)
![is saved for the first time, and left untouched thereafter. If the current user is not signed
in, the value is set to None.
class BookReview(db.Model):
created_by_user = db.UserProperty(auto_current_user_add=True)
last_edited_by_user = db.UserProperty(auto_current_user=True)
br = BookReview()
br.put() # created_by_user and last_edited_by_user set
# ...
br.put() # last_edited_by_user set again
At first glance, it might seem reasonable to set a default for a
db.UserProperty this way:
from google.appengine.api import users
class BookReview(db.Model):
created_by_user = db.UserProperty(
default=users.get_current_user())
# WRONG
This would set the default value to be the user who is signed in when
the class is imported. Requests that use a cached instance of the appli-
cation may end up using a previous user instead of the current user as
the default.
To guard against this mistake, db.UserProperty does not accept
the default argument. You can use only auto_current_user or
auto_current_user_add to set an automatic value.
List Properties
The data modeling API provides a property declaration class for multivalued properties,
called db.ListProperty. This class ensures that every value for the property is of the
same type. You pass this type to the property declaration, like so:
class Book(db.Model):
tags = db.ListProperty(basestring)
b = Book()
b.tags = ['python', 'app engine', 'data']
The type argument to the db.ListProperty constructor must be the Python represen-
tation of one of the native datastore types. Refer back to Table 4-1 for a complete list.
The datastore does not distinguish between a multivalued property with no elements
and no property at all. As such, an undeclared property on a db.Expando object can’t
store the empty list. If it did, when the entity is loaded back into an object, the property
simply wouldn’t be there, potentially confusing code that’s expecting to find an empty
190 | Chapter 7: Data Modeling with Python](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-211-2048.jpg)
![list. To avoid confusion, db.Expando disallows assigning an empty list to an undeclared
property.
The db.ListProperty declaration makes it possible to keep an empty list value on a
multivalued property. The declaration interprets the state of an entity that doesn’t have
the declared property as the property being set to the empty list, and maintains that
distinction on the object. This also means that you cannot assign None to a declared list
property—but this isn’t of the expected type for the property anyway.
The datastore does distinguish between a property with a single value and a multivalued
property with a single value. An undeclared property on a db.Expando object can store
a list with one element, and represent it as a list value the next time the entity is loaded.
The example above declares a list of string values. (basestring is the Python base
type for str and unicode.) This case is so common that the API also provides
db.StringListProperty.
You can provide a default value to db.ListProperty using the default argument. If you
specify a nonempty list as the default, a shallow copy of the list value is made for each
new object that doesn’t have an initial value for the property.
db.ListProperty does not support the required validator, since every list property tech-
nically has a list value (possibly empty). If you wish to disallow the empty list, you can
provide your own validator function that does so:
def is_not_empty(lst):
if len(lst) == 0:
raise db.BadValueError
class Book(db.Model):
tags = db.ListProperty(basestring, validator=is_not_empty)
b = Book(tags=[]) # db.BadValueError
b = Book() # db.BadValueError, default "tags" is empty
b = Book(tags=['awesome']) # OK
db.ListProperty does not allow None as an element in the list because it doesn’t match
the required value type. It is possible to store None as an element in a list for an unde-
clared property.
Models and Schema Migration
Property declarations prevent the application from creating an invalid data object, or
assigning an invalid value to a property. If the application always uses the same model
classes to create and manipulate entities, then all entities in the datastore will be con-
sistent with the rules you establish using property declarations.
Property Declarations | 191](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-212-2048.jpg)
![b.put()
br = BookReview()
br.book = b # sets br's 'book' property to b's key
br.book = b.key() # same thing
We’ll explain what collection_name does in a moment.
The referenced object must have a “complete” key before it can be assigned to a refer-
ence property. A key is complete when it has all of its parts, including the string name
or the system-assigned numeric ID. If you create a new object without a key name, the
key is not complete until you save the object. When you save the object, the system
completes the key with a numeric ID. If you create the object (or a db.Key) with a key
name, the key is already complete, and you can use it for a reference without saving it
first.
b = Book()
br = BookReview()
br.book = b # db.BadValueError, b's key is not complete
b.put()
br.book = b # OK, b's key has system ID
b = Book(key_name='The_Grapes_of_Wrath')
br = BookReview()
br.book = b # OK, b's key has a name
db.put([b, br])
A model class must be defined before it can be the subject of a db.ReferenceProperty.
To declare a reference property that can refer to another instance of the same class, you
use a different declaration, db.SelfReferenceProperty.
class Book(db.Model):
previous_edition = db.SelfReferenceProperty()
b1 = Book()
b2 = Book()
b2.previous_edition = b1
Reference properties have a powerful and intuitive syntax for accessing referenced ob-
jects. When you access the value of a reference property, the property fetches the entity
from the datastore using the stored key, then returns it as an instance of its model class.
A referenced entity is loaded “lazily”: it is not fetched from the datastore until the
property is dereferenced.
br = db.get(book_review_key)
# br is a BookReview instance
title = br.book.title # fetches book, gets its title property
Modeling Relationships | 193](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-214-2048.jpg)
![The key list method is best suited for situations where there are fewer objects on one
side of the relationship than on the other, and the short list is small enough to store
directly on an entity. In this example, many players each belong to a few guilds; each
player has a short list of guilds, while each guild may have a long list of players. We
put the list property on the Player side of the relationship to keep the entity small, and
use queries to produce the long list when it is needed.
The link model method
The link model method represents each relationship as an entity. The relationship entity
has reference properties pointing to the related classes. You traverse the relationship
by going through the relationship entity via the back-references.
class Player(db.Model):
name = db.StringProperty()
class Guild(db.Model):
name = db.StringProperty()
class GuildMembership(db.Model):
player = db.ReferenceProperty(Player, collection_name='guild_memberships')
guild = db.ReferenceProperty(Guild, collection_name='player_memberships')
p = Player()
g = Guild()
db.put([p, g])
gm = GuildMembership(player=p, guild=g)
db.put(gm)
# Guilds to which a player belongs:
for gm in p.guild_memberships:
guild_name = gm.guild.name
# ...
# Players that belong to a guild:
for gm in g.player_memberships:
player_name = gm.player.name
# ...
This technique is similar to how you’d use “join tables” in a SQL database. It’s a good
choice if either side of the relationship may get too large to store on the entity itself.
You can also use the relationship entity to store metadata about the relationship (such
as when the player joined the guild), or model more complex relationships between
multiple classes.
The link model method is more expensive than the key list method. It requires fetching
the relationship entity to access the related object.
Modeling Relationships | 197](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-218-2048.jpg)
![self.first_name = first_name
self.surname = surname
def is_valid(self):
return (isinstance(self.first_name, unicode)
and isinstance(self.surname, unicode)
and len(self.surname) >= 6)
class PlayerNameProperty(db.Property):
data_type = basestring
def validate(self, value):
value = super(PlayerNameProperty, self).validate(value)
if value is not None:
if not isinstance(value, PlayerName):
raise db.BadValueError('Property %s must be a PlayerName.' %
(self.name))
# Let the data class have a say in validity.
if not value.is_valid():
raise db.BadValueError('Property %s must be a valid PlayerName.' %
self.name)
# Disallow the serialization delimiter in the first field.
if value.surname.find('|') != -1:
raise db.BadValueError(('PlayerName surname in property %s cannot ' +
'contain a "|".') % self.name)
return value
def get_value_for_datastore(self, model_instance):
# Convert the data object's PlayerName to a unicode.
return (getattr(model_instance, self.name).surname + u'|'
+ getattr(model_instance, self.name).first_name)
def make_value_for_datastore(self, value):
# Convert a unicode to a PlayerName.
i = value.find(u'|')
return PlayerName(first_name=value[i+1:],
surname=value[:i])
And here’s how you’d use it:
p = Player()
p.player_name = PlayerName('Ned', 'Nederlander')
p.player_name = PlayerName('Ned', 'Neder|lander')
# db.BadValueError, surname contains serialization delimiter
p.player_name = PlayerName('Ned', 'Neder')
# db.BadValueError, PlayerName.is_valid() == False
Here, the application value type is a PlayerName instance, and the datastore value type
is that value encoded as a Unicode string. The encoding format is the surname field,
followed by a delimiter, followed by the first_name field. We disallow the delimiter
character in the surname using the validate() method. (Instead of disallowing
Creating Your Own Property Classes | 203](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-224-2048.jpg)
![selects the requested fields and assembles the results. This is only true if one of the
requested fields is a datastore property, and is not true if the only field is a key field (@Id).
If the query is for one field, each result is a value of the type of that field. If the query
is for multiple fields, each result is an Object[] whose elements are the field values in
the order specified in the query.
Query query = em.createQuery("SELECT isbn, title, author FROM Book");
// Fetch complete Book objects, then
// produce result objects from 3 fields
// of each result
@SuppressWarnings("unchecked")
List<Object[]> results = (List<Object[]>) query.getResultList();
for (Object[] result : results) {
String isbn = result[0];
String title = result[1];
String author = result[2];
// ...
}
You specify filters on fields using a WHERE clause and one or more conditions:
SELECT b FROM Book b WHERE author = "John Steinbeck"
AND copyrightYear >= 1940
To filter on the entity key, refer to the field that represents the key in the data class (the
@Id field):
SELECT b FROM Book b WHERE author = "John Steinbeck"
AND isbn > :firstKeyToFetch
You can perform an ancestor filter by establishing a parent key field (as we did in the
previous section) and referring to that field in the query:
SELECT br FROM BookReview br WHERE bookKey = :pk
As with find(), you can use any of the four key types with parameterized queries, but
the most portable way is to use the type used in the class.
You specify sort orders using an ORDER BY clause. Multiple sort orders are comma-
delimited. Each sort order can have a direction of ASC (the default) or DESC.
SELECT b FROM Book b ORDER BY rating DESC title
The App Engine implementation of JPQL includes a couple of additional tricks that
the datastore can support natively. One such trick is the string prefix trick:
SELECT b FROM Book b WHERE title LIKE 'The Grape%'
The implementation translates this to WHERE title >= 'The Grape', which does the
same thing: it returns all books with a title that begins with the string The Grape, in-
cluding "The Grape", "The Grapefruit", and "The Grapes of Wrath".
Queries and JPQL | 219](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-240-2048.jpg)
![Keys are strings. The API also accepts a memcached-style tuple of a hash code and a string
as a key for compatibility, but it ignores the hash code. The string can be any size;
strings larger than 250 bytes are hashed to 250 bytes.
Values can be any Python value that can be serialized using the pickle module. This
includes most simple values, and data structures and objects whose members are
pickleable values. Pretty much any db.Model instance falls into this category. You can
check the documentation for the pickle module for a more complete description.
Setting and Getting Values in Python
To store a value in memcache, you call the set() function (or method, if you’re using
the Client class) with the key and the value:
from google.appengine.api import memcache
# headlines is a pickle-able object
headlines = ['...', '...', '...']
success = memcache.set("headlines", headlines)
if not success:
# Problem accessing memcache...
set() returns True on success, or False if the value could not be stored.
If there already exists a value with the given key, set() overwrites it. If there is no value
for the key, set() sets it.
To set a value only if there is no value for the key, you call the add() function with
similar arguments:
success = memcache.add("headlines", headlines)
if not success:
# Key already set, or another problem setting...
add() works like set(), but does not update the value and returns False if a value is
currently set for the given key.
To update a value only if it’s already set, you call the replace() function, also with
similar arguments:
success = memcache.replace("headlines", headlines)
if not success:
# Key not set, or another problem setting...
To get a value, you call the get() method with the key:
headlines = memcache.get("headlines")
if headlines is None:
# Key not set, or another problem getting...
get() returns the value, or None if there is no value set for the given key.
The Python Memcache API | 229](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-250-2048.jpg)
![Setting and Getting Multiple Values
You can set and get multiple values at a time using the batch interface. A batch call to
memcache is performed as a single call to the service, which is faster than multiple
individual calls for many items. Like the datastore batch operations, the total amount
of data sent to or received from memcache in a single call must not exceed one
megabyte.
set_multi(), add_multi(), and replace_multi() are similar to their singular counter-
parts, except instead of taking a key and a value as arguments, they take a Python
mapping (such as a dict) of keys to values.
article_summaries = {'article00174': '...',
'article05234': '...',
'article15820': '...',
}
failed_keys = memcache.set_multi(article_summaries)
if failed_keys:
# One or more keys failed to update...
These methods return a list of keys that failed to be set. A completely successful batch
update returns the empty list.
To get multiple values in a single batch, you call the get_multi() method with a list of
keys:
article_summary_keys = ['article00174',
'article05234',
'article15820',
]
article_summaries = memcache.get_multi(article_summary_keys)
get_multi() returns a dict of keys and values that were present and retrieved
successfully.
All of the batch operations accept a key_prefix keyword argument that prepends every
key with the given prefix string. When used with get_multi(), the keys on the returned
values do not include the prefix.
article_summaries = {'00174': '...',
'05234': '...',
'15820': '...',
}
failed_keys = memcache.set_multi(article_summaries,
key_prefix='article')
if failed_keys:
# ...
# ...
article_summary_keys = ['00174',
230 | Chapter 9: The Memory Cache](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-251-2048.jpg)
!['05234',
'15820',
]
article_summaries = memcache.get_multi(article_summary_keys,
key_prefix='article')
Batch calls to the memcache service are not atomic actions.
set_multi() may be successful for some items and not others.
get_multi() will return the data seen at the time of each individual get,
and does not guarantee that all retrieved data is read at the same time.
Memcache Namespaces
You can specify an optional namespace for a memcache key with any setter or getter
call. Namespaces let you segment the key space by the kind of thing you are storing,
so you don’t have to worry about key collisions between disparate uses of the cache.
All of the setter and getter functions accept a namespace argument, a string that declares
the namespace for all keys involved in the call. The batch functions use one namespace
for all keys in the batch.
article_summaries = {'article00174': '...',
'article05234': '...',
'article15820': '...',
}
failed_keys = memcache.set_multi(article_summaries,
namespace='News')
if failed_keys:
# ...
# ...
summary = memcache.get('article05234',
namespace='News')
Namespaces are different from key prefixes in batch calls. Key prefixes merely prepend
keys in a batch with a common string, as a convenience for batch processing. The
namespace is a separate part of the key, and must be specified with an explicit
namespace argument in all calls.
Cache Expiration
You can tell the memcache to delete (or evict) a value automatically after a period of
time when you set or update the value.
All of the setter methods accept a time argument. Its value is either a number of seconds
from the current time up to one month (up to 2,678,400 seconds), or it is a Unix epoch
time (seconds since January 1, 1970, such as returned by time.time()) in the future.
The Python Memcache API | 231](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-252-2048.jpg)
![The callback function is called without arguments. But a common use for a callback
function is to process the results of the fetch, so the function needs access to the RPC
object. There are several ways to give the callback function access to the object.
One way is to use a bound method, a feature of Python that lets you refer to a method
of an instance of a class as a callable object. Define a class with a method that processes
the results of the fetch, using an RPC object stored as a member of the class. Create an
instance of the class, then create the RPC object, assigning the bound method as the
callback by passing it as the callback argument to the create_rpc() function. Exam-
ple 10-6 demonstrates this technique.
Example 10-6. Using an object method as a callback to access the RPC object
from google.appengine.api import urlfetch
# ...
class CatalogUpdater(object):
def prepare_urlfetch_rpc(self):
self.rpc = urlfetch.create_rpc(callback=self.process_results)
urlfetch.make_fetch_call(self.rpc, 'http://api.example.com/catalog_feed')
return self.rpc
def process_results(self):
try:
results = self.rpc.get_result()
# Process results.content...
except urlfetch.Error, e:
# Handle urlfetch errors...
class MainHandler(webapp.RequestHandler):
def get(self):
rpcs = []
catalog_updater = CatalogUpdater(self.response)
rpcs.append(catalog_updater.prepare_urlfetch_rpc())
# ...
for rpc in rpcs:
rpc.wait()
Another way to give the callback access to the RPC object is to use a nested function
(sometimes called a closure). If the callback function is defined in the same scope as a
variable whose value is the RPC object, the function can access the variable when it is
called. To allow for this style of calling, the URL Fetch API lets you assign the callback
function to the RPC object’s callback property after the object has been created.
Example 10-7 demonstrates the use of a nested function as a callback. The
create_callback() function creates a function object, a lambda expression, that calls
another function with the RPC object as an argument. This function object is assigned
to the callback property of the RPC object.
248 | Chapter 10: Fetching URLs and Web Resources](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-269-2048.jpg)
![Example 11-1. An example of sending an email message in Python, using several features
from google.appengine.api import mail
def send_registration_key(user_addr, software_key_data):
message_body = '''
Thank you for purchasing The Example App, the best
example on the market! Your registration key is attached
to this email.
To install your key, download the attachment, then select
"Register..." from the Help menu. Select the key file, then
click "Register".
You can download the app at any time from:
http://www.example.com/downloads/
Thanks again!
The Example Team
'''
html_message_body = '''
<p>Thank you for purchasing The Example App, the best
example on the market! Your registration key is attached
to this email.</p>
<p>To install your key, download the attachment, then select
<b>Register...</b> from the <b>Help</b> menu. Select the key file, then
click <b>Register</b>.</p>
<p>You can download the app at any time from:</p>
<p>
<a href="http://www.example.com/downloads/">
http://www.example.com/downloads/
</a>
</p>
<p>Thanks again!</p>
<p>The Example Team<br />
<img src="http://www.example.com/images/logo_email.gif" /></p>
'''
message = mail.EmailMessage(
sender='The Example Team <admin@example.com>',
to=user_addr,
subject='Your Example Registration Key',
body=message_body,
html=html_message_body,
attachments=[('example_key.txt', software_key_data)])
message.send()
260 | Chapter 11: Sending and Receiving Mail and Instant Messages](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-281-2048.jpg)
![// There was an error contacting the Mail service.
// ...
}
As shown here, you call methods on the MimeMessage to set fields and to add recipients
and content. The simplest message has a sender (setFrom()), one “To” recipient
(addRecipient()), a subject (setSubject()), and a plain-text message body (setText()).
The setFrom() method takes an InternetAddress. You can create an InternetAddress
with just the email address (a String) or the address and a human-readable name as
arguments to the constructor. The email address of the sender must meet the require-
ments described earlier. You can use any string for the human-readable name.
The addRecipient() method takes a recipient type and an InternetAddress. The
allowed recipient types are Message.RecipientType.TO (“To,” a primary recipient),
Message.RecipientType.CC (“Cc” or “carbon-copy,” a secondary recipient), and
Message.RecipientType.BCC (“Bcc” or “blind carbon-copy,” where the recipient is sent
the message but the address does not appear in the message content). You can call
addRecipient() multiple times to add multiple recipients of any type.
The setText() method sets the plain-text body for the message. To include an
HTML version of the message body for mail readers that support HTML, you create a
MimeMultipart object, then create a MimeBodyPart for the plain-text body and another
for the HTML body and add them to the MimeMultipart. You then make the
MimeMultipart the content of the MimeMessage:
import javax.mail.Multipart;
import javax.mail.internet.MimeBodyPart;
import javax.mail.internet.MimeMultipart;
// ...
String textBody = "...text...";
String htmlBody = "...HTML...";
Multipart multipart = new MimeMultipart();
MimeBodyPart textPart = new MimeBodyPart();
textPart.setContent(textBody, "text/plain");
multipart.addBodyPart(htmlPart);
MimeBodyPart htmlPart = new MimeBodyPart();
htmlPart.setContent(htmlBody, "text/html");
multipart.addBodyPart(htmlPart);
message.setContent(multipart);
You attach files to the email message in a similar way:
Multipart multipart = new MimeMultipart();
// ...
byte[] fileData = getBrochureData();
String fileName = "brochure.pdf";
262 | Chapter 11: Sending and Receiving Mail and Instant Messages](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-283-2048.jpg)
![The development server console (http://localhost:8080/_ah/admin/) includes a fea-
ture for simulating incoming email by submitting a web form. The development server
cannot receive actual email messages.
If the app has the incoming mail service enabled but does not have a
request handler for the appropriate URL, or if the request handler re-
turns an HTTP response code other than 200 for the request, the mes-
sage gets “bounced” and the sender receives an error email message.
Receiving Email in Python
To receive email in Python, you map the incoming email URL path to a script handler
in app.yaml file:
handlers:
- url: /_ah/mail/.+
script: handle_email.py
The app address used for the message is included in the URL path, so you can set up
separate handlers for different addresses directly in the configuration:
handlers:
- url: /_ah/mail/support%40.*app-id.appspotmail.com
script: support_contact.py
- url: /_ah/mail/.+
script: handle_email.py
Email addresses are URL-encoded in the final URL, so this pattern uses %40 to represent
an @ symbol. Also notice you must include a .* before the application ID when using
this technique, so the pattern works for messages sent to version-specific addresses
(such as support@dev.latest.app-id.appspotmail.com).
The Python SDK includes a class for parsing the POST content into a convenient object,
called InboundEmailMessage (in the google.appengine.api.mail package). It takes the
multipart MIME data (the POST body) as an argument to its constructor. Here’s an
example using the webapp framework:
from google.appengine.api import mail
from google.appengine.ext import webapp
from google.appengine.ext.webapp.util import run_wsgi_app
class IncomingMailHandler(webapp.RequestHandler):
def post(self):
message = mail.InboundEmailMessage(self.request.body)
# sender = message.sender
# recipients = message.to
# body = list(message.bodies(content_type='text/plain'))[0]
# ...
application = webapp.WSGIApplication([('/_ah/mail/.+', IncomingMailHandler)],
debug=True)
264 | Chapter 11: Sending and Receiving Mail and Instant Messages](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-285-2048.jpg)
![def main():
run_wsgi_app(application)
if __name__ == '__main__':
main()
The InboundEmailMessage object includes attributes for the fields of the message, similar
to EmailMessage. sender is the sender’s email address, possibly with a displayable name
in the standard format (Mr. Sender <sender@example.com>). to is a list of primary
recipient addresses, and cc is a list of secondary recipients. (There is no bcc on an
incoming message, because blind-carbon-copied recipients are not included in the
message content.) subject is the message’s subject.
The InboundEmailMessage object may have more than one message body: an HTML
body and a plain-text body. You can iterate over the MIME multipart parts of the types
text/html and text/plain using the bodies() method. Without arguments, this method
returns an iterator that returns the HTML parts first, then the plain-text parts. You can
limit the parts returned to just the HTML or plain-text parts by setting the
content_type parameter. For example, to get just the plain-text bodies:
for text_body in message.bodies(content_type='text/plain'):
# ...
In the example earlier, we extracted the first plain-text body by passing the iterator to
the list() type, then indexing its first argument (which assumes one exists):
text = list(message.bodies(content_type='text/plain'))[0]
If the incoming message has file attachments, then these are accessible on the
attachments attribute. As with using EmailMessage for sending, this attribute is a list of
tuples whose first element is the filename and whose second element is the data byte
string. InboundEmailMessage allows all file types for incoming attachments, and does
not require that the filename accurately represent the file type. Be careful when using
files sent to the app by users, since they may not be what they say they are, and have
not been scanned for viruses.
The Python SDK includes a convenient webapp handler base class
for processing incoming email, called InboundMailHandler in the
google.appengine.ext.webapp.mail_handlers package. You use the handler by creating
a subclass that overrides the receive() method, then installing it like any other handler.
When the handler receives an email message, the receive() method is called with an
InboundEmailMessage object as its argument.
from google.appengine.ext.webapp import mail_handlers
class MyMailHandler(mail_handlers.InboundMailHandler):
def receive(self, message):
# ...
application = webapp.WSGIApplication([('/_ah/mail/.+', MyMailHandler)],
Receiving Email Messages | 265](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-286-2048.jpg)
![The body content of the HTTP POST request is a MIME multipart message, with a
part for each field of the message:
from
The sender’s JID.
to
The app JID to which this message was sent.
body
The message body content (with characters as they were originally typed).
stanza
The full XML stanza of the message, including the previous fields (with XML spe-
cial characters escaped); useful for communicating with a custom client using
XML.
The Python and Java SDKs include classes for parsing the request data into objects.
When the service is enabled, all incoming messages are accepted and routed to the app.
All incoming chat invitations are accepted automatically. The app always appears as
connected and available when someone inquires for the app’s presence.
The development server console (http://localhost:8080/_ah/admin/) includes a fea-
ture for simulating incoming XMPP messages by submitting a web form. The devel-
opment server cannot receive actual XMPP messages.
Receiving XMPP Messages in Python
In Python, you map the URL path to a script handler in app.yaml file, as usual:
handlers:
- url: /_ah/xmpp/message/chat/
script: handle_xmpp.py
You can parse the incoming message into a Message object by passing a mapping of the
POST parameters to its constructor. With the webapp framework, this is a simple
matter of passing the parsed POST data (a mapping of the POST parameter names to
values) in directly:
from google.appengine.api import xmpp
from google.appengine.ext import webapp
from google.appengine.ext.webapp.util import run_wsgi_app
class IncomingXMPPHandler(webapp.RequestHandler):
def post(self):
message = xmpp.Message(self.request.POST)
message.reply('I got your message!')
application = webapp.WSGIApplication([('/_ah/xmpp/message/chat/',
IncomingXMPPHandler)],
debug=True)
Receiving XMPP Messages | 273](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-294-2048.jpg)
![def main():
run_wsgi_app(application)
if __name__ == '__main__':
main()
The Message object has attributes for each message field: from, to, and body. It also
includes a convenience method for replying to the message: reply(), which takes the
body of the reply as its first argument.
The Python SDK includes an extension for webapp that makes it easy to implement
chat interfaces that perform user-issued commands. Here’s an example that responds
to the commands /stats and /score username, and ignores all other messages:
from google.appengine.api import xmpp
from google.appengine.ext import webapp
from google.appengine.ext.webapp import xmpp_handlers
from google.appengine.ext.webapp.util import run_wsgi_app
def get_stats():
# ...
def get_score_for_user(username):
# ...
class UnknownUserError(Exception):
pass
class ScoreBotHandler(xmpp_handlers.CommandHandler):
def stats_command(self, message):
stats = get_stats()
if stats:
message.reply('The latest stats: %s' % stats)
else:
message.reply('Stats are not available right now.')
def score_command(self, message):
try:
score = get_score_for_user(message.arg)
message.reply('Score for user %s: %d' % (message.arg, score))
except UnknownUserError, e:
message.reply('Unknown user %s' % message.arg)
application = webapp.WSGIApplication([('/_ah/xmpp/message/chat/', ScoreBotHandler)],
debug=True)
def main():
run_wsgi_app(application)
if __name__ == '__main__':
main()
274 | Chapter 11: Sending and Receiving Mail and Instant Messages](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-295-2048.jpg)
![public class XMPPReceiverServlet extends HttpServlet {
public void doPost(HttpServletRequest req,
HttpServletResponse resp)
throws IOException {
XMPPService xmpp = XMPPServiceFactory.getXMPPService();
Message message = xmpp.parseMessage(req);
// ...
}
}
You access the fields of a Message using methods:
getFromJid()
The sender JID.
getMessageType()
The MessageType of the message (MessageType.CHAT or MessageType.NORMAL).
getRecipientJids()
The app JID used, as a single-element JID[].
getBody()
The String content of the message.
getStanza()
The String raw XML of the message.
276 | Chapter 11: Sending and Receiving Mail and Instant Messages](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-297-2048.jpg)
![the sqlite3 module installed by starting a Python shell and importing it. If nothing
happens, the module is installed:
% python
Python 2.6.1 (r261:67515, Jul 7 2009, 23:51:51)
[GCC 4.2.1 (Apple Inc. build 5646)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import sqlite3
>>>
If you do not have the module, you can get it from the SQLite website. Note that if you
compiled your Python installation from source, you will need to recompile it with
SQLite support.
http://www.sqlite.org/download.html
If you’d prefer not to install SQLite, you can disable the use of the progress datafile in
the bulk loader tool by providing this command-line argument when running the tool:
--db_filename=skip.
Backup and Restore
One of the easiest features of the bulk loader to use is the backup and restore feature.
The bulk loader can download every entity of a given kind to a datafile, and can upload
that datafile back to the app to restore those entities. It can also upload the datafile to
another app, effectively moving all of the entities of a kind from one app to another.
You can also use these features to download data from the live app, and “upload” it to
the development server for testing (using the --url=... argument mentioned earlier).
These features require no additional configuration or code beyond the installation of
the remote API handler.
To download every entity of a given kind, run the bulkloader.py tool with the --dump,
--kind=..., and --filename=... arguments:
bulkloader.py --dump
--app_id=app-id
--url=http://app-id.appspot.com/remote_api
--kind=kind
--filename=backup.dat
(This is a single command line. The characters are line continuation markers, sup-
ported by some command shells. You can remove these characters and type all of the
arguments on one line.)
To upload a datafile created by --dump, thereby re-creating every entity in the file, use
the --restore argument instead of --dump:
bulkloader.py --restore
--app_id=app-id
--url=http://app-id.appspot.com/remote_api
--kind=kind
--filename=backup.dat
Using the Bulk Loader Tool | 281](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-302-2048.jpg)
![return datetime.datetime.strptime(s, '%m/%d/%Y').date()
else:
return None
class BookLoader(bulkloader.Loader):
def __init__(self):
bulkloader.Loader.__init__(self, 'Book',
[('title', str),
('author', str),
('copyright_year', int),
('author_birthdate', get_date_from_str),
])
loaders = [BookLoader]
The loader definition file must do several things:
1. It must import or define a data model class for each kind to import, using the
Python data modeling API. The loader file does not have the app directory in the
module load path, so the loader file must add it. The modules should be imported
directly into the local namespace.
2. It must define (or import) a loader class for each kind to import, a subclass of the
Loader class in the package google.appengine.tools.bulkloader. This class must
have a constructor (an __init__() method) that calls the parent class’s constructor
with arguments that describe the CSV file’s columns.
3. It must define a module-global variable named loaders, whose value is a list of
loader classes.
The loader class’s __init__() method calls its parent class’s __init__() method with
three arguments: self (the object being initialized), the kind of the entity to create as
a string, and a list of tuples. Each tuple represents a column from the CSV file, in order
from left to right. The first element of each tuple is the property name to use. The second
element is a Python function that takes the CSV file value as a string and returns the
desired datastore value in the appropriate Python value type. (Refer to Table 4-1 for a
list of the Python value types for the datastore.)
For simple data types, you can use the constructor for the type as the type conversion
function. In the example above, we use str to leave the value from the CSV file as a
string, and int to parse the CSV string value as an integer. Depending on the output of
your spreadsheet program, you can use this technique for long, float, db.Text, and
other types whose constructors accept strings as initial arguments. For instance,
users.User takes an email address as a string argument.
The bool type constructor returns False for the empty string, and True in all other cases.
If your CSV file contains Boolean values in other forms, you’ll need to write a converter
function, like this:
def bool_from_string(s):
return s == 'TRUE'
Using the Bulk Loader Tool | 283](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-304-2048.jpg)
![Dates, times, and date-times usually require a custom converter function to make an
appropriate datetime.datetime value from the string. As shown previously, you can use
the datetime.datetime.strptime() function to parse the string in a particular format.
Tailor this to the output of your spreadsheet program.
To create a multivalued property, you can use a type conversion function that returns
a list value. However, the function must get all of its data from a single cell (though you
can work around this by tweaking the file reader routine, described below). It cannot
reference other cells in the row, or other rows in the table.
Storing a key value (a reference to another entity) is usually a straightforward matter
of defining a type conversion function that uses db.Key.from_path() with the desired
kind and a key name derived from a CSV file value.
By default, the bulk loader creates each entity using a key with no ancestors, the given
kind, and a system-assigned numeric ID. You can tell the bulk loader to use a custom
key name instead of a numeric ID by defining a method of your loader class named
generate_key(). This method takes as arguments a loop counter (i) and the row’s col-
umn values as a list of strings (values), and returns either a string key name or a
db.Key object. The db.Key object can include an entity group parent key. If the function
returns None, a system-assigned numeric ID will be used.
class BookLoader(bulkloader.Loader):
# ...
def generate_key(self, i, values):
# Use the first column as the key name.
return values[0]
If an entity is uploaded with the same key as an entity that already exists
in the datastore, the new entity will replace the existing entity. You can
use this effect to download, modify, then reupload data to update ex-
isting entities instead of creating new ones.
You can “skip” an unused column in the CSV file by giving it a property name that
begins with an underscore. The Python data modeling API does not save attributes
beginning with an underscore as properties, so the final entity will not have a property
with this name. You can reuse the same dummy property name for all columns you
wish to skip:
class BookLoader(bulkloader.Loader):
def __init__(self):
bulkloader.Loader.__init__(self, 'Book',
[('_UNUSED', lambda x: None),
('title', str),
('author', str),
('_UNUSED', lambda x: None),
('_UNUSED', lambda x: None),
('copyright_year', int),
284 | Chapter 12: Bulk Data Operations and Remote Access](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-305-2048.jpg)
![# ...
])
If you have a column that is to be the entity key name (via generate_key()) and you do
not want this value to also become a property, you must “skip” the column in the
column list using this technique. The bulk loader will complain if it finds more columns
than are declared in the list.
The bulk loader uses the Python data modeling API to construct, validate, and save
entities. If your app is written in Python, you just use the model classes of your app for
this purpose, and get the validation for free.
If your app is written in Java, you have a choice to make. You can either implement
Python data model classes that resemble your application’s data model, or you can
define Python data models that allow arbitrary properties to be set by the loader without
validation. For information on defining data models in Python, see Chapter 7. To define
a model class that allows all properties and does no validation, simply define a class
like the following for each kind in the loader file, above the definition of the loader
classes:
from google.appengine.ext import db
class Book(db.Expando):
pass
The loader class can do postprocessing on an entity after it is created but before it is
saved to the datastore. This gives you the opportunity to inspect the entity, add or
remove properties, and even create additional entities based on each row of the datafile.
To perform postprocessing, implement a method named handle_entity() on the loader
class. This takes an instance of the model class as its only argument (entity). It should
return the instance to create, a list of model instances if more than one entity should
be created, or None if no entity should be created.
You can extend the loader class to read file formats other than CSV files. To do
this, you override the generate_records() method. This method takes a filename
(filename), and returns a Python generator. Each iteration of the generator should yield
a list of strings for each entity to create, where each string corresponds to a property
defined in the loader’s constructor. You can use a generate_records() routine to mas-
sage the columns processed by the loader so that all the information needed for a single
property appears in the column’s value, such as to assemble what will eventually be-
come a multivalued property from multiple columns in the datafile.
You can override the initialize() method of the loader to perform actions at the be-
ginning of an upload. This method takes the input file’s name (filename) and a string
given on the command line as the --loader_opts=... argument (loader_opts), which
you can use to pass options to this method. Similarly, you can override the
finalize() method to do something after the upload is complete.
Using the Bulk Loader Tool | 285](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-306-2048.jpg)
![specifies which property goes in which column of the resulting CSV file, and can include
value transformation functions to get the entity data into a format accepted by your
spreadsheet program.
Here is a simple exporter definition file that is symmetric with the loader definition
example in the previous section. Data downloaded with this exporter can be loaded
back in with the loader. Let’s call this file exporters.py:
from google.appengine.tools import bulkloader
import datetime
# Import the app's data models directly into
# this namespace. We must add the app
# directory to the path explicitly.
import sys
import os.path
sys.path.append(
os.path.abspath(
os.path.dirname(
os.path.realpath(__file__))))
from models import *
def make_str_for_date(d):
if d:
return d.strftime('%m/%d/%Y')
else:
return ''
class BookExporter(bulkloader.Exporter):
def __init__(self):
bulkloader.Exporter.__init__(self, 'Book',
[('title', str, None),
('author', str, ''),
('copyright_year', int, ''),
('author_birthdate', make_str_for_date, ''),
])
exporters = [BookExporter]
The exporter definition file must do several things:
1. It must import or define a data model class for each kind to export, using the Python
data modeling API. The exporter file does not have the app directory in the module
load path, so the exporter file must add it. The modules should be imported directly
into the local namespace.
2. It must define (or import) an exporter class for each kind to export, a subclass of
the Exporter class in the package google.appengine.tools.bulkloader. This class
must have a constructor (an __init__() method) that calls the parent class’s con-
structor with arguments that describe the CSV file’s columns.
3. It must define a module-global variable named exporters, whose value is a list of
exporter classes.
Using the Bulk Loader Tool | 287](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-308-2048.jpg)
![Python 2.5.1 (r251:54863, Feb 6 2009, 19:02:12)
[GCC 4.0.1 (Apple Inc. build 5465)]
The db, users, urlfetch, and memcache modules are imported.
clock> import os.path
clock> import sys
clock> sys.path.append(os.path.realpath('.'))
clock> import models
clock> books = models.Book.all().fetch(6)
clock> books
[<models.Book object at 0x7a2c30>, <models.Book object at 0x7a2bf0>,
<models.Book object at 0x7a2cd0>, <models.Book object at 0x7a2cb0>,
<models.Book object at 0x7a2d30>, <models.Book object at 0x7a2c90>]
clock> books[0].title
u'The Grapes of Wrath'
clock> from google.appengine.api import mail
clock> mail.send_mail('juliet@example.com', 'test@example.com',
'Test email', 'This is a test message.')
clock>
To exit the shell, press Ctrl-D.
Using the Remote API from a Script
You can call the remote API directly from your own Python scripts using a library from
the Python SDK. This configures the Python API to use the remote API handler for your
application for all service calls, so you can use the service APIs as you would from a
request handler directly in your scripts.
Here’s a simple example script that prompts for a developer account email address and
password, then accesses the datastore of a live application:
#!/usr/bin/python
import getpass
import sys
# Add the Python SDK to the package path.
# Adjust these paths accordingly.
sys.path.append('~/google_appengine')
sys.path.append('~/google_appengine/lib/yaml/lib')
from google.appengine.ext.remote_api import remote_api_stub
from google.appengine.ext import db
import models
# Your app ID and remote API URL path go here.
APP-ID = 'app-id'
REMOTE_API_PATH = '/remote_api'
def auth_func():
email_address = raw_input('Email address: ')
password = getpass.getpass('Password: ')
Using the Remote API from a Script | 291](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-312-2048.jpg)
![return email_address, password
def initialize_remote_api(app-id=APP-ID,
path=REMOTE_API_PATH):
remote_api_stub.ConfigureRemoteApi(
app-id,
path,
auth_func)
remote_api_stub.MaybeInvokeAuthentication()
def main(args):
initialize_remote_api()
books = models.Book.all().fetch(10)
for book in books:
print book.title
return 0
if __name__ == '__main__':
sys.exit(main(sys.argv[1:]))
The ConfigureRemoteApi() function (yes, it has a TitleCase name) sets up the remote
API access. It takes as arguments the application ID, the remote API handler URL path,
and a callable that returns a tuple containing the email address and password to use
when connecting. In this example, we define a function that prompts for the email
address and password, and pass the function to ConfigureRemoteApi().
The function also accepts an optional fourth argument specifying an alternate domain
name for the connection. By default, it uses app-id.appspot.com, where app-id is the
application ID in the first argument.
The MaybeInvokeAuthentication() function sends an empty request to verify that the
email address and password are correct, and raises an exception if they are not. (With-
out this, the script would wait until the first remote call to verify the authentication.)
Remember that every call to an App Engine library that performs a service call does so
over the network via an HTTP request to the application. This is inevitably slower than
running within the live application. It also consumes application resources like web
requests do, including bandwidth and request counts, which are not normally con-
sumed by service calls in the live app.
On the plus side, since your code runs on your local computer, it is not constrained by
the App Engine runtime sandbox or the 30-second request deadline. You can run long
jobs and interactive applications on your computer without restriction, using any Py-
thon modules you like—at the expense of consuming app resources to marshal service
calls over HTTP.
292 | Chapter 12: Bulk Data Operations and Remote Access](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-313-2048.jpg)
![And here’s the request handler script:
from google.appengine.ext import webapp
from google.appengine.ext.webapp.util import run_wsgi_app
from google.appengine.ext import deferred
import invitation
class SendInvitationHandler(webapp.RequestHandler):
def get(self):
# recipient = ...
deferred.defer(invitation.send_invitation, recipient)
# ...
application = webapp.WSGIApplication([
('/sendinvite', SendInvitationHandler),
], debug=True)
def main():
run_wsgi_app(application)
if __name__ == '__main__':
main()
You set up the request handler for deferred work using the following configuration in
your app.yaml file:
handlers:
- url: /_ah/queue/deferred
script: $PYTHON_LIB/google/appengine/ext/deferred/
login: admin
This maps the URL path /_ah/queue/deferred to the deferred module in the API.
($PYTHON_LIB is replaced with the path to the Python directory containing the SDK on
App Engine.) Tasks created with defer() use this URL path by default. login: admin
restricts this handler to task queues and application developers.
The defer() function enqueues a task on the default queue that calls the given callable
object with the given arguments. (Despite its use of the URL /_ah/queue/deferred for
deferred tasks, this utility uses the default queue, not a queue named deferred.) The
arguments are serialized and deserialized using Python’s pickle module; all argument
values must be pickle-able.
Most Python callable objects can be used with defer(), including functions and classes
defined at the top level of a module, methods of objects, class methods, instances of
classes that implement __call__(), and built-in functions and methods. defer() does
not work with lambda functions, nested functions, nested classes, or instances of nested
classes. The task handler must be able to access the callable object by name, possibly
via a serializable object, since it does not preserve the scope of the call to defer().
You also can’t use a function or class in the same module as the request handler class
from which you call defer(). This is because pickle believes the module of the request
handler class to be __main__ while it is running, and so it doesn’t save the correct
Task Queues | 303](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-324-2048.jpg)
![The url() method sets a custom URL for the task. If no custom URL is set, the task
calls the default URL for the queue.
The param() method adds a request parameter to be included with the task’s HTTP
request. The method takes the name of the parameter, and either a String or a
byte[] as the value of the parameter. Parameters are added to the HTTP request body
URL-encoded, similar to a web form. The removeParam() method removes a parameter
from the TaskOptions object; it takes the parameter name as its argument.
You can also set the HTTP request body with the payload() method. This method takes
the request body as a String. You can also include a character set specification (such
as "utf-8") as the second argument. You can’t set both a payload and parameters for
the same task.
You can control other aspects of the HTTP request as well. The method() method sets
the HTTP method; the default is "POST". You can also add arbitrary HTTP headers by
calling the header() method (which takes a String key and String value as arguments)
for each header, or by calling the headers() method with a Map<String, String>.
You can set a delay for the task. The countdownMillis() method sets the delay as a
number of milliseconds (a long) from the time the task is enqueued. The etaMillis()
method sets the delay to be an absolute date and time, as a number of milliseconds
from midnight, January 1, 1970, UTC.
To set the name for a task, you call the name() method. As described earlier, the name
ensures that only one task with that name can be enqueued for a period of time, at least
seven days with the current implementation. If you do not specify a name, the system
assigns a unique name that has not been used in the past seven days.
An app can have up to 10 named queues, each with its own task execution rate (token
replenishment rate and bucket size). In a Java app, you configure queues using a con-
figuration file named queue.xml, in the WEB-INF/ directory, which looks like this:
<queue-entries>
<queue>
<name>fastest</name>
<rate>10/s</rate>
<bucket-size>10</bucket-size>
</queue>
<queue>
<name>limitedemail</name>
<rate>1000/d</rate>
<bucket-size>5</bucket-size>
</queue>
</queue-entries>
The file includes one <queue> element per queue to be configured. Each queue has a
<name>, a token replenishment <rate>, and a <bucket-size>.
306 | Chapter 13: Task Queues and Scheduled Tasks](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-327-2048.jpg)
![Because App Engine promises not to make backward-incompatible changes to the
runtime environment, the default version of Django on App Engine (in version 1 of the
Python runtime) is still 0.96. However, App Engine includes a mechanism for explicitly
selecting a newer version, such that all subsequent attempts to import Django modules
will use the new version. You can use this mechanism in your main request handler
script to use Django 1.1 for your app.
Before you do that, you must download and install Django 1.1 on your local computer.
You won’t need to add it to your application directory (it’s available when running on
App Engine), but you will need it locally while building and testing your application.
Go to the Django project website and download Django 1.1:
http://www.djangoproject.com/download/1.1/tarball/
Follow the instructions to install the Python modules in your local Python environment
by running the following commands. (The sudo command will require your computer’s
administrator password.)
tar xzvf Django-1.1.tar.gz
cd Django-1.1
sudo python setup.py install
Django 1.1 was the latest release when this book was published. If you
see a newer version available on the Django website, be sure to check
the App Engine documentation to confirm that the new version has been
added to the Python runtime environment before using it locally.
If you’d prefer to use your own copy of Django instead of the one pro-
vided by the Python runtime environment (such as a newer version not
yet in the environment), you can add it directly to your application di-
rectory. Doing so requires an additional step to remove the version of
Django that’s part of the runtime environment from the Python module
search path. To do this, put this in your main request handler script
prior to importing django modules:
for k in [k for k in sys.modules if k.startswith('django')]:
del sys.modules[k]
Creating a Django Project
In Django, the complete set of code, configuration, and static files for a website is called
a project. You create a new project by running a command that was installed with
Django, called django-admin.py, like so:
django-admin.py startproject myproject
This command creates in your current working directory a new directory named after
the project, in this case myproject/, with several starter files:
Creating a Django Project | 315](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-336-2048.jpg)
![__init__.py
A file that tells Python that code files in this directory can be imported as modules
(this directory is a Python package).
manage.py
A command-line utility you will use to build and manage this project, with many
features.
settings.py
Configuration for this project, in the form of a Python source file.
urls.py
Mappings of URL paths to Python code, as a Python source file.
If you’re following along with a Django tutorial or book, the next step is usually to start
the Django development server using the manage.py command. If you did so now, you
would be running the Django server, but it would know nothing of App Engine. To
run the project in the App Engine development server, we must connect the Django
project to the App Engine runtime environment using a request handler script and
configuration in the app.yaml file.
The Request Handler Script
All that’s left to do is to invoke the Django project from a handler script. Exam-
ple 14-1 is a minimal script that hands the request off to Django 1.1 using the WSGI
runner function from the webapp library. Create a file named main.py in the Django
project directory you created in the previous step (e.g., myproject/).
Example 14-1. A simple request handler for invoking the Django framework
from google.appengine.dist import use_library
use_library('django', '1.1')
import os
os.environ['DJANGO_SETTINGS_MODULE'] = 'settings'
import django.core.handlers.wsgi
from google.appengine.ext.webapp import util
def main():
# Run Django via WSGI.
application = django.core.handlers.wsgi.WSGIHandler()
util.run_wsgi_app(application)
if __name__ == '__main__':
main()
The use_library() function is an App Engine utility that tells the runtime environment
that version 1.1 of Django should be used when importing libraries from the django
package. Without this call, importing from django would use Django 0.96.
316 | Chapter 14: The Django Web Application Framework](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-337-2048.jpg)
![InstallAppengineHelperForDjango(), and adding the version identifier as an argument
to that function call. For instance, to select Django 1.0:
InstallAppengineHelperForDjango('1.0')
Now that the Helper is installed, you can start the App Engine development server
using manage.py as you would with any other Django project:
python manage.py runserver
Thanks to the Helper, this is no different than running the dev_appserver.py command
or using the Launcher to start the server. If you’d prefer to use those methods, you may,
without losing any functionality of the Helper.
Test that the default project is working locally by visiting the server’s URL in a web
browser. By default, Django uses port 8000, not 8080, so this URL is:
http://localhost:8000/
The Helper adds several features to manage.py that work with the App Engine devel-
opment server. One especially useful feature is the shell command, which starts a
Python shell in the context of the Django project—and the development server. You
can use this shell to import App Engine modules and interact with the development
server, the simulated sandbox, and the simulated services.
python manage.py shell
Here is a simple shell session that demonstrates calling the development server’s
simulated Mail service (several harmless warning messages have been elided):
% python manage.py shell
Python 2.6.1 (r261:67515, Jul 7 2009, 23:51:51)
[GCC 4.2.1 (Apple Inc. build 5646)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
(InteractiveConsole)
>>> from google.appengine.api import mail
>>> mail.send_mail('sender@example.com', 'to@example.com', 'my subject',
'message body')
INFO:root:MailService.Send
INFO:root: From: sender@example.com
INFO:root: To: to@example.com
INFO:root: Subject: my subject
INFO:root: Body:
INFO:root: Content-type: text/plain
INFO:root: Data length: 12
INFO:root:You are not currently sending out real email. If you have sendmail
installed you can use it by using the server with --enable_sendmail
>>>
You can open a Python shell that connects to the live application via remote_api using
the console command. This requires that you set up remote_api as described in Chap-
ter 12 using a URL path of /remote_api.
python manage.py console
The Django App Engine Helper | 319](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-340-2048.jpg)
![from appengine_django.models import BaseModel
from google.appengine.ext import db
class Book(BaseModel):
title = db.StringProperty()
author = db.StringProperty()
copyright_year = db.IntegerProperty()
author_birthdate = db.DateProperty()
class BookReview(BaseModel):
book = db.ReferenceProperty(Book, collection_name='reviews')
review_author = db.UserProperty()
review_text = db.TextProperty()
rating = db.StringProperty(choices=['Poor', 'OK', 'Good', 'Very Good', 'Great'],
default='Great')
create_date = db.DateTimeProperty(auto_now_add=True)
Edit views.py to query the database for books and pass the query object to the template,
like this:
from django.shortcuts import render_to_response
from google.appengine.ext import db
from bookstore import models
def home(request):
q = models.Book.all().order('title')
return render_to_response('bookstore/index.html',
{ 'books': q })
Finally, edit templates/bookstore/index.html to display the list of books:
<html>
<body>
<p>Welcome to The Book Store!</p>
<p>Books in our catalog:</p>
<ul>
{% for book in books %}
<li>{{ book.title }}, by {{ book.author }} ({{ book.copyright_year }})</li>
{% endfor %}
</ul>
</body>
</html>
Notice that we’re passing the db.Query object directly to the template. The template
uses the query object’s iterator interface to fetch the Book entities from the datastore.
If you load the page right now (and feel free to do so to test the code), you’ll see an
empty list of books. Let’s create some books from the Python shell. Stop the server (hit
Ctrl-C), then start the shell:
python manage.py shell
You can create objects in the development server’s datastore by typing Python code
directly into the shell. Here is a complete shell session, including output, that creates
several Book entities:
Using App Engine Models With Django | 323](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-344-2048.jpg)
![% python manage.py shell
Python 2.6.1 (r261:67515, Jul 7 2009, 23:51:51)
[GCC 4.2.1 (Apple Inc. build 5646)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
(InteractiveConsole)
>>> from bookstore import models
>>> import datetime
>>> b = models.Book(title='Grapes of Wrath, The', author='Steinbeck, John',
copyright_year=1939, author_birthdate=datetime.date(1902, 2, 27))
>>> b.put()
datastore_types.Key.from_path(u'Book', 1,
_app_id_namespace=u'google-app-engine-django')
>>> b = models.Book(title='Cat in the Hat, The', author='Dr. Seuss',
copyright_year=1957, author_birthdate=datetime.date(1904, 3, 2))
>>> b.put()
datastore_types.Key.from_path(u'Book', 2,
_app_id_namespace=u'google-app-engine-django')
>>>
Exit the shell (press Ctrl-D), then restart the server and load the page.
Using Django Unit Tests and Fixtures
The Django App Engine Helper includes support for unit tests using Django’s unit test
runner and the App Engine development server. The unit test runner runs tests inside
the development server’s environment, so tests can access the simulated services, in-
cluding the simulated datastore. The test runner supports tests defined with the
standard Python libraries unittest or doctest, or other test frameworks with test runner
customizations. Django provides its own class for defining unittest test cases, called
django.test.TestCase, which adds several Django-specific features.
For unittest tests, the runner looks in files in the Django application directory (e.g.,
bookstore/) named models.py and tests.py for classes that inherit from
unittest.TestCase or django.test.TestCase. Each method of a test case class whose
name begins with test is a separate test. For each test, the runner initializes the test
environment, instantiates the TestCase class, and calls the method for the test. If the
class defines a setUp() method, it calls it before each test. If the class defines a tear
Down() method, it calls it after each test. You can use setUp() and tearDown() to create
and clean up test data and preparation for every test in a test case.
The TestCase class provides methods that cause the test to report failure if the condi-
tions tested are not met. self.assertEqual() takes two arguments, and causes the test
to fail if the arguments are not equal. self.assert_() takes one argument, and fails the
test if the argument is a false value (False, None, 0, the empty string, and so forth).
self.assertRaises() takes an exception class, a callable object, and arguments for the
callable object, calls the callable, then fails the test if the callable does not raise the given
exception. There are several other test methods, as well. Each of these methods takes
an optional message argument that is printed in the test report if the test fails.
324 | Chapter 14: The Django Web Application Framework](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-345-2048.jpg)
![Enter Python statements to create test data in the datastore. Here is a shell session that
creates a few Book and BookReview entities:
% python manage.py shell
Python 2.6.1 (r261:67515, Jul 7 2009, 23:51:51)
[GCC 4.2.1 (Apple Inc. build 5646)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
(InteractiveConsole)
>>> from google.appengine.ext import db
>>> from bookstore import models
>>> b1 = models.Book(title="one", author="")
>>> b1.put()
datastore_types.Key.from_path(u'Book', 1,
_app_id_namespace=u'google-app-engine-django')
>>> br1 = models.BookReview(book=b1, review_text="review1")
>>> br2 = models.BookReview(book=b1, review_text="review2")
>>> br3 = models.BookReview(book=b1, review_text="review3")
>>> db.put(br1)
datastore_types.Key.from_path(u'BookReview', 2,
_app_id_namespace=u'google-app-engine-django')
>>> db.put(br2)
datastore_types.Key.from_path(u'BookReview', 3,
_app_id_namespace=u'google-app-engine-django')
>>> db.put(br3)
datastore_types.Key.from_path(u'BookReview', 4,
_app_id_namespace=u'google-app-engine-django')
>>> b2 = models.Book(title="two", author="")
>>> b2.put()
datastore_types.Key.from_path(u'Book', 5,
_app_id_namespace=u'google-app-engine-django')
>>> b2br1 = models.BookReview(book=b2, review_text="review1")
>>> b2br2 = models.BookReview(book=b2, review_text="review2")
>>> db.put(b2br1)
datastore_types.Key.from_path(u'BookReview', 6,
_app_id_namespace=u'google-app-engine-django')
>>> db.put(b2br2)
datastore_types.Key.from_path(u'BookReview', 7,
_app_id_namespace=u'google-app-engine-django')
>>>
Exit the shell (Ctrl-D).
Create a directory named fixtures/ in the bookstore/ app directory. Now run the
dumpdata command, and redirect its output to a file in this new directory:
python manage.py dumpdata bookstore >bookstore/fixtures/bookstore.json
The resulting file contains representations of all of the datastore entities for the kinds
whose models are defined in the bookstore app, in the JSON format.
To load this fixture for every test in a test case, you create a class attribute for the
django.test.TestCase named fixtures, whose value is a list of the names of the fixture
files to load from the fixtures/ directory. Here’s another (trivial) test case that uses
fixtures, which you can put in models.py:
326 | Chapter 14: The Django Web Application Framework](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-347-2048.jpg)
![# ...
class TestBooksAndReviews(test.TestCase):
fixtures = ['bookstore.json']
def testBookReviewOrder(self):
d = Book.all().filter('title =', 'one').get()
self.assert_(d, 'Book "one" must be in test fixture')
q = BookReview.all().filter('book =', d)
q.order('-create_date')
reviews = q.fetch(3)
self.assertEqual('review3', reviews[0].review_text)
self.assertEqual('review2', reviews[1].review_text)
self.assertEqual('review1', reviews[2].review_text)
Run the new test using the test command, as before:
python manage.py test bookstore
For each test in the TestBooksAndReviews test case, Django flushes the simulated data-
store, loads the fixtures, calls the test case’s setUp() method (if any), then calls the test
method, and finally calls the tearDown() method (if any).
You can also load fixtures into the project’s development datastore manually using the
python manage.py loaddata command:
python manage.py loaddata bookstore/fixtures/bookstore.json
See the Django documentation for more information about unit testing and fixtures.
Using Django Forms
Django includes a powerful feature for building web forms based on data model defi-
nitions. The Django forms library can generate HTML for forms, validate that submit-
ted data meets the requirements of the model, and redisplay the form to the user with
error messages. The default appearance is useful, and you can customize the appearance
extensively.
Django’s version of the form library only works with the Django data modeling API,
and does not work directly with App Engine models. Thankfully, the App Engine Py-
thon runtime environment includes an App Engine–compatible implementation of
Django forms, in the package google.appengine.ext.db.djangoforms.
We won’t go into the details of how Django forms work—see the Django documen-
tation for a complete explanation—but let’s walk through a quick example to see how
the pieces fit together. Our example will use the following behavior for creating and
editing Book entities:
• An HTTP GET request to /books/book/ displays an empty form for creating a new
Book.
Using Django Forms | 327](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-348-2048.jpg)
![To display the form, we simply pass the BookForm object to a template. There are several
methods on the object for rendering it in different ways; we’ll use the as_p() method
to display the form fields in <p> elements.
Create that template now, named templates/bookstore/bookform.html:
<html>
<body>
{% if book_id %}
<p>Edit book {{ book_id }}:</p>
<form action="/books/book/{{ book_id }}" method="POST">
{% else %}
<p>Create book:</p>
<form action="/books/book/" method="POST">
{% endif %}
{{ form.as_p }}
<input type="submit" />
</form>
</body>
</html>
The template is responsible for outputting the <form> tag and the Submit button. The
BookForm does the rest.
Before we test this, let’s make one more change to the book-listing template to add
some “create” and “edit” links. Edit templates/bookstore/index.html:
<html>
<body>
<p>Welcome to The Book Store!</p>
<p>Books in our catalog:</p>
<ul>
{% for book in books %}
<li>{{ book.title }}, by {{ book.author }} ({{ book.copyright_year }})
[<a href="/books/book/{{ book.key.id }}">edit</a>]</li>
{% endfor %}
</ul>
<p>[<a href="/books/book/">add a book</a>]</p>
</body>
</html>
Restart the development server, then load the book list URL (/books/) in the browser.
Click “add a book” to show the book creation form. Enter some data for the book, then
submit it to create it.
The default form widget for a date field is just a text field, and it’s finicky
about the format. In this case, the “author birthdate” field expects input
in the form YYYY-MM-DD, such as 1902-02-27.
330 | Chapter 14: The Django Web Application Framework](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-351-2048.jpg)
![Figure 15-3. The Administration Console log browser
The following command downloads request logs for the app in the development di-
rectory clock, and saves them to a file named logs.txt:
appcfg.py request_logs clock logs.txt
Request data appears in the file in a common format known as the Apache Combined
(or “NCSA Combined”) logfile format, one request per line (shown here as two lines
to fit on the page):
127.0.0.1 - - [05/Jul/2009:06:46:30 -0700] "GET /blog/ HTTP/1.1" 200 14598 -
"Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10_5_8; en-us)...,gzip(gfe)"
From left to right, the fields are:
• The IP address of the client
• A - (an unused field retained for backward compatibility)
• The email address of the user who made the request, if the user is signed in using
Google Accounts; - otherwise
• The date and time of the request
• The HTTP command string in double quotes, including the method and URL path
• The HTTP response code returned by the server
• The size of the response, as a number of bytes
• The “Referrer” header provided by the client, usually the URL of the page that
linked to this URL
• The “User-Agent” header provided by the client, usually identifying the browser
and its capabilities
340 | Chapter 15: Deploying and Managing Applications](https://image.slidesharecdn.com/programminggoogleappengine-100608200404-phpapp01/75/Programming-google-app-engine-361-2048.jpg)
Programming google app engine
- 2. Programming Google AppEngine
- 4. Programming Google AppEngine Dan Sanderson Beijing • Cambridge • Farnham • Köln • Sebastopol • Taipei • Tokyo
- 5. Programming Google AppEngine by Dan Sanderson Copyright © 2010 Dan Sanderson. All rights reserved. Printed in the United States of America. Published by O’Reilly Media, Inc., 1005 Gravenstein Highway North, Sebastopol, CA 95472. O’Reilly books may be purchased for educational, business, or sales promotional use. Online editions are also available for most titles (http://my.safaribooksonline.com). For more information, contact our corporate/institutional sales department: (800) 998-9938 or corporate@oreilly.com. Editor: Mike Loukides Indexer: Ellen Troutman Zaig Production Editor: Sumita Mukherji Cover Designer: Karen Montgomery Proofreader: Sada Preisch Interior Designer: David Futato Illustrator: Robert Romano Printing History: November 2009: First Edition. Nutshell Handbook, the Nutshell Handbook logo, and the O’Reilly logo are registered trademarks of O’Reilly Media, Inc. Programming Google App Engine, the image of a waterbuck, and related trade dress are trademarks of O’Reilly Media, Inc. Many of the designations used by manufacturers and sellers to distinguish their products are claimed as trademarks. Where those designations appear in this book, and O’Reilly Media, Inc., was aware of a trademark claim, the designations have been printed in caps or initial caps. While every precaution has been taken in the preparation of this book, the publisher and author assume no responsibility for errors or omissions, or for damages resulting from the use of the information con- tained herein. TM This book uses RepKover™, a durable and flexible lay-flat binding. ISBN: 978-0-596-52272-8 [M] 1257864694
- 6.
- 8. Table of Contents Preface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii 1. Introducing Google App Engine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 The Runtime Environment 2 The Static File Servers 4 The Datastore 4 Entities and Properties 5 Queries and Indexes 6 Transactions 6 The Services 8 Google Accounts 9 Task Queues and Cron Jobs 9 Developer Tools 10 The Administration Console 11 Things App Engine Doesn’t Do...Yet 12 Getting Started 13 2. Creating an Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 Setting Up the SDK 15 Installing the Python SDK 16 Installing the Java SDK 20 Developing the Application 24 The User Preferences Pattern 24 Developing a Python App 25 Developing a Java App 39 The Development Console 54 Registering the Application 55 The Application ID and Title 57 Setting Up a Domain Name 58 Google Apps and Authentication 59 Uploading the Application 60 vii
- 9. Introducing the AdministrationConsole 61 3. Handling Web Requests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 The App Engine Architecture 64 Configuring the Frontend 66 Configuring a Python App 66 Configuring a Java App 68 Domain Names 69 App IDs and Versions 70 Request Handlers 72 Static Files and Resource Files 75 Secure Connections 81 Authorization with Google Accounts 83 How the App Is Run 85 The Python Runtime Environment 86 The Java Runtime Environment 87 The Sandbox 88 App Caching 89 Logging 93 Quotas and Limits 96 Request Limits 96 CPU Limits 97 Service Limits 98 Deployment Limits 98 Billable Quotas 100 Resource Usage Headers 101 4. Datastore Entities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103 Entities, Keys, and Properties 104 Introducing the Python Datastore API 105 Introducing the Java Datastore API 108 Property Values 110 Strings, Text, and Blobs 112 Unset Versus the Null Value 112 Multivalued Properties 113 Keys and Key Objects 114 Using Entities 116 Getting Entities Using Keys 116 Inspecting Entity Objects 117 Saving Entities 118 Deleting Entities 119 viii | Table of Contents
- 10. 5. Datastore Queries. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121 Queries and Kinds 122 Query Results and Keys 122 GQL 123 The Python Query API 126 The Query Class 127 GQL in Python 128 Retrieving Results 129 Keys-Only Queries 131 The Java Query API 132 Keys-Only Queries in Java 133 Introducing Indexes 134 Automatic Indexes and Simple Queries 136 All Entities of a Kind 137 One Equality Filter 137 Greater-Than and Less-Than Filters 138 One Sort Order 139 Queries on Keys 141 Kindless Queries 142 Custom Indexes and Complex Queries 143 Multiple Sort Orders 143 Filters on Multiple Properties 144 Multiple Equality Filters 147 Not-Equal and IN Filters 150 Unset and Nonindexed Properties 150 Sort Orders and Value Types 152 Queries and Multivalued Properties 153 A Simple Example 153 MVPs in Python 154 MVPs and Equality Filters 155 MVPs and Inequality Filters 156 MVPs and Sort Orders 157 Exploding Indexes 159 Configuring Indexes 159 Index Configuration for Python 160 Index Configuration for Java 161 6. Datastore Transactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163 Entities and Entity Groups 165 Keys, Paths, and Ancestors 166 Ancestor Queries 167 What Can Happen in a Transaction 168 Transactional Reads 169 Table of Contents | ix
- 11. Transactions in Python 169 Transactions in Java 172 How Entities Are Updated 175 How Entities Are Read 178 Batch Updates 179 How Indexes Are Updated 180 7. Data Modeling with Python . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183 Models and Properties 184 Property Declarations 185 Property Value Types 186 Property Validation 187 Nonindexed Properties 188 Automatic Values 189 List Properties 190 Models and Schema Migration 191 Modeling Relationships 192 One-to-Many Relationships 195 One-to-One Relationships 195 Many-to-Many Relationships 196 Model Inheritance 198 Queries and PolyModels 199 Creating Your Own Property Classes 200 Validating Property Values 201 Marshaling Value Types 202 Customizing Default Values 204 Accepting Arguments 205 8. The Java Persistence API . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207 Setting Up JPA 208 Entities and Keys 209 Entity Properties 212 Embedded Objects 213 Saving, Fetching, and Deleting Objects 214 Transactions in JPA 216 Queries and JPQL 217 Relationships 220 For More Information 225 9. The Memory Cache . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227 The Python Memcache API 228 Setting and Getting Values in Python 229 Setting and Getting Multiple Values 230 x | Table of Contents
- 12. Memcache Namespaces 231 Cache Expiration 231 Deleting Keys 232 Memcache Counters 233 Cache Statistics 233 The Java Memcache API 234 10. Fetching URLs and Web Resources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239 Fetching URLs in Python 240 Fetching URLs in Java 242 Asynchronous Requests in Python 244 RPC Objects 246 Processing Results with Callbacks 247 11. Sending and Receiving Mail and Instant Messages . . . . . . . . . . . . . . . . . . . . . . . . . 251 Enabling Inbound Services 253 Sending Email Messages 254 Sender Addresses 255 Recipients 256 Attachments 257 Sending Email in Python 258 Sending Email in Java 261 Receiving Email Messages 263 Receiving Email in Python 264 Receiving Email in Java 266 Sending XMPP Messages 267 Sending a Chat Invitation 269 Sending a Chat Message 270 Checking a Google Talk User’s Status 271 Receiving XMPP Messages 272 Receiving XMPP Messages in Python 273 Receiving XMPP Messages in Java 275 12. Bulk Data Operations and Remote Access . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277 Setting Up the Remote API for Python 278 Setting Up the Remote API for Java 279 Using the Bulk Loader Tool 280 Installing SQLite 280 Backup and Restore 281 Uploading Data 282 Downloading Data 286 Controlling the Bulk Loader 289 Using the Remote Shell Tool 290 Table of Contents | xi
- 13. Using the RemoteAPI from a Script 291 13. Task Queues and Scheduled Tasks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293 Task Queues 294 Processing Rates and Token Buckets 295 Elements of a Task 296 Task Handlers and Retries 297 Testing and Managing Tasks 299 Using Task Queues in Python 299 Using Task Queues in Java 304 Transactional Task Enqueueing 307 Scheduled Tasks 308 14. The Django Web Application Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313 Installing Django 314 Creating a Django Project 315 The Request Handler Script 316 The Django App Engine Helper 317 Creating a Django Application 320 Using App Engine Models With Django 322 Using Django Unit Tests and Fixtures 324 Using Django Forms 327 15. Deploying and Managing Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333 Uploading an Application 334 Using Versions 335 Managing Service Configuration 337 Managing Indexes 337 Browsing and Downloading Logs 339 Inspecting the Datastore 342 Application Settings 342 Managing Developers 343 Quotas and Billing 344 Getting Help 345 Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 347 xii | Table of Contents
- 14. Preface On the Internet,popularity is swift and fleeting. A mention of your website on a popular blog can bring 300,000 potential customers your way at once, all expecting to find out who you are and what you have to offer. But if you’re a small company just starting out, your hardware and software aren’t likely to be able to handle that kind of traffic. Chances are, you’ve sensibly built your site to handle the 30,000 visits per hour you’re actually expecting in your first 6 months. Under heavy load, such a system would be incapable of showing even your company logo to the 270,000 others that showed up to look around. And those potential customers are not likely to come back after the traffic has subsided. The answer is not to spend time and money building a system to serve millions of visitors on the first day, when those same systems are only expected to serve mere thousands per day for the subsequent months. If you delay your launch to build big, you miss the opportunity to improve your product using feedback from your customers. Building big before allowing customers to use the product risks building something your cus- tomers don’t want. Small companies usually don’t have access to large systems of servers on day one. The best they can do is to build small and hope meltdowns don’t damage their reputation as they try to grow. The lucky ones find their audience, get another round of funding, and halt feature development to rebuild their product for larger capacity. The unlucky ones, well, don’t. But these days, there are other options. Large Internet companies such as Amazon.com, Google, and Microsoft are leasing parts of their high-capacity systems using a pay-per-use model. Your website is served from those large systems, which are plenty capable of handling sudden surges in traffic and ongoing success. And since you pay only for what you use, there is no up-front investment that goes to waste when traffic is low. As your customer base grows, the costs grow proportionally. xiii
- 15. Google App Engine,Google’s application hosting service, does more than just provide access to hardware. It provides a model for building applications that grow automati- cally. App Engine runs your application so that each user who accesses it gets the same experience as every other user, whether there are dozens of simultaneous users or thousands. The application uses the same large-scale services that power Google’s ap- plications for data storage and retrieval, caching, and network access. App Engine takes care of the tasks of large-scale computing, such as load balancing, data replication, and fault tolerance, automatically. The App Engine model really kicks in at the point where a traditional system would outgrow its first database server. With such a system, adding load-balanced web servers and caching layers can get you pretty far, but when your application needs to write data to more than one place, you have a hard problem. This problem is made harder when development up to that point has relied on features of database software that were never intended for data distributed across multiple machines. By thinking about your data in terms of App Engine’s model up front, you save yourself from having to rebuild the whole thing later, without much additional effort. Running on Google’s infrastructure means you never have to set up a server, replace a failed hard drive, or troubleshoot a network card. And you don’t have to be woken up in the middle of the night by a screaming pager because an ISP hiccup confused a service alarm. And with automatic scaling, you don’t have to scramble to set up new hardware as traffic increases. Google App Engine lets you focus on your application’s functionality and user expe- rience. You can launch early, enjoy the flood of attention, retain customers, and start improving your product with the help of your users. Your app grows with the size of your audience—up to Google-sized proportions—without having to rebuild for a new architecture. Meanwhile, your competitors are still putting out fires and configuring databases. With this book, you will learn how to develop applications that run on Google App Engine, and how to get the most out of the scalable model. A significant portion of the book discusses the App Engine scalable datastore, which does not behave like the re- lational databases that have been a staple of web development for the past decade. The application model and the datastore together represent a new way of thinking about web applications that, while being almost as simple as the model we’ve known, requires reconsidering a few principles we often take for granted. This book introduces the major features of App Engine, including the scalable services (such as for sending email and manipulating images), tools for deploying and managing applications, and features for integrating your application with Google Accounts and Google Apps using your own domain name. The book also discusses techniques for optimizing your application, using task queues and offline processes, and otherwise getting the most out of Google App Engine. xiv | Preface
- 16. Using This Book Asof this writing, App Engine supports two technology stacks for building web applications: Java and Python. The Java technology stack lets you develop web appli- cations using the Java programming language (or most other languages that compile to Java bytecode or have a JVM-based interpreter) and Java web technologies such as servlets and JSPs. The Python technology stack provides a fast interpreter for the Python programming language, and is compatible with several major open source web appli- cation frameworks such as Django. This book covers concepts that apply to both technology stacks, as well as important language-specific subjects. If you’ve already decided which language you’re going to use, you probably won’t be interested in information that doesn’t apply to that lan- guage. This poses a challenge for a printed book: how should the text be organized so information about one technology doesn’t interfere with information about the other? Foremost, we’ve tried to organize the chapters by the major concepts that apply to all App Engine applications. Where necessary, chapters split into separate sections to talk about specifics for each language. In cases where an example in one language illustrates a concept equally well for other languages, the example is given in Python. If Python is not your language of choice, hopefully you’ll be able to glean the equivalent infor- mation from other parts of the book or from the official App Engine documentation on Google’s website. The datastore is a large enough subject that it gets multiple chapters to itself. Starting with Chapter 4, datastore concepts are introduced alongside Python and Java APIs related to those concepts. Note that we’ve taken an unconventional approach to in- troducing the datastore APIs by starting with the low-level APIs that map directly to datastore concepts. In your applications, you are most likely to prefer the higher level APIs of the data modeling interfaces. Data modeling is discussed separately, in Chap- ter 7 for Python, and in Chapter 8 for Java. Google may release additional technology stacks for other languages in the future. If they’ve done so by the time you read this, the concepts described here should still be relevant. Check this book’s website for information about future editions. This book has the following chapters: Chapter 1, Introducing Google App Engine A high-level overview of Google App Engine and its components, tools, and major features. This chapter also includes a brief discussion of features you might expect App Engine to have but that it doesn’t have yet. Chapter 2, Creating an Application An introductory tutorial for both Python and Java, including instructions on setting up a development environment, setting up accounts and domain names, and de- ploying the application to App Engine. The tutorial application demonstrates Preface | xv
- 17. the use ofseveral App Engine features—Google Accounts, the datastore, and memcache—to implement a pattern common to many web applications: storing and retrieving user preferences. Chapter 3, Handling Web Requests Contains details about App Engine’s architecture, the various features of the frontend, app servers, and static file servers, and details about the app server run- time environments for Python and Java. The frontend routes requests to the app servers and the static file servers, and manages secure connections and Google Accounts authentication and authorization. This chapter also discusses quotas and limits, and how to raise them by setting a budget. Chapter 4, Datastore Entities The first of several chapters on the App Engine datastore, a strongly consistent scalable object data storage system with support for local transactions. This chapter introduces data entities, keys and properties, and Python and Java APIs for creat- ing, updating, and deleting entities. Chapter 5, Datastore Queries An introduction to datastore queries and indexes, and the Python and Java APIs for queries. The App Engine datastore’s query engine uses prebuilt indexes for all queries. This chapter describes the features of the query engine in detail, and how each feature uses indexes. The chapter also discusses how to define and manage indexes for your application’s queries. Chapter 6, Datastore Transactions How to use transactions to keep your data consistent. The App Engine datastore uses local transactions in a scalable environment. Your app arranges its entities in units of transactionality known as entity groups. This chapter attempts to provide a complete explanation of how the datastore updates data, and how to design your data and your app to best take advantage of these features. Chapter 7, Data Modeling with Python How to use the Python data modeling API to enforce invariants in your data schema. The datastore itself is schemaless, a fundamental aspect of its scalability. You can automate the enforcement of data schemas using App Engine’s data mod- eling interface. This chapter covers Python exclusively, though Java developers may wish to skim it for advice related to data modeling. Chapter 8, The Java Persistence API A brief introduction to the Java Persistence API (JPA), how its concepts translate to the datastore, how to use it to model data schemas, and how using it makes your application easier to port to other environments. JPA is a Java EE standard inter- face. App Engine also supports another standard interface known as Java Data Objects (JDO), though JDO is not covered in this book. This chapter covers Java exclusively. xvi | Preface
- 18. Chapter 9, TheMemory Cache App Engine’s memory cache service (aka “memcache”), and its Python and Java APIs. Aggressive caching is essential for high-performance web applications. Chapter 10, Fetching URLs and Web Resources How to access other resources on the Internet via HTTP using the URL Fetch service. This chapter covers the Python and Java interfaces, including implemen- tations of standard URL fetching libraries. It also describes the asynchronous URL Fetch interface, which as of this writing is exclusive to Python. Chapter 11, Sending and Receiving Mail and Instant Messages How to use App Engine services to send email and instant messages to XMPP-compatible services (such as Google Talk). This chapter covers receiving email and XMPP chat messages relayed by App Engine using request handlers. It also discusses creating and processing messages using tools in the API. Chapter 12, Bulk Data Operations and Remote Access How to perform large maintenance operations on your live application using scripts running on your computer. Tools included with the SDK make it easy to back up, restore, load, and retrieve data in your app’s datastore. You can also write your own tools using the remote access API for data transformations and other jobs. You can also run an interactive Python command shell that uses the remote API to manipulate a live Python or Java app. Chapter 13, Task Queues and Scheduled Tasks How to perform work outside of user requests using task queues. Task queues perform tasks in parallel by running your code on multiple application servers. You control the processing rate with configuration. Tasks can also be executed on a regular schedule with no user interaction. Chapter 14, The Django Web Application Framework How to use the Django web application framework with the Python runtime en- vironment. This chapter discusses setting up a Django project, using the Django App Engine Helper, and taking advantage of features of Django via the Helper such as using the App Engine data modeling interface with forms and test fixtures. Chapter 15, Deploying and Managing Applications How to upload and run your app on App Engine, how to update and test an application using app versions, and how to manage and inspect the running ap- plication. This chapter also introduces other maintenance features of the Admin- istrator Console, including billing. We conclude with a list of places to go for help and further reading. Preface | xvii
- 19. Conventions Used inThis Book The following typographical conventions are used in this book: Italic Indicates new terms, URLs, email addresses, filenames, and file extensions. Constant width Used for program listings, as well as within paragraphs to refer to program elements such as variable or function names, databases, data types, environment variables, statements, and keywords. Constant width bold Shows commands or other text that should be typed literally by the user. Constant width italic Shows text that should be replaced with user-supplied values or by values deter- mined by context. This icon signifies a tip, suggestion, or general note. Using Code Samples This book is here to help you get your job done. In general, you may use the code in this book in your programs and documentation. You do not need to contact us for permission unless you’re reproducing a significant portion of the code. For example, writing a program that uses several chunks of code from this book does not require permission. Selling or distributing a CD-ROM of examples from O’Reilly books does require permission. Answering a question by citing this book and quoting example code does not require permission. Incorporating a significant amount of example code from this book into your product’s documentation does require permission. We appreciate, but do not require, attribution. An attribution usually includes the title, author, publisher, and ISBN. For example: “Programming Google App Engine by Dan Sanderson. Copyright 2010 Dan Sanderson, 978-0-596-52272-8.” If you feel your use of code examples falls outside fair use or the permission given above, feel free to contact us at permissions@oreilly.com. Safari® Books Online Safari Books Online is an on-demand digital library that lets you easily search over 7,500 technology and creative reference books and videos to find the answers you need quickly. xviii | Preface
- 20. With a subscription,you can read any page and watch any video from our library online. Read books on your cell phone and mobile devices. Access new titles before they are available for print, and get exclusive access to manuscripts in development and post feedback for the authors. Copy and paste code samples, organize your favorites, down- load chapters, bookmark key sections, create notes, print out pages, and benefit from tons of other time-saving features. O’Reilly Media has uploaded this book to the Safari Books Online service. To have full digital access to this book and others on similar topics from O’Reilly and other pub- lishers, sign up for free at http://my.safaribooksonline.com. How to Contact Us Please address comments and questions concerning this book to the publisher: O’Reilly Media, Inc. 1005 Gravenstein Highway North Sebastopol, CA 95472 800-998-9938 (in the United States or Canada) 707-829-0515 (international or local) 707-829-0104 (fax) We have a web page for this book, where we list errata, examples, and any additional information. You can access this page at: http://oreilly.com/catalog/9780596522728 You can also download the examples from the author’s website: http://www.dansanderson.com/appengine To comment or ask technical questions about this book, send email to: bookquestions@oreilly.com For more information about our books, conferences, Resource Centers, and the O’Reilly Network, see our website at: http://www.oreilly.com Acknowledgments I owe a great deal of thanks to the App Engine team, of which I’ve been a proud member since 2008. This book would not exist without the efforts and leadership of Paul McDonald, Pete Koomen, and App Engine’s fearless tech lead, Kevin Gibbs. I am especially indebted to the App Engine datastore team, who have made significant contributions to the datastore chapters. Ryan Barrett, lead datastore engineer, provided many hours of conversation and detailed technical review. Max Ross, implementor of Preface | xix
- 21. the Java datastoreinterfaces and the JDO and JPA adapters, wrote major portions of Chapter 8. Rafe Kaplan, designer of the Python data modeling library, contributed portions of Chapter 7. My thanks to them. Thanks to Matthew Blain, Michael Davidson, Alex Gaysinsky, Peter McKenzie, Don Schwarz, and Jeffrey Scudder for reviewing portions of the book in detail. Thanks also to Andy Smith for making last-minute improvements to the Django Helper in time to be included here. Many other App Engine contributors had a hand, directly or indi- rectly, in making this book what it is: Freeland Abbott, Mike Aizatsky, Ken Ashcraft, Anthony Baxter, Chris Beckmann, Andrew Bowers, Matthew Brown, Ryan Brown, Hannah Chen, Lei Chen, Jason Cooper, Mark Dalrymple, Pavni Diwanji, Brad Fitzpatrick, Alfred Fuller, David Glazer, John Grabowski, Joe Gregorio, Raju Gulabani, Justin Haugh, Jeff Huber, Kevin Jin, Erik Johnson, Nick Johnson, Mickey Kataria, Scott Knaster, Marc Kriguer, Alon Levi, Sean Lynch, Gianni Mariani, Mano Marks, Jon McAlister, Sean McBride, Marzia Niccolai, Alan Noble, Brandon Nutter, Karsten Petersen, George Pirocanac, Alexander Power, Mike Repass, Toby Reyelts, Fred Sauer, Jens Scheffler, Robert Schuppenies, Lindsey Simon, John Skidgel, Brett Slatkin, Graham Spencer, Amanda Surya, David Symonds, Joseph Ternasky, Eric Tholomé, Troy Trimble, Guido van Rossum, Nicholas Verne, Michael Winton, and Wenbo Zhu. Thanks also to Dan Morrill, Mark Pilgrim, Steffi Wu, Karen Wickre, Jane Penner, Jon Murchinson, Tom Stocky, Vic Gundotra, Bill Coughran, and Alan Eustace. At O’Reilly, I’m eternally grateful to Michael Loukides, who had nothing but good advice and an astonishing amount of patience for a first-time author. Let’s do another one! xx | Preface
- 22. CHAPTER 1 Introducing Google App Engine Google App Engine is a web application hosting service. By “web application,” we mean an application or service accessed over the Web, usually with a web browser: storefronts with shopping carts, social networking sites, multiplayer games, mobile applications, survey applications, project management, collaboration, publishing, and all of the other things we’re discovering are good uses for the Web. App Engine can serve tradi- tional website content too, such as documents and images, but the environment is especially designed for real-time dynamic applications. In particular, Google App Engine is designed to host applications with many simulta- neous users. When an application can serve many simultaneous users without degrading performance, we say it scales. Applications written for App Engine scale automatically. As more people use the application, App Engine allocates more resour- ces for the application and manages the use of those resources. The application itself does not need to know anything about the resources it is using. Unlike traditional web hosting or self-managed servers, with Google App Engine, you only pay for the resources you use. These resources are measured down to the gigabyte, with no monthly fees or up-front charges. Billed resources include CPU usage, storage per month, incoming and outgoing bandwidth, and several resources specific to App Engine services. To help you get started, every developer gets a certain amount of re- sources for free, enough for small applications with low traffic. Google estimates that with the free resources, an app can accommodate about 5 million page views a month. App Engine can be described as three parts: the runtime environment, the datastore, and the scalable services. In this chapter, we’ll look at each of these parts at a high level. We’ll also discuss features of App Engine for deploying and managing web applications, and for building websites integrated with other Google offerings such as Google Apps and Google Accounts. 1
- 23. The Runtime Environment AnApp Engine application responds to web requests. A web request begins when a client, typically a user’s web browser, contacts the application with an HTTP request, such as to fetch a web page at a URL. When App Engine receives the request, it identifies the application from the domain name of the address, either an .appspot.com subdo- main (provided for free with every app) or a subdomain of a custom domain name you have registered and set up with Google Apps. App Engine selects a server from many possible servers to handle the request, making its selection based on which server is most likely to provide a fast response. It then calls the application with the content of the HTTP request, receives the response data from the application, and returns the response to the client. From the application’s perspective, the runtime environment springs into existence when the request handler begins, and disappears when it ends. App Engine provides at least two methods for storing data that persists between requests (discussed later), but these mechanisms live outside of the runtime environment. By not retaining state in the runtime environment between requests—or at least, by not expecting that state will be retained between requests—App Engine can distribute traffic among as many servers as it needs to give every request the same treatment, regardless of how much traffic it is handling at one time. Application code cannot access the server on which it is running in the traditional sense. An application can read its own files from the filesystem, but it cannot write to files, and it cannot read files that belong to other applications. An application can see envi- ronment variables set by App Engine, but manipulations of these variables do not nec- essarily persist between requests. An application cannot access the networking facilities of the server hardware, though it can perform networking operations using services. In short, each request lives in its own “sandbox.” This allows App Engine to handle a request with the server that would, in its estimation, provide the fastest response. There is no way to guarantee that the same server hardware will handle two requests, even if the requests come from the same client and arrive relatively quickly. Sandboxing also allows App Engine to run multiple applications on the same server without the behavior of one application affecting another. In addition to limiting access to the operating system, the runtime environment also limits the amount of clock time, CPU use, and memory a single request can take. App Engine keeps these limits flexible, and applies stricter limits to applications that use up more resources to protect shared resources from “runaway” applications. A request has up to 30 seconds to return a response to the client. While that may seem like a comfortably large amount for a web app, App Engine is optimized for applications that respond in less than a second. Also, if an application uses many CPU cycles, App Engine may slow it down so the app isn’t hogging the processor on a machine serving multiple apps. A CPU-intensive request handler may take more clock time to complete 2 | Chapter 1: Introducing Google App Engine
- 24. than it wouldif it had exclusive use of the processor, and clock time may vary as App Engine detects patterns in CPU usage and allocates accordingly. Google App Engine provides two possible runtime environments for applications: a Java environment and a Python environment. The environment you choose depends on the language and related technologies you want to use for developing the application. The Java environment runs applications built for the Java 6 Virtual Machine (JVM). An app can be developed using the Java programming language, or most other lan- guages that compile to or otherwise run in the JVM, such as PHP (using Quercus), Ruby (using JRuby), JavaScript (using the Rhino interpreter), Scala, and Groovy. The app accesses the environment and services using interfaces based on web industry standards, including Java servlets and the Java Persistence API (JPA). Any Java tech- nology that functions within the sandbox restrictions can run on App Engine, making it suitable for many existing frameworks and libraries. Notably, App Engine fully sup- ports Google Web Toolkit (GWT), a framework for rich web applications that lets you write all of the app’s code—including the user interface—in the Java language, and have your rich graphical app work with all major browsers without plug-ins. The Python environment runs apps written in the Python 2.5 programming language, using a custom version of CPython, the official Python interpreter. App Engine invokes a Python app using CGI, a widely supported application interface standard. An appli- cation can use most of Python’s large and excellent standard library, as well as rich APIs and libraries for accessing services and modeling data. Many open source Python web application frameworks work with App Engine, such as Django, web2py, and Pylons, and App Engine even includes a simple framework of its own. The Java and Python environments use the same application server model: a request is routed to an app server, the application is started on the app server (if necessary) and invoked to handle the request to produce a response, and the response is returned to the client. Each environment runs its interpreter (the JVM or the Python interpreter) with sandbox restrictions, such that any attempt to use a feature of the language or a library that would require access outside of the sandbox fails with an exception. While using a different server for every request has advantages for scaling, it’s time- consuming to start up a new instance of the application for every request. App Engine mitigates startup costs by keeping the application in memory on an application server as long as possible and reusing servers intelligently. When a server needs to reclaim resources, it purges the least recently used app. All app servers have the runtime envi- ronment (JVM or Python interpreter) preloaded before the request reaches the server, so only the app itself needs to be loaded on a fresh server. Applications can exploit the app caching behavior to cache data directly on the app server using global (static) variables. Since an app can be evicted between any two requests (and low-traffic apps are evicted frequently), and there is no guarantee that a The Runtime Environment | 3
- 25. given user’s requestswill be handled by a given server, global variables are mostly useful for caching startup resources, like parsed configuration files. I haven’t said anything about which operating system or hardware configuration App Engine uses. There are ways to figure out what operating system or hardware a server is using, but in the end it doesn’t matter: the runtime environment is an abstraction above the operating system that allows App Engine to manage resource allocation, computation, request handling, scaling, and load distribution without the application’s involvement. Features that typically require knowledge of the operating system are either provided by services outside of the runtime environment, provided or emulated using standard library calls, or restricted in logical ways within the definition of the sandbox. The Static File Servers Most websites have resources they deliver to browsers that do not change during the regular operation of the site. The images and CSS files that describe the appearance of the site, the JavaScript code that runs in the browser, and HTML files for pages without dynamic components are examples of these resources, collectively known as static files. Since the delivery of these files doesn’t involve application code, it’s unnecessary and inefficient to serve them from the application servers. Instead, App Engine provides a separate set of servers dedicated to delivering static files. These servers are optimized for both internal architecture and network topology to handle requests for static resources. To the client, static files look like any other resource served by your app. You upload the static files of your application right alongside the application code. You can configure several aspects of how static files are served, including the URLs for static files, content types, and instructions for browsers to keep copies of the files in a cache for a given amount of time to reduce traffic and speed up rendering of the page. The Datastore Most useful web applications need to store information during the handling of a request for retrieval during a later request. A typical arrangement for a small website involves a single database server for the entire site, and one or more web servers that connect to the database to store or retrieve data. Using a single central database server makes it easy to have one canonical representation of the data, so multiple users accessing mul- tiple web servers all see the same and most recent information. But a central server is difficult to scale once it reaches its capacity for simultaneous connections. By far the most popular kind of data storage system for web applications in the past decade has been the relational database, with tables of rows and columns arranged for space efficiency and concision, and with indexes and raw computing power for 4 | Chapter 1: Introducing Google App Engine
- 26. performing queries, especially“join” queries that can treat multiple related records as a queryable unit. Other kinds of data storage systems include hierarchical datastores (filesystems, XML databases) and object databases. Each kind of database has pros and cons, and which type is best suited for an application depends on the nature of the application’s data and how it is accessed. And each kind of database has its own tech- niques for growing past the first server. Google App Engine’s database system most closely resembles an object database. It is not a join-query relational database, and if you come from the world of relational-database-backed web applications (as I did), this will probably require changing the way you think about your application’s data. As with the runtime envi- ronment, the design of the App Engine datastore is an abstraction that allows App Engine to handle the details of distributing and scaling the application, so your code can focus on other things. Entities and Properties An App Engine application stores its data as one or more datastore entities. An entity has one or more properties, each of which has a name, and a value that is of one of several primitive value types. Each entity is of a named kind, which categorizes the entity for the purpose of queries. At first glance, this seems similar to a relational database: entities of a kind are like rows in a table, and properties are like columns (fields). However, there are two major dif- ferences between entities and rows. First, an entity of a given kind is not required to have the same properties as other entities of the same kind. Second, an entity can have a property of the same name as another entity has, but with a different type of value. In this way, datastore entities are “schemaless.” As you’ll soon see, this design provides both powerful flexibility as well as some maintenance challenges. Another difference between an entity and a table row is that an entity can have multiple values for a single property. This feature is a bit quirky, but can be quite useful once understood. Every datastore entity has a unique key that is either provided by the application or generated by App Engine (your choice). Unlike a relational database, the key is not a “field” or property, but an independent aspect of the entity. You can fetch an entity quickly if you know its key, and you can perform queries on key values. A entity’s key cannot be changed after the entity has been created. Neither can its kind. App Engine uses the entity’s kind and key to help determine where the entity is stored in a large collection of servers—though neither the key nor the kind ensure that two entities are stored on the same server. The Datastore | 5
- 27. Queries and Indexes Adatastore query returns zero or more entities of a single kind. It can also return just the keys of entities that would be returned for a query. A query can filter based on conditions that must be met by the values of an entity’s properties, and can return entities ordered by property values. A query can also filter and sort using keys. In a typical relational database, queries are planned and executed in real time against the data tables, which are stored as they were designed by the developer. The developer can also tell the database to produce and maintain indexes on certain columns to speed up certain queries. App Engine does something dramatically different. With App Engine, every query has a corresponding index maintained by the datastore. When the application performs a query, the datastore finds the index for that query, scans down to the first row that matches the query, then returns the entity for each consecutive row in the index until the first row that doesn’t match the query. Of course, this requires that App Engine know ahead of time which queries the appli- cation is going to perform. It doesn’t need to know the values of the filters in advance, but it does need to know the kind of entity to query, the properties being filtered or sorted, and the operators of the filters and the orders of the sorts. App Engine provides a set of indexes for simple queries by default, based on which properties exist on entities of a kind. For more complex queries, an app must include index specifications in its configuration. The App Engine SDK helps produce this con- figuration file by watching which queries are performed as you test your application with the provided development web server on your computer. When you upload your app, the datastore knows to make indexes for every query the app performed during testing. You can also edit the index configuration manually. When your application creates new entities and updates existing ones, the datastore updates every corresponding index. This makes queries very fast (each query is a simple table scan) at the expense of entity updates (possibly many tables may need updating for a single change). In fact, the performance of an index-backed query is not affected by the number of entities in the datastore, only the size of the result set. It’s worth paying attention to indexes, as they take up space and increase the time it takes to update entities. We discuss indexes in detail in Chapter 5. Transactions When an application has many clients attempting to read or write the same data si- multaneously, it is imperative that the data always be in a consistent state. One user should never see half-written data or data that doesn’t make sense because another user’s action hasn’t completed. 6 | Chapter 1: Introducing Google App Engine
- 28. When an applicationupdates the properties of a single entity, App Engine ensures that either every update to the entity succeeds all at once, or the entire update fails and the entity remains the way it was prior to the beginning of the update. Other users do not see any effects of the change until the change succeeds. In other words, an update of a single entity occurs in a transaction. Each transaction is atomic: the transaction either succeeds completely or fails completely, and cannot suc- ceed or fail in smaller pieces. An application can read or update multiple entities in a single transaction, but it must tell App Engine which entities will be updated together when it creates the entities. The application does this by creating entities in entity groups. App Engine uses entity groups to control how entities are distributed across servers, so it can guarantee a transaction on a group succeeds or fails completely. In database terms, the App Engine datastore natively supports local transactions. When an application calls the datastore API to update an entity, control does not return to the application until the transaction succeeds or fails, and the call returns with knowledge of success or failure. For updates, this means the application waits for all entities and indexes to be updated before doing anything else. If a user tries to update an entity while another user’s update of the entity is in progress, the datastore returns immediately with a concurrency failure exception. It is often ap- propriate for the app to retry a bounced transaction several times before declaring the condition an error, usually retrieving data that may have changed within the transaction before calculating new values and updating it. In database terms, App Engine uses optimistic concurrency control. Reading the entity never fails due to concurrency; the application just sees the entity in its most recent stable state. You can also perform multiple reads in a transaction to ensure that all of the data read in the transaction is current and consistent with itself. In most cases, retrying a transaction on a contested entity will succeed. But if an application is designed such that many users might update a single entity, the more popular the application gets, the more likely users will get concurrency failures. It is important to design entity groups to avoid concurrency failures even with a large num- ber of users. An application can bundle multiple datastore operations in a single transaction. For example, the application can start a transaction, read an entity, update a property value based on the last read value, save the entity, then commit the transaction. In this case, the save action does not occur unless the entire transaction succeeds without conflict with another transaction. If there is a conflict and the app wants to try again, the app should retry the entire transaction: read the (possibly updated) entity again, use the new value for the calculation, and attempt the update again. The Datastore | 7
- 29. With indexes andoptimistic concurrency control, the App Engine datastore is designed for applications that need to read data quickly, ensure that the data it sees is in a con- sistent form, and scale the number of users and the size of the data automatically. While these goals are somewhat different from those of a relational database, they are espe- cially well suited to web applications. The Services The datastore’s relationship with the runtime environment is that of a service: the ap- plication uses an API to access a separate system that manages all of its own scaling needs separately from the runtime environment. Google App Engine includes several other self-scaling services useful for web applications. The memory cache (or memcache) service is a short-term key-value storage service. Its main advantage over the datastore is that it is fast, much faster than the datastore for simple storage and retrieval. The memcache stores values in memory instead of on disk for faster access. It is distributed like the datastore, so every request sees the same set of keys and values. However, it is not persistent like the datastore: if a server goes down, such as during a power failure, memory is erased. It also has a more limited sense of atomicity and transactionality than the datastore. As the name implies, the memcache service is best used as a cache for the results of frequently performed queries or calcu- lations. The application checks for a cached value, and if the value isn’t there, it per- forms the query or calculation and stores the value in the cache for future use. App Engine applications can access other web resources using the URL Fetch service. The service makes HTTP requests to other servers on the Internet, such as to retrieve pages or interact with web services. Since remote servers can be slow to respond, the URL Fetch API supports fetching URLs in the background while a request handler does other things, but in all cases the fetch must start and finish within the request handler’s lifetime. The application can also set a deadline, after which the call is canceled if the remote host hasn’t responded. App Engine applications can send messages using the Mail service. Messages can be sent on behalf of the application or on behalf of the user who made the request that is sending the email (if the message is from the user). Many web applications use email to notify users, confirm user actions, and validate contact information. An application can also receive email messages. If an app is configured to receive email, a message sent to the app’s address is routed to the Mail service, which delivers the message to the app in the form of an HTTP request to a request handler. App Engine applications can send and receive instant messages to and from chat serv- ices that support the XMPP protocol, including Google Talk. An app sends an XMPP chat message by calling the XMPP service. As with incoming email, when someone sends a message to the app’s address, the XMPP service delivers it to the app by calling a request handler. 8 | Chapter 1: Introducing Google App Engine
- 30. The image processingservice can do lightweight transformations of image data, such as for making thumbnail images of uploaded photos. The image processing tasks are performed using the same infrastructure Google uses to process images with some of its other products, so the results come back quickly. We won’t be covering the image service API in this book because Google’s official documentation says everything there is to say about this easy-to-use service. Google Accounts App Engine features integration with Google Accounts, the user account system used by Google applications such as Google Mail, Google Docs, and Google Calendar. You can use Google Accounts as your app’s account system, so you don’t have to build your own. And if your users already have Google accounts, they can sign in to your app using their existing accounts, with no need to create new accounts just for your app. Of course, there is no obligation to use Google Accounts. You can always build your own account system, or use an OpenID provider. Google Accounts is especially useful for developing applications for your company or organization using Google Apps. With Google Apps, your organization’s members can use the same account to access your custom applications as well as their email, calendar, and documents. Task Queues and Cron Jobs A web application has to respond to web requests very quickly, usually in less than a second and preferably in just a few dozen milliseconds, to provide a smooth experience to the user sitting in front of the browser. This doesn’t give the application much time to do work. Sometimes, there is more work to do than there is time to do it. In such cases it’s usually OK if the work gets done within a few seconds, minutes, or hours, instead of right away, as the user is waiting for a response from the server. But the user needs a guarantee that the work will get done. For this kind of work, App Engine uses task queues. Task queues let request handlers describe work to be done at a later time, outside the scope of the web request. Queues ensure that every task gets done eventually. If a task fails, the queue retries the task until it succeeds. You can configure the rate at which queues are processed to spread the workload throughout the day. A queue performs a task by calling a request handler. It can include a data payload provided by the code that created the task, delivered to the task’s handler as an HTTP request. The task’s handler is subject to the same limits as other request handlers, including the 30-second time limit. An especially powerful feature of task queues is the ability to enqueue a task within a datastore transaction. This ensures that the task will be enqueued only if the rest of the Task Queues and Cron Jobs | 9
- 31. datastore transaction succeeds.You can use transactional tasks to perform additional datastore operations that must be consistent with the transaction eventually, but that do not need the strong consistency guarantees of the datastore’s local transactions. App Engine has another service for executing tasks at specific times of the day. Sched- uled tasks are also known as “cron jobs,” a name borrowed from a similar feature of the Unix operating system. The scheduled tasks service can invoke a request handler at a specified time of the day, week, or month, based on a schedule you provide when you upload your application. Scheduled tasks are useful for doing regular maintenance or sending periodic notification messages. We’ll look at these features and some powerful uses for them in Chapter 13. Developer Tools Google provides free tools for developing App Engine applications in Java or Python. You can download the software development kit (SDK) for your chosen language and your computer’s operating system from Google’s website. Java users can get the Java SDK in the form of a plug-in for the Eclipse integrated development environment. Py- thon users using Windows or Mac OS X can get the Python SDK in the form of a GUI application. Both SDKs are also available as ZIP archives of command-line tools, for using directly or integrating into your development environment or build system. Each SDK includes a development web server that runs your application on your local computer and simulates the runtime environment, the datastore, and the services. The development server automatically detects changes in your source files and reloads them as needed, so you can keep the server running while you develop the application. If you’re using Eclipse, you can run the Java development server in the interactive de- bugger, and can set breakpoints in your application code. You can also use Eclipse for Python app development using PyDev, an Eclipse extension that includes an interactive Python debugger. (Using PyDev is not covered in this book, but there are instructions on Google’s site.) The development version of the datastore can automatically generate configuration for query indexes as the application performs queries, which App Engine will use to pre- build indexes for those queries. You can turn this feature off for testing whether queries have appropriate indexes in the configuration. The development web server includes a built-in web application for inspecting the contents of the (simulated) datastore. You can also create new datastore entities using this interface for testing purposes. Each SDK also includes a tool for interacting with the application running on App Engine. Primarily, you use this tool to upload your application code to App Engine. You can also use this tool to download log data from your live application, or manage the live application’s indexes. 10 | Chapter 1: Introducing Google App Engine
- 32. The Python andJava SDKs include a feature you can install in your app for secure remote programmatic access to your live application. The Python SDK includes tools that use this feature for bulk data operations, such as uploading new data from a text file and downloading large amounts of data for backup or migration purposes. The SDK also includes a Python interactive command-line shell for testing, debugging, and manually manipulating live data. (These tools are in the Python SDK, but also work with Java apps using the Java version of the remote access feature.) You can write your own scripts and programs that use the remote access feature for large-scale data trans- formations or other maintenance. The Administration Console When your application is ready for its public debut, you create an administrator ac- count and set up the application on App Engine. You use your administrator account to create and manage the application, view its access and resource usage statistics and message logs, and more, all with a web-based interface called the Administration Console. You sign in to the Administration Console using your Google account. You can use your current Google account if you have one, though you may also want to create a Google account just for your application, which you might use as the “from” address on email messages. Once you have created an application using the Administration Console, you can add additional Google accounts as administrators. Any administrator can access the Console, and can upload a new version of the application. The Console gives you access to real-time performance data about how your application is being used, as well as access to log data emitted by your application. You can also query the datastore for the live application using a web interface, and check on the status of datastore indexes. (Newly created indexes with large data sets take time to build.) When you upload new code for your application using the SDK, the uploaded version is assigned a version identifier, which you specify in the application’s configuration file. The version used for the live application is whichever major version is selected as the “default.” You control which version is the “default” using the Administration Console. You can access nondefault versions using a special URL containing the version identi- fier. This allows you to test a new version of an app running on App Engine before making it official. You use the Console to set up and manage the billing account for your application. When you’re ready for your application to consume more resources beyond the free amounts, you set up a billing account using a credit card and Google Accounts. The owner of the billing account sets a budget, a maximum amount of money that can be charged per calendar day. Within that budget, you can allocate how much additional The Administration Console | 11
- 33. CPU time, bandwidth,storage, and email recipients the app can consume. You are only charged for what the application actually uses beyond the free amounts. Things App Engine Doesn’t Do...Yet When people first start using App Engine, there are several things they ask about that App Engine doesn’t do. Some of these are things Google may implement in the near future, and others run against the grain of the App Engine design and aren’t likely to be added. Listing such features in a book is difficult, because by the time you read this, Google may have already implemented them. But it’s worth noting these features here, especially to note workaround techniques. App Engine supports secure connections (HTTPS) to .appspot.com subdomains, but does not yet support secure connections to custom domains. Google Accounts sign- ins always use secure connections. An application can use the URL Fetch service to make an HTTPS request to another site, but App Engine does not verify the certificate used on the remote server. An app can receive incoming email and XMPP chat messages at several addresses. As of this writing, none of these addresses can use a custom domain name. See Chap- ter 11 for information on incoming email and XMPP addresses. An app can accept web requests on a custom domain using Google Apps. Google Apps maps a subdomain of your custom domain to an app, and this subdomain can be www if you choose. This does not yet support requests for “naked” domains, such as http:// example.com/. It also does not support arbitrary tertiary domains on custom domains (http://foo.www.example.com). App Engine does support arbitrary subdomains on appspot.com URLs, such as foo.app-id.appspot.com. App Engine does not host long-running background processes. Task queues and sched- uled tasks can invoke request handlers outside of a user request, and can drive some kinds of batch processing. But processing large chores in small batches is different in character and range from full-scale distributed computing tasks. We will discuss batch processing later in Chapter 12. App Engine does not support streaming or long-term connections. If the client supports it, the app can use XMPP and an XMPP service (such as Google Talk) to deliver state updates to the client. You could also do this using a polling technique, where the client asks the application for updates on a regular basis, but polling is difficult to scale (5,000 simultaneous users polling every 5 seconds = 1,000 queries per second), and is not appropriate for all applications. Also note that request handlers cannot communicate with the client while performing other calculations. The server sends a response to the client’s request only after the handler has returned control to the server. 12 | Chapter 1: Introducing Google App Engine
- 34. App Engine onlysupports web requests via HTTP or HTTPS, and email and XMPP messages via the services. It does not support other kinds of network connections. For instance, a client cannot connect to an App Engine application via FTP. The App Engine datastore does not support full-text search queries, such as for imple- menting a search engine for a content management system. Long text values are not indexed, and short text values are only indexed for equality and inequality queries. It is possible to implement text search by building search indexes within the application, but this is difficult to do in a scalable way for large amounts of dynamic data. Getting Started You can start developing applications for Google App Engine without creating an ac- count. All you need to get started is the App Engine SDK appropriate for your choice of language, which is a free download from the App Engine website: http://code.google.com/appengine/ While you’re there, check out the official “Getting Started Guide” for your language, which demonstrates how to create an application and use several of App Engine’s features. In the next chapter, we’ll describe how to create a new project from start to finish, including how to create an account, upload the application, and run it on App Engine. Getting Started | 13
- 36. CHAPTER 2 Creating an Application The App Engine development model is as simple as it gets: 1. Create the application. 2. Test the application on your own computer using the web server software included with the App Engine development kit. 3. Upload the finished application to App Engine. In this chapter, we will walk through the process of creating a new application, testing it with the development server, registering a new application ID and setting up a domain name, and uploading the app to App Engine. We will look at some of the features of the Python and Java software development kits (SDKs) and the App Engine Adminis- tration Console. We’ll also discuss the workflow for developing and deploying an app. We will take this opportunity to demonstrate a common pattern in web applications: managing user preferences data. This pattern uses several App Engine services and features. Setting Up the SDK All the tools and libraries you need to develop an application are included in the App Engine SDK. There are separate SDKs for Python and Java, each with features useful for developing with each language. The SDKs work on any platform, including Win- dows, Mac OS X, and Linux. The Python and Java SDKs each include a web server that runs your app in a simulated runtime environment on your computer. The development server enforces the sandbox restrictions of the full runtime environment and simulates each of the App Engine services. You can start the development server and leave it running while you build your app, reloading pages in your browser to see your changes in effect. Both SDKs include a multifunction tool for interacting with the app running on App Engine. You use this tool to upload your app’s code, static files, and configuration. 15
- 37. The tool canalso manage datastore indexes, task queues, and scheduled tasks, and can download messages logged by the live application so you can analyze your app’s traffic and behavior. Because Google launched Python support before Java, the Python SDK has a few tools not available in the Java SDK. Most notably, the Python SDK includes tools for up- loading and downloading data to and from the datastore. This is useful for making backups, changing the structure of existing data, and for processing data offline. If you are using Java, you can use the Python-based data tools with a bit of effort. The Python SDKs for Windows and Mac OS X include a “launcher” application that makes it especially easy to create, edit, test, and upload an app using a simple graphical interface. Paired with a good programming text editor (such as Notepad++ for Win- dows, or TextMate for Mac OS X), the launcher provides a fast and intuitive Python development experience. For Java developers, Google provides a plug-in for the Eclipse integrated development environment that implements a complete App Engine development workflow. The plug-in includes a template for creating new App Engine Java apps, as well as a debug- ging profile for running the app and the development web server in the Eclipse debug- ger. To deploy a project to App Engine, you just click a button on the Eclipse toolbar. Both SDKs also include cross-platform command-line tools that provide these features. You can use these tools from a command prompt, or otherwise integrate them into your development environment as you see fit. We’ll discuss the Python SDK first, then the Java SDK in “Installing the Java SDK” on page 20. Feel free to skip the section that does not apply to your chosen language. Installing the Python SDK The App Engine SDK for the Python runtime environment runs on any computer that runs Python 2.5. If you are using Mac OS X or Linux, or if you have used Python previously, you may already have Python on your system. You can test whether Python is installed on your system and check which version is installed by running the following command at a command prompt (in Windows, Command Prompt; in Mac OS X, Terminal): python -V (That’s a capital “V.”) If Python is installed, it prints its version number, like so: Python 2.5.2 You can download and install Python 2.5 for your platform from the Python website: http://www.python.org/ 16 | Chapter 2: Creating an Application
- 38. Be sure toget Python version 2.5 (such as 2.5.4) from the “Download” section of the site. As of this writing, the latest major version of Python is 3.1, and the latest 2.x- compatible release is 2.6. The App Engine Python SDK works with Python 2.6, but it’s better to use the same version of Python that’s used on App Engine for development so you are not surprised by obscure compatibility issues. App Engine Python does not yet support Python 3. Python 3 includes several new language and library features that are not backward com- patible with earlier versions. When App Engine adds support for Python 3, it will likely be in the form of a new runtime environment, in addition to the Python 2 environment. You control which runtime environment your application uses with a setting in the app’s configuration file, so your application will continue to run as intended when new runtime environments are released. You can download the App Engine Python SDK bundle for your operating system from the Google App Engine website: http://code.google.com/appengine/downloads.html Download and install the file appropriate for your operating system: • For Windows, the Python SDK is an .msi (Microsoft Installer) file. Click on the appropriate link to download it, then double-click on the file to start the installation process. This installs the Google App Engine Launcher application, adds an icon to your Start menu, and adds the command-line tools to the command path. • For Mac OS X, the Python SDK is a Mac application in a .dmg (disk image) file. Click on the link to download it, then double-click on the file to mount the disk image. Drag the GoogleAppEngineLauncher icon to your Applications folder. To install the command-line tools, double-click the icon to start the Launcher, then allow the Launcher to create the “symlinks” when prompted. • If you are using Linux or another platform, the Python SDK is available as a .zip archive. Download and unpack it (typically with the the unzip command) to create a directory named google_appengine. The command-line tools all reside in this directory. Adjust your command path as needed. To test that the App Engine Python SDK is installed, run the following command at a command prompt: dev_appserver.py --help The command prints a helpful message and exits. If instead you see a message about the command not being found, check that the installer completed successfully, and that the location of the dev_appserver.py command is on your command path. Windows users, if when you run this command a dialog box opens with the message “Windows cannot open this file... To open this file, Windows needs to know what Setting Up the SDK | 17
- 39. program created it,”you must tell Windows to use Python to open the file. In the dialog box, choose “Select the program from a list,” and click OK. Click Browse, then locate your Python installation (such as C:Python25). Select python from this folder, then click Open. Select “Always use the selected program to open this kind of file.” Click OK. A window will open and attempt to run the command, then immediately close. You can now run the command from the Command Prompt. A brief tour of the Launcher The Windows and Mac OS X versions of the Python SDK include an application called the Google App Engine Launcher (hereafter just “Launcher”). With the Launcher, you can create and manage multiple App Engine Python projects using a graphical interface. Figure 2-1 shows an example of the Launcher window in Mac OS X. Figure 2-1. The Google App Engine Launcher for Mac OS X main window, with a project selected To create a new project, select New Project... from the File menu (or click the plus-sign button at the bottom of the window). Browse to where you want to keep your project files, then enter a name for the project. The Launcher creates a new directory at that location, named after the project, to hold the project’s files, and creates several starter files. The project appears in the project list in the main launcher window. 18 | Chapter 2: Creating an Application
- 40. To start thedevelopment web server, make sure the project is selected, then click the Run button. You can stop the server with the Stop button. To open the home page of the running app in a browser, click the Browse button. The Logs button displays mes- sages logged by the app in the development server. The SDK Console button opens a web interface for the development server with several features for inspecting the running application, including tools to inspect the contents of the (simulated) datastore and memory cache, and an interactive console that executes Python statements and displays the results. The Edit button opens the project’s files in your default text editor. In the Mac OS X version, this is especially useful with text editors that can open a directory’s worth of files, such as TextMate or Emacs. In the Windows version, this just opens app.yaml for editing. The Deploy button uploads the project to App Engine. Before you can deploy a project, you must register an application ID with App Engine and edit the application’s con- figuration file with the registered ID. The Dashboard button opens a browser window with the App Engine Administration Console for the deployed app. We’ll look at the configuration file, the registration process, and the Administration Console later in this chapter. The complete App Engine Python SDK, including the command-line tools, resides in the Launcher’s application directory. In the Windows version, the installer adds the appropriate directory to the command path, so you can run these tools from a Com- mand Prompt. In Mac OS X, when you start the Launcher for the first time it asks for permission to create “symlinks.” This creates symbolic links in the directory /usr/local/bin/ that refer to the command-line tools in the application bundle. With the links in this directory, you can type just the name of a command at a Terminal prompt to run it. If you didn’t create the symlinks, you can do so later by selecting the Make Symlinks... item from the GoogleAppEngineLauncher menu. You can set command-line flags for the development server within the Launcher. To do so, select the application, then go to the Edit menu and select Application Settings.... Add the desired command-line options to the “Extra Flags” field, then click Update. The Mac OS X version of the Launcher installs Google’s software update facility to check for new versions of the App Engine SDK. When a new version is released, this feature notifies you and offers to upgrade. Immediately after you upgrade, you’ll notice the symlinks stop working. To fix the symlinks, reopen the Launcher app and follow the prompts. The upgrade can’t do this automatically because it needs your permis- sion to create new symlinks. Setting Up the SDK | 19
- 41. Installing the JavaSDK The App Engine SDK for the Java runtime environment runs on any computer that runs the Java SE Development Kit (JDK). The App Engine for Java SDK supports JDK 5 and JDK 6. When running on App Engine, the Java runtime environment uses the Java 6 JVM. If you don’t already have it, you can download and install the Java 6 JDK for most platforms from Sun’s website. (Mac users, see the next section.) http://java.sun.com/javase/downloads/index.jsp You can test whether the Java development kit is installed on your system and check which version it is by running the following command at a command prompt (in Win- dows, Command Prompt; in Mac OS X, Terminal): javac -version If you have the Java 6 JDK installed, the command will print a version number similar to javac 1.6.0. If you have the Java 5 JDK installed, the command will print a version number similar to javac 1.5.0. The actual output varies depending on which specific version you have. App Engine Java apps use interfaces and features from Java Enterprise Edition (Java EE). The App Engine SDK includes implementations for the relevant Java EE features. You do not need to install a separate Java EE implementation. The steps for installing the App Engine SDK for Java depend on whether you wish to use the Google Plugin for the Eclipse IDE. We’ll cover these situations separately. Java on Mac OS X If you are using Mac OS X, you already have Java and the JDK installed. How you’d use it depends on the version of the operating system, and whether your computer has a 32-bit processor (such as the Intel Core Duo) or a 64-bit processor (Intel Core 2 Duo, Intel Xeon). You can check which processor you have by selecting the Apple menu, About This Mac. Mac OS X 10.6 Snow Leopard includes Java 6 and its JDK, and it includes separate versions for 32-bit processors and for 64-bit processors. If you have a 64-bit processor, the 64-bit Java 6 is the default. If you have a 32-bit processor, the 32-bit Java 6 is the default. Mac OS X 10.5 Leopard includes both Java 5 and Java 6. However, in Leopard, Java 5 is the default. This is because Leopard’s version of Java 6 only works with 64-bit pro- cessors. If you have a 64-bit processor, you can change the default version to the 64- bit Java 6. 20 | Chapter 2: Creating an Application
- 42. To change theversion of Java used by the system, open the Java Preferences utility, which you can find under /Applications/Utilities/. In the “Java Applications” list, drag the desired version (such as “Java SE 6, 64-bit”) to the top of the list. OS X uses the topmost version in the list that is compatible with your system. If you have a 32-bit Mac running Leopard, you’re stuck using Java 5. The App Engine SDK works fine under Java 5, and apps built with Java 5 run fine on App Engine. Just be aware that you’re using Java 5, and code samples you might find for App Engine may assume Java 6. If you are using Eclipse, make sure you get the version that corresponds with your processor and selected version of Java. Separate versions of the “Eclipse IDE for Java EE Developers” bundle are available for 32-bit and 64-bit processors. For more information about Java and Mac OS X, see Apple’s developer website: http://developer.apple.com/java/ Installing the Java SDK with the Google Plugin for Eclipse One of the easiest ways to develop App Engine applications in Java is to use the Eclipse IDE and the Google Plugin for Eclipse. The plug-in works with Eclipse 3.3 (Europa), Eclipse 3.4 (Ganymede), and Eclipse 3.5 (Galileo). You can get Eclipse for your platform for free at the Eclipse website: http://www.eclipse.org/ If you’re getting Eclipse specifically for App Engine development, get the “Eclipse IDE for Java EE Developers” bundle. This bundle includes several useful components for developing web applications, including the Eclipse Web Tools Platform (WTP) package. You can tell Eclipse to use the JDK you have installed in the Preferences window. In Eclipse 3.5, select Preferences (Windows and Linux, in the Window menu; Mac OS X, in the Eclipse menu). In the Java category, select “Installed JREs.” If necessary, add the location of the SDK to the list, and make sure the checkbox is checked. To install the App Engine Java SDK and the Google Plugin, use the software installation feature of Eclipse. In Eclipse 3.5, select Install New Software... from the Help menu, then type the following URL in the “Work with” field and click the Add... button: http://dl.google.com/eclipse/plugin/3.5 (This URL does not work in a browser; it only works with the Eclipse software installer.) In the dialog window that opens, enter “Google” for the name, then click OK. Two items are added to the list, one for the plug-in (“Plugin”) and a set for the App Engine and Google Web Toolkit SDKs (“SDKs”). Figure 2-2 shows the Install Software window with the appropriate items selected. Setting Up the SDK | 21
- 43. Figure 2-2. TheEclipse 3.5 (Galileo) Install Software window, with the Google Plugin selected Check the boxes for these two items. Click the Next > button and follow the prompts. For more information on installing the Google Plugin for Eclipse, including instructions for Eclipse 3.3 or 3.4, see the website for the plug-in: http://code.google.com/eclipse/ After installation, the Eclipse toolbar has three new buttons, as shown in Figure 2-3. Figure 2-3. The Eclipse 3.5 toolbar with the Google Plugin installed, with three new buttons: New Web Application Project, GWT Compile Project, and Deploy App Engine Project The plug-in adds several features to the Eclipse interface: • The three buttons in the toolbar: New Web Application Project, GWT Compile Project, and Deploy App Engine Project • A Web Application Project item under New in the File menu 22 | Chapter 2: Creating an Application
- 44. • A WebApplication debug profile, for running an app in the development web server under the Eclipse debugger You can use Eclipse to develop your application, and to deploy it to App Engine. To use other features of the SDK, like downloading log data, you must use the command- line tools from the App Engine SDK. Eclipse installs the SDK in your Eclipse application directory, under eclipse/plugins/. The actual directory name depends on the specific version of the SDK installed, but it looks something like this: com.google.appengine.eclipse.sdkbundle_1.2.5.v200909021031/appengine-java-sdk-1.2.5/ This directory contains command-line tools in a subdirectory named bin/. In Mac OS X or Linux, you may need to change the permissions of these files to be executable in order to use the tools from the command line: chmod 755 bin/* You can add the bin/ directory to your command path, but keep in mind that the path will change each time you update the SDK. Installing the Java SDK without Eclipse If you are not using the Eclipse IDE or otherwise don’t wish to use the Google Plugin, you can download the App Engine Java SDK as a .zip archive from the App Engine website: http://code.google.com/appengine/downloads.html The archive unpacks to a directory with a name such as appengine-java-sdk-1.2.5. The SDK contains command-line launch scripts in the bin/ subdirectory. You can add this directory to your command path to make the commands easier to run. Both the AppCfg tool and the development web server execute Java classes to perform their functions. You can integrate these tools into your IDE or build scripts by calling the launch scripts, or by calling the Java classes directly. Look at the contents of the launch scripts to see the syntax. The App Engine SDK includes a plug-in for Apache Ant that lets you perform functions of the SDK from an Ant build script. See the App Engine documentation for more information about using Ant with App Engine. Test that the App Engine Java SDK is installed properly by running the following com- mand at a command prompt: dev_appserver --help Mac OS X and Linux users, use dev_appserver.sh as the command name. Setting Up the SDK | 23
- 45. The command printsa helpful message and exits. If instead you see a message about the command not being found, check that the archive unpacked successfully, and that the SDK’s bin/ directory is on your command path. Developing the Application An App Engine application responds to web requests. It does so by calling request handlers, routines that accept request parameters and return responses. App Engine determines which request handler to use for a given request from the request’s URL, using a configuration file included with the app that maps URLs to handlers. An app can also include static files, such as images, CSS stylesheets, and browser JavaScript. App Engine serves these files directly to clients in response to requests for corresponding URLs without invoking any code. The app’s configuration specifies which of its files are static, and which URLs to use for those files. The application configuration includes metadata about the app, such as its application ID and version number. When you deploy the app to App Engine, all of the app’s files, including the code, configuration files, and static files, are uploaded and associated with the application ID and version number mentioned in the configuration. An app can also have configuration files specific to the services, such as for datastore indexes, task queues, and scheduled tasks. These files are associated with the app in general, not a specific version of the app. The structure and format of the code and configuration files differ for Python apps and for Java apps, but the concepts are similar. In the next few sections, we will create the files needed for a simple application in both Python and Java, and will look at how to use the tools and libraries included with each SDK. The User Preferences Pattern The application we will create in this section is a simple clock. When a user visits the site, the app displays the current time of day according to the server’s system clock. By default, the app shows the current time in the Coordinated Universal Time (UTC) time zone. The user can customize the time zone by signing in using Google Accounts and setting a preference. This app demonstrates three App Engine features: • The datastore, primary storage for user settings data that is persistent, reliable, and scalable • The memory cache (or memcache), secondary storage that is faster than the data- store, but is not necessarily persistent in the long term • Google Accounts, the ability to use Google’s user account system for authenticat- ing and identifying users 24 | Chapter 2: Creating an Application
- 46. Google Accounts workssimilarly to most user account systems. If the user is not signed in to the clock application, she sees a generic view with default settings (the UTC time zone) and a link to sign in or create a new account. If the user chooses to sign in or register, the application directs her to a sign-in form managed by Google Accounts. Signing in or creating an account redirects the user back to the application. Of course, you can implement your own account mechanism instead of using Google Accounts. Using Google Accounts has advantages and disadvantages—the chief ad- vantage being that you don’t have to implement your own account mechanism. If a user of your app already has a Google account, the user can sign in with that account without creating a new account for your app. If the user accesses the application while signed in, the app loads the user’s preferences data and uses it to render the page. The app retrieves the preferences data in two steps. First, it attempts to get the data from the fast secondary storage, the memory cache. If the data is not present in the memory cache, the app attempts to retrieve it from the primary storage (the datastore), and if successful, it puts it into the memory cache to be found by future requests. This means that for most requests, the application can get the user’s preferences from the memcache without accessing the datastore. While reading from the datastore is reasonably fast, reading from the memcache is much faster. The difference is substantial when the same data must be accessed every time the user visits a page. Our clock application has two request handlers. One handler displays the current time of day, along with links for signing in and out. It also displays a web form for adjusting the time zone when the user is signed in. The second request handler processes the time zone form when it is submitted. When the user submits the preferences form, the app saves the changes and redirects the browser back to the main page. The application gets the current time from the application server’s system clock. It’s worth noting that App Engine makes no guarantees that the system clocks of all of its web servers are synchronized. Since two requests for this app may be handled by dif- ferent servers, different requests may see different clocks. The server clock is not con- sistent enough as a source of time data for a real-world application, but it’s good enough for this example. In the next section, we implement this app using Python. We do the same thing with Java in the section “Developing a Java App” on page 39. As before, feel free to skip the section that doesn’t apply to you. Developing a Python App The simplest Python application for App Engine is a single directory with two files: a configuration file named app.yaml, and a file of Python code for a request handler. The directory containing the app.yaml file is the application root directory. You’ll refer to this directory often when using the tools. Developing the Application | 25
- 47. If you areusing the Launcher, you can start a new project by selecting the File menu, New Application.... The Launcher creates a new project with several files, which you may wish to edit to follow along with the example. Alternatively, you can create the project directory and files by hand, then add the project to the Launcher by selecting the File menu, Add Existing Application.... Create a directory named clock to contain the project. Using your favorite text editor, create a file inside this directory named app.yaml similar to Example 2-1. Example 2-1. The app.yaml configuration file for a simple application application: clock version: 1 runtime: python api_version: 1 handlers: - url: /.* script: main.py This configuration file is in a format called YAML, an open format for configuration files and network messages. You don’t need to know much about the format beyond what you see here. In this example, the configuration file tells App Engine that this is version 1 of an application called clock, which uses version 1 of the Python runtime environment (the “API version”). Every request for this application (every URL that matches the regular expression /.*) is to be handled by a Python script named main.py. Create a file named main.py similar to Example 2-2, in the same directory as app.yaml. Example 2-2. A simple Python request handler script import datetime print 'Content-Type: text/html' print '' print '<p>The time is: %s</p>' % str(datetime.datetime.now()) This simple Python program imports the datetime module from the Python standard library, prints an HTTP header that indicates the type of the document (HTML), then prints a message containing the current time according to the web server’s clock. Python request handler scripts use the CGI protocol for communicating with App En- gine. When App Engine receives a request for your Python application, App Engine establishes a runtime environment with the request data in environment variables, de- termines which handler script to run using the configuration file, then runs the script with the request body (if any) on the standard input stream. The handler script is ex- pected to perform all necessary actions for the request, then print a response to the 26 | Chapter 2: Creating an Application
- 48. standard output stream,including a valid HTTP header. This simple example ignores the request data, prints a header indicating the type of the response data, then prints a message with the current time to be displayed in the browser. Let’s test what we have so far. Start the development server by running the dev_appserver.py command, specifying the path to the project directory (clock) as an argument: dev_appserver.py clock If your current working directory is the clock directory you just created, you can run the command using a dot (.) as the path to the project: dev_appserver.py . The server starts up and prints several messages to the console. If this is the first time you’re running the server from the command line, it may ask whether you want it to check for updates; type your answer, then hit Enter. You can safely ignore warnings that say “Could not read datastore data” and “Could not initialize images API.” These are expected if you have followed the installation steps so far. The last message should look something like this: INFO ... Running application clock on port 8080: http://localhost:8080 This message indicates the server started successfully. If you do not see this message, check the other messages for hints, and double-check that the syntax of your app.yaml file is correct. Test your application by visiting the server’s URL in a web browser: http://localhost:8080/ The browser displays a page similar to Figure 2-4. Figure 2-4. The first version of the clock application viewed in a browser Developing the Application | 27
- 49. You can leavethe web server running while you develop your application. The web server notices when you make changes to your files, and reloads them automatically as needed. Using the Launcher, you can start the development web server by click- ing the Run button. The icon next to the project turns green when the server starts successfully. To open a browser to view the project, click the Browse button. Introducing the webapp framework The code in Example 2-2 attempts to implement the CGI protocol directly, but you’d never do this for a real application. The Python standard library includes modules, such as the aptly named cgi module, that implement the CGI protocol and perform other common web application tasks. These implementations are complete, fast, and thor- oughly tested, and it’s nearly always better to use modules like these than to implement your own from scratch. Web application frameworks go beyond libraries of modules to implement the best practices of web application development as a coherent suite of tools, components, and patterns: data modeling interfaces, template systems, request handling mechanisms, project management tools, and development environments that work together to re- duce the amount of code you have to write and maintain. There are dozens of web frameworks written in Python, and several are mature, well documented, and have active developer communities. Django, web2py, and Pylons are examples of well-established Python web frameworks. Not every Python web application framework works completely with the App Engine Python runtime environment. Constraints imposed by App Engine’s sandboxing logic, especially the restriction on modules that use compiled C code, limit which frameworks work out of the box. Django (http://www.djangoproject.com/) is known to work well, and others have been adapted with additional software. We’ll discuss how to use Django with App Engine in Chapter 14. To make it easy to get started, App Engine includes a simple web framework called “webapp.” The webapp framework is intended to be small and easy to use. It doesn’t have the features of more established frameworks, but it’s good enough for small projects. If you created the “clock” project using the Launcher, you may have noticed that the starter files use the webapp framework. For simplicity, most of the Python examples in this book use the webapp framework. We won’t cover webapp in detail in this book, but we’ll introduce some of its features here. For larger applications, you may want to use a more featureful framework such as Django. 28 | Chapter 2: Creating an Application
- 50. Let’s upgrade theclock app to use the webapp framework. Replace the contents of main.py with the version shown in Example 2-3. Reload the page in your browser to see the new version in action. (You won’t notice a difference other than an updated time. This example is equivalent to the previous version.) Example 2-3. A simple request handler using the webapp framework from google.appengine.ext import webapp from google.appengine.ext.webapp.util import run_wsgi_app import datetime class MainPage(webapp.RequestHandler): def get(self): time = datetime.datetime.now() self.response.headers['Content-Type'] = 'text/html' self.response.out.write('<p>The time is: %s</p>' % str(time)) application = webapp.WSGIApplication([('/', MainPage)], debug=True) def main(): run_wsgi_app(application) if __name__ == '__main__': main() Example 2-3 imports the module google.appengine.ext.webapp, then defines a request handler class called MainPage, a subclass of webapp.RequestHandler. The class defines methods for each HTTP method supported by the handler, in this case one method for HTTP GET called get(). When the application handles a request, it instantiates the handler class, sets self.request and self.response to values the handler method can access and modify, then calls the appropriate handler method, in this case get(). When the handler method exits, the application uses the value of self.response as the HTTP response. The application itself is represented by an instance of the class webapp.WSGIApplication. The instance is initialized with a list of mappings of URLs to handler classes. The debug parameter tells the application to print error messages to the browser window when a handler returns an exception if the application is running under the development web server. webapp detects whether it is running under the development server or running as a live App Engine application, and will not print errors to the browser when running live even if debug is True. You can set it to False to have the development server emulate the live server when errors occur. The script defines a main() function that runs the application using a utility method. Lastly, the script calls main() using the Python idiom of if __name__ == '__main__': ..., a condition that is always true when the script is run by the web server. This idiom allows you to import the script as a module for other code, including the classes and functions defined in the script, without running the main() routine. Developing the Application | 29
- 51. Defining a main()function this way allows App Engine to cache the compiled handler script, making subsequent requests faster to execute. For more information on app caching, see Chapter 3. A single WSGIApplication instance can handle multiple URLs, routing the request to different RequestHandler classes based on the URL pattern. But we’ve already seen that the app.yaml file maps URL patterns to handler scripts. So which URL patterns should appear in app.yaml, and which should appear in the WSGIApplication? Many web frameworks include their own URL dispatcher logic, and it’s common to route all dy- namic URLs to the framework’s dispatcher in app.yaml. With webapp, the answer mostly depends on how you’d like to organize your code. For the clock application, we will create a second request handler as a separate script to take advantage of a feature of app.yaml for user authentication, but we could also put this logic in main.py and route the URL with the WSGIApplication object. Users and Google Accounts So far, our clock shows the same display for every user. To allow each user to customize the display and save her preferences for future sessions, we need a way to identify the user making a request. An easy way to do this is with Google Accounts. Let’s add something to the page that indicates whether the user is signed in, and pro- vides links for signing in and signing out of the application. Edit main.py to resemble Example 2-4. Example 2-4. A version of main.py that displays Google Accounts information and links from google.appengine.api import users from google.appengine.ext import webapp from google.appengine.ext.webapp.util import run_wsgi_app import datetime class MainPage(webapp.RequestHandler): def get(self): time = datetime.datetime.now() user = users.get_current_user() if not user: navbar = ('<p>Welcome! <a href="%s">Sign in or register</a> to customize.</p>' % (users.create_login_url(self.request.path))) else: navbar = ('<p>Welcome, %s! You can <a href="%s">sign out</a>.</p>' % (user.email(), users.create_logout_url(self.request.path))) self.response.headers['Content-Type'] = 'text/html' self.response.out.write(''' <html> <head> <title>The Time Is...</title> </head> 30 | Chapter 2: Creating an Application
- 52. <body> %s <p>The time is: %s</p> </body> </html> ''' % (navbar, str(time))) application = webapp.WSGIApplication([('/', MainPage)], debug=True) def main(): run_wsgi_app(application) if __name__ == '__main__': main() In a real application, you would use a templating system for the output, separating the HTML and display logic from the application code. While many web application frameworks include a templating system, webapp does not. Since the clock app only has one page, we’ll put the HTML in the handler code, using Python string formatting to keep things organized. The Python runtime environment includes a version of Django, whose templating system can be used with webapp. When Google released version 1 of the Python runtime environment, the latest version of Django was 0.96, so this is what the runtime includes. For more infor- mation on using Django templates with webapp, see the App Engine documentation. Reload the page in your browser. The new page resembles Figure 2-5. Figure 2-5. The clock app with a link to Google Accounts when the user is not signed in Developing the Application | 31
- 53. The Python API for Google Accounts is provided by the module google.appengine.api.users. The get_current_user() function in this module returns None if the user is not signed in, or an object of the class User with the user’s account information. The email() method on the User object returns the user’s email address. The create_login_url() and create_logout_url() methods generate URLs that go to Google Accounts. Each of these methods takes a URL path for the app where the user should be redirected after performing the desired task. The login URL goes to the Google Accounts page where the user can sign in or register for a new account. The logout URL visits Google Accounts to sign out the current user, then immediately re- directs back to the given application URL. If you click on the “Sign in or register” link with the app running in the development server, the link goes to the development server’s simulated version of the Google Ac- counts sign-in screen, as shown in Figure 2-6. At this screen, you can enter any email address, and the development server will proceed as if you are signed in with an account that has that address. Figure 2-6. The development server’s simulated Google Accounts sign-in screen If this app were running on App Engine, the login and logout URLs would go to the actual Google Accounts locations. Once signed in or out, Google Accounts redirects back to the given URL path for the live application. Click on “Sign in or register,” then click on the Login button on the simulated Google Accounts screen, using the default test email address (test@example.com). The clock app now looks like Figure 2-7. To sign out again, click the “sign out” link. 32 | Chapter 2: Creating an Application
- 54. Figure 2-7. Theclock app, with the user signed in Web forms and the datastore Now that we know who the user is, we can ask her for her preferred time zone, remember her preference, and use it on future visits. First, we need a way to remember the user’s preferences so future requests can access them. The App Engine datastore provides reliable, scalable storage for this purpose. The Python API includes a data modeling interface that maps Python objects to data- store entities. We can use it to write a UserPrefs class. Create a file named models.py, as shown in Example 2-5. Example 2-5. The file models.py, with a class for storing user preferences in the datastore from google.appengine.api import users from google.appengine.ext import db class UserPrefs(db.Model): tz_offset = db.IntegerProperty(default=0) user = db.UserProperty(auto_current_user_add=True) def get_userprefs(user_id=None): if not user_id: user = users.get_current_user() if not user: return None user_id = user.user_id() key = db.Key.from_path('UserPrefs', user_id) userprefs = db.get(key) if not userprefs: Developing the Application | 33
- 55. userprefs = UserPrefs(key_name=user_id) return userprefs The Python data modeling interface is provided by the module google.appengine.ext.db. A data model is a class whose base class is db.Model. The model subclass defines the structure of the data in each object using class properties. This structure is enforced by db.Model when values are assigned to instance properties. For our UserPrefs class, we define two properties: tz_offset, an integer, and user, a User object returned by the Google Accounts API. Every datastore entity has a primary key. Unlike a primary key in a relational database table, an entity key is permanent and can only be set when the entity is created. A key is unique across all entities in the system, and consists of several parts, including the entity’s kind (in this case 'UserPrefs'). An app can set one component of the key to an arbitrary value, known in the API as the key name. The clock application uses the user’s unique ID, provided by the user_id() method of the User object, as the key name of a UserPrefs entity. This allows the app to fetch the entity by key, since it knows the user’s ID from the Google Accounts API. Fetching the entity by key is faster than performing a datastore query. In models.py, we define a function named get_userprefs() that gets the UserPrefs ob- ject for the user. After determining the user ID, the function constructs a datastore key for an entity of the kind 'UserPrefs' with a key name equivalent to the user ID. If the entity exists in the datastore, the function returns the UserPrefs object. If the entity does not exist in the datastore, the function creates a new UserPrefs object with default settings and a key name that corresponds to the user. The new object is not saved to the datastore automatically. The caller must invoke the put() method on the UserPrefs instance to save it. Now that we have a mechanism for getting a UserPrefs object, we can make two up- grades to the main page. If the user is signed in, we can get the user’s preferences (if any) and adjust the clock’s time zone. Let’s also display a web form so the user can set a time zone preference. Edit main.py to resemble Example 2-6 to implement these features. Example 2-6. A new version of main.py that adjusts the clock to the user’s time zone and displays a preferences form from google.appengine.api import users from google.appengine.ext import webapp from google.appengine.ext.webapp.util import run_wsgi_app import datetime import models class MainPage(webapp.RequestHandler): def get(self): time = datetime.datetime.now() user = users.get_current_user() 34 | Chapter 2: Creating an Application
- 56. if not user: navbar = ('<p>Welcome! <a href="%s">Sign in or register</a> to customize.</p>' % (users.create_login_url(self.request.path))) tz_form = '' else: userprefs = models.get_userprefs() navbar = ('<p>Welcome, %s! You can <a href="%s">sign out</a>.</p>' % (user.nickname(), users.create_logout_url(self.request.path))) tz_form = ''' <form action="/prefs" method="post"> <label for="tz_offset"> Timezone offset from UTC (can be negative): </label> <input name="tz_offset" id="tz_offset" type="text" size="4" value="%d" /> <input type="submit" value="Set" /> </form> ''' % userprefs.tz_offset time += datetime.timedelta(0, 0, 0, 0, 0, userprefs.tz_offset) self.response.headers['Content-Type'] = 'text/html' self.response.out.write(''' <html> <head> <title>The Time Is...</title> </head> <body> %s <p>The time is: %s</p> %s </body> </html> ''' % (navbar, str(time), tz_form)) application = webapp.WSGIApplication([('/', MainPage)], debug=True) def main(): run_wsgi_app(application) if __name__ == '__main__': main() To enable the preferences form, we need a request handler to parse the form data and update the datastore. Let’s implement this as a new request handler script. Create a file named prefs.py with the contents shown in Example 2-7. Example 2-7. A new handler script, prefs.py, for the preferences form from google.appengine.ext import webapp from google.appengine.ext.webapp.util import run_wsgi_app import models class PrefsPage(webapp.RequestHandler): def post(self): Developing the Application | 35
- 57. userprefs = models.get_userprefs() try: tz_offset = int(self.request.get('tz_offset')) userprefs.tz_offset = tz_offset userprefs.put() except ValueError: # User entered a value that wasn't an integer. Ignore for now. pass self.redirect('/') application = webapp.WSGIApplication([('/prefs', PrefsPage)], debug=True) def main(): run_wsgi_app(application) if __name__ == '__main__': main() This request handler handles HTTP POST requests to the URL /prefs, which is the URL (“action”) and HTTP method used by the form. The handler calls the get_userprefs() function from models.py to get the UserPrefs object for the current user, which is either a new unsaved object with default values, or the object for an existing entity. The handler parses the tz_offset parameter from the form data as an integer, sets the property of the UserPrefs object, then saves the object to the datastore by calling its put() method. The put() method creates the object if it doesn’t exist, or updates the existing object. If the user enters a noninteger in the form field, we don’t do anything. It’d be appro- priate to return an error message, but we’ll leave this as is to keep the example simple. Finally, edit app.yaml to map the handler script to the URL /prefs in the handlers: section, as shown in Example 2-8. Example 2-8. A new version of app.yaml mapping the URL /prefs, with login required application: clock version: 1 runtime: python api_version: 1 handlers: - url: /prefs script: prefs.py login: required - url: /.* script: main.py The login: required line says that the user must be signed in to Google Accounts to access the /prefs URL. If the user accesses the URL while not signed in, App Engine automatically directs the user to the Google Accounts sign-in page, then redirects her 36 | Chapter 2: Creating an Application
- 58. back to thisURL afterward. This makes it easy to require sign-in for sections of your site, and to ensure that the user is signed in before the request handler is called. Be sure to put the /prefs URL mapping before the /.* mapping. URL patterns are tried in order, and the first pattern to match determines the handler used for the request. Since the pattern /.* matches all URLs, /prefs must come first or it will be ignored. Reload the page to see the customizable clock in action. Try changing the time zone by submitting the form. Also try signing out, then signing in again using the same email address, and again with a different email address. The app remembers the time zone preference for each user. Caching with memcache The code that gets user preferences data in Example 2-5 fetches an entity from the datastore every time a signed-in user visits the site. User preferences are often read and seldom changed, so getting a UserPrefs object from the datastore with every request is more expensive than it needs to be. We can mitigate the costs of reading from primary storage using a caching layer. We can use the memcache service as secondary storage for user preferences data. We can add caching with just a few changes to models.py. Edit this file as shown in Exam- ple 2-9. Example 2-9. A new version of models.py that caches UserPrefs objects in memcache from google.appengine.api import memcache from google.appengine.api import users from google.appengine.ext import db class UserPrefs(db.Model): tz_offset = db.IntegerProperty(default=0) user = db.UserProperty(auto_current_user_add=True) def cache_set(self): memcache.set(self.key().name(), self, namespace=self.key().kind()) def put(self): self.cache_set() db.Model.put(self) def get_userprefs(user_id=None): if not user_id: user = users.get_current_user() if not user: return None user_id = user.user_id() userprefs = memcache.get(user_id, namespace='UserPrefs') if not userprefs: key = db.Key.from_path('UserPrefs', user_id) userprefs = db.get(key) Developing the Application | 37
- 59. if userprefs: userprefs.cache_set() else: userprefs = UserPrefs(key_name=user_id) return userprefs The Python API for the memcache service is provided by the module google.appengine.api.memcache. The memcache stores key-value pairs, with an op- tional namespace for the key. The value can be of any type that can be converted to and from a flat data representation (serialized) using the Python pickle module, in- cluding most data objects. The new version of the UserPrefs class overrides the put() method. When the put() method is called on an instance, the instance is saved to the memcache, then it is saved to the datastore using the original put() method. (db.Model.put(self) is one way to call the overridden superclass method in Python.) A new UserPrefs method called cache_set() makes the call to memcache.set(). memcache.set() takes a key, a value, and an optional namespace for the key. Here, we use the entity’s key name as the key, the full object (self) as the value, and the entity’s kind ('UserPrefs') as the namespace. The API takes care of serializing the UserPrefs object, so we can put in and take out fully formed objects. The new version of get_userprefs() checks the memcache for the UserPrefs object before going to the datastore. If it finds it in the cache, it uses it. If it doesn’t, it checks the datastore, and if it finds it there, it stores it in the cache and uses it. If the object is in neither the memcache nor the datastore, get_userprefs() returns a fresh UserPrefs object with default values. Reload the page to see the new version work. To make the caching behavior more visible, you can add logging statements in the appropriate places, like so: import logging class UserPrefs(db.Model): # ... def cache_set(self): logging.info('cache set') # ... The development server prints logging output to the console. If you are using the Launcher, you can open a window of development server output by clicking the Logs button. Next, we’ll take a look at the same example using the Java runtime environment. If you’re not interested in Java, you can skip ahead to “Registering the Applica- tion” on page 55. 38 | Chapter 2: Creating an Application
- 60. Developing a JavaApp Java web applications for App Engine use the Java Servlet standard interface for inter- acting with the application server. An application consists of one or more classes that extend a servlet base class. Servlets are mapped to URLs using a standard configuration file called a “deployment descriptor.” When App Engine receives a request for a Java application, it determines which servlet class to use based on the URL and the deploy- ment descriptor, instantiates the class, then calls an appropriate method on the servlet object. All of the files for a Java application, including the compiled Java classes, configuration files, and static files, are organized in a standard directory structure called a Web Ap- plication Archive, or “WAR.” Everything in the WAR directory gets deployed to App Engine. It’s common to have your development workflow build the contents of the WAR from a set of source files, either using an automated build process or WAR-aware development tools. If you are using the Eclipse IDE with the Google Plugin, you can create a new project using the Web Application wizard. From the File menu, select New, then Web Appli- cation Project. In the window that opens, enter a project name (such as Clock) and package name (such as clock). Uncheck the “Use Google Web Toolkit” checkbox, and make sure the “Use Google App Engine” checkbox is checked. (If you leave the GWT checkbox checked, the new project will be created with GWT starter files.) Click Finish to create the project. If you are not using the Google Plugin for Eclipse, you will need to create the directories and files another way. If you are already familiar with Java web development, you can use your existing tools and processes to produce the final WAR. For the rest of this section, we’ll assume the directory structure created by the Eclipse plug-in. Figure 2-8 shows the project file structure, as depicted in the Eclipse Package Explorer. The project root directory (Clock) contains two major subdirectories: src and war. The src/ directory contains all of the project’s class files in the usual Java package structure. With a package path of clock, Eclipse created source code for a servlet class named ClockServlet in the file clock/ClockServlet.java. The war/ directory contains the complete final contents of the application. Eclipse compiles source code from src/ automatically and puts the compiled class files in war/ WEB-INF/classes/, which is hidden from Eclipse’s Package Explorer by default. Eclipse copies the contents of src/META-INF/ to war/WEB-INF/classes/META-INF/ automat- ically, as well. Everything else must be created in the war/ directory in its intended location. Let’s start our clock application with a simple servlet that displays the current time. Open the file src/clock/ClockServlet.java for editing (creating it if necessary), and give it contents similar to Example 2-10. Developing the Application | 39
- 61. Figure 2-8. Anew Java project structure, as shown in the Eclipse Package Explorer Example 2-10. A simple Java servlet package clock; import java.io.IOException; import java.io.PrintWriter; import java.text.SimpleDateFormat; import java.util.Date; import java.util.SimpleTimeZone; import javax.servlet.http.*; @SuppressWarnings("serial") public class ClockServlet extends HttpServlet { public void doGet(HttpServletRequest req, HttpServletResponse resp) throws IOException { SimpleDateFormat fmt = new SimpleDateFormat("yyyy-MM-dd hh:mm:ss.SSSSSS"); fmt.setTimeZone(new SimpleTimeZone(0, "")); resp.setContentType("text/html"); PrintWriter out = resp.getWriter(); out.println("<p>The time is: " + fmt.format(new Date()) + "</p>"); } } The servlet class extends javax.servlet.http.HttpServlet, and overrides methods for each of the HTTP methods it intends to support. This servlet overrides the doGet() 40 | Chapter 2: Creating an Application
- 62. method to handleHTTP GET requests. The server calls the method with an HttpServletRequest object and an HttpServletResponse object as parameters. The HttpServletRequest contains information about the request, such as the URL, form parameters, and cookies. The method prepares the response using methods on the HttpServletResponse, such as setContentType() and getWriter(). App Engine sends the response when the servlet method exits. To tell App Engine to invoke this servlet for requests, we need a deployment descriptor. Open or create the file war/WEB-INF/web.xml, and give it contents similar to Exam- ple 2-11. Example 2-11. The web.xml file, also known as the deployment descriptor, mapping all URLs to ClockServlet <?xml version="1.0" encoding="utf-8"?> <!DOCTYPE web-app PUBLIC "-//Sun Microsystems, Inc.//DTD Web Application 2.3//EN" "http://java.sun.com/dtd/web-app_2_3.dtd"> <web-app xmlns="http://java.sun.com/xml/ns/javaee" version="2.5"> <servlet> <servlet-name>clock</servlet-name> <servlet-class>clock.ClockServlet</servlet-class> </servlet> <servlet-mapping> <servlet-name>clock</servlet-name> <url-pattern>/*</url-pattern> </servlet-mapping> </web-app> Eclipse may open this file in its XML “Design” view, a table-like view of the elements and values. Select the “Source” tab at the bottom of the editor pane to edit the XML source. web.xml is an XML file with a root element of <web-app>. To map URL patterns to servlets, you declare each servlet with a <servlet> element, then declare the mapping with a <servlet-mapping> element. The <url-pattern> of a servlet mapping can be a full URL path, or a URL path with a * at the beginning or end to represent a part of a path. In this case, the URL pattern /* matches all URLs. Be sure that each of your <url-pattern> values starts with a forward slash (/). Omitting the starting slash may have the intended behavior on the development web server but unintended behavior on App Engine. App Engine needs one additional configuration file that isn’t part of the servlet standard. Open or create the file war/WEB-INF/appengine-web.xml, and give it con- tents similar to Example 2-12. Developing the Application | 41
- 63. Example 2-12. Theappengine-web.xml file, with App Engine-specific configuration for the Java app <?xml version="1.0" encoding="utf-8"?> <appengine-web-app xmlns="http://appengine.google.com/ns/1.0"> <application>clock</application> <version>1</version> </appengine-web-app> In this example, the configuration file tells App Engine that this is version 1 of an application called clock. You can also use this configuration file to control other be- haviors, such as static files and sessions. For more information, see Chapter 3. The WAR for the application must include several JARs from the App Engine SDK: the Java EE implementation JARs, and the App Engine API JAR. The Eclipse plug-in installs these JARs in the WAR automatically. If you are not using the Eclipse plug-in, you must copy these JARs manually. Look in the SDK directory in the lib/user/ and lib/shared/ subdirectories. Copy every .jar file from these directories to the war/WEB-INF/lib/ directory in your project. Finally, the servlet class must be compiled. Eclipse compiles all of your classes auto- matically, as needed. If you are not using Eclipse, you probably want to use a build tool such as Apache Ant to compile source code and perform other build tasks. See the official App Engine documentation for information on using Apache Ant to build App Engine projects. I suppose it’s traditional to explain how to compile a Java project from the command line using the javac command. You can do so by putting each of the JARs from war/ WEB-INF/lib/ and the war/WEB-INF/classes/ directory in the classpath, and making sure the compiled classes end up in the classes/ directory. But in the real world, you want your IDE or an Ant script to take care of this for you. Also, when we introduce the datastore in the next few sections, we will need to add a step to building the project that makes this even more impractical to do by hand. It’s time to test this application with the development web server. The Eclipse plug-in can run the application and the development server inside the Eclipse debugger. To start it, select the Run menu, Debug As, and Web Application. The server starts, and prints the following message to the Console panel: The server is running at http://localhost:8080/ If you are not using Eclipse, you can start the development server using the dev_appserver command (dev_appserver.sh for Mac OS X or Linux). The command takes the path to the WAR directory as an argument, like so: dev_appserver war Test your application by visiting the server’s URL in a web browser: http://localhost:8080 The browser displays a page similar to the Python example, shown earlier in Figure 2-4. 42 | Chapter 2: Creating an Application
- 64. Users and GoogleAccounts Right now, our clock displays the time in the UTC time zone. We’d like for our appli- cation to let the user customize the time zone, and to remember the user’s preference for future visits. To do that, we will use Google Accounts to identify which user is using the application. Edit ClockServlet.java to resemble Example 2-13. Example 2-13. Code for ClockServlet.java that displays Google Accounts information and links package clock; import java.io.IOException; import java.io.PrintWriter; import java.text.SimpleDateFormat; import java.util.Date; import java.util.SimpleTimeZone; import javax.servlet.http.*; import com.google.appengine.api.users.User; import com.google.appengine.api.users.UserService; import com.google.appengine.api.users.UserServiceFactory; @SuppressWarnings("serial") public class ClockServlet extends HttpServlet { public void doGet(HttpServletRequest req, HttpServletResponse resp) throws IOException { SimpleDateFormat fmt = new SimpleDateFormat("yyyy-MM-dd hh:mm:ss.SSSSSS"); fmt.setTimeZone(new SimpleTimeZone(0, "")); UserService userService = UserServiceFactory.getUserService(); User user = userService.getCurrentUser(); String navBar; if (user != null) { navBar = "<p>Welcome, " + user.getNickname() + "! You can <a href="" + userService.createLogoutURL("/") + "">sign out</a>.</p>"; } else { navBar = "<p>Welcome! <a href="" + userService.createLoginURL("/") + "">Sign in or register</a> to customize.</p>"; } resp.setContentType("text/html"); PrintWriter out = resp.getWriter(); out.println(navBar); out.println("<p>The time is: " + fmt.format(new Date()) + "</p>"); } } In a real application, you wouldn’t mix HTML and Java source code like this. You can use JavaServer Pages (JSPs) to represent the page, or you can use a templating engine to render output. To keep things simple, we will continue to write HTML directly from the servlet code for the rest of this example, but keep in mind this is not a best practice. Developing the Application | 43
- 65. Using Eclipse, youcan leave the development web server running while you edit code. When you save changes to code, Eclipse compiles the class, and if it compiles success- fully, Eclipse injects the new class into the already-running server. In most cases, you can simply reload the page in your browser, and it will use the new code. If you are not using Eclipse, shut down the development server by hitting Ctrl-C. Re- compile your project, then start the server again. Reload the new version of the clock app in your browser. The new page resembles the Python example, shown previously in Figure 2-5. This version of the clock app uses the interface for Google Accounts provided by the com.google.appengine.api.users package. The app gets a UserService instance by call- ing the getUserService() method of the UserServiceFactory class. Then it calls the getCurrentUser() method of the UserService, which returns a User object, or null if the current user is not signed in. The getEmail() method of the User object returns the email address for the user. The createLoginURL() and createLogoutURL() methods of the UserService generate URLs that go to Google Accounts. Each of these methods takes a URL path for the app where the user should be redirected after performing the desired task. The login URL goes to the Google Accounts page where the user can sign in or register for a new account. The logout URL visits Google Accounts to sign out the current user, then immediately redirects back to the given application URL without displaying anything. If you click on the “Sign in or register” link with the app running in the development server, the link goes to the development server’s simulated version of the Google Ac- counts sign-in screen, similar to the Python version shown earlier in Figure 2-6. At this screen, you can enter any email address, and the development server will proceed as if you are signed in with an account that has that address. If this app were running on App Engine, the login and logout URLs would go to the actual Google Accounts locations. Once signed in or out, Google Accounts redirects back to the given URL path for the live application. Click on “Sign in or register,” then enter an email address (such as test@example.com) and click on the Login button on the simulated Google Accounts screen. The clock app now looks like Figure 2-7 (shown earlier). To sign out again, click the “sign out” link. In addition to the UserService API, an app can also get information about the current user with the servlet “user principal” interface. The app can call the getUserPrincipal() method on the HttpServletRequest object to get a java.security.Principal object, or null if the user is not signed in. This object has a getName() method, which in App Engine is equivalent to calling the getEmail() method of a User object. The main advantage to getting user information from the servlet interface is that the servlet interface is a standard. Coding an app to use standard interfaces makes the app easier to port to alternate implementations, such as other servlet-based web application 44 | Chapter 2: Creating an Application
- 66. environments or privateservers. As much as possible, App Engine implements standard interfaces for its services and features. The disadvantage to the standard interfaces is that not all standard interfaces represent all of App Engine’s features, and in some cases the App Engine services don’t implement every feature of an interface. All of the services include a nonstandard “low-level” API, which you can use directly or use to implement adapters to other interfaces. Web forms and the datastore Now that we can identify the user, we can prompt for the user’s preferences and re- member them for future requests. We can store preferences data in the App Engine datastore. The App Engine SDK supports two major standard interfaces for accessing the datastore: Java Data Objects (JDO) 2.3 and the Java Persistence API (JPA) 1.0. As with the other services, the datastore also has a low-level API. Let’s use the JPA interface to store the user’s time zone setting. JPA requires a config- uration file that specifies the JPA implementation to use, and other options. The final location of this file is war/WEB-INF/classes/META-INF/persistence.xml. If you are using Eclipse, you can create this file as src/META-INF/persistence.xml, and Eclipse will copy it to the final location automatically. Create the file src/META-INF/persistence.xml with the contents shown in Exam- ple 2-14. Example 2-14. The JPA configuration file, persistence.xml, with several useful options <?xml version="1.0" encoding="UTF-8" ?> <persistence xmlns="http://java.sun.com/xml/ns/persistence" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://java.sun.com/xml/ns/persistence http://java.sun.com/xml/ns/persistence/persistence_1_0.xsd" version="1.0"> <persistence-unit name="transactions-optional"> <provider>org.datanucleus.store.appengine.jpa.DatastorePersistenceProvider</provider> <properties> <property name="datanucleus.NontransactionalRead" value="true"/> <property name="datanucleus.NontransactionalWrite" value="true"/> <property name="datanucleus.ConnectionURL" value="appengine"/> </properties> </persistence-unit> </persistence> The application interacts with the datastore using an EntityManager object. It gets this object from an EntityManagerFactory. For efficiency, it’s best to instantiate the factory only once per servlet. You can store this instance in a static member of a wrapper class. Create a new class named EMF in the clock package (src/clock/EMF.java) resembling Example 2-15. Developing the Application | 45
- 67. Example 2-15. Thefile EMF.java, a static wrapper class for an EntityManagerFactory instance package clock; import javax.persistence.EntityManagerFactory; import javax.persistence.Persistence; public final class EMF { private static final EntityManagerFactory emfInstance = Persistence.createEntityManagerFactory("transactions-optional"); private EMF() {} public static EntityManagerFactory get() { return emfInstance; } } JPA makes your Java data objects persistent. As far as the rest of your application is concerned, the data objects are just plain old Java objects (POJOs), with members and methods. When you create an instance of a data class, you declare it as persistent by passing it to the EntityManager. From that point on, JPA ensures that changes to the object are saved to the datastore. When you retrieve the object from the datastore later, it still has all of its data, and it retains its persistent behavior. When you define a JPA data class, you declare it as a persistent-capable class, and optionally tell JPA how to save and restore instances of the class. You do this using Java annotations in the class definition. Example 2-16 shows the code for a user preferences data class called UserPrefs, using annotations to declare it as persistent-capable by JPA. Create this class (src/clock/UserPrefs.java). Example 2-16. Code for UserPrefs.java, a data class using JPA annotations to make instances persistent package clock; import javax.persistence.Basic; import javax.persistence.Entity; import javax.persistence.EntityManager; import javax.persistence.Id; import com.google.appengine.api.users.User; import clock.EMF; import clock.UserPrefs; @Entity(name = "UserPrefs") public class UserPrefs { @Id private String userId; private int tzOffset; 46 | Chapter 2: Creating an Application
- 68. @Basic private User user; public UserPrefs(String userId) { this.userId = userId; } public String getUserId() { return userId; } public int getTzOffset() { return tzOffset; } public void setTzOffset(int tzOffset) { this.tzOffset = tzOffset; } public User getUser() { return user; } public void setUser(User user) { this.user = user; } public static UserPrefs getPrefsForUser(User user) { UserPrefs userPrefs = null; EntityManager em = EMF.get().createEntityManager(); try { userPrefs = em.find(UserPrefs.class, user.getUserId()); if (userPrefs == null) { userPrefs = new UserPrefs(user.getUserId()); userPrefs.setUser(user); } } finally { em.close(); } return userPrefs; } public void save() { EntityManager em = EMF.get().createEntityManager(); try { em.persist(this); } finally { em.close(); } } } Developing the Application | 47
- 69. The UserPrefs classhas three members: the user ID, the user’s time zone preference, and the User object representing the current user (which contains the user’s email ad- dress). The class is declared as persistence-capable using the @Entity annotation. Its name argument specifies the name to be used in JPA queries, typically the same as the class name. By default, the name of the underlying datastore entity kind is derived from the simple name of the class, in this case UserPrefs. The user field gets an @Basic annotation because JPA does not recognize its field type (User) as persistence-capable by default. String and int are understood by JPA as persistence-capable by default. The userId field is the primary key for the object, as declared by the @Id annotation. Unlike a relational database, the key is not a field of a record, but a permanent aspect of the underlying datastore entity, set when the object is first saved. When the data class uses a String member as the primary key, JPA expects the member to be set to the key name, a value unique across all objects of this class, when the object is saved for the first time. This value cannot be changed once the object is saved. This application creates a UserPrefs object for each user with preferences set, using the unique user ID provided by Google Accounts as the key name of the object’s key. When a user visits the clock, the app attempts to get the UserPrefs object via the key con- structed from the user ID, and adjusts the clock display accordingly if such an object is found. There are other ways to declare the primary key using JPA and App Engine, including a way to let the datastore assign a unique ID automatically. These are discussed in Chapter 8. JPA attaches its plumbing to the data class after the class is compiled, in a step called “enhancement.” The annotations tell the enhancement process how to modify the compiled class bytecode, adding calls to the JPA API in places that ensure the object’s persistent members are saved to the datastore. If you are using Eclipse and the plug-in, JPA enhancement happens automatically when data classes are compiled. The App Engine SDK includes an Ant plug-in that performs this step, and you can also run the enhancement process from your own build process by running a tool. See the official App Engine documentation for more information on performing the JPA class en- hancement step from a build script. In anticipation of adding caching logic later on, we’ve included two methods on this class for getting and saving UserPrefs objects. The static method getPrefsForUser() takes a User object, as returned by the Google Accounts API, and attempts to fetch the UserPrefs object from the datastore for that user. The instance method save() stores the object in the datastore, creating a new datastore entity if one does not already exist for this key, or updating the existing entity if one does. (This save() method goes against JPA’s notion of automatic object persistence, but is a convenient way to integrate memcache later with very little code.) 48 | Chapter 2: Creating an Application
- 70. It’s time toupgrade the clock application to allow the user to customize the time zone of the clock. Example 2-17 shows a new version of the ClockServlet class that retrieves the UserPrefs object for the currently signed-in user, if any, and uses it to customize the clock display. It also displays a web form that the user can submit to change her time zone preference. Example 2-17. A new version of ClockServlet.java that adjusts the clock to the user’s time zone and displays a preferences form // ... import clock.UserPrefs; @SuppressWarnings("serial") public class ClockServlet extends HttpServlet { public void doGet(HttpServletRequest req, HttpServletResponse resp) throws IOException { SimpleDateFormat fmt = new SimpleDateFormat("yyyy-MM-dd hh:mm:ss.SSSSSS"); UserService userService = UserServiceFactory.getUserService(); User user = userService.getCurrentUser(); String navBar; String tzForm; if (user == null) { navBar = "<p>Welcome! <a href="" + userService.createLoginURL("/") + "">Sign in or register</a> to customize.</p>"; tzForm = ""; fmt.setTimeZone(new SimpleTimeZone(0, "")); } else { UserPrefs userPrefs = UserPrefs.getPrefsForUser(user); int tzOffset = 0; if (userPrefs != null) { tzOffset = userPrefs.getTzOffset(); } navBar = "<p>Welcome, " + user.getEmail() + "! You can <a href="" + userService.createLogoutURL("/") + "">sign out</a>.</p>"; tzForm = "<form action="/prefs" method="post">" + "<label for="tz_offset">" + "Timezone offset from UTC (can be negative):" + "</label>" + "<input name="tz_offset" id="tz_offset" type="text" size="4" " + "value="" + tzOffset + "" />" + "<input type="submit" value="Set" />" + "</form>"; fmt.setTimeZone(new SimpleTimeZone(tzOffset * 60 * 60 * 1000, "")); } resp.setContentType("text/html"); PrintWriter out = resp.getWriter(); out.println(navBar); out.println("<p>The time is: " + fmt.format(new Date()) + "</p>"); Developing the Application | 49
- 71. out.println(tzForm); } } To enable the preferences form, we need a servlet to parse the form data and update the datastore. Let’s implement this as a new servlet class. Create a class named PrefsServlet (src/clock/PrefsServlet.java) with the contents shown in Example 2-18. Example 2-18. Code for PrefsServlet.java, a servlet that handles the preferences form package clock; import java.io.IOException; import javax.servlet.http.HttpServlet; import javax.servlet.http.HttpServletRequest; import javax.servlet.http.HttpServletResponse; import com.google.appengine.api.users.User; import com.google.appengine.api.users.UserService; import com.google.appengine.api.users.UserServiceFactory; import clock.UserPrefs; @SuppressWarnings("serial") public class PrefsServlet extends HttpServlet { public void doPost(HttpServletRequest req, HttpServletResponse resp) throws IOException { UserService userService = UserServiceFactory.getUserService(); User user = userService.getCurrentUser(); UserPrefs userPrefs = UserPrefs.getPrefsForUser(user); try { int tzOffset = new Integer(req.getParameter("tz_offset")).intValue(); userPrefs.setTzOffset(tzOffset); userPrefs.save(); } catch (NumberFormatException nfe) { // User entered a value that wasn't an integer. Ignore for now. } resp.sendRedirect("/"); } } Next, we need to change web.xml to map this servlet to the /prefs URL. Edit this file and add the XML shown in Example 2-19. Example 2-19. Mapping the URL /prefs to the PrefsServlet using a security constraint in web.xml (excerpt) <servlet> <servlet-name>prefs</servlet-name> <servlet-class>clock.PrefsServlet</servlet-class> </servlet> <servlet-mapping> 50 | Chapter 2: Creating an Application
- 72. <servlet-name>prefs</servlet-name> <url-pattern>/prefs</url-pattern> </servlet-mapping> <security-constraint> <web-resource-collection> <web-resource-name>prefs</web-resource-name> <url-pattern>/prefs</url-pattern> </web-resource-collection> <auth-constraint> <role-name>*</role-name> </auth-constraint> </security-constraint> The order in which the URL mappings appear in the file does not matter. Longer pat- terns (not counting wildcards) match before shorter ones. The <security-constraint> block tells App Engine that only users signed in with a Google Account can access the URL /prefs. If a user who is not signed in attempts to access this URL, App Engine redirects the user to Google Accounts to sign in. When the user signs in, she is directed back to the URL she attempted to access. A security constraint is a convenient way to implement Google Accounts authentication for a set of URLs. In this case, it means that PrefsServlet does not need to handle the case where someone tries to submit data to the URL without being signed in. The servlet accesses the form data using the HttpServletRequest object passed to the doPost() method. For now, if the user enters a noninteger in the form field, we don’t do anything. Later, we can implement an error message. If the form data is valid, the servlet sets the value on the UserPrefs object obtained from our getPrefsForUser() method, then calls our save() method. The save() method opens a PersistenceManager, attaches the UserPrefs object (perhaps making it persis- tent for the first time), then closes the PersistenceManager to save the object in the datastore. Finally, PrefsServlet redirects the user back to the main page. Redirecting after the form submission allows the user to reload the main page without resubmitting the form. Restart your development server, then reload the page to see the customizable clock in action. Try changing the time zone by submitting the form. Also try signing out, then signing in again using the same email address, and again with a different email address. The app remembers the time zone preference for each user. Caching with memcache So far, our application fetches the object from the datastore every time a signed-in user visits the site. Since user preferences data doesn’t change very often, we can speed up the per-request data access using the memory cache as secondary storage. Developing the Application | 51
- 73. As with theother services, the App Engine SDK includes two interfaces to the memcache service: a featureful proprietary API, and an API that conforms to a proposed Java standard known as JCache (JSR 107). We could use either for this example; for now, we’ll use the proprietary API. Because we limited fetching and saving UserPrefs objects to two methods, we can im- plement the caching of UserPrefs objects with minimal changes. Example 2-20 shows the needed changes to the UserPrefs class. Example 2-20. Changes for UserPrefs.java to implement caching of UserPrefs objects import java.io.Serializable; import com.google.appengine.api.memcache.MemcacheService; import com.google.appengine.api.memcache.MemcacheServiceException; import com.google.appengine.api.memcache.MemcacheServiceFactory; // ... @SuppressWarnings("serial") @Entity(name = "UserPrefs") public class UserPrefs implements Serializable { // ... @SuppressWarnings("unchecked") public static UserPrefs getPrefsForUser(User myUser) { UserPrefs userPrefs = null; String cacheKey = "UserPrefs:" + myUser.getUserId(); try { MemcacheService memcache = MemcacheServiceFactory.getMemcacheService(); if (memcache.contains(cacheKey)) { userPrefs = (UserPrefs) memcache.get(cacheKey); return userPrefs; } // If the UserPrefs object isn't in memcache, // fall through to the datastore. } catch (MemcacheServiceException e) { // If there is a problem with the cache, // fall through to the datastore. } EntityManager em = EMF.get().createEntityManager(); try { userPrefs = em.find(UserPrefs.class, myUser.getUserId()); if (userPrefs == null) { userPrefs = new UserPrefs(myUser.getUserId()); userPrefs.setUser(myUser); } else { try { MemcacheService memcache = MemcacheServiceFactory.getMemcacheService(); memcache.put(cacheKey, userPrefs); } catch (MemcacheServiceException e) { // Ignore cache problems, nothing we can do. } 52 | Chapter 2: Creating an Application
- 74. } } finally { em.close(); } return userPrefs; } public void save() { EntityManager em = EMF.get().createEntityManager(); try { em.persist(this); } finally { em.close(); } } } Any object you store in the memcache must be serializable; that is, it must implement the Serializable interface from the java.io package. For UserPrefs, it suffices to de- clare that the class implements the interface, since all the relevant members are already serializable. The new version of the getPrefsForUser() static method checks to see whether the UserPrefs object for the given user is present in the cache before going to the datastore. Each cache value is stored with a key, which itself can be any serializable object. For UserPrefs objects, we use a cache key equivalent to the string "UserPrefs:" followed by the email address from the User object. If a value with that key is not in the cache, or if there is a problem accessing the cache, the method proceeds to get the object from the datastore, then stores it in the cache by calling a new helper method, cacheSet(). Similarly, the new version of the save() method stores the object in the datastore, then also stores it in the cache. There is no way to guarantee that the cache and the datastore contain the same value if one of the services fails and the other succeeds, but it’s usually sufficient to save to the datastore first, then save to the cache. For further safety, we could set the cache values to expire after a period of time, so if they do get out of sync, it won’t be for long. As written, the cache values persist as long as possible in memory. Reload the page to see the new version work. To make the caching behavior more visible, you can add logging statements in the appropriate places, like so: import java.util.logging.*; import javax.persistence.Transient; // ... public class UserPrefs implements Serializable { @Transient private static Logger logger = Logger.getLogger(UserPrefs.class.getName()); // ... if (cache.containsKey(cacheKey)) { logger.warning("CACHE HIT"); Developing the Application | 53
- 75. userPrefs = (UserPrefs)cache.get(cacheKey); return userPrefs; } logger.warning("CACHE MISS"); The development server prints logging output to the console. If you are using Eclipse, these messages appear in the Console pane. The Development Console The Python and Java development web servers include a handy feature for inspecting and debugging your application while testing on your local machine: a web-based de- velopment console. With your development server running, visit the following URL in a browser to access the console: http://localhost:8080/_ah/admin In the Python Launcher, you can also click the SDK Console button to open the console in a browser window. The Java development console is currently behind the Python console in features, but it’s catching up. Figure 2-9 shows the datastore viewer in the Python console. Figure 2-9. The development console’s datastore viewer, Python version The Python development console’s datastore viewer lets you list and inspect entities by kind, edit entities, and create new ones. You can edit the values for existing prop- erties, but you cannot delete properties or add new ones, nor can you change the type of the value. For new entities, the console makes a guess as to which properties belong on the entity based on existing entities of that kind, and displays a form to fill in those properties. Similarly, you can only create new entities of existing kinds, and cannot create new kinds from the console. 54 | Chapter 2: Creating an Application
- 76. The Python consolealso has a viewer for the memcache. You can see cache statistics, and inspect, create, and delete keys. Values are displayed and edited in their serialized (“pickled”) form. An especially powerful feature of the Python console is the “Interactive Console.” This feature lets you type arbitrary Python code directly into a web form and see the results displayed in the browser. You can use this to write ad hoc Python code to test and manipulate the datastore, memcache, and global data within the local development server. Here’s an example: run your clock application, sign in with an email address, then set a time zone preference, such as -8. Now open the Python development console, then select “Interactive Console.” In the lefthand text box, enter the following, where -8 is the time zone preference you used: from google.appengine.ext import db import models q = models.UserPrefs.gql("WHERE tz_offset = -8") for prefs in q: print prefs.user Click the Run Program button. The code runs, and the email address you used appears in the righthand box. Code run in the development console behaves just like application code. If you perform a datastore query that needs a custom index, the development server adds configuration for that index to the application’s index.yaml configuration file. Datastore index con- figuration is discussed in Chapter 5. You can use the Python console to inspect the application’s task queue and cron job configuration in the browser. You can also use the task queue inspector to see tasks currently on the queue (in the local instance of the app), run them, and flush them. (The development server does not run task queues in the background; you must run them from the console. See Chapter 13.) Lastly, you can test how your app receives email and XMPP messages by sending it mock messages through the console. The Java development server also has a console. It includes a datastore viewer that lets you list and inspect datastore entities by kind, the ability to run task queues, and the ability to send email and XMPP messages to the app. Registering the Application Before you can upload your application to App Engine and share it with the world, you must first create a developer account, then register an application ID. If you intend to use a custom domain name (instead of the free appspot.com domain name included Registering the Application | 55
- 77. with every app),you must also set up the Google Apps service for the domain. You can do all of this from the App Engine Administration Console. To access the Administration Console, visit the following URL in your browser: https://appengine.google.com/ Sign in using the Google account you intend to use as your developer account. If you don’t already have a Google account (such as a Gmail account), you can create one using any email address. Once you have signed in, the Console displays a list of applications you have created, if any, and a button to “Create an Application,” similar to Figure 2-10. From this screen, you can create and manage multiple applications, each with its own URL, configura- tion, and resource limits. Figure 2-10. The Administration Console application list, with one app When you register your first application ID, the Administration Console prompts you to verify your developer account using an SMS message sent to your mobile phone. After you enter your mobile phone number, Google sends an SMS to your phone with a confirmation code. Enter this code to continue the registration process. You can verify only one account per phone number, so if you have only one mobile number (like most people), be sure to use it with the account you intend to use with App Engine. If you don’t have a mobile phone number, you can apply to Google for manual verifi- cation by filling out a web form. This process takes about a week. For information on applying for manual verification, see the official App Engine website. You can have up to 10 active application IDs created by a given developer account. If you decide you do not want an app ID, you can disable it using the Administration Console to reclaim one of your 10 available apps. Disabling an app makes the app inaccessible by the public, and disables portions of the Console for the app. Disabling an app does not free the application ID for someone else to register. You can request that a disabled app be deleted permanently. To disable or request deletion of an app, go to “Application Settings” in the Adminis- tration Console, and click the Disable Application... button. When you request 56 | Chapter 2: Creating an Application
- 78. deletion, all developersof the app are notified by email, and if nobody cancels the request, the app is deleted after 24 hours. The Application ID and Title When you click the “Create an Application” button, the Console prompts for an ap- plication identifier. The application ID must be unique across all App Engine applica- tions, just like an account username. The application ID identifies your application when you interact with App Engine using the developer tools. The tools get the application ID from the application configuration file. For Python applications, you specify the app ID in the app.yaml file, on the application: line. For Java applications, you enter it in the <application> element of the appengine-web.xml file. In the example earlier in this chapter, we chose the application ID “clock” arbitrarily. If you’d like to try uploading this application to App Engine, remember to edit the appropriate configuration file after you register the application to change the application ID to the one you chose. The application ID is part of the domain name you can use to test the application running on App Engine. Every application gets a free domain name that looks like this: app-id.appspot.com The application ID is also part of the email and XMPP addresses the app can use to receive incoming messages. See Chapter 11. Because the application ID is used in the domain name, an ID can contain only low- ercase letters, numbers, or hyphens, and must be shorter than 32 characters. Addi- tionally, Google reserves every Gmail username as an application ID that only the corresponding Gmail user can register. As with usernames on most popular websites, a user-friendly application ID may be hard to come by. When you register a new application, the Console also prompts for an “application title.” This title is used to represent your application throughout the Console and the rest of the system. In particular, it is displayed to a user when the application directs the user to sign in with a Google account. Make sure the title is as you would want your users to see it. Once you have registered an application, its ID cannot be changed, though you can delete the application and create a new one. You can change the title for an app at any time from the Administration Console. Registering the Application | 57
- 79. Setting Up aDomain Name If you are developing a professional or commercial application, you probably want to use your own domain name instead of the appspot.com domain as the official location of your application. You can set up a custom domain name for your App Engine app using Google’s “software as a service” service, Google Apps. Google Apps provides hosted applications for your business or organization, including email (with Gmail and POP/IMAP interfaces), calendaring (Google Calendar), chat (Google Talk), hosted word processing, spreadsheets and presentations (Google Docs), easy-to-edit websites (Google Sites), video hosting, and so forth. You associate these services with your organization’s domain name by mapping the domain to Google’s servers in its DNS record, either by letting Google manage the DNS for the domain or by pointing subdomains to Google in your own DNS configuration. Your organiza- tion’s members access the hosted services using your domain name. With App Engine, you can add your own applications to subdomains of your domain. Even if you do not intend to use the other Google Apps services, you can use Google Apps to associate your own domain with your App Engine application. The website for Google Apps indicates that Standard Edition accounts are “ad-supported.” This refers to ads that appear on Google products such as Gmail. It does not refer to App Engine: Google does not place ads on the pages of App Engine applications, even those using free ac- counts. Of course, you can put ads on your own sites, but that’s your choice—and your ad revenue. If you have not set up Google Apps for your domain already, you can do so during the application ID registration process. You can also set up Google Apps from the Admin- istration Console after you have registered the app ID. If you haven’t yet purchased a domain name, you can do so while setting up Google Apps, and you can host the domain on Google’s name servers for free. To use a domain you purchased previously, follow the instructions on the website to point the domain to Google’s servers. Once you have set up Google Apps for a domain, you can access the Google Apps dashboard at a URL similar to the following: http://www.google.com/a/example.com To add an App Engine application as a service, click the “Add more services” link, then find Google App Engine in the list. Enter the application ID for your app, then click “Add it now.” On the following settings screen, you can configure a subdomain of your domain name for the application. All web traffic to this subdomain will go to the application. 58 | Chapter 2: Creating an Application
- 80. Google Apps doesnot support routing web traffic for the top-level do- main (such as http://example.com/) directly to an App Engine app. If you bought the domain name through Google, an HTTP request to the top-level domain will redirect to http://www.example.com, and you can assign the “www” subdomain to your App Engine app. If Google does not maintain the DNS record for your domain, you will need to set up the redirect yourself using a web server associated with the top-level domain. By default, the subdomain “www” is assigned to Google Sites, even if you do not have the Sites app activated. To release this subdomain for use with App Engine, first enable the Sites service, then edit the settings for Sites and remove the “www” subdomain. Google Apps and Authentication Google Apps allows your organization’s members (employees, contractors, volunteers) to create user accounts with email addresses that use your domain name (such as juliet@example.com). Members can sign in with these accounts to access services that are private to your organization, such as email or word processing. Using Apps ac- counts, you can restrict access to certain documents and services to members of the organization, like a hosted intranet that members can access from anywhere. You can also limit access to your App Engine applications to just those users with accounts on the domain. This lets you use App Engine for internal applications such as project management or sales reporting. When an App Engine application is restricted to an organization’s domain, only members of the organization can sign in to the ap- plication’s Google Accounts prompt. Other Google accounts are denied access. The authentication restriction must be set when the application is registered, in the “Authentication Options” section of the registration form. The default setting allows any user with a Google account to sign in to the application, leaving it up to the appli- cation to decide how to respond to each user. When the app is restricted to a Google Apps domain, only users with Google accounts on the domain can sign in. After the application ID has been registered, the authentication options cannot be changed. If you want different authentication options for your application, you must register a new application ID. The restriction applies only to the application’s use of Google Accounts. If the appli- cation has any URLs that can be accessed without signing in to Google Accounts (such as a welcome page), those URLs will still be accessible by everyone. One of the simplest ways to restrict access to a URL is with application configuration. For example, a Py- thon application can require sign-in for all URLs with the following in the app.yaml file: handlers: - url: /.* Registering the Application | 59
- 81. script: main.py login: required A Java app can do something similar in the application’s deployment descriptor (web.xml). See Chapter 3. The sign-in restriction applies even when the user accesses the app using the appspot.com domain. The user does not need to be accessing the app with the Apps domain for the authentication restriction to be enforced. If you or other members of your organization want to use Google Apps accounts as developer accounts, you must access the Administration Console using a special URL. For example, if your Apps domain is example.com, you would use the following URL to access the Administration Console: https://appengine.google.com/a/example.com You sign in to the domain’s Console with your Apps account (for instance, juliet@example.com). If you create an app using a non-Apps account, then restrict its authentication to the domain, you will still be able to access the Administration Console using the non-Apps account. However, you will not be able to sign in to the app with that account, including when accessing URLs restricted to administrators. Uploading the Application In a traditional web application environment, releasing an application to the world can be a laborious process. Getting the latest software and configuration to multiple web servers and backend services in the right order and at the right time to minimize down- time and prevent breakage is often difficult and delicate. With App Engine, deployment is as simple as uploading the files with a single click or command. You can upload and test multiple versions of your application, and set any uploaded version to be the current public version. For Python apps, you can upload an app from the Launcher, or from a command prompt. From the Launcher, select the app to deploy, then click the “Deploy” button. From a command prompt, run the appcfg.py command as follows, substituting the path to your application directory for clock: appcfg.py update clock As with dev_appserver.py, clock is just the path to the directory. If the current working directory is the clock/ directory, you can use the relative path, a dot (.). For Java apps, you can upload from Eclipse using the Google plug-in, or from a com- mand prompt. In Eclipse, click the “Deploy to App Engine” button (the little App Engine logo) in the Eclipse toolbar. Or from a command prompt, run the appcfg (or 60 | Chapter 2: Creating an Application
- 82. appcfg.sh) command fromthe SDK’s bin/ directory as follows, using the path to your application’s WAR directory for war: appcfg update war When prompted by these tools, enter your developer account’s email address and password. The tools remember your credentials for subsequent runs so you don’t have to enter them every time. The upload process determines the application ID and version number from the app configuration file—app.yaml for Python apps, appengine-web.xml for Java apps—then uploads and installs the files and configuration as the given version of the app. After you upload an application for the first time, you can access the application immediately using either the .appspot.com subdomain or the custom Google Apps domain you set up earlier. For example, if the application ID is clock, you can access the application with the following URL: http://clock.appspot.com/ There is no way to download an application’s files from App Engine after they have been uploaded. Make sure you are retaining copies of your application files, such as with a revision control system and regular backups. Introducing the Administration Console You manage your live application from your browser using the App Engine Adminis- tration Console. You saw the Console when you registered the application, but as a reminder, you can access the Console at the following URL: https://appengine.google.com/ If your app uses a Google Apps domain name and you are using an Apps account on the domain as your developer account, you must use the Apps address of the Admin- istration Console: https://appengine.google.com/a/example.com Select your application (click its ID) to go to the Console for the app. The first screen you see is the “dashboard,” shown in Figure 2-11. The dashboard summarizes the current and past status of your application, including traffic and load, resource usage, and error rates. You can view charts for the request rate, the amount of time spent on each request, error rates, bandwidth and CPU usage, and whether your application is hitting its resource limits. If you’ve already tested your new application, you should see a spike in the requests- per-second chart. The scale of the chart goes up to the highest point in the chart, so the Introducing the Administration Console | 61
- 83. Figure 2-11. TheAdministration Console dashboard for a new app spike reaches the top of the graph even though you have only accessed the application a few times. The Administration Console is your home base for managing your live application. From here, you can examine how the app is using resources, browse the application’s request and message logs, and query the datastore and check the status of its indexes. You can also manage multiple versions of your app, so you can test a newly uploaded version before making it the live “default” version. You can invite other people to be developers of the app, allowing them to access the Administration Console and upload new files. And when you’re ready to take on large amounts of traffic, you can establish a billing account, set a daily budget, and monitor expenses. Take a moment to browse the Console, especially the Dashboard, Quota Details, and Logs sections. Throughout this book, we will discuss how an application consumes system resources, and how you can optimize an app for speed and cost effectiveness. The Administration Console is your main resource for tracking resource consumption and diagnosing problems. 62 | Chapter 2: Creating an Application
- 84. CHAPTER 3 Handling Web Requests A web application is an application that responds to requests over the Web. Ideally, a web application is an application that responds to web requests quickly, doing the smallest amount of work required to return a response. Most web apps serve users interacting with the application in real time, and a fast response means less time the user is waiting for an action to be completed or information to be displayed. With user- facing web apps, milliseconds matter. A less obvious advantage to an app that responds quickly is that it’s easier to scale. The less work the app does in response to a request, the more efficiently those requests can be distributed across multiple servers. It’s like scheduling meetings on a busy day: the shorter the meeting, the more likely you’ll be able to fit it in. Apps with faster responses are also more tolerant of system faults. An app receiving 100 queries per second of traffic will have fewer simultaneous requests in progress at a given moment in time if each request takes 10 milliseconds than if each request takes 100 milliseconds. If a machine goes down and a portion of the requests in progress must be canceled, fewer users will be affected, and more subsequent new requests will be routed to other machines. App Engine is designed for web applications that respond to requests quickly. An app that can respond within tens of milliseconds is doing pretty well. Occasionally, an app must take hundreds of milliseconds, such as to save data to the datastore or contact a remote server. If an app routinely takes a long time to respond to requests, App Engine triages the slow requests to make room in the schedule for faster ones. In this chapter, we’ll take a look at App Engine’s request handling architecture, and follow the path of a web request through the system. We’ll discuss how to configure the system to handle different kinds of requests, including requests for static content, requests for the application to perform work, and secure connections (HTTP over SSL, also known as HTTPS). Finally, we’ll take a close look at the application runtime en- vironments for Python and Java, how App Engine invokes an application to respond to requests, and how to take advantage of the environment to speed up request handling. 63
- 85. The App EngineArchitecture The architecture of App Engine—and therefore an App Engine application—can be summarized as shown in Figure 3-1. Figure 3-1. The App Engine request handling architecture The first stop for an incoming request is the App Engine frontend. A load balancer, a dedicated system for distributing requests optimally across multiple machines, routes the request to one of many frontend servers. The frontend determines the app for which the request is intended from the request’s domain name, either the Google Apps domain and subdomain or the appspot.com subdomain. It then consults the app’s configuration to determine the next step. The app’s configuration describes how the frontends should treat requests based on their URL paths. A URL path may map to a static file that should be served to the client directly, such as an image or a file of JavaScript code. Or, a URL path may map to a request handler, application code that is invoked to determine the response for the request. You upload this configuration data along with the rest of your application. We’ll look at how to configure URL paths and static files in the next section. If the URL path for a request does not match anything in the app’s configuration, the frontends return an HTTP 404 “Not Found” error response to the client. The frontends return a generic error response. If you want clients to receive a custom response when accessing your app (such as an HTTP 404 error code with a friendly HTML message), you can map a URL pattern that matches all URLs to a request handler that returns the custom response. If the URL path of the request matches the path of one of the app’s static files, the frontend routes the request to the static file servers. These servers are dedicated to the task of serving static files, with network topology and caching behavior optimized for 64 | Chapter 3: Handling Web Requests
- 86. fast delivery ofresources that do not change often. You tell App Engine about your app’s static files in the app’s configuration. When you upload the app, these files are pushed to the static file servers. If the URL path of the request matches a pattern mapped to one of the application’s request handlers, the frontend sends the request to the app servers. The app server pool starts up an instance of the application on a server, or reuses an existing instance if there is one already running from a previous request. The server invokes the app by calling the request handler that corresponds with the URL path of the request, accord- ing to the app configuration. You can configure the frontend to authenticate the user with Google Accounts. The frontend can restrict access to URL paths with several levels of authorization: all users, users who have signed in, and users who are application administrators. With a Google Apps domain, you can also set your application to allow only users on the domain to access URLs. The frontend checks whether the user is signed in, and redirects the user to the Google Accounts sign-in screen if needed. We’ll look at authentication and au- thorization later in this chapter. The app servers use one of several strategies for distributing requests and starting up instances, depending on the app’s traffic and resource usage patterns. As of this writing, the specifics of these strategies are still being developed and tuned, but they are all intended to work best with request handlers that return quickly. The goal is to maximize the throughput of app instances, so as many instances are running as needed to handle the current traffic levels, but not so many that instances are sitting around doing noth- ing. The app servers are also designed so that one request handler cannot interfere with the behavior or performance of another handler running on the same server. When an app server receives a request for your application, the server checks the URL configuration to determine which of the application’s request handlers should process the request. The server invokes the request handler and awaits its response. The server manages the local resources available to the app, including CPU cycles, memory, and execution time, and ensures that apps do not consume system resources in a way that interferes with other apps. Your application code executes in a runtime environment, an abstraction above the server hardware and operating system that provides access to system resources and other services. The runtime environment is a “sandbox,” a walled arena that lets the application use only the features of the server that can scale without interfering with other apps. For instance, the sandbox prevents the application from writing to the server’s filesystem, or from making arbitrary network connections to other hosts. Applications can access various services to perform tasks outside of the runtime envi- ronment. For instance, the URL Fetch service allows an app to make HTTP requests to remote machines, using Google’s infrastructure for fetching web pages. These services are the same scalable services used by Google’s own applications, such as Gmail, Google Reader, and Picasa. They provide a scalable alternative to performing The App Engine Architecture | 65
- 87. similar tasks directlyon the app server. All app servers use the same services, so data saved to the datastore by one request handler can be accessed by another. The request handler prepares the response, then returns it and terminates. The app server does not send any data to the client until the request handler has terminated, so it cannot stream data or keep a connection open for a long time. When the handler terminates, the app server returns the response, and the request is complete. The frontend takes the opportunity to tailor the response to the client. Most notably, the frontend will compress the response data using the “gzip” format if the client gives some indication that it supports compressed responses. This applies to both app re- sponses and static file responses, and is done automatically. The frontend uses several techniques to determine when it is appropriate to compress responses, based on web standards and known browser behaviors. (If you are using a custom client that does not support compressed content, simply omit the “Accept-Encoding” request header to disable the automatic gzip behavior.) The frontends, app servers, and static file servers are governed by an “app master.” Among other things, the app master is responsible for deploying new versions of ap- plication software and configuration, and updating the “default” version served on an app’s user-facing domain. Updates to an app propagate quickly, but are not atomic in the sense that only code from one version of an app is running at any one time. If you switch the default version to new software, all requests that started before the switch are allowed to complete using their version of the software. (An app that makes an HTTP request to itself might find itself in a pickle, but you can manage that situation in your own code, if you really need to.) Configuring the Frontend You control how the frontend routes requests for your application using configuration files. These files reside alongside your application’s code and static files in your appli- cation directory. When you upload your application, all of these files are uploaded together as a single logical unit. Let’s take a look at how to configure the frontend for an application. First, we’ll look at the overall layout and syntax for the configuration files for a Python app and for a Java app. Then we’ll discuss each frontend feature, with examples for each runtime environment. Configuring a Python App A Python application consists of files, including Python code for request handlers and libraries, static files, and configuration files. On your computer, these files reside in the application root directory. Static files and application code may reside in the root 66 | Chapter 3: Handling Web Requests
- 88. directory or insubdirectories. Configuration files always reside in fixed locations in the root directory. You configure the frontend for a Python application using a file named app.yaml in the application root directory. This file is in a format called YAML, a concise human-readable data format with support for nested structures like sequences and mappings. Example 3-1 shows an example of a simple app.yaml file. We’ll discuss these features in the following sections. For now, notice a few things about the structure of the file: • The file is a mapping of values to names. For instance, the value python is associated with the name runtime. • Values can be scalars (python, 1), sequences of other values, or mappings of values to names. The value of handlers is a sequence of two values, each of which is a mapping containing two name-value pairs. • Order is significant in sequences, but not mappings. • YAML uses indentation to indicate scope. • YAML supports all characters in the Unicode character set. The encoding is as- sumed to be UTF-8 unless the file uses a byte order mark signifying UTF-16. • A YAML file can contain comments. All characters on a line after a # character are ignored, unless the # is in a quoted string value. Example 3-1. An example of an app.yaml configuration file application: ae-book version: 1 runtime: python api_version: 1 handlers: - url: /css static_dir: css - url: /.* script: main.py Runtime versions Among other things, this configuration file declares that this application (or, specifi- cally, this version of this application) uses the Python runtime environment. It also declares which version of the Python runtime environment to use. As of this writing, there is only one version of the Python runtime environment. If Google ever makes changes to the runtime environment that may be incompatible with existing applica- tions, the changes will be released using a new version number. Your app will continue to use the version of the runtime environment specified in your configuration file, giving you a chance to test your app with the new version before uploading the new configuration. Configuring the Frontend | 67
- 89. You specify thename and version of the runtime environment in app.yaml using the runtime and api_version elements, like so: runtime: python api_version: 1 Configuring a Java App A Java application consists of files bundled in a standard format called WAR (short for “Web application archive”). The WAR standard specifies the layout of a directory structure for a Java web application, including the locations of several standard con- figuration files, compiled Java classes, JAR files, static files, and other auxiliary files. Some tools that manipulate WARs support compressing the directory structure into a single file similar to a JAR. App Engine’s tools generally expect the WAR to be a di- rectory on your computer’s filesystem. Java servlet applications use a file called a “deployment descriptor” to specify how the server invokes the application. This file uses an XML format, and is part of the servlet standard specification. In a WAR, the deployment descriptor is a file named web.xml that resides in a directory named WEB-INF/, which itself is in the WAR’s root directory. Example 3-2 shows a very simple deployment descriptor. Example 3-2. An example of a web.xml deployment descriptor file <?xml version="1.0" encoding="utf-8"?> <!DOCTYPE web-app PUBLIC "-//Sun Microsystems, Inc.//DTD Web Application 2.5//EN" "http://java.sun.com/dtd/web-app_2_5.dtd"> <web-app xmlns="http://java.sun.com/xml/ns/javaee" version="2.5"> <servlet> <servlet-name>ae-book</servlet-name> <servlet-class>aebook.MainServlet</servlet-class> </servlet> <servlet-mapping> <servlet-name>ae-book</servlet-name> <url-pattern>/*</url-pattern> </servlet-mapping> </web-app> The deployment descriptor tells the App Engine frontend most of what it needs to know, but not all. For the rest, App Engine uses a file named appengine-web.xml, also in the WEB-INF/ directory and also using XML syntax. If your code editor supports XML validation, you can find the schema definition for this file in the App Engine Java SDK. Example 3-3 shows a brief example. Example 3-3. An example of an appengine-web.xml configuration file <?xml version="1.0" encoding="utf-8"?> <appengine-web-app xmlns="http://appengine.google.com/ns/1.0"> <application>ae-book</application> 68 | Chapter 3: Handling Web Requests
- 90. <version>1</version> </appengine-web-app> When Google releasesmajor new features for the Java API, the release includes a new version of the SDK with an updated appengine-api-... .jar file. App Engine knows which version of the API the app is expecting by examining the API JAR included in the app’s WAR. The server may replace the JAR with a different but compatible implementation when the app is run. Domain Names Every app gets a free domain name on appspot.com, based on the application ID: app-id.appspot.com Requests for URLs that use your domain name are routed to your app by the frontend. http://app-id.appspot.com/url/path... You can register your own domain name (such as example.com) and set it up with Google Apps to point to your app. You assign a subdomain of your top-level domain to the app. For instance, if your registered domain is example.com and you assign the www subdomain, the domain name for the app is: www.example.com Google Apps does not support routing requests for the top-level domain without a subdomain. If you want users to see something when they visit http://example.com/, you must use your own domain name service (DNS) and web server to handle traffic to that domain name, and point subdomains to Google Apps in the DNS record. If you use the Google Apps DNS service for the domain, Google Apps will automatically redirect web requests for the bare domain to the www subdomain. The appspot.com domain has a couple of useful features. One such feature is the ability to accept an additional domain name part: anything.app-id.appspot.com Requests for domain names of this form, where anything is any valid single domain name part (that cannot contain a dot, .), are routed to the application. This is useful for accepting different kinds of traffic on different domain names, such as for allowing your users to serve content from their own subdomains. Only appspot.com domains support the additional part. Google Apps domains do not. You can determine which domain name was used for the request in your application code by checking the Host header on the request. Here’s how you check this header using Python and webapp: Configuring the Frontend | 69
- 91. class MainHandler(webapp.RequestHandler): def get(self): host = self.request.headers['Host'] self.response.out.write('Host: %s' % host) App IDs and Versions Every App Engine application has an application ID that uniquely distinguishes the app from all other applications. As described in Chapter 2, you can register an ID for a new application using the Administration Console. Once you have an ID, you add it to the app’s configuration so the developer tools know that the files in the app root directory belong to the app with that ID. The app’s configuration also includes a version identifier. Like the app ID, the version identifier is associated with the app’s files when the app is uploaded. App Engine retains one set of files and frontend configuration for each distinct version identifier used dur- ing an upload. If you do not change the app version in the configuration when you upload, the existing files for that version of the app are replaced. Each distinct version of the app is accessible at its own domain name, of the following form: version-id.latest.app-id.appspot.com When you have multiple versions of an app uploaded to App Engine, you can use the Administration Console to select which version is the one you want the public to access. The Console calls this the “default” version. When a user visits your Google Apps domain (and configured subdomain), or the appspot.com domain without the version ID, he sees the default version. The appspot.com domain containing the version ID supports an additional domain part, just like the default appspot.com domain: anything.version-id.latest.app-id.appspot.com Unless you explicitly prevent it, anyone who knows your application ID and version identifiers can access any uploaded version of your appli- cation using the appspot.com URLs. You can restrict access to nondefault versions of the application using code that checks the domain of the request and only allows authorized users to access the versioned do- mains. You can’t restrict access to static files this way. Another way to restrict access to nondefault versions is to use Google Accounts authorization, described later in this chapter. You can restrict access to app administrators while a version is in development, then replace the configuration to remove the restriction just before making that version the default version. 70 | Chapter 3: Handling Web Requests
- 92. All versions ofan app access the same datastore, memcache, and other services, and all versions share the same set of resources. Later on, we’ll discuss other configuration files that control these backend services. These files are separate from the configuration files that control the frontend because they are not specific to each app version. There are several ways to use app versions. For instance, you can have just one version, and always update it in place. Or you can have a “dev” version for testing and a “live” version that is always the public version, and do separate uploads for each. Some de- velopers generate a new app version identifier for each upload based on the version numbers used by a source code revision control system. You can have up to 10 active versions. You can delete previous versions using the Administration Console. Application IDs and version identifiers can contain numbers, lowercase letters, and hyphens. App IDs and versions in Python For a Python app, the application ID and version identifier appear in the app.yaml file. The app ID is specified with the name application. The version ID is specified as version. Here is an example of app.yaml using dev as the version identifier: application: ae-book version: dev This would be accessible using this domain name: http://dev.latest.ae-book.appspot.com App IDs and versions in Java The app ID and version identifier of a Java app appear in the appengine-web.xml file. The app ID is specified with the XML element <application>, and the version identifier is specified with <version>. For example: <?xml version="1.0" encoding="utf-8"?> <appengine-web-app xmlns="http://appengine.google.com/ns/1.0"> <application>ae-book</application> <version>dev</version> </appengine-web-app> As in the Python example, this version of this app would be accessible using this domain name: http://dev.latest.ae-book.appspot.com Configuring the Frontend | 71
- 93. Request Handlers The appconfiguration tells the frontend what to do with each request, routing it to either the application servers or the static file servers. The destination is determined by the URL path of the request. For instance, an app might send all requests whose URL paths start with /images/ to the static file server, and all requests for the site’s home page (the path /) to the app servers. The configuration specifies a list of patterns that match URL paths, with instructions for each pattern. For requests intended for the app servers, the configuration also specifies the request handler responsible for specific URL paths. A request handler is an entry point into the application code. In Python, a request handler is a script of Python code. In Java, a request handler is a servlet class. Each runtime environment has its own interface for invoking the application. Request handlers in Python All URL paths for Python apps are described in the app.yaml file using the handlers element. The value of this element is a sequence of mappings, where each item includes a pattern that matches a set of URL paths and instructions on how to handle requests for those paths. Here is an example with four URL patterns: handlers: - url: /profile/.* script: userprofile.py - url: /css static_dir: css - url: /info/(.*.xml) static_files: /datafiles/1 - url: /.* script: main.py The url element in a handler description is a regular expression that matches URL paths. Every path begins with a forward slash (/), so a pattern can match the beginning of a path by also starting with this character. This URL pattern matches all paths: url: /.* If you are new to regular expressions, here is the briefest of tutorials: the . character matches any single character, and the * character says the previous symbol, in this case any character, can occur zero or more times. There are several other characters with special status in regular expressions. All other characters, like /, match literally. So this pattern matches any URL that begins with a / followed by zero or more of any character. If a special character is preceded by a backslash (), it is treated as a literal character in the pattern. Here is a pattern that matches the exact path /home.html: url: /home.html 72 | Chapter 3: Handling Web Requests
- 94. See the Pythondocumentation for the re module for an excellent introduction to regular expressions. The actual regular expression engine used for URL patterns is not Py- thon’s, but it’s similar. App Engine attempts to match the URL path of a request to each handler pattern in the order the handlers appear in the configuration file. The first pattern that matches determines the handler to use. If you use the catchall pattern /.*, make sure it’s the last one in the list, since a later pattern will never match. To map a URL path pattern to application code, you provide a script element. The value is the path to a Python source file, relative to the application root directory. If the frontend gets a request whose path matches a script handler, it routes the request to an application server to invoke the script and produce the response. In the previous example, the following handler definition routes all URL paths that begin with /profile/ to a script named userprofile.py: - url: /profile/.* script: userprofile.py We’ll take a closer look at how App Engine invokes a script handler later in this chapter. Request handlers in Java A Java web application maps URL patterns to servlets in the deployment descriptor (web.xml). You set up a servlet in two steps: the servlet declaration, and the servlet mapping. The <servlet> element declares a servlet. It includes a <servlet-name>, a name for the purposes of referring to the servlet elsewhere in the file, and the <servlet-class>, the name of the class that implements the servlet. Here’s a simple example: <servlet> <servlet-name>ae-book</servlet-name> <servlet-class>aebook.MainServlet</servlet-class> </servlet> The servlet declaration can also define initialization parameters for the servlet. This is useful if you want to use the same servlet class in multiple servlet declarations, with different parameters for each one. For example: <servlet> <servlet-name>ae-book</servlet-name> <servlet-class>aebook.MainServlet</servlet-class> <init-param> <param-name>colorscheme</param-name> <param-value>monochrome</param-value> </init-param> <init-param> <param-name>background</param-name> <param-value>dark</param-value> Configuring the Frontend | 73
- 95. </init-param> </servlet> To map a servlet to a URL path pattern, you use the <servlet-mapping> element. A mapping includes the <servlet-name> that matches a servlet declaration, and a <url-pattern>. <servlet-mapping> <servlet-name>ae-book</servlet-name> <url-mapping>/home/*</url-mapping> </servlet-mapping> The URL pattern matches the URL path. It can use a * character at the beginning or end of the pattern to represent zero or more of any character. Note that this wildcard can only appear at the beginning or end of the pattern, and you can only use one wildcard per pattern. The order in which URL mappings appear is not significant. The “most specific” matching pattern wins, determined by the number of nonwildcard characters in the pattern. The pattern /* matches all URLs, but will only match if none of the other patterns in the deployment descriptor match the URL. We’ll take a closer look at servlets and the Java runtime environment later in this chapter. JavaServer Pages (JSPs). App Engine includes support for JavaServer Pages (JSPs). JSPs are dynamic web pages defined using a mix of HTML (or other output text) and Java code. JSPs are compiled to Java classes equivalent to servlets that output the static content and evaluate the Java code. You can build large dynamic websites using JSPs for all user-facing servlets. From the developer’s point of view, working with JSPs is a lot like working with files of static HTML content. Like other JSP-capable web servers, App Engine compiles JSP files automatically, so an additional compilation step is usually not necessary. When you upload the app, App Engine compiles the JSPs, and stores the compiled servlet classes. No compilation occurs on the app servers themselves. JSPs reside in the application directory, outside of the WEB-INF/ directory. The file- name of a JSP must end with the characters .jsp. By default, each JSP is mapped au- tomatically to a URL path equivalent to the path to the JSP file from the application root. So if a JSP file’s path is forum/home.jsp, it can be accessed with the URL path /forum/home.jsp. As we’ll see later in this chapter, this is also how URLs for static files work for Java. This lets you store JSPs and static files together in the application directory, and refer to them using intuitive paths in the HTML. You can also set an explicit URL mapping for a JSP in the deployment descriptor by declaring it as a servlet. Instead of a <servlet-class> element, use a <jsp-file> element that contains the path to the file from the application root. 74 | Chapter 3: Handling Web Requests
- 96. <servlet> <servlet-name>forum-home</servlet-name> <jsp-file>/forum/home.jsp</jsp-file> </servlet> <servlet-mapping> <servlet-name>forum-home</servlet-name> <url-pattern>/forum</url-pattern> <servlet-mapping> App Engine includes the JavaServer Pages Standard Tag Library (JSTL), a standard library of extensions for use with JSPs. You do not need to add it to your app or install it—and in fact doing so might conflict with the one included with App Engine. For more information on JSPs and the JSTL, see Head First Servlets and JSP by Brian Basham et al. (O’Reilly) and JavaServer Pages by Hans Bergsten (O’Reilly). Static Files and Resource Files Most web applications have a set of files that are served verbatim to all users, and do not change as the application is used. These can be media assets like images used for site decoration, CSS stylesheets that describe how the site should be drawn to the screen, JavaScript code to be downloaded and executed by a web browser, or HTML for full pages with no dynamic content. To speed up the delivery of these files and improve page rendering time, App Engine uses dedicated servers for static content. Using dedicated servers also means the app servers don’t have to spend resources on requests for static files. Both the deployment process and the frontend must be told which of the application’s files are static files. The deployment process delivers static files to the dedicated servers. The frontend remembers which URL paths refer to static files, so it can route requests for those paths to the appropriate servers. The static file configuration can also include a recommendation for a cache expiration interval. App Engine returns the cache instructions to the client in the HTTP header along with the file. If the client chooses to heed the recommendation, it will retain the file for up to that amount of time, and use its local copy instead of asking for it again. This reduces the amount of bandwidth used, but at the expense of browsers retaining old copies of files that may have changed. To save space and reduce the amount of data involved when setting up new app in- stances, static files are not pushed to the application servers. This means application code cannot access the contents of static files using the filesystem. The files that do get pushed to the application servers are known as “resource files.” These can include app-specific configuration files, web page templates, or other static data that is read by the app but not served directly to clients. Application code can access these files by reading them from the filesystem. The code itself is also accessible this way. Configuring the Frontend | 75
- 97. There are waysto specify that a file is both a resource file and a static file, depending on which runtime environment you are using. Static files in Python We’ve seen how request handlers defined in the app.yaml file can direct requests to scripts that run on the app servers. Handler definitions can also direct requests to the static file servers. There are two ways to specify static file handlers. The easiest is to declare a directory of files as static, and map the entire directory to a URL path. You do this with the static_dir element, as follows: handlers: - url: /images static_dir: myimgs This says that all of the files in the directory myimgs/ are static files, and the URL path for each of these files is /images/ followed by the directory path and filename of the file. If the app has a file at the path myimgs/people/frank.jpg, App Engine pushes this file to the static file servers, and serves it whenever someone requests the URL path /images/people/frank.jpg. Notice that with static_dir handlers, the url pattern does not include a regular ex- pression to match the subpath or filename. The subpath is implied: whatever appears in the URL path after the URL pattern becomes the subpath to the file in the directory. The other way to specify static files is with the static_files element. With static_files, you use a full regular expression for the url. The URL pattern can use regular expression groups to match pieces of the path, then use those matched pieces in the path to the file. The following is equivalent to the static_dir handler above: - url: /images/(.*) static_files: myimgs/1 upload: myimgs/.* The parentheses in the regular expression identify which characters are members of the group. The 1 in the file path is replaced with the contents of the group when looking for the file. You can have multiple groups in a pattern, and refer to each group by number in the file path. Groups are numbered in the order they appear in the pattern from left to right, where 1 is the leftmost pattern, 2 is the next, and so on. When using static_files, you must also specify an upload element. This is a regular expression that matches paths to files in the application directory on your computer. App Engine needs this pattern to know which files to upload as static files, since it cannot determine this from the static_files pattern alone (as it can with static_dir). While developing a Python app, you keep the app’s static files in the application di- rectory along with the code and configuration files. When you upload the app, App Engine determines which files are static files from the handler definitions in 76 | Chapter 3: Handling Web Requests
- 98. app.yaml. Files mentionedin static file handler definitions are pushed to the static file servers. All other files in the application directory are considered resource files, and are pushed to the application servers. As such, static files are not accessible to the appli- cation code via the filesystem. The Python SDK treats every file as either a resource file or a static file. If you have a file that you want treated as both a resource file (available to the app via the filesystem) and a static file (served verbatim from the static file servers), you can create a symbolic link in the project directory to make the file appear twice to the deployment tool under two separate names. The file will be uploaded twice, and count as two files toward the file count limit. MIME types. When the data of an HTTP response is of a particular type, such as a JPEG image, and the web server knows the type of the data, the server can tell the client the type of the data using an HTTP header in the response. The type can be any from a long list of standard type names, known as MIME types. If the server doesn’t say what the type of the data is, the client has to guess, and may guess incorrectly. By default, for static files, App Engine makes its own guess of the file type based on the last few characters of the filename (such as .jpeg). If the filename does not end in one of several known extensions, App Engine serves the file as the MIME type application/ octet-stream, a generic type most web browsers treat as generic binary data. If this is not sufficient, you can specify the MIME type of a set of static files using the mime_type element in the static file handler configuration. For example: - url: docs/(.*).ps static_files: psoutput/1.dat upload: psoutput/.*.dat mime_type: application/postscript This says that the application has a set of datafiles in a directory named psoutput/ whose filenames end in .dat, and these should be served using URL paths that consist of docs/, followed by the filename with the .dat replaced with .ps. When App Engine serves one of these files, it declares that the file is a PostScript document. You can also specify mime_type with a static_dir handler. All files in the directory are served with the declared type. Cache expiration. It’s common for a static file to be used on multiple web pages of a site. Since static files seldom change, it would be wasteful for a web browser to download the file every time the user visits a page. Instead, browsers can retain static files in a cache on the user’s hard drive, and reuse the files when they are needed. To do this, the browser needs to know how long it can safely retain the file. The server can suggest a maximum cache expiration in the HTTP response. You can configure the cache expiration period App Engine suggests to the client. Configuring the Frontend | 77
- 99. To set adefault cache expiration period for all static files for an app, you specify a default_expiration value. This value applies to all static file handlers, and belongs at the top level of the app.yaml file, like so: application: ae-book version: 1 runtime: python api_version: 1 default_expiration: "5d 12h" handlers: # ... The value is string that specifies a number of days, hours, minutes, and seconds. As shown here, each number is followed by a unit (d, h, m, or s), and values are separated by spaces. You can also specify an expiration value for static_dir and static_files handlers individually, using an expiration element in the handler definition. This value overrides the default_expiration value, if any. For example: handlers: - url: /docs/latest static_dir: /docs expiration: "12h" If the configuration does not suggest a cache expiration period for a set of static files, App Engine does not give an expiration period when serving the files. Browsers will use their own caching behavior in this case, and may not cache the files at all. Sometimes you want a static file to be cached in the browser as long as possible, but then replaced immediately when the static file changes. A common technique is to add a version number for the file to the URL, then use a new version number from the app’s HTML when the file changes. The browser sees a new URL, assumes it is a new re- source, and fetches the new version. You can put the version number of the resource in a fake URL parameter, such as /js/code.js?v=19, which gets ignored by the static file server. Alternatively, in Py- thon, you can use regular expression matching to match all versions of the URL and route them to the same file in the static file server, like so: - handlers: url: /js/(.*)/code.js static_files: js/code.js expiration: "90d" This handler serves the static file js/code.js for all URLs such as /js/v19/code.js, using a cache expiration of 90 days. 78 | Chapter 3: Handling Web Requests
- 100. If you’d likebrowsers to reload a static file resource automatically every time you launch a new major version of the app, you can use the mul- tiversion URL handler just discussed, then use the CURRENT_VERSION_ID environment variable as the “version” in the static file URLs: self.response.out('<script src="/js/' + os.environ['CURRENT_VERSION_ID'] + '/code.js" />') Static files in Java As we saw earlier, the WAR directory structure for a Java web application keeps all application code, JARs, and configuration in a subdirectory named WEB-INF/. Typi- cally, files outside of WEB-INF/ represent resources that the user can access directly, including static files and JSPs. The URL paths to these resources are equivalent to the paths to these files within the WAR. Say an app’s WAR has the following files: main.jsp forum/home.jsp images/logo.png images/cancelbutton.png images/okbutton.png terms.html WEB-INF/classes/com/example/Forum.class WEB-INF/classes/com/example/MainServlet.class WEB-INF/classes/com/example/Message.class WEB-INF/classes/com/example/UserPrefs.class WEB-INF/lib/appengine-api.jar This app has four static files: three PNG images and an HTML file named terms.html. When the app is uploaded, these four files are pushed to the static file servers. The frontends know to route requests for URL paths equivalent to these file paths (such as /images/logo.png) to the static file servers. The two .jsp files are assumed to be JSPs, and are compiled to servlet classes and mapped to the URL paths equivalent to their file paths. Since these are application code, they are handled by the application servers. The JSP source files themselves are not pushed to the static file servers. By default, all files in the WAR are pushed to the application servers, and are accessible by the application code via the filesystem. This includes the files that are identified as static files and pushed to the static file servers. In other words, all files are considered resource files, and all files except for JSPs and the WEB-INF/ directory are considered static files. Configuring the Frontend | 79
- 101. You can specifywhich files are considered resource files and which are considered static files using the appengine-web.xml file, with the <resource-files> and <static-files> elements, respectively. These elements can contain an <include> element and an <exclude> element that modify the default behavior of including all files. For example: <resource-files> <exclude path="/images/**" /> </resource-files> This excludes the contents of the images/ directory and all subdirectories from the set of resource files. This reduces the amount of data that is pushed to the application servers when starting up a new application instance, at the expense of not being able to access those files from within the application (probably fine for site images). The ** pattern matches any number of characters in file and directory names, including subdirectories. Another example: <static-files> <exclude path="/**.xml" /> <include path="/sitemap.xml" /> </static-files> This excludes all files with names ending in .xml from the set of static files, except for sitemap.xml. Perhaps the XML files are intended for the application’s eyes only, but we want to make sure search engines can see the site map. Files in the WEB-INF/ directory are always considered resource files. They cannot be included as static files or excluded from the set of resource files. Browsers rely on the web server to tell them the type of the file being served. The static file server determines the MIME content type of a file from the extension on the file- name. For instance, a file whose name ends in .jpeg is served with a MIME type of image/jpeg. The server has a built-in set of mappings from filename extensions to MIME types. You can specify additional mappings using <mime-mapping> elements in the de- ployment descriptor (web.xml). See a web.xml reference or the App Engine documen- tation for more information. Browsers also need to know if a file is safe to cache, and for how long. The static file server can suggest a cache expiration duration when it serves a file (though a browser is not obligated to honor it). You can specify that a set of static files should be cached for a particular duration by including an expiration attribute on the <include> element in appengine-web.xml: <static-files> <include path="images/**" expiration="30d" /> </static-files> The value of expiration is a duration specified as numbers and units, where d is days, h is hours, m is minutes, and s is seconds. You can add values of multiple units by specifying them separated with spaces: 3d 12h. 80 | Chapter 3: Handling Web Requests
- 102. Secure Connections When aclient requests and retrieves a web page over an HTTP connection, every aspect of the interaction is transmitted over the network in its final intended form, including the URL path, request parameters, uploaded data, and the complete content of the server’s response. For web pages, this usually means human-readable text is flying across the wire, or through the air if the user is using a wireless connection. Anyone else privy to the network traffic can capture and analyze this data, and possibly glean sensitive information about the user and the service. Websites that deal in sensitive information, such as banks and online retailers, can use a secure alternative for web traffic. With servers that support it, the client can make an HTTPS connection (HTTP over the Secure Socket Layer, or SSL). All data sent in either direction over the connection is encrypted by the sender and decrypted by the recipient, so only the participants can understand what is being transmitted even if the encrypted messages are intercepted. Web browsers usually have an indicator that tells the user when a connection is secure. App Engine has limited support for secure connections. As of this writing, secure con- nections are only supported for the appspot.com domain, and not custom Google Apps domains. This limitation has to do with how the server proves to the client that he is who he says he is, and not an impostor. Site identification is based on the domain name of the site and the security certificate returned by the App Engine frontend. Because the frontend does not know the domain name this early in the connection process— the browser connects directly to an IP address shared by multiple apps, and only sends the domain name as part of the HTTP header after the connection is made—it cannot verify that the certificate matches the domain. A newer version of the SSL protocol will allow App Engine to identify the site via other means (using Server Name Indication, or SNI), but not all browsers support the newer protocol yet. When browser support improves (or the use of older browsers dimin- ishes), App Engine will be able to support HTTPS for custom domains. Since an App Engine app can use both a custom domain and an appspot.com domain simultaneously, secure connections are still useful in cases where the domain name of the connection doesn’t matter. And you can always redirect users to the appspot.com domain to use the secure features of the app. It’s not the ideal user experience—most users rightly get suspicious when the domain name for a secure connection is not the one they expect—but it may be appropriate for some apps. With App Engine, you can enable secure connections for each URL path in the frontend configuration. Secure connections take a bit more time to decrypt requests and encrypt responses. The application code itself doesn’t need to know the difference between a secure connection and a standard connection: it just consumes the decrypted request and provides a response that is encrypted by App Engine. Configuring the Frontend | 81
- 103. All URL pathscan be configured to use secure connections, including those mapped to application code and those mapped to static files. The frontend takes care of the secure connection on behalf of the app servers and static file servers. App Engine only supports secure connections over TCP port 443, the standard port used by browsers for https:// URLs. Similarly, App Engine only supports standard connections over port 80. The App Engine frontend returns an error for URLs that specify a port other than the standard port for the given connection method. The development server does not support secure connections, and ignores the security settings in the configuration. You can test these URLs during development using the nonsecure equivalent URLs. Because HTTPS uses the domain name to validate the secure connection, requests to versioned appspot.com URLs, such as https://3.latest.ae-book.appspot.com/, will display a security warning in the browser saying that the domain does not match the security certificate. You can accept the warning to bypass this check against imposters (which guards against “man-in-the-middle” attacks) and continue loading the page. Secure connections in Python To configure a URL handler in a Python application to accept secure connections, add a secure element to the handler’s properties in the app.yaml file: handler: - url: /profile/.* script: userprofile.py secure: always The value of the secure element can be either always, never, or optional. If you don’t specify a secure element for a URL path, the default is optional. always says that requests to this URL path should always use a secure connection. If a user attempts to request the URL path over a nonsecure connection, the App Engine frontend issues an HTTP redirect code telling it to try again using a secure HTTP con- nection. Browsers follow this redirect automatically. never says that requests to this URL path should never use a secure connection, and requests for an HTTPS URL should be redirected to the HTTP equivalent. Note that browsers often display a warning when a user follows a link from a secure page to a nonsecure page. optional allows either connection method for the URL path, without redirects. The app can use the HTTPS environment variable to determine which method was used for the request, and produce a custom response. Secure connections in Java For a Java application, anyone can access any URL path with either HTTP or HTTPS. You can use the deployment descriptor to require secure connections for certain URL 82 | Chapter 3: Handling Web Requests
- 104. paths. In theweb.xml file, you declare a security constraint for a URL path or set of URL paths as follows: <security-constraint> <web-resource-collection> <web-resource-name>home</web-resource-name> <url-pattern>/home/*</url-pattern> </web-resource-collection> <user-data-constraint> <transport-guarantee>CONFIDENTIAL</transport-guarantee> </user-data-constraint> </security-constraint> A security constraint, indicated by the <security-constraint> element, describes the minimum security requirements a request must meet to access a resource. You identify the resource using a URL pattern in a <web-resource-collection> element containing a <url-pattern> element. (According to the spec, <web-resource-collection> must have a <web-resource-name>, though this name is not used for anything.) As with URL pat- terns in servlet mappings, the URL pattern can be a single URL path, or a partial URL path with a * wildcard at the beginning or at the end. You specify a security constraint requiring a secure connection with a <user-data-constraint> element containing a <transport-guarantee> element, itself containing the value CONFIDENTIAL. (The transport guarantee INTEGRAL is also supported as a synonym for CONFIDENTIAL.) App Engine does not support other transport guarantee constraints. If you do not want users accessing your app with HTTPS, you can disable secure con- nections by adding this to your appengine-web.xml configuration file: <ssl-enabled>false</ssl-enabled> You can only enable or disable SSL for the entire application. The deployment descrip- tor standard does not have a concept of accepting secure connections for some URL paths and not others. Authorization with Google Accounts Back in Chapter 2, we discussed how an App Engine application can integrate with Google Accounts to identify and authenticate users. We saw how an app can use library calls to check whether the user making a request is signed in, access the user’s email address, and calculate the sign-in and sign-out URLs of the Google Accounts system. With this API, application code can perform fine-grained access control and customize displays. Another way to do access control is to leave it to the frontend. With just a little con- figuration, you can instruct the frontend to protect access to specific URL handlers such that only signed-in users can request them. If a user who is not signed in requests such a URL, the frontend redirects the user to the Google Accounts sign-in and registration Configuring the Frontend | 83
- 105. screen. Upon successfullysigning in or registering a new account, the user is redirected back to the URL. You can also tell the frontend that only the registered developers of the application can access certain URL handlers. This makes it easy to build administrator-only sections of your website, with no need for code that confirms the user is an administrator. You can manage which accounts have developer status in the Administration Console, un- der Developers. If you revoke an account’s developer status, that user is no longer able to access administrator-only resources, effective immediately. Later on, we will discuss App Engine services that call your application in response to events. For example, the scheduled tasks service (the “cron” service) can be configured to trigger a request to a URL at certain times of the day. Typically, you want to restrict access to these URLs so not just anybody can call them. For the purposes of access control enforced by the frontend, these services act as app administrators, so restricting these URLs to administrators effectively locks out meddling outsiders while allowing the services to call the app. If your app runs on a Google Apps domain for the purposes of serving your organiza- tion’s members, you can set an access policy for the app such that only Google Apps accounts on the domain can sign in to the app. You set this policy when you register the application ID; it cannot be changed after the app ID has been registered. This coarse-grained access control is easy to set up in the frontend configuration. And unlike access control in the application code, frontend authentication can restrict access to static files as well as application request handlers. Authorization in Python For a Python app, you establish frontend access control for a URL handler with the login element in app.yaml, like so: handlers: - url: /myaccount/.* script: account.py login: required The login element has two possible values: required and admin. If login is required, then the user must be signed in to access URLs for this handler. If the user is not signed in, the frontend returns an HTTP redirect code to send the user to the Google Accounts sign-in and registration form. If login is admin, then the user must be signed in and must be a registered developer for the application. If no login is provided, the default policy is to allow anyone to access the resource, whether or not the client represents a signed-in user, and whether or not the app is set to use a members-only access policy. You can use the login element with both script handlers and static file handlers. 84 | Chapter 3: Handling Web Requests
- 106. Authorization in Java ForJava, you establish a frontend access control policy using a security constraint in the deployment descriptor. We introduced security constraints earlier when we dis- cussed secure connections. Authentication constraints are similar: they specify the minimum level of a condition required to access a resource. Here’s what an authentication constraint looks like in web.xml: <security-constraint> <web-resource-collection> <url-pattern>/myaccount/*</url-pattern> </web-resource-collection> <auth-constraint> <role-name>*</role-name> </auth-constraint> </security-constraint> As before, the security constraint identifies a resource with a URL pattern, then specifies the constraint to apply. An <auth-constraint> element contains a <role-name> element that specifies the minimum level of authentication. <role-name> can be one of two values: * (a single asterisk) or admin. If the role name is *, then any user that has signed in can access the resource. If the user is not signed in, the frontend sends an HTTP redirect code with the Google Accounts sign-in and reg- istration page as the destination. If the role name is admin, then only a user who is both signed in and a registered developer for the application can access the resource. If a URL does not have an authentication constraint, then anyone can access the URL, whether or not the client represents a signed-in user, and whether or not the app is set to use a members-only access policy. How the App Is Run The frontend routes a request whose URL is mapped to an application handler to one of many application servers. When the app server receives the request, it loads the app’s code, resource files, and configuration into a runtime environment. The server invokes the request handler for the request, gathers the response data, and sends the response data to the user. The runtime environment includes the interpreter that executes the code (either the Python interpreter or the Java virtual machine), the application’s memory, a read-only filesystem with the application’s resource files, and APIs to the App Engine services. The runtime environment also includes the request handler API, which defines how the server and the handler communicate information about the request and the re- sponse, and how the handler emits messages to the App Engine logging system. The details differ between the Python and Java environments, but can involve environment variables and the system input and output streams. How the App Is Run | 85
- 107. The Python RuntimeEnvironment When an app server receives a request intended for a Python application, it compares the URL path of the request to the URL patterns in in the app’s app.yaml file. Each URL pattern is associated with a file of Python code (a “script”), and the first pattern to match the path identifies the script that will handle the request. The server uses the Common Gateway Interface (CGI) standard to communicate with the Python code. CGI is one of the earliest building blocks of the Web, and if you did any web application programming in the 20th century, you may currently be experi- encing flashbacks of a slow and cumbersome application model that was common in the days of Apache servers and mod_cgi, with a fresh Perl interpreter started from scratch for every request. Thankfully, that execution model has nothing to do with the inter- face, and CGI is just as useful for modern, high-performance web apps. The CGI standard specifies how the server communicates the request to the application and how the application returns the response to the server. It uses a combination of environment variables and the application process’s input and output streams to handle this communication. While you could write your own code that implements this in- terface, it’s far more productive to let an application framework take care of this for you. Python developers have many web application frameworks to choose from. Django, Pylons, and web2py are several “full stack” frameworks currently known to work well with App Engine, and App Engine includes Django as part of the runtime environment. You can include other frameworks and libraries with your application simply by adding them to your application directory. As we saw in Chapter 2, App Engine also includes a simple framework of its own, called webapp. By the time an app server receives the request, it has already fired up the Python inter- preter, ready to handle requests. If the server served a request for the application re- cently, it may have the application in memory as well, but if it hasn’t, it loads the handler script for the request. The server evaluates the script with the data for the request, and returns the response the script emits. If the server already has the application in memory from a previous request, it keeps the application memory, including imported modules and global variables. In this sense, it behaves more like FastCGI than CGI. The Python environment does flush environment variables at the end of each request. We’ll discuss app caching later in this chapter. App Engine’s Python interpreter is a modified version of the official Python 2.5 inter- preter, sometimes referred to as “CPython” to distinguish it from other Python inter- preters. The application code must run entirely within the Python interpreter. That is, the code must be purely Python code, and cannot include or depend upon extensions to the Python interpreter. Python modules that include extensions written in C cannot be uploaded with your app or otherwise added to the runtime environment. The “pure 86 | Chapter 3: Handling Web Requests
- 108. Python” requirement canbe problematic for some third-party libraries, so be sure that libraries you want to use operate without extensions. The handler script produces a response by writing to the standard output stream (sys.stdout, or however your framework of choice marshals output). The runtime en- vironment buffers the output of the script while it is running. When the handler exits, the server returns the response to the client. (An app cannot stream content to the client as it is doing other work.) The server runs the handler script as if it were the main program. The __name__ variable is set to '__main__' in the scope of the handler script, so the Python idiom if __name__ == '__main__': ... works as expected. App Engine sets the following environment variables at the beginning of each request (in addition to the CGI variables), which you can access using os.environ: APPLICATION_ID The ID of the application. CURRENT_VERSION_ID The ID of the version of the app serving this request. AUTH_DOMAIN This is set to gmail.com if the user is signed in using a Google Account, or the domain of the app if signed in with a Google Apps account; not set otherwise. SERVER_SOFTWARE The version of the runtime environment; starts with the word Development when running on the development server. The Java Runtime Environment The Java runtime environment behaves like a servlet container. When the app server receives a request, it determines the servlet class to call by comparing the URL path to the servlet mappings in the deployment descriptor. The server uses the standard servlet interface to invoke the servlet, passing it a populated request object and an empty response object. The application’s servlet code populates the response object and exits, and the server returns the response to the client. The Java runtime environment uses the Java 6 virtual machine (JVM). The JVM runs Java bytecode, which is what you get from passing Java code to a Java compiler. It’s also what you get from compilers for other languages that produce Java bytecode, such as Scala, and from interpreters for other languages implemented in Java bytecode, such as JRuby (Ruby), Rhino (JavaScript), Groovy, and even Jython (a Python interpreter implemented in Java). You can use any language that compiles to or has an interpreter for the JVM to write applications for App Engine, as long as the result implements a servlet interface. How the App Is Run | 87
- 109. Having a completeJVM also means you can use many third-party libraries with your application. Some restrictions apply—we’ll look at a few in a moment—but in most cases, using a library is a simple matter of including the JAR or class files in the appli- cation’s WAR. An app can determine whether it is running on App Engine or in the development server from the servlet info string in the servlet context, returned by this.getServletContext().getServletInfo(). This string starts with Google App Engine Development when running in the development server. It also contains the release num- ber of the runtime environment. An app can get its application ID and version ID using the ApiProxy class: import com.google.apphosting.api.ApiProxy; // ... String appId = ApiProxy.getCurrentEnvironment().getAppId(); String versionId = ApiProxy.getCurrentEnvironment().getVersionId(); The Sandbox The runtime environment does not expose the complete operating system to the ap- plication. Some functions, such as the ability to create arbitrary network connections, are restricted. This “sandbox” is necessary to prevent other applications running on the same server from interfering with your application (and vice versa). Instead, an app can perform some of these functions using App Engine’s scalable services, such as the URL Fetch service. The most notable sandbox restrictions include the following: • An app cannot spawn additional processes or threads. All processing for a request must be performed by the request handler’s process. • An app cannot make arbitrary network connections. Some networking features are provided by the App Engine services, such as URL Fetch and Mail. • An app can only read from the filesystem, and can only read its own code and resource files. It cannot create or modify files. Instead of files, an app can use the datastore to save data. • An app cannot see or otherwise know about other applications or processes that may be running on the server. This includes other request handlers from the same application that may be running simultaneously. These restrictions are implemented on multiple levels, both to ensure that the restric- tions are enforced and to make it easier to troubleshoot problems that may be related to the sandbox. 88 | Chapter 3: Handling Web Requests
- 110. The Python sandbox Inthe Python runtime environment, the Python interpreter prevents the app from ac- cessing illegal system resources at a low level. Since a Python app can only consist of Python code, an app must perform all processing within the Python interpreter. For convenience, portions of the Python standard library whose only use is to access restricted system resources have been disabled. If you attempt to import a disabled module or call a disabled function, the interpreter raises an ImportError. The Python development server enforces the standard module import restrictions, so you can test imports on your computer. Some standard library modules have been replaced with alternate versions for speed or compatibility. Other modules have custom implementations, such as zipimport. You can find out more about modules that have been replaced or added from the official documentation. The Java sandbox In the Java runtime environment, sandbox restrictions are enforced within the JVM. These restrictions are implemented using a combination of JVM permissions, a Java Runtime Environment (JRE) class whitelist, and alternate implementations for specific functions. This fine-grained approach allows more third-party libraries to work and makes other code easier to port than relying on JVM permissions alone. The Java runtime environment includes a subset of the JRE classes. You can find a complete list of supported JRE classes in the official documentation. Not every feature of every class in this list works with App Engine. For instance, the java.lang.Thread class can be imported successfully, but can only be used to examine the current thread. It cannot be used to create new threads. Reflection is supported for all of the app’s own classes. Custom class loaders are sup- ported, but with all classes granted the same permissions. Native JNI code is not supported. App Caching Once an application server loads an app’s code and resource files, it keeps them in memory to serve multiple requests. The behavior of this “app cache” differs slightly between the Python and Java runtime environments, but in general, it’s as if the ap- plication instance is “running” on the server and responding to requests as events. The app can store code and data in memory, and expects it to be available for subsequent requests handled by the app instance. App caching is one of the main mechanisms for speeding up requests, and it’s worthwhile to use it as much as possible, when appropriate. How the App Is Run | 89
- 111. App Engine makesno guarantee that the requests from the same client reach the same server, nor does it guarantee that an app instance will stay in memory for any particular amount of time. If the app is receiving many requests, App Engine may start new in- stances on additional servers to spread the load. If an app instance does not receive a request in a while, the server may purge it from memory to free resources for other apps. Each server frees the least recently used app when it needs more resources. The best uses of the app cache are for caching code (Python module imports, Java classes) and processed data that is not specific to a given user or request (such as parsed configuration files). All of the application’s memory is retained in the cache, so using the app cache is a simple matter of storing values in global variables. Because there are no guarantees for how apps are cached or how requests are distrib- uted, the app cache is not a good choice for persisting data that isn’t local to the app instance. Instead, use the datastore for permanent reliable storage, and the memcache for app-visible cached data and temporary values. App caching in Python The Python runtime environment keeps the global variable space in memory as long as the environment exists. This includes imported modules, as well as the script han- dler’s global variables and function and class definitions. The first time a Python module is imported, the contents of the module are evaluated. If the module is imported again in the same runtime environment, nothing happens. Typically, evaluating a module’s code defines classes, functions, and global variables, and does nothing else, so doing nothing on subsequent tries saves time and otherwise does the right thing. This is how imports normally work during a run of a Python program. With App Engine, this is also how imports work across multiple requests handled by the same application instance. Once a module has been imported in an instance of an app, subsequent attempts to import it in every request handled by that instance will do nothing. When to Import Modules Import caching might suggest that you ought to do something clever with imports. If you import all the modules you would ever need at the top of the request handler, then only the first request for the instance incurs the cost of the imports, and all subsequent requests are faster. Alternatively, if you wait to import a module until just before you use the module, you spread the cost of importing modules across multiple request handlers. (In Python, you can put an import statement anywhere, including inside conditionals.) 90 | Chapter 3: Handling Web Requests
- 112. In practice, neithermethod of importing offers much advantage. Late importing may actually be slower if the import is inside a loop or frequent code path, since importing an already-imported module takes some time, even if it doesn’t load files or evaluate module code. It’s best to import modules in the way that makes the most sense for code clarity, typically at the top of each module (source file) that uses the module. Once you’ve developed your app, if you notice you have a significant slowdown during im- ports of the first request, you could consider importing selected modules in their code paths if you can demonstrate that doing so would skip a lot of unneeded code in a significant number of cases. App Engine includes a feature similar to import caching to achieve similar savings for the handler scripts themselves. In general, when the app server invokes a handler script, it evaluates all of the script’s code, regardless of whether the script has been invoked by the app instance in the past. But if the script defines a function named main() that takes no arguments (or arguments with default values), App Engine treats the handler script like a module. The first invocation evaluates the complete script as usual, but for all subsequent invocations, App Engine just calls the main() function. Consider an early version of the clock application we built in Chapter 2, repeated here as Example 3-4. This handler script imports three modules, defines three global ele- ments (a class, a global variable, and the main() function), then calls the main() function. Example 3-4. A simple request handler illustrating the use of app caching from google.appengine.ext import webapp from google.appengine.ext.webapp.util import run_wsgi_app import datetime class MainPage(webapp.RequestHandler): def get(self): time = datetime.datetime.now() self.response.headers['Content-Type'] = 'text/html' self.response.out.write('<p>The time is: %s</p>' % str(time)) application = webapp.WSGIApplication([('/', MainPage)], debug=True) def main(): run_wsgi_app(application) if __name__ == '__main__': main() The first time an app instance invokes this script, the full script is evaluated, thereby importing the modules and defining the global elements in the script’s namespace within the app instance. Evaluating the script also calls the main() function. How the App Is Run | 91
- 113. The if __name__... condition is a Python idiom that allows you to import the script as if it were a module without side effects, such as to test the code in a separate test script. App Engine ensures this idiomatic condition is true when the script is invoked the first time. If the app instance receives a request for this handler after the script has already been evaluated, the instance simply calls its main() function directly. This avoids the unnec- essary overhead of constituting the values of the global elements, since they’re already defined and aren’t expected to change. This includes the global variable application, whose value is an object representing the webapp application run by the main() func- tion. This object can be reused by multiple requests, so creating it only on the first invocation and storing it in a global saves time during subsequent requests. App caching is also useful for reusing calculations that are not specific to the user. When you use regular expressions with the re module, each regular expression is compiled the first time it is used. The module remembers compiled versions of up to 100 regular expressions, so subsequent uses of a pattern do not incur processing time to compile the pattern. With app caching, only a fraction of requests spend time compiling regular expressions. You can store your own data in the app cache using global variables. For instance, if your app loads and parses configuration or datafiles that are uploaded with the app (and therefore do not change), you can store the parsed objects in globals. As long as it’s reasonable for two requests coming from different users to use the same data, you can store it in the memory of the app instance for reuse. Remember to call your main() function in the body of your script han- dler. If you do not, the first invocation of the handler will appear to do nothing, but subsequent invocations will succeed, because they are call- ing main() directly. This behavior will appear erratic as App Engine cre- ates, reuses, and destroys app instances. There is no reason not to take advantage of app caching, so make sure all of your Python script handlers define a main() function. App caching in Java When an app server receives a request for a Java application for the first time, the server creates a new instance of the app environment, then loads the servlet classes. Subse- quent requests for the app routed to that server are handled by the same JVM instance, with the app’s memory preserved between requests, including imported classes and static variables. You can reduce the amount of time it takes to handle requests by storing values that apply to all requests in statics, such as data structures parsed from config- uration files. 92 | Chapter 3: Handling Web Requests
- 114. When an appaccesses data that is global to the JVM, it should do so in a thread-safe manner. App Engine may run multiple request handlers with the same JVM instance simultaneously, each in its own thread. As per the sandbox restriction, request handlers cannot spawn their own threads, but App Engine may use multiple threads to initiate request handlers. Using thread safety with global data in a servlet container is a common practice, and this best practice is no different for App Engine. If your servlets must be loaded in a particular order for initialization to succeed, you can control this order using the <load-on-startup> element in the deployment descrip- tor. This element appears inside the servlet declaration (<servlet>) and contains an integer that describes its load order relative to the other servlets (1 for the first, 2 for the second, and so on). Note that App Engine loads the servlet classes during the first request, not beforehand. If possible, you may want to delay initialization tasks until the latest moment they are needed to spread out application warm-up across multiple requests. Logging Activity and message logs are an essential part of a web application. They are your view into what happens with your application over time as it is being used, who is using it and how, and what problems, if any, your users are having. App Engine logs all incoming requests for your application, including application re- quests, static file requests, and requests for invalid URLs (so you can determine whether there is a bad link somewhere). For each request, App Engine logs the date and time, the IP address of the client, the URL requested (including the path and parameters), the domain name requested, the browser’s identification string (the “user agent”), the referring URL if the user followed a link, and the HTTP status code in the response returned by the app or by the frontend. App Engine also logs several important statistics about each request: the amount of time it took to handle each request, the amount of “CPU time” that was spent handling the request, and the size of the response. The CPU time measurement is particularly important to watch because requests that consistently consume a large amount of CPU may be throttled, such that the CPU use is spread over more clock time. Your application code can log the occurrence of notable events and data using a logging API. Logging a message associates a line of text with the request that emitted it, in- cluding all of the data logged for the request. Each message has a log level indicating the severity of the message to make it easier to find important messages during analysis. App Engine supports five log levels: debug, info, warning, error, and critical. You can browse your application’s request and message logs using the Administration Console, under Logs. You can also download your log data for offline analysis and recordkeeping. We’ll look at these features in detail in Chapter 15. How the App Is Run | 93
- 115. Logging in Python Pythonapplications can use the logging module from the standard library to log mes- sages. App Engine hooks into this module to relay messages to the logging system, and to get the log level for each message. Example 3-5 shows this module in action. Example 3-5. The use of the logging Python module to emit messages at different log levels import logging # ... logging.debug('debug level') logging.info('info level') logging.warning('warning level') logging.error('error level') logging.critical('critical level') sys.stderr.write('stderr write, logged at the error leveln') In addition to messages logged with the logging module, each line of text written to the standard error stream (sys.stderr) is logged at the “error” level. (Because Python uses CGI, anything written to the standard output stream becomes part of the response data.) In a traditional application using the logging module, you would configure the module to output only messages above a given level of severity. When running on App Engine, the level of the output is always the “debug” level, and it cannot be changed. You can filter messages by severity after the fact in the Administration Console, or when down- loading logs using appcfg.py. When running in the development web server, log messages are written to the Console, and data written to sys.stderr is written to the server’s error stream. If you are running your server in the Launcher, you can open a window to view log messages by clicking the Logs button. The development server sets its log level to INFO by default. You can change this to DEBUG by giving the server the command-line argument --debug. Logging in Java For Java applications, App Engine supports the java.util.logging library from the JRE. App Engine recognizes log levels of messages logged using this library. Example 3-6 illustrates the use of the Logger class and its convenience methods. Example 3-6. The use of the java.util.logging package to emit messages at different log levels import java.io.IOException; import javax.servlet.http.*; import java.util.logging.Logger; public class LoggingServlet extends HttpServlet { 94 | Chapter 3: Handling Web Requests
- 116. private static finalLogger log = Logger.getLogger(TestServlet.class.getName()); public void doGet(HttpServletRequest req, HttpServletResponse resp) throws IOException { log.finest("finest level"); log.finer("finer level"); log.fine("fine level"); log.config("config level"); log.info("info level"); log.warning("warning level"); log.severe("severe level"); System.out.println("stdout level"); System.err.println("stderr level"); } } The seven log levels of java.util.logging correspond to four of App Engine’s log levels: “finest,” “finer,” “fine,” and “config” all correspond to the App Engine debug level; “info” is info, “warning” is warning, and “severe” is error. The “critical” log level is reserved for exceptions that are not caught by the servlet; when this happens, the run- time environment logs a message at this level. If the application writes any data to the standard output or error streams (System.out and System.err), App Engine adds that data to the log. Each line of text written to standard output becomes a log message at the “info” level, and each line written to standard error is logged at the “warning” level. You can control which level of message should be written to the log using configuration for java.util.logging. This allows you to leave detailed low-level logging statements in your code without having all of that information clutter up the logs unnecessarily in a high-traffic app. Configuring the log level requires two things: a configuration file and a system property that identifies the configuration file. For the configuration, create a resource file, such as war/WEB-INF/logging.properties, containing a line like this: .level=INFO You can configure the log level on a per-class basis by adding lines like this with the package path before the .level. Be sure to use the logging level name (such as FINEST) and not the App Engine level name. Next, set a system property telling the logging library where to find its configuration file. You do this by including a <system-properties> element in your appengine- web.xml file, like so: <system-properties> <property name="java.util.logging.config.file" value="WEB-INF/logging.properties" /> </system-properties> How the App Is Run | 95
- 117. If you createdyour Java project using the Google Plugin for Eclipse, your app already has this configuration file and this system property. This configuration comes pre- loaded with log levels for the DataNucleus interface (an interface for the datastore), so you can leave those set to the “warning” level while the rest of your app uses another level. The java.util.logging and the standard output and error streams are the only ways to log messages at specific log levels. If you or a component of your app prefers a different logging library, such as log4j, messages emitted by that library will work as long as the library can write to the standard streams. If you want to be able to use the Administration Console to filter logs by levels other than “info” and “warning,” you will need an adapter of some kind that calls java.util.logging behind the scenes. You get complete log messages when you download log data, so you can always analyze alternate log formats in downloaded data. When running in the development web server, log messages are written to the console, and text written to the standard streams is written to the corresponding streams for the server. In Eclipse, these messages appear in the Console pane. Quotas and Limits The runtime environment monitors the system resources used by the application and limits how much the app can consume. For the resources you pay for, such as CPU time and storage, you set these limits yourself by allocating a budget in the Adminis- tration Console. App Engine also enforces several system-wide limits that protect the integrity of the servers and their ability to serve multiple apps. In App Engine parlance, “quotas” are resource limits that refresh at the beginning of each calendar day (at midnight, Pacific Time). You can monitor your application’s daily consumption of quotas using the Administration Console, in the Quota Details section. Since Google may change how the limits are set as the system is tuned for performance, we won’t state some of the specific values of these limits in this book. You can find the actual values of these limits in the official App Engine documentation. Google has said it will give 90 days’ notice before changing limits in a way that would affect existing apps. Request Limits Several system-wide limits specify how requests can behave. These include the size and number of requests over a period of time, and the bandwidth consumed by inbound and outbound network traffic. One important request limit is the request timer. An application has 30 seconds to respond to a request. 96 | Chapter 3: Handling Web Requests
- 118. Near the endof the 30 seconds, the server raises an exception that the application can catch for the purposes of exiting cleanly or returning a user-friendly error message. For Python, the request timer raises a google.appengine.runtime.DeadlineExceededError. For Java, the request timer throws a com.google.apphosting.api.DeadlineExceededException. If the request handler has not returned a response or otherwise exited after 30 seconds, the server terminates the process and returns a generic system error (HTTP code 500) to the client. The 30-second limit applies to every request, including user web requests, enqueued tasks, scheduled tasks, and other web hooks such as incoming XMPP and email requests. The size of a request is limited to 10 megabytes, as is the size of the request handler’s response. CPU Limits Naturally, when application code is executing, it uses the CPU of the server. The CPU is shared by multiple request handlers running on a given server at the same time. App Engine prevents one request handler’s use of the CPU from affecting the performance of other apps by allocating more resources as needed, such as serving apps from alter- nate servers. Use of the CPU is an important commodity in App Engine, so it is a billable quota. It is also limited to prevent bursts of usage from causing problems. Use of a CPU could be measured by the number of processor cycles it takes to do the work, in units such as megacycles (millions of CPU operations per second). However, the speed of a CPU may vary from one app server to the next. While the same amount of work gets done on any CPU for a given number of megaflops, the total impact to performance may vary (though not in a way you’d notice). So instead of megaflops, App Engine measures CPU usage in “CPU minutes.” A CPU minute is the number of megaflops that can be performed by a standard processor in one minute. The standard processor in this case is a 1.2 GHz Intel x86 processor. In other words, if your app were running by itself on a single 1.2 GHz processor, it would do a CPU minute’s worth of work in one clock minute. The actual amount of clock time to perform that work can vary, depending on how App Engine allocates resources. App Engine tries to balance the load so each app gets a consistent rate of CPU attention. But if a request handler uses too much CPU time, the app server may throttle the allocation of CPU. In that case, the request handler consumes the same number of CPU minutes, but it takes more clock time to do it. The CPU time quota includes two sets of resources: CPU time used on the app server, and CPU rime used in the datastore. Other services do not contribute to the CPU time Quotas and Limits | 97
- 119. quota; they manageCPU resources in other ways. An app server does not consume CPU time while it is waiting for a service call to return. The Python runtime environment provides a function you can call to analyze the CPU usage in a request handler. The get_request_cpu_usage() function in the google.appengine.api.quota package returns the number of megacycles (not CPU minutes) used by the request handler so far. You can call this function before and after a calculation to determine how much CPU time the calculation used. Service Limits Each service has its own set of quotas and limits. As with system-wide limits, some can be raised using a billing account and a budget, such as the number of recipients the application has sent emails to. Other limits are there to protect the integrity of the service, such as the number of image transformations performed by the image service. We won’t list all of the service limits here. See the official documentation for a complete list, including values. In Python, when an app exceeds a service-specific limit or quota, the runtime environ- ment raises a ...runtime.apiproxy_errors.OverQuotaError. In Java, the service call throws a com.google.apphosting.api.ApiProxy.OverQuotaException (note the apphosting package name, not appengine). Two particular limits apply to all services: the size of a service call and the size of the service response are each limited to 1 megabyte. This imposes an inherent limit on the size of datastore entities and memcache values. Even though an incoming request can contain up to 10 megabytes, only 1 megabyte of that data can be stored using a single call to the datastore or memcache. The service call size limits also apply to the datastore’s batch API. Batch calls fetch or store multiple entities in one service call. The total size of a batch call must fit within the 1-megabyte service call limit. Deployment Limits Two limits affect the size and structure of your application’s files. An application file cannot be larger than 10 megabytes. This applies to resource files (code, configuration) as well as static files. Also, the total number of files for an application cannot be larger than 3,000, including resource files and static files. These limits aren’t likely to cause problems in most cases, but some common tasks can push on these numbers. Some third-party libraries or frameworks can be many hun- dreds of files. Sites consisting of many pages of text or images can reach 3,000 pretty quickly. A site offering video or software for download might have difficulty with the 10-megabyte limit. 98 | Chapter 3: Handling Web Requests
- 120. The Python runtimeoffers two ways to mitigate the file count limit. If you have many files of Python code, you can store the code files in a ZIP archive file, then add the path to the ZIP archive to sys.path at the top of your request handler scripts. The request handler scripts themselves must not be in a ZIP archive. Thanks to zipimport, a feature built into Python, the Python interpreter will recognize the ZIP file automatically and unpack it as needed when importing modules. Unpacking takes additional CPU time, but since imports are cached, the app only incurs this cost the first time the module is imported in a given app instance. import sys sys.path.insert(0, 'locales.zip') import locales.es The Python App Engine runtime includes a similar mechanism for serving static files from a ZIP archive file, called zipserve. Unlike zipimport, this feature is specific to App Engine. To serve static files from a ZIP archive, add the zipserve request handler to your app.yaml, associated with a URL path that represents the path to the ZIP file: - url: /static/images/.* script: $PYTHON_LIB/google/appengine/ext/zipserve This declares that all requests for a URL starting with /static/images/ should resolve to a path in the ZIP file /static/images.zip. The string $PYTHON_LIB in the script path refers to the location of the App Engine li- braries, and is the only such substitution available (so far). It’s useful precisely for this purpose, to set up a request handler whose code is in the App Engine Python modules included with the runtime environment. When using zipserve, keep in mind that the ZIP archive is uploaded as a resource file, not a static file. Files are served by application code, not the static file infrastructure. By default, the handler advises browsers to cache the files for 20 minutes. You can customize the handler’s cache duration using the wrapper WSGIApplication. See the source code for google/appengine/ext/zipserve/__init__.py in the SDK for details. Java applications have a common solution for reducing the file count for application code: JARs. If your app has too many .class files, simply put them in a JAR file using the jar utility included with the Java development kit. As with Python, app caching reduces the overhead of unpacking JARs. There is no equivalent to zipserve for static files in Java included with the SDK, but it’s easy to write something that behaves sim- ilarly using JAR files. Also, with Java, make sure to use <static-files> and <resource-files> directives in your appengine-web.xml file to exclude the appropriate files. By default, all files outside of WEB-INF/ belong to both groups, and so are counted twice, once for each group. The file count limit is the total count for both groups. An application can only be uploaded a limited number of times per day, currently 250. You may not notice this limit during normal application development. If you are using Quotas and Limits | 99
- 121. app deployments toupload data to the application on a regular schedule, you may want to keep this limit in mind. See Chapter 12 for a better way to upload data. Billable Quotas Every application gets a limited amount of computing resources for free, so you can start developing and testing your application right away. You can purchase additional computing resources at competitive rates. You only pay for what you actually use: you specify the maximum amount of money you want to spend, and allocate those funds to the various resources to increase those limits. The amount App Engine bills is for the resources your application actually consumes, and it never exceeds your declared budget. The first step is to establish a billing account for the app. To enable billing, sign in to the Administration Console with the developer account that is to become the billing account, then select Billing Settings from the sidebar. Click the Enable Billing button, and follow the prompts. Billing is managed through Google Checkout, Google’s payment service. When you set a budget, you are prompted to go through Google Checkout to approve the maxi- mum charge. Setting a budget does not apply a charge to your account, it only sets the maximum. Billing occurs once a day. When you set up billing, you are prompted to specify a budget and the resource allo- cations for that budget. You can accept the default settings, pick another preset, or set custom allocations for each of the resources. The billable resources include: • CPU time, in CPU hours • Outgoing bandwidth, in gigabytes sent • Incoming bandwidth, in gigabytes received • Data stored, in gigabytes per day • Recipients emailed (the number of total recipients in all email messages sent) The bandwidth limits apply to incoming requests, outgoing responses, bandwidth used in a URL Fetch, and other network activity in and out of Google’s data centers. The data storage limit comprises the datastore, including automatic and custom indexes, as well as the memcache. As mentioned earlier, CPU time includes CPU usage on the app servers and on the datastore servers. Each billable resource has a maximum amount you can set in the Console. If your application needs more resources than the allowed maximum settings, you can request an increase of these maximums. See the official documentation for the link to the re- quest form. The owner of the billing account can change the budget and the resource allocations for the app at any time using the Administration Console. A change to your budget 100 | Chapter 3: Handling Web Requests
- 122. takes about 10minutes to complete, and you will not be able to change the setting again during those 10 minutes. The resources you purchase with your budget are in addition to the free resources included with every app. In fact, when you enable billing, App Engine increases the free resource levels for the app. You can enable billing, then set the budget to $0 to claim the free quota increase without additional charges, though you still need to as- sociate a method of payment with the billing account. The official documentation includes a complete list of the free quota limits, the in- creased free quota limits with billing enabled, the maximum allocation amounts, and the latest billing rates. Resource Usage Headers If you are signed in to your app via Google Accounts using your developer account when you request an app URL, App Engine includes HTTP headers in the response that give you information about resources consumed by that request. You can view these headers using a browser plug-in, such as the Web Developer extension for Firefox (Information menu, View Response Headers). Figure 3-2 shows an example of viewing these response headers using the Web Developer extension. Figure 3-2. Response headers for a request from a developer account, including resource usage, as viewed with the Firefox Web Developer extension Quotas and Limits | 101
- 123. The X-AppEngine-Estimated-CPM-US-Dollars headerrepresents the monetary cost of the request times 1,000, at the current price in U.S. dollars. That is, it’s an estimate of what 1,000 requests similar to this one would cost. The X-AppEngine-Resource-Usage header summarizes various resource usage metrics for the request, including clock time and CPU time spent on the app server and APIs. Each number is in milliseconds. If your app does not use Google Accounts for authentication, you can create a request handler restricted to administrators for the sole purpose of signing in. After you visit that handler and sign in, subsequent requests to the app will return the resource usage headers. 102 | Chapter 3: Handling Web Requests
- 124. CHAPTER 4 Datastore Entities Most scalable web applications use separate systems for handling web requests and for storing data. The request handling system routes each request to one of many machines, each of which handles the request without knowledge of other requests going to other machines. Each request handler behaves as if it is stateless, acting solely on the content of the request to produce the response. But most web applications need to maintain state, whether it’s remembering that a customer ordered a product, or just remembering that the user who made the current request is the same user who made an earlier request handled by another machine. For this, request handlers must interact with a central database to fetch and update the latest information about the state of the application. Just as the request handling system distributes web requests across many machines for scaling and robustness, so does the database. But unlike the request handlers, databases are by definition stateful, and this poses a variety of questions. Which machine re- members which piece of data? How does the system route a data query to the machine— or machines—that can answer the query? When a client updates data, how long does it take for all machines that know that data to get the latest version, and what does the system return for queries about that data in the meantime? What happens when two clients try to update the same data at the same time? What happens when a machine goes down? As with request handling, Google App Engine manages the scaling and maintenance of data storage automatically. Your application interacts with an abstract model that hides the details of managing and growing a pool of data servers. This model and the service behind it provide answers to the questions of scalable data storage specifically designed for web applications. 103
- 125. App Engine’s abstractionfor data is easy to understand, but it is not obvious how to best take advantage of its features. In particular, it is surprisingly different from the kind of database with which most of us are most familiar, the relational database. It’s different enough, in fact, that Google doesn’t call it a “database,” but a “datastore.” We will dedicate the next several chapters to this important subject. Entities, Keys, and Properties The App Engine datastore is best understood as an object database. An object in the datastore is known as an entity. An entity has a key that uniquely identifies the object across the entire system. If you have a key, you can fetch the entity for the key quickly. Keys can be stored as data in entities, such as to create a reference from one entity to another. A key has several parts, some of which we’ll discuss here and some of which we’ll cover later. One part of the key is the application’s ID, which ensures that nothing else about the key can collide with the entities of any other application. It also ensures that no other app can access your app’s data, and that your app cannot access data for other apps. You won’t see the app ID mentioned in the datastore API; this is automatic. An important part of the key is the kind. An entity’s kind categorizes the entity for the purposes of queries, and for ensuring the uniqueness of the rest of the key. For example, a shopping cart application might represent each customer order with an entity of the kind “Order.” The application specifies the kind when it creates the entity. The key also contains an entity ID. This can be an arbitrary string specified by the app, or it can be generated automatically by the datastore. The API calls an entity ID given by the app a key name, and an entity ID generated by the datastore an ID. An entity has either a key name or an ID, but not both. Once an entity has been created, its key cannot be changed. This applies to all parts of its key, including the kind and the key name or ID. The data for the entity is stored in one or more properties. Each property has a name and at least one value. Each value is of one of several supported data types, such as a string, an integer, a date-time, or a null value. We’ll look at property value types in detail later in this chapter. A property can have multiple values, and each value can be of a different type. As you will see in “Multivalued Properties” on page 113, multivalued properties have unusual behavior, but are quite useful for modeling some kinds of data, and surprisingly efficient. 104 | Chapter 4: Datastore Entities
- 126. It’s tempting tocompare these concepts with similar concepts in rela- tional databases: kinds are tables; entities are rows; properties are fields or columns. That’s a useful comparison, but watch out for differences. Unlike a table in a relational database, there is no relationship between an entity’s kind and its properties. Two entities of the same kind can have different properties set or not set, and can each have a property of the same name but with values of different types. You can (and often will) enforce a data schema in your own code, and App Engine includes libraries to make this easy, but this is not required by the datastore. Also unlike relational databases, keys are not properties. You can per- form queries on key names just like properties, but you cannot change a key name after the entity has been created. And of course, a relational database cannot store multiple values in a single cell, while an App Engine property can have multiple values. Introducing the Python Datastore API In the Python API for the App Engine datastore, Python objects represent datastore entities. The class of the object corresponds to the entity’s kind, where the name of the class is the name of the kind. You define kinds by creating classes that extend one of the provided base classes. Each attribute of the object corresponds with a property of the entity. To create a new entity in the datastore, you call the class constructor, set attributes on the object, then call a method to save it. To update an existing entity, you call a method that returns the object for the entity (such as via a query), modify its attributes, then save it. Example 4-1 defines a class named Book to represent entities of the kind Book. It creates an object of this class by calling the class constructor, then sets several property values. Finally, it calls the put() method to save the new entity to the datastore. The entity does not exist in the datastore until it is put() for the first time. Example 4-1. Python code to create an entity of the kind Book from google.appengine.ext import db import datetime class Book(db.Expando): pass obj = Book() obj.title = 'The Grapes of Wrath' obj.author = 'John Steinbeck' obj.copyright_year = 1939 obj.author_birthdate = datetime.date(1902, 2, 27) obj.put() Introducing the Python Datastore API | 105
- 127. The Book classinherits from the class Expando in App Engine’s db package. The Expando base class says Book objects can have any of their properties assigned any value. The entity “expands” to accommodate new properties as they are assigned to attributes of the object. Python does not require that an object’s member variables be declared in a class definition, and this example takes advantage of this using an empty class definition—the pass keyword indicates the empty definition—and assigns values to attributes of the object after it is created. The Expando base class knows to use the object’s attributes as the values of the corresponding entity’s properties. The Expando class has a funny name because this isn’t the way the API’s designers expect us to create new classes in most cases. Instead, you’re more likely to use the Model base class with a class definition that ensures each instance conforms to a structure, so a mistake in the code doesn’t accidentally create entities with malformed properties. Here is how we might implement the Book class using Model: class Book(db.Model): title = db.StringProperty() author = db.StringProperty() copyright_year = db.IntegerProperty() author_birthdate = db.DateProperty() The Model version of Book specifies a structure for Book objects that is enforced while the object is being manipulated. It ensures that values assigned to an object’s properties are of appropriate types, such as string values for title and author properties, and raises a runtime error if the app attempts to assign a value of the wrong type to a property. With Model as the base class, the object does not “expand” to accommodate other entities: an attempt to assign a value to a property not mentioned in the class definition raises a runtime error. Model and the various Property definitions also provide other features for managing the structure of your data, such as automatic values, required values, and the ability to add your own validation and serialization logic. It’s important to notice that these validation features are provided by the Model class and your application code, not the datastore. Even if part of your app uses a Model class to ensure a property’s value meets certain conditions, another part of your app can still retrieve the entity without using the class and do whatever it likes to that value. The bad value won’t raise an error until the app tries to load the changed entity into a new instance of the Model class. This is both a feature and a burden: your app can manage entities flexibly and enforce structure where needed, but it must also be careful when those structures need to change. Data modeling and the Model class are discussed in detail in Chapter 7. The Book constructor accepts initial values for the object’s properties as keyword ar- guments. The constructor code earlier could also be written like this: 106 | Chapter 4: Datastore Entities
- 128. obj = Book(title='TheGrapes of Wrath', author='John Steinbeck', copyright_year=1939, author_birthdate=datetime.date(1902, 2, 27)) As written, this code does not set a key name for the new entity. Without a key name, the datastore generates a unique ID when the object is saved for the first time. If you prefer to use a key name generated by the app, you call the constructor with the key_name parameter: obj = Book(key_name='0143039431', title='The Grapes of Wrath', author='John Steinbeck', copyright_year=1939, author_birthdate=datetime.date(1902, 2, 27)) Because the Python API uses keyword arguments, object attributes, and object methods for purposes besides entity properties, there are several property names that are off-limits. For instance, you cannot use the Py- thon API to set a property named key_name, because this could get con- fused with the key_name parameter for the object constructor. Names reserved by the Python API are enforced in the API, but not in the da- tastore itself. Google’s official documentation lists the reserved property names. The datastore reserves all property names beginning and ending with two underscores (such as __internal__). This is true for the Python API and the Java API, and will be true for future APIs as well. The Python API ignores all object attributes whose names begin with a single underscore (such as _counter). You can use such attributes to attach data and functionality to an object that should not be saved as properties for the entity. The complete key of an entity, including the key name and kind, must be unique. (We’ll discuss another part to keys that contributes to a key’s uniqueness, called ancestors, in Chapter 6.) If you build a new object with a key that is already in use, then try to save it, the save will replace the existing object. For when you don’t want to overwrite ex- isting data, the datastore API provides an alternate way to create an object. The get_or_insert() class method takes a key name and either returns an existing entity with that key name, or creates a new entity with that key name and no properties and returns it. Either way, the method is guaranteed to return an object that represents an entity in the datastore: obj = Book.get_or_insert('0143039431') if obj.title: # Book already exists. # ... else: Introducing the Python Datastore API | 107
- 129. obj.title = 'TheGrapes of Wrath' obj.author = 'John Steinbeck' obj.copyright_year = 1939 obj.author_birthdate = datetime.date(1902, 2, 27) obj.put() Introducing the Java Datastore API App Engine for Java includes support for two major standard interfaces for databases: Java Data Objects (JDO) and the Java Persistence API (JPA). Like the other standards- based interfaces in the App Engine Java API, using one of these interfaces makes it easier to move your application from and to another platform. JDO and JPA support different kinds of databases, including object databases and relational databases. They provide an object-oriented interface to your data, even if the underlying database is not an object store. Many of the concepts of these interfaces translate directly to App Engine datastore concepts: classes are kinds, objects are entities, fields are properties. App Engine’s im- plementation also supports several advanced features of these interfaces, such as object relationships. Inevitably, some concepts do not translate directly and have behaviors that are specific to App Engine. We’ll discuss one of these interfaces, JPA, in Chapter 8. For now, here is a simple example of a data class using JPA: import java.util.Date; import javax.persistence.Entity; import javax.persistence.GeneratedValue; import javax.persistence.GenerationType; import javax.persistence.Id; @Entity public class Book { @Id @GeneratedValue(strategy = GenerationType.IDENTITY) private Long id; private String title; private String author; private int copyrightYear; private Date authorBirthdate; public Long getId() { return id; } public String getTitle() { return title; } public void setTitle(String title) { this.title = title; 108 | Chapter 4: Datastore Entities
- 130. } public String getAuthor() { return author; } public void setAuthor(String author) { this.author = author; } public int getCopyrightYear() { return copyrightYear; } public void setCopyrightYear(int copyrightYear) { this.copyrightYear = copyrightYear; } public Date getAuthorBirthdate() { return authorBirthdate; } public void setAuthorBirthdate(Date authorBirthdate) { this.authorBirthdate = authorBirthdate; } } The JDO and JPA implementations are built on top of a low-level API for the App Engine datastore. The low-level API exposes all of the datastore’s features, and corre- sponds directly to datastore concepts. For instance, you must use the low-level API to manipulate entities with properties of unknown names or value types. You can also use the low-level API directly in your applications out of preference, or to implement your own data management layer. The following code creates a Book entity using the low-level API: import java.io.IOException; import java.util.Calendar; import java.util.Date; import java.util.GregorianCalendar; import javax.servlet.http.HttpServlet; import javax.servlet.http.HttpServletRequest; import javax.servlet.http.HttpServletResponse; import com.google.appengine.api.datastore.DatastoreService; import com.google.appengine.api.datastore.DatastoreServiceFactory; import com.google.appengine.api.datastore.Entity; // ... DatastoreService ds = DatastoreServiceFactory.getDatastoreService(); Entity book = new Entity("Book"); book.setProperty("title", "The Grapes of Wrath"); book.setProperty("author", "John Steinbeck"); book.setProperty("copyrightYear", 1939); Date authorBirthdate = Introducing the Java Datastore API | 109
- 131. new GregorianCalendar(1902, Calendar.FEBRUARY,27).getTime(); book.setProperty("authorBirthdate", authorBirthdate); ds.put(book); // ... Notice that the application code, not the datastore, is responsible for managing the structure of the data. JDO and JPA impose this structure using classes whose fields are persisted to the datastore behind the scenes. This can be both a benefit and a burden when you need to change the structure of existing data. To illustrate the datastore concepts, we will use the low-level API for Java examples in the next few chapters. In Chapter 8, we will reintroduce JPA, and discuss how JPA concepts correspond with App Engine concepts. For more information on the Java Data Objects interface, see the official App Engine documentation. Property Values Each value data type supported by the datastore is represented by a primitive type in the language for the runtime or a class provided by the API. The data types and their language-specific equivalents are listed in Table 4-1. In this table, db is the Python package google.appengine.ext.db, and datastore is the Java package com.google.appengine.api.datastore. Table 4-1. Datastore property value types and equivalent language types Data type Python type Java type Unicode text string (up to unicode or str (converted to uni java.lang.String 500 bytes, indexed) code as ASCII) Long Unicode text string db.Text datastore.Text (not indexed) Short byte string (up to 500 db.ByteString datastore.ShortBlob bytes, indexed) Long byte string (not db.Blob datastore.Blob indexed) Boolean bool boolean Integer (64-bit) int or long (converted to 64-bit long) byte, short, int, or long (converted to long) Float (double precision) float float or double (converted to double) Date-time datetime.datetime java.util.Date Null value None null Entity key db.Key datastore.Key A Google account users.User ...api.users.User A category (GD) db.Category datastore.Category 110 | Chapter 4: Datastore Entities
- 132. Data type Python type Java type A URL (GD) db.Link datastore.Link An email address (GD) db.Email datastore.Email A geographical point (GD) db.GeoPt datastore.GeoPt An instant messaging han- db.IM datastore.IMHandle dle (GD) A phone number (GD) db.PhoneNumber datastore.PhoneNumber A postal address (GD) db.PostalAddress datastore.PostalAddress A user rating (GD) db.Rating datastore.Rating The datastore types in this table labeled “(GD)” are types borrowed from the Google Data protocol. These are supported as distinct native data types in the datastore, though most of them are implemented as text strings. Notable exceptions are GeoPt, which is a pair of floating-point values for latitude (–90 to +90) and longitude (–180 to +180), and Rating, which is an integer between 1 and 100. Example 4-2 demonstrates the use of several of these data types in Python. Example 4-2. Python code to set property values of various types from google.appengine.ext import webapp from google.appengine.ext import db from google.appengine.api import users import datetime class Comment(db.Expando): pass class CommentHandler(webapp.RequestHandler): def post(self): c = Comment() c.commenter = users.get_current_user() # returns a users.User object c.message = db.Text(self.request.get('message')) c.date = datetime.datetime.now() c.put() # Display the result page... When you use Python’s db.Expando or Java’s low-level datastore API, types that are widened to other types when stored come back as the wider datastore types when you retrieve the entity. For instance, a Java Integer comes back as a Long. If you use these APIs in your app, it’s best to use the native datastore types, so the value types stay consistent. The data modeling interfaces offer a way to store values in these alternate types and convert them back automatically when retrieving the entity. See Chapters 7 and 8. Property Values | 111
- 133. Strings, Text, andBlobs The datastore has two distinct data types for storing strings of text: short strings and long strings. Short strings are indexed; that is, they can be the subject of queries, such as a search for every Person entity with a given value for a last_name property. Short string values must be less than 500 bytes in length. Long strings can be longer than 500 bytes, but are not indexed. Text strings, short and long, are strings of characters from the Unicode character set. Internally, the datastore stores Unicode strings using the UTF-8 encoding, which rep- resents some characters using multiple bytes. This means that the 500-byte limit for short strings is not necessarily the same as 500 Unicode characters. The actual limit on the number of characters depends on which characters are in the string. The Python API distinguishes between short strings and long strings using Python data types. The Python built-in types unicode and str represent short string values. str val- ues are assumed to be text encoded as ASCII, and are treated as UTF-8 (which is equiv- alent to ASCII for the first 128 characters in the character set). For long strings, the Python API includes a db.Text class, which takes a unicode or str value as an argument for its constructor. # Short strings. e.prop = "a short string, as an ASCII str" e.prop = unicode("a short string, as a unicode value") # A long string. e.prop = db.Text("a long string, can be longer than 500 bytes") The Java API makes a similar distinction, treating String values as short strings, and using the datastore.Text class to represent long text strings. The datastore also supports two additional classes for strings of bytes, or “blobs.” Blobs are not assumed to be of any particular format, and their bytes are preserved. This makes them good for nontext data, such as images, movies, or other media. As with text strings, the blob types come in indexed and nonindexed varieties. The Python API provides the db.Blob class to represent blob values, which takes a str value as an ar- gument for its constructor. # A blob. self.request.body is the body of the request in a # webapp request handler, such as an uploaded file. e.prop = db.Blob(self.request.body) In Java, the blob types are datastore.ShortBlob and datastore.Blob. Unset Versus the Null Value One possible value of a property is the null value. In Python, the null value is represented by the Python built-in value None. In Java, this value is null. 112 | Chapter 4: Datastore Entities
- 134. A property withthe null value is not the same as an unset property. Consider the fol- lowing Python code: class Entity(db.Expando): pass a = Entity() a.prop1 = 'abc' a.prop2 = None a.put() b = Entity() b.prop1 = 'def' b.put() This creates two entities of the kind Entity. Both entities have a property named prop1. The first entity has a property named prop2; the second does not. Of course, an unset property can be set later: b.prop2 = 123 b.put() # b now has a property named "prop2." Similarly, a set property can be made unset. In the Python API, you delete the property by deleting the attribute from the object, using the del keyword: del b.prop2 b.put() # b no longer has a property named "prop2." In Java, the low-level datastore API’s Entity class has methods to set properties (setProperty()) and unset properties (removeProperty()). Multivalued Properties As we mentioned earlier, a property can have multiple values. We’ll discuss the more substantial aspects of multivalued properties when we talk about queries and data modeling. But for now, it’s worth a brief mention. A property can have one or more values. A property cannot have zero values; a property without a value is simply unset. Each value for a property can be of a different type, and can be the null value. The datastore preserves the order of values as they are assigned. The Python API returns the values in the same order as they were set. In Python, a property with multiple values is represented as a single Python list value: e.prop = [1, 2, 'a', None, 'b'] Property Values | 113
- 135. Because a propertymust have at least one value, it is an error to assign an empty list ([] in Python) to a property on an entity whose Python class is based on the Expando class. class Entity(db.Expando): pass e = Entity() e.prop = [] # ERROR In contrast, the Model base class includes a feature that automatically translates between the empty list value and “no property set.” You’ll see this feature in Chapter 7. In the Java low-level datastore API, you can store multiple values for a property using a Collection type. The low-level API returns the values as a java.util.List. The items are stored in the order provided by the Collection type’s iterator. For many types, such as SortedSet or TreeSet, this order is deterministic. For others, such as HashSet, it is not. If the app needs the original data structure, it must convert the List returned by the datastore to the appropriate type. Keys and Key Objects The key for an entity is a value that can be retrieved, passed around, and stored like any other value. If you have the key for an entity, you can retrieve the entity from the datastore quickly, much more quickly than with a datastore query. Keys can be stored as property values, as an easy way for one entity to refer to another. The Python API represents an entity key value as an instance of the Key class, in the db package. To get the key for an entity, you call the entity object’s key() method. The Key instance provides access to its several parts using accessor methods, including the kind, key name (if any), and system-assigned ID (if the entity does not have a key name). The Java low-level API is similar: the getKey() method of the Entity class returns an instance of the Key class. When you construct a new entity object and do not provide a key name, the entity object has a key, but the key does not yet have an ID. The ID is populated when the entity object is saved to the datastore for the first time. You can get the key object prior to saving the object, but it will be incomplete. e = Entity() e.prop = 123 k = e.key() # key is incomplete, has neither key name nor ID kind = k.kind() # 'Entity' e.put() # ID is assigned 114 | Chapter 4: Datastore Entities
- 136. k = e.key() # key is complete, has ID id = k.id() # the system-assigned ID If the entity object was constructed with a key name, the key is complete before the object is saved—though if the entity has not been saved, the key name is not guaranteed to be unique. (In Python, the entity class method get_or_insert(), mentioned earlier, always returns a saved entity, either one that was saved previously or a new one created by the call.) You can test whether a key is complete using a method on the Key object. In Python, this is the has_id_or_name() method. The id_or_name() method returns either the ob- ject’s key name or its ID, whichever one it has. In Java, you can call isComplete() to test the Key for completeness, and getId() or getName() to get the numeric ID or the string name. Once you have a complete key, you can assign it as a property value on another entity to create a reference. e2 = Entity() e2.ref = k e2.put() If you know the kind and either the key name or ID of an entity in the datastore, you can construct the key for that entity without its object. In Python, you use the from_path() class method of the Key class. A complete explanation of this feature in- volves another feature we haven’t mentioned yet (paths), but the following suffices for the examples you’ve seen so far: e = Entity(key_name='alphabeta') e.prop = 123 e.put() # ... k = db.Key.from_path('Entity', 'alphabeta') In Java, you can build a Key object from parts using KeyFactory. The static method KeyFactory.createKey() takes the kind and the ID or name as arguments: Key k = KeyFactory.createKey("Entity", "alphabeta"); Paths are related to how the datastore does transactions. We’ll get to them in Chap- ter 6. For the entities we have created so far, the path is just the kind followed by the ID or name. Keys can be converted to string representations for the purposes of passing around as textual data, such as in a web form or cookie. The string representation avoids char- acters considered special in HTML or URLs, so it is safe to use without escaping char- acters. The encoding of the value to a string is simple and easily reversed, so make sure you do not use any information that ought to be secret as part of a key whose string form will be sent to users, such as a key name or kind. In Python: Keys and Key Objects | 115
- 137. k_str = str(k) # ... k = db.Key(k_str) And in Java: String k_str = KeyFactory.keyToString(k); // ... Key k = KeyFactory.stringToKey(k_str); The Java Key class’s toString() method does not return the key’s string encoding. You must use KeyFactory.keyToString() to get the string encoding of a key. Using Entities Let’s look briefly at how to retrieve entities from the datastore using keys, how to inspect the contents of entities, and how to update and delete entities. The API methods for these features are straightforward. Getting Entities Using Keys Given a complete key for an entity, you can retrieve the entity from the datastore. In the Python API, you can call the get() function in the db package with the Key object as an argument: from google.appengine.ext import db k = db.Key('Entity', 'alphabeta') e = db.get(k) If you know the kind of the entity you are fetching, you can also use the get() class method on the appropriate entity class. This does a bit of type checking, ensuring that the key you provide is of the appropriate kind: class Entity(db.Expando): pass e = Entity.get(k) To fetch multiple entities in a batch, you can pass the keys to get() as a list. Given a list, the method returns a list containing entity objects, with None values for keys that do not have a corresponding entity in the datastore. entities = db.get([k1, k2, k3]) 116 | Chapter 4: Datastore Entities
- 138. Getting a batchof entities in this way performs a single service call to the datastore for the entire batch. This is faster than getting each entity in a separate call. It is also subject to the service call size limit (one megabyte), so make sure the total size of all of the entities will not exceed this limit when doing a batch get. Batch gets are also subject to an entity count limit. For convenience, entity classes include methods that take just the IDs or key names and retrieve the corresponding entities, inferring the kind from the class name. See get_by_id() and get_by_key_name() in the official reference documentation. In the Java low-level API, you get an entity by its key using a DatastoreService instance (returned by DatastoreServiceFactory.getDatastoreService()). The instance provides a get() method that takes a Key for a single entity get, or an Iterable<Key> for a batch get. If given an iterable of keys, get() returns a Map of Key to Entity: DatastoreService ds = DatastoreServiceFactory.getDatastoreService(); Map<Key, Entity> entities = ds.get(new ArrayList(Arrays.asList(k1, k2, k3))); Entity e1 = entities.get(k1); Of course, you won’t always have the keys for the entities you want to fetch from the datastore. To retrieve entities that meet other criteria, you use datastore queries. Quer- ies are discussed in Chapter 5. Inspecting Entity Objects Entity objects have methods for inspecting various aspects of the entity. In the Java API, the methods of the Entity class provide straightforward access to the key (getKey()) and kind (getKind()) of the entity. The getProperty() method returns the value of a property given its name. The hasProperty() method tests whether a property is set. setProperty() takes a name and a value and sets the property, replacing any existing value. The Python API has several features for inspecting entities worth mentioning here. You’ve already seen the key() method of an entity object, which returns the db.Key. The is_saved() method returns False if the object has not been saved to the datastore since the object was constructed. If the object has been saved since it was constructed, or if the object was retrieved from the datastore, the method returns True. The method continues to return True even if the object’s properties have been modified, so do not rely on this method to track changes to properties of previously saved entities. e = Entity() # e.is_saved() == False e.put() # e.is_saved() == True Using Entities | 117
- 139. The Java APIdoes not have an equivalent to is_saved(). In Python, entity properties can be accessed and modified just like object attributes: e.prop1 = 1 e.prop2 = 'two' print "prop2 has the value " + e.prop2 You can use Python built-in functions for accessing object attributes to access entity properties. For instance, to test that an entity has a property with a given name, use the hasattr() built-in: if hasattr(e, 'prop1'): # ... To get or set a property whose name is a string, use getattr() and setattr(), respectively: # Set prop1, prop2, ..., prop9. for n in range(1, 10): value = n * n setattr(e, 'prop' + str(n), value) value = getattr(e, 'prop' + str(7)) While entity objects support accessing properties using these methods, the objects do not actually store property values as object attributes. For instance, you cannot use Python’s dir() built-in to get a list of an entity’s properties. Instead, entity objects provide their own method, instance_properties(), for this purpose: for name in e.instance_properties(): value = getattr(e, name) Saving Entities In Python, calling the put() method on an entity object saves the entity to the datastore. If the entity does not yet exist in the datastore, put() creates the entity. If the entity exists, put() updates the entity so that it matches the object. e = Entity() e.prop = 123 e.put() When you update an entity, the app sends the complete contents of the entity to the datastore. The update is all or nothing: there is no way to send just the properties that have changed to the datastore. There is also no way to update a property on an entity without retrieving the complete entity, making the change, then sending the new entity back. You use the same API to create an entity as you do to update an entity. The datastore does not make a distinction between creates and updates. If you save an entity with a 118 | Chapter 4: Datastore Entities
- 140. complete key (suchas a key with a kind and a key name) and an entity already exists with that key, the datastore replaces the existing entity with the new one. If you want to test that an entity with a given key does not exist before you create it, you can do so using the transaction API. You must use a transaction to ensure that another process doesn’t create an entity with that key after you test for it and before you create it. For more informa- tion on transactions, see Chapter 6. If you have several entity objects to save, you can save them all in one call using the put() function in the db package. The put() function can also take a single entity object. db.put(e) db.put([e1, e2, e3]) As with a batch get, a batch put performs a single call to the service. The total size of the call is subject to the API call limits for the datastore. The entity count is also subject to a limit. In Java, you can save entities using the put() method of a DatastoreService instance. As with get(), the method takes a single Entity for a single put, or an Iterable<Entity> for a batch put. When the call to put() returns, the datastore is up to date, and all future queries in the current request handler and other handlers will see the new data. The specifics of how the datastore gets updated are discussed in detail in Chapter 6. Deleting Entities Deleting entities works similarly to putting entities. In Python, you can call the delete() method on the entity object, or you can pass entity objects or Key objects to the delete() function. e = db.get('Entity', 'alphabeta') e.delete() db.delete(e) db.delete([e1, e2, e3]) # Deleting without first fetching the entity: k = db.Key('Entity', 'alphabeta') db.delete(k) In Java, you call the delete() method of the DatastoreService with either a single Key or an Iterable<Key>. As with gets and puts, a delete of multiple entities occurs in a single batch call to the service, and is faster than making multiple service calls. Using Entities | 119
- 142. CHAPTER 5 Datastore Queries Inevitably, an application that manages data must do more than store and retrieve that data one record at a time. It must also answer questions about that data: which records meet certain criteria, how records compare to one another, what a set of records rep- resents in aggregate. Web applications in particular are expected not only to know the answers to questions about large amounts of data, but to provide them quickly in re- sponse to web requests. Most database systems provide a mechanism for executing queries, and the App Engine datastore is no exception. But App Engine’s technique differs significantly from that of traditional database systems. When the application asks a question, instead of rifling through the original records and performing calculations to determine the answer, App Engine simply finds the answer in a list of possible answers prepared in advance. App Engine can do this because it knows which questions are going to be asked. This kind of list, or index, is common to many database technologies, and some rela- tional databases can be told to maintain a limited set of indexes to speed up some kinds of queries. But App Engine is different: it maintains an index for every query the ap- plication is going to perform. Since the datastore need only do a simple scan of an index for every query, the application gets results back quickly. And for large amounts of data, App Engine can spread the data and the indexes across many machines, and get results back from all of them without an expensive aggregate operation. This indexing strategy has significant drawbacks. The datastore’s built-in query engine is downright weak compared to some relational databases, and is not suited to sophis- ticated data processing applications that would prefer slow but powerful runtime quer- ies to fast simple ones. But most web applications need fast results, and the dirty secret about those powerful query engines is that they can’t perform at web speeds with large amounts of data distributed across many machines. App Engine uses a model suited to scalable web applications: calculate the answers to known questions when the data is written, so reading is fast. 121
- 143. In this chapter,we will explain how queries and indexes work, how the developer tools help you configure indexes automatically, and how to manage indexes as your appli- cation evolves. By understanding indexes, you will have an intuition for how to design your application and your data to make the most of the scalable datastore. Queries and Kinds You’ve seen how to retrieve an entity from the datastore given its key. But in most cases, the application does not know the keys of the entities it needs; it has only a general idea that it needs entities that meet certain criteria. For example, a leader board for the game app would need to retrieve the 10 Player entities with the highest score property values. To retrieve entities this way, the app performs a query. A query includes: • The kind of the entities to query • Zero or more filters, criteria that property values must meet for an entity to be returned by the query • Zero or more sort orders that determine the order in which results are returned based on property values A query based on property values can only return entities of a single kind. This is the primary purpose of kinds: to determine which entities are considered together as pos- sible results for a query. In practice, kinds correspond to the intuitive notion that each entity of the same nominal kind represents the same kind of data. But unlike other database systems, it’s up to the app to enforce this consistency if it is desired, and the app can diverge from it if it’s useful. It is also possible to perform a limited set of queries on entities regardless of kind. Kindless queries can use a filter on the ID or key name, or on ancestors. We’ll discuss ancestors and kindless queries in Chapter 6. Query Results and Keys When retrieving results for a query, the datastore returns the full entity for each result to the application. There is no way to query for a subset of the properties, as you might with fields in a relational database. For large entities, this may mean more data is transmitted between the datastore and the app than is needed. If you have objects with large property values or very many properties and you only need to access a subset of this data most of the time, you can speed up access by storing the data across multiple entities. One entity stores the oft- accessed properties and properties used in queries, as well as a reference to another entity carrying the larger data (the key of the other entity). When the app needs the larger data, it queries for the small entity, then follows the reference to the large entity. 122 | Chapter 5: Datastore Queries
- 144. The datastore canreturn just the keys for the entities that match your query instead of fetching the full entities. A keys-only query is useful for determining the entities that match query criteria separately from when the entities are used. Keys can be remem- bered in the memcache or in a datastore property, and vivified as full entities when needed. We’ll look at keys-only queries when we discuss each query API. GQL The Python and Java runtime environments provide several ways to formulate queries. They all do the same thing: they ask the datastore to return the entities (or keys for entities) whose keys and properties meet the given filter criteria, returned in an order determined by sorting the given properties. One way to formulate a query is with GQL. GQL is a text-based query language that is intended to resemble SQL, the query language of relational databases. It supports only the features of the datastore’s query engine, and therefore lacks many features common to SQL. But it is expressive and concise enough to be useful for datastore queries. Say we have entities of the kind Player representing players in an online role-playing game. The following GQL query retrieves all Player entities whose “level” property is an integer between 5 and 20, sorted by level in ascending order, then by score in de- scending order: SELECT * FROM Player WHERE level > 5 AND level < 20 ORDER BY level ASC, score DESC You can use GQL to browse the contents of the datastore of your live application using the Administration Console. To do this, select Data Viewer from the sidebar. You can browse entities by kind, or enter a GQL query. You can also create new entities through the Console, though new entities are limited to the kinds, properties, and value types of existing entities. You can also edit the properties of an entity by clicking on its ID. Figure 5-1 shows the results of running a GQL query in the Console. You can only perform a query in the Administration Console if the query does not require a custom index, or if the app already has the requisite index for the query. We discuss indexes in depth later in this chapter. Python apps can also perform queries using GQL from code. We’ll look at this API in a moment. The Java datastore API does not yet support GQL. As you’ll see later, Java apps using JDO or JPA have other textual query languages at their disposal. GQL | 123
- 145. Figure 5-1. TheData Viewer panel of the Administration Console The first part of the SQL-like syntax in the GQL query above, SELECT * FROM Player, says that the datastore should return complete entities as results, and that it should only consider entities of the kind Player. This is the most common kind of query. Despite its similarity with SQL, GQL can only represent queries, and cannot perform updates, inserts, or deletes. In other words, every GQL query begins with SELECT. The rest of the query syntax also resembles SQL, and translates directly to filters and sort orders. The WHERE clause, if present, represents one or more filter conditions, sep- arated by AND. The ORDER BY clause, if present, represents one or more sort orders, separated by commas, applied from left to right. Each condition in a WHERE clause consists of the name of a property, a comparison operator, and a value. You can also specify a condition that matches an entity’s ID or key name by using the reserved name __key__ (that’s two underscores, the word “key,” and two underscores) like a property name. GQL does not support expressions fancier than this: the lefthand side can only be a property name or __key__, and the righthand side can only be a simple value. The datastore query engine supports five comparison operators for filters: =, <, <=, >, and >=. GQL also supports two additional operators: !=, meaning “not equal to,” and IN, which tests that the value equals any in a set of values. 124 | Chapter 5: Datastore Queries
- 146. For IN, thevalues are represented by a comma-delimited list of values surrounded by parentheses: SELECT * FROM Player WHERE level IN (5, 6, 7) Internally, GQL translates the != and IN operators into multiple datastore queries that have the same effect. If a query contains an IN clause, the query is evaluated once for each value in the clause, using an = filter. The results of all of the queries are aggregated. The != operator performs the query once using a < filter and once using a >, then ag- gregates the results. Using these operators more than once in a GQL statement requires a query for each combination of the required queries, so be judicious in their use. The WHERE clause is equivalent to one or more filters. It is not like SQL’s WHERE clause, and does not support arbitrary logical expressions. In particular, it does not support testing the logical-OR of two conditions. The value on the righthand side of a condition can be a literal value that appears inside the query string. Seven of the datastore value types have string literal representations, as shown in Table 5-1. Table 5-1. GQL value literals for datastore types Type Literal syntax Examples String Single-quoted string; escape the quote by doubling it 'Haven''t You Heard' Integer or Sign, digits; float uses decimal point -7 3.14 float Boolean True or false keywords TRUE FALSE Date-time, Type, and value as numbers or a string DATETIME(1999, 12, 31, 23, 59, 59) DATETIME(year, month, day, hour, minute, second) DATETIME('1999-12-31 23:59:59') date or time DATE(1999, 12, 31) DATETIME('YYYY-MM-DD HH:MM:SS') DATE(year, month, day) DATE('1999-12-31') DATE('YYYY-MM-DD') TIME(23, 59, 59) TIME(hour, minute, second) TIME('23:59:59') TIME('HH:MM:SS') Entity key Entity kind, and name or ID; can be a path KEY('Player', 1287) KEY('kind', 'name'/id) KEY('kind', 'name'/id, 'kind', 'name'/id, ...) User object User('email-address') User('edward@example.com') GeoPt GEOPT(lat, long) GEOPT(37.4219, -122.0846) object As we’ll see later, the type of a filter’s value is significant in a datastore query. If you use the wrong type for a value, such as by specifying a 'string_key' when you meant to specify a KEY('kind', 'key_name'), your query may not return the results you intend, or may unexpectedly return no results. GQL | 125
- 147. A GQL querycan specify a LIMIT, a maximum number of results to return. A query can also specify an OFFSET, a number of results to skip before returning the remaining results. The following example returns the third, fourth, and fifth results, and may return fewer if the query has fewer than five results: SELECT * FROM Player LIMIT 3 OFFSET 2 GQL keywords, shown here in uppercase as is traditional with SQL, are case-insensitive. Kind and property names, however, are case-sensitive. The datastore can return results as complete entities, or just the keys for those entities. Unlike SQL’s ability to select specific columns from tables, the datastore cannot return a subset of an entity’s properties, nor can it “join” properties from entities of multiple kinds into a single set of results. In GQL, SELECT * says to return full entities. SELECT __key__ says to return just the keys. The Python Query API Recall that the Python datastore API represents entities using objects of classes named after kinds. Example 5-1 shows Python code that creates three Player entities for an online role-playing game. Example 5-1. Python code to create several entities of the kind Player from google.appengine.ext import db import datetime class Player(db.Expando): pass player1 = Player(name='wizard612', level=1, score=32, charclass='mage', create_date=datetime.datetime.now()) player1.put() player2 = Player(name='druidjane', level=10, score=896, charclass='druid', create_date=datetime.datetime.now()) player2.put() player3 = Player(name='TheHulk', level=7, score=500, charclass='warrior', create_date=datetime.datetime.now()) player3.put() 126 | Chapter 5: Datastore Queries
- 148. Once again, we’lluse Expando to keep the examples simple. As we start talking about queries, the importance of using a consistent layout, or schema, for entities of a kind will become apparent. The Python API provides two ways to formulate queries, one using an object-oriented interface and one based on GQL. The Query Class The first way to formulate a query is with an instance of the Query class. A Query object can be constructed in one of two ways, with equivalent results: q = db.Query(Player) q = Player.all() In both cases, q is assigned a new Query instance that represents all entities of the kind Player. The query is not executed right away; right now it’s just a question waiting to be asked. Without filters or sort orders, the object represents a query for all objects of the given kind. To apply a filter to the query, you call the filter() method on the Query object. It takes two arguments. The first argument is a string containing the name of a property, a space character, and a comparison operator. The second argument is the value for the com- parison, of a type appropriate for the situation. q.filter("level >", 5) You specify multiple filters by calling the filter() method multiple times. An entity must meet all filter criteria in order to be a result for the query. That is, filters have a logical-AND relationship with one another. q.filter("level >", 5) q.filter("level <", 20) For convenience, the filter() method returns the Query object, so you can chain mul- tiple calls to filter() in a single line: q.filter("level >", 5).filter("level <", 20) The filter() method supports the equality (=) operator, and the four inequality oper- ators (<, <=, >, and >=). The Query class also supports the != (not-equal) and IN operators from GQL, and does so in the same way, using multiple datastore queries and aggre- gating the results. To apply a sort order, you call the order() method. This method takes one argument, a string containing the name of a property to sort by. If the string begins with a hyphen character (-), the sort will be in descending order; otherwise, the sort will be in as- cending order. Like filter(), order() returns the Query object, for chaining. q.order("-score") The Python Query API | 127
- 149. The datastore cansort query results by multiple properties. First, it sorts by the property and order from the first call to order(), then from the next call, and so on. For example, the following sorts first by level ascending, then by score descending: q.order("level").order("-score") The query engine supports limiting the number of results returned, and skipping ahead a number of results. You specify the limit and offset when you retrieve the results, described later. GQL in Python The Python environment includes a rich API for preparing queries using GQL. In ad- dition to the pure textual syntax, the Python API supports additional syntax for para- meterized substitution of filter values. To use GQL from Python, you instantiate an instance of the GqlQuery class with the text of the query: q = db.GqlQuery("SELECT * FROM Player WHERE level > 5 AND level < 20 ORDER BY level ASC, score DESC") You can also instantiate a GqlQuery from the kind class directly, using its gql() class method. When using this method, omit the SELECT * FROM Kind from the string, since this is implied by the use of the method: q = Player.gql("WHERE level > 5 AND level < 20 ORDER BY level ASC, score DESC") You can specify values for conditions using parameter substitution. With parameter substitution, the query string contains a placeholder with either a number or a name, and the actual value is passed to GqlQuery() or gql() as either a positional argument or a keyword argument. q = db.GqlQuery("SELECT * FROM Player WHERE level > :1 AND level < :2", 5, 20) q = db.GqlQuery("SELECT * FROM Player WHERE level > :min_level AND level < :max_level", min_level=5, max_level=20) One advantage to parameter substitution is that each argument value is of the appro- priate datastore type, so you don’t need to bother with the string syntax. For the da- tastore types that do not have string literal syntax (mostly the Google Data value types mentioned in Table 4-1), parameter substitution is the only way to use values of those types in a GQL query. 128 | Chapter 5: Datastore Queries
- 150. You can rebindnew values to a GQL query using parameter substitution after the GqlQuery object has been instantiated using the bind() method. This means you can reuse the GqlQuery object for multiple queries that have the same structure but different values. This saves time because the query string only needs to be parsed once. The bind() method takes the new values as either positional arguments or keyword arguments. You can save more time by caching parameterized GqlQuery objects in global variables. This way, the query string is parsed only when a server loads the application for the first time, and the application reuses the object for all subsequent calls handled by that server. _LEADERBOARD_QUERY = db.GqlQuery( "SELECT * FROM Player WHERE level > :min_level AND level < :max_level ORDER BY level ASC, score DESC") class LeaderboardHandler(webapp.RequestHandler): def get(self): _LEADERBOARD_QUERY.bind(min_level=5, max_level=20) # ... Retrieving Results Once you have a Query or GqlQuery object configured with filters, sort orders, and value bindings (in the case of GqlQuery), you can execute the query using one of several methods. The query is not executed until you call one of these methods, and you can call these methods repeatedly on the same query object to reexecute the query and get new results. The fetch() method returns a number of results, up to a specified limit. fetch() returns a list of entity objects, which are instances of the kind class. q = db.Query(Player).order('-score') results = q.fetch(10) When serving web requests, it’s always good to limit the amount of data the datastore might return, so an unexpectedly large result set doesn’t cause the request handler to exceed its deadline. This is so often the case that App Engine imposes a maximum limit of 1,000 results for all queries, which is the default if no smaller limit is given. The fetch() method also accepts an optional offset parameter. If provided, the method skips that many results, then returns subsequent results, up to the specified limit. q = db.Query(Player).order('-score') results = q.fetch(10, offset=20) The Python Query API | 129
- 151. For a GqlQuery,the fetch result limit and offset are equivalent to the LIMIT and OFFSET that can be specified in the query itself. The arguments to fetch() override those speci- fied in the GQL statement. The limit argument to fetch() is required, so it always overrides a GQL LIMIT clause. In order to perform a fetch with an offset, the datastore must collect the complete result set from all of its servers, then scan down the compiled list to the offset. The amount of time this takes depends on the size of the offset. To prevent the transient result set from getting too large and the query from taking too long, App Engine imposes the same maximum on the offset as it does on the limit: 1,000 results. This makes fetch offsets unsuitable for retrieving records in batches, such as for a paginated display. Later pages will take much longer to retrieve results than earlier pages, and when the offset gets too large, pages won’t display at all. If you’re used to using a single-server data- base’s result offset feature to power paginated displays, you’ll want to consider another strategy. One viable strategy is to maintain a property that can be used with an inequality filter to find the entity at the top of each page. As you’ll see later in this chapter, index scans are much more efficient than result offsets, and can operate on very large data sets. You can also perform inequality queries on entity keys. It’s common for an app to perform a query expecting to get just one result, or nothing. You can do this with fetch(1), then test whether the list it returns is empty. For con- venience, the query object also provides a get() method, which returns either the first result, or the Python value None if the query did not return any results. q = db.Query(Player).filter('name =', 'django97') player = q.get() if player: # ... The count() method executes the query, then returns the number of results that would be returned instead of the results themselves. count() must perform the query to count the results, and so it is subject to the same 1,000 result maximum as fetch(). count() accepts a limit as a parameter, and providing one causes the method to return imme- diately once that many results have been counted. q = db.Query(Player).filter('level >', 10) if q.count(100) == 100: # 100 or more players are above level 10. Query objects provide one other mechanism for retrieving results, and it’s quite pow- erful. In the Python API, query objects are iterable. If you use the object in a context that accepts iterables, such as in a for loop, the object executes the query, then provides 130 | Chapter 5: Datastore Queries
- 152. a Python standarditerator that returns each result one at a time, starting with the first result. q = db.Query(Player).order('-score') for player in q: # ... Unlike fetch(), the iterator interface does not impose a maximum limit. As the app uses the iterator, the iterator fetches the results in small batches. The iterator could return results beyond 1,000, though in practice, the app request handler will probably reach its deadline before that happens. It’s up to the app to stop requesting results from the iterator (to break from the for loop) when it has had enough. If you want to use the iterator interface for fetching results but also want to use a limit or an offset, you can specify the limit or offset in a GQL query: q = db.GqlQuery("SELECT * FROM Player LIMIT 10 OFFSET 5") # Iterate over Players 5 through 10. for player in q: # ... You can pass the iterator to something that expects one, such as a templating engine, and the iterator will honor the limit and offset specified in the GQL query. Keys-Only Queries As we mentioned earlier, instead of returning complete entities, the datastore can return just the keys of the entities that match the query. In the Python query API, you can query for keys using either the Query interface or the GqlQuery interface. To request just the keys with Query, provide the keys_only argument set to True to the Query constructor: q = db.Query(Player, keys_only=True) To specify a keys-only query using GQL, begin the GQL query with SELECT __key__ instead of SELECT *: q = db.GqlQuery('SELECT __key__ FROM Player') When performing a keys-only query, each result returned by the query object is a Key object instead of an instance of the model class. q = db.Query(Player, keys_only=True) for result_key in q: # result_key is a Key... There is no way to perform keys-only queries using the Model class methods all() and gql(). This makes sense: the return values of the Model methods represent collections of instances of the model class, not keys. The Python Query API | 131
- 153. The Java QueryAPI If you are using the JPA interface or the JDO interface, you will use the query facilities of those interfaces to perform datastore queries: JPQL or JDOQL, respectively. The concepts of those interfaces map nicely to the concepts of datastore queries: a query has a kind (a class), filters, and sort orders. We’ll look at the calling conventions for JPQL when we look at JPA in Chapter 8. Naturally, the low-level Java datastore API includes a query interface as well. Here is a brief example: import com.google.appengine.api.datastore.DatastoreService; import com.google.appengine.api.datastore.DatastoreServiceFactory; import com.google.appengine.api.datastore.Entity; import com.google.appengine.api.datastore.PreparedQuery; import com.google.appengine.api.datastore.Query; // ... DatastoreService ds = DatastoreServiceFactory.getDatastoreService(); Query q = new Query("Book"); q.addFilter("copyrightYear", Query.FilterOperator.LESS_THAN_OR_EQUAL, 1950); q.addSort("title"); PreparedQuery pq = ds.prepare(q); for (Entity result : pq.asIterable()) { String title = result.getProperty("title"); // ... } To perform a query, you instantiate the Query class (from the com.google.appengine.api.datastore package), providing the name of the kind of the entities to query as an argument to the constructor. You call methods on the query object to add filters and sort orders. To perform the query, you pass the query object to a method of the DatastoreService instance. This method returns a PreparedQuery object, which you can manipulate to retrieve the results. The query is not actually performed until you attempt to access results using the PreparedQuery object. If you access the results using an iterator via the asIterable() or asIterator() methods, the act of iterating causes the API to fetch the results in batches. When these methods are called without arguments, the resulting iterator will keep going until all results for the query have been returned. If a query has a large number of results, this may take longer than the time allowed for the request. The asIterable() and asIterator() methods accept an optional argument, a FetchOptions object, that controls which results are returned. Options can include an offset, a number of results to skip prior to returning any, and a limit, a maximum number of results to return. FetchOptions uses a builder-style interface, as follows: 132 | Chapter 5: Datastore Queries
- 154. import com.google.appengine.api.datastore.Entity; import com.google.appengine.api.datastore.FetchOptions; import com.google.appengine.api.datastore.PreparedQuery; import com.google.appengine.api.datastore.Query; // ... // Query q = ... PreparedQuery pq = ds.prepare(q); Iterable<Entity> results = pq.asIterable(FetchOptions.Builder.limit(10).offset(20)); for (Entity result : results) { String title = result.getProperty("title"); // ... } This tells the datastore to skip the first 20 results, and return up to the next 10 results (if any). As with the Python interface, offsets require calculation within the service after the results have been gathered from across the datastore. This means the performance of an offset is proportional to the size of the offset, which makes offsets unsuitable for rendering paginated lists. See the similar note in the description of the Python query API, within the section “Retrieving Results” on page 129. Instead of fetching results in batches, you can get all results in a list by calling the asList() method of the PreparedQuery class. Unlike the iterator interface, which gets results in batches, this method retrieves all results with a single service call. The method requires that a limit be specified using FetchOptions, up to the maximum of 1,000 entities. If a query is likely to have only one result, or if only the first result is desired, calling the asSingleEntity() method retrieves the result and returns an Entity object, or null. If you just want a count of the results and not the entities themselves, you can call the countEntities() method of the PreparedQuery. Because the datastore has to perform the query to get the count, the speed of this call is proportional to the count, though faster than actually fetching the results by a constant factor. Like an asList() fetch, the count is also limited to a maximum of 1,000 entities. Keys-Only Queries in Java You can fetch just the keys for the entities that match a query instead of the full entities using the low-level Java datastore API. To declare that a query should return just the keys, call the setKeysOnly() method on the Query object: The Java Query API | 133
- 155. Query q =new Query("Book"); q.setKeysOnly(); When a query is set to return only keys, the results of the query are Entity objects without any properties set. You can get the key from these objects using the getKey() method: PreparedQuery pq = ds.prepare(q); for (Entity result : pq.asIterable()) { Key k = result.getKey(); // ... } You can also perform keys-only queries using the JDO and JPA interfaces. See Chap- ter 8. Introducing Indexes For every query an application performs, App Engine maintains an index, a single table of possible answers for the query. Specifically, it maintains an index for a set of queries that use the same filters and sort orders, possibly with different values for the filters. Consider the following simple query: SELECT * FROM Player WHERE name = 'druidjane' To perform this query, App Engine uses an index containing the keys of every Player entity and the value of each entity’s name property, sorted by the name property values in ascending order. Such an index is illustrated in Figure 5-2. Figure 5-2. An index of Player entity keys and “name” property values, sorted by name in ascending order, with the result for WHERE name = ‘druidjane’ 134 | Chapter 5: Datastore Queries
- 156. To find allentities that meet the conditions of the query, App Engine finds the first row in the index that matches, then it scans down to the first row that doesn’t match. It returns the entities mentioned on all rows in this range (not counting the nonmatching row), in the order they appear in the index. Because the index is sorted, all results for the query are guaranteed to be on consecutive rows in the table. App Engine would use this same index to perform other queries with a similar structure but different values, such as the following query: SELECT * FROM Player WHERE name = 'duran89' This query mechanism is fast, even with a very large number of entities. Entities and indexes are distributed across multiple machines, and each machine scans its own index in parallel with the others. Each machine returns results to App Engine as it scans its own index, and App Engine delivers the final result set to the app, in order, as if all of the results were in one large index. Another reason queries are fast has to do with how the datastore finds the first matching row. Because indexes are sorted, the datastore can use an efficient algorithm to find the first matching row. In the common case, finding the first row takes approximately the same amount of time regardless of the size of the index. In other words, the speed of a query is not affected by the size of the data set. App Engine updates all relevant indexes when property values change. In this example, if an application retrieves a Player entity, changes the name, then saves the entity with a call to the put() method, App Engine updates the appropriate row in the previous index. It also moves the row if necessary so the ordering of the index is preserved. The call to put() does not return until all appropriate indexes are updated. Similarly, if the application creates a new Player entity with a name property, or deletes a Player entity with a name property, App Engine updates the index. In contrast, if the application updates a Player but does not change the name property, or creates or deletes a Player that does not have a name property, App Engine does not update the name index because no update is needed. App Engine maintains two indexes like the previous example above for every property name and entity kind, one with the property values sorted in ascending order and one with values in descending order. App Engine also maintains an index of entities of each kind. These indexes satisfy some simple queries, and App Engine also uses them in- ternally for bookkeeping purposes. For other queries, you must tell App Engine which indexes to prepare. You do this using a configuration file, which gets uploaded along with your application’s code. For Python apps, this file is named index.yaml. For Java, this file is WEB-INF/datastore- indexes.xml. It’d be a pain to write this file by hand, but thankfully you don’t have to. While you’re testing your application using the development web server from the SDK, when the app performs a datastore query, the server checks that the configuration file has an Introducing Indexes | 135
- 157. appropriate entry forthe needed index. If it doesn’t find one, it adds one. As long as the app performs each of its queries at least once during testing, the resulting config- uration file will be complete. The index configuration file must be complete, because when the app is running on App Engine, if the application performs a query for which there is no index, the query returns an error. You can tell the development web server to behave similarly if you want to test for these error conditions. (How to do this depends on which SDK you are using; see “Configuring Indexes” on page 159.) Indexes require a bit of discipline to maintain. While the development tools can help add index configuration, they cannot know when an index is unused and can be deleted from the file. Extra indexes consume storage space and slow down updates of properties mentioned in the index. And while the version of the app you’re developing may not need a given index, the version of the app still running on App Engine may still need it. The App Engine SDK and the Administration Console include tools for inspecting and maintaining indexes. We’ll look at these tools in Chapter 15. Before we discuss index configuration, let’s look more closely at how indexes support queries. We just saw an example where the results for a simple query appear on con- secutive rows in a simple index. In fact, this is how most queries work: the results for every query that would use an index appear on consecutive rows in the index. This is both surprisingly powerful in some ways and surprisingly limited in others, and it’s worth understanding why. Automatic Indexes and Simple Queries As we mentioned, App Engine maintains two indexes for every single property of every entity kind, one with values in ascending order and one with values in descending order. App Engine builds these indexes automatically, whether or not they are mentioned in the index configuration file. These automatic indexes satisfy the following kinds of queries using consecutive rows: • A simple query for all entities of a given kind, no filters or sort orders • One filter on a property using the equality (=) operator • Filters using greater-than or less-than operators (>, >=, <, <=) on a single property • One sort order, ascending or descending, and no filters, or with filters only on the same property used with the sort order • Filters or a sort order on the entity key • Kindless queries with or without key filters Let’s look at each of these in action. 136 | Chapter 5: Datastore Queries
- 158. All Entities ofa Kind The simplest datastore query asks for every entity of a given kind, in any order. Stated in GQL, a query for all entities of the kind Player looks like this: SELECT * FROM Player App Engine maintains an index mapping kinds to entity keys. This index is sorted using a deterministic ordering for entity keys, so this query returns results in “key order.” The kind of an entity cannot be changed after it is created, so this index is updated only when entities are created and deleted. Since a query can only refer to one kind at a time, you can imagine this index as simply a list of entity keys for each kind. Figure 5-3 illustrates an example of this index. Figure 5-3. An index of all Player entity keys, with results for SELECT * FROM Player When the query results are fetched, App Engine uses the entity keys in the index to find the corresponding entities, and returns the full entities to the application. One Equality Filter Consider the following query, which asks for every Player entity with a level property with a value of the integer 10: SELECT * FROM Player WHERE level = 10 This query uses an index of Player entities with the level property, ascending—one of the automatic indexes. It uses an efficient algorithm to find the first row with a level equal to 10. Then it scans down the index until it finds the first row with a level not equal to 10. The consecutive rows from the first matching to the last matching represent all of the Player entities with a level property equal to the integer 10. This is illustrated in Figure 5-4. Automatic Indexes and Simple Queries | 137
- 159. Figure 5-4. Anindex of the Player entity “level” properties, sorted by level then by key, with results for WHERE level = 10 Greater-Than and Less-Than Filters The following query asks for every Player entity with a score property whose value is greater than the integer 500: SELECT * FROM Player WHERE score > 500 This uses an index of Player entities with the score property, ascending, also an auto- matic index. As with the equality filter, it finds the first row in the index whose score is greater than 500. In the case of greater-than, since the table is sorted by score in ascending order, every row from this point to the bottom of the table is a result for the query. See Figure 5-5. Similarly, consider a query that asks for every Player with a score less than 1,000: SELECT * FROM Player WHERE score < 1000 App Engine uses the same index (score, ascending), and the same strategy: it finds the first row that matches the query, in this case the first row. Then it scans to the next row that doesn’t match the query, the first row whose score is greater than or equal to 1000. The results are represented by everything above that row. Finally, consider a query for score values between 500 and 1,000: SELECT * FROM Player WHERE score > 500 AND score < 1000 138 | Chapter 5: Datastore Queries
- 160. Figure 5-5. Anindex of the Player entity “score” properties, sorted by “score” then by key, with results for WHERE score > 500 Once again, the same index and strategy prevail: App Engine scans from the top down, finding the first matching and next nonmatching rows, returning the entities represen- ted by everything in between. This is shown in Figure 5-6. If the values used with the filters do not represent a valid range, such as score < 500 AND score > 1000, the query planner notices this and doesn’t bother performing a query, since it knows the query has no results. One Sort Order The following query asks for every Player entity, arranged in order by level, from lowest to highest: SELECT * FROM Player ORDER BY level As before, this uses an index of Player entities with level properties in ascending order. If both this query and the previous equality query were performed by the application, both queries would use the same index. This query uses the index to determine the order in which to return Player entities, starting at the top of the table and moving down until the application stops fetching results, or until the bottom of the table. Recall that every Player entity with a level property is mentioned in this table. See Figure 5-7. Automatic Indexes and Simple Queries | 139
- 161. Figure 5-6. Anindex of the Player entity “score” properties, sorted by score, with results for WHERE score > 500 AND score < 1000 Figure 5-7. An index of the Player entity “level” properties sorted by level in ascending order, with results for ORDER BY level 140 | Chapter 5: Datastore Queries
- 162. The following queryis similar to the previous one, but asks for the entities arranged by level from highest to lowest: SELECT * FROM Player ORDER BY level DESC This query cannot use the same index as before, because the results are in the wrong order. For this query, the results should start at the entity with the highest level, so the query needs an index where this result is in the first row. App Engine provides an automatic index for single properties in descending order for this purpose. See Fig- ure 5-8. Figure 5-8. An index of the Player entity “level” properties sorted by level in descending order, with results for ORDER BY level DESC If a query with a sort order on a single property also includes filters on that property, and no other filters, App Engine still needs only the one automatic index to fulfill the query. In fact, you may have noticed that for these simple queries, the results are re- turned sorted by the property in ascending order, whether or not the query specifies the sort order explicitly. In these cases, the ascending sort order is redundant. Queries on Keys In addition to filters and sort orders on properties, you can also perform queries with filters and sort orders on entity keys. You can refer to an entity’s key in a filter or sort order using the special name __key__. An equality filter on the key isn’t much use. Only one entity can have a given key, and if the key is known, it’s faster to perform a get() than a query. But an inequality filter on the key is especially useful, and in fact it’s a good way to perform a common task: to fetch all of the entities of a given kind in batches. Automatic Indexes and Simple Queries | 141
- 163. Notice that givenan entity key and an index of keys sorted by key, an app can fetch a batch of entities using an inequality filter. Python code to fetch a batch in this way might look like this: # ... # next_key is the db.Key object of the first entity in the batch. if next_key: q = Player.gql("WHERE __key__ >= :1", next_key) else: q = Player.all() players = q.fetch(11) if len(players) == 11: # Remove the last entity, then use its key # as the start of the next batch. next_key = players.pop().key else: next_key = None # Process the players, and give next_key to the # client for the next call. # ... The query in this example uses the built-in index of keys in ascending order. App Engine provides an automatic indexes of kinds and keys, sorted by key in ascending order. The query returns the results sorted in key order. This order isn’t useful for display purposes, but it’s deterministic, making it useful for processing entities in batches. A query that sorts keys in descending order requires a custom index. App Engine uses indexes for filters on keys in the same way as filters on properties, with a minor twist: a query using a key filter in addition to other filters can use an automatic index if a similar query without the key filter could use an automatic index. Automatic indexes for properties already include the keys, so such queries can just use the same indexes. And of course, if the query has no other filters beyond the key filter, it can use the automatic key index. Kindless Queries In addition to performing queries on entities of a given kind, the datastore lets you perform a limited set of queries on entities of all kinds. Kindless queries cannot use filters or sort orders on properties. They can, however, use equality and inequality filters on keys (IDs or names). Kindless queries are mostly useful in combination with ancestors, which we’ll discuss in Chapter 6. They can also be used to get every entity in the datastore, using an in- equality filter on keys and a limit to get them a batch at a time. 142 | Chapter 5: Datastore Queries
- 164. Using the PythonQuery class, you perform a kindless query by omitting the model class argument from the constructor: q = db.Query() q.filter('__key__ >', last_key) Similarly in the Java low-level API, you perform a kindless query by instantiating the Query class using the no-argument constructor. In GQL, you specify a kindless query by omitting the FROM Kind part of the statement: q = db.GqlQuery('SELECT * WHERE __key__ > :1', last_key) The results of a kindless query are returned in key order, ascending. Kindless queries use an automatic index. Custom Indexes and Complex Queries All queries not covered by the automatic indexes must have a corresponding index defined in the app’s index configuration file. We’ll refer to these as “custom indexes,” in contrast with “automatic indexes.” App Engine needs these hints because building every possible index for every combination of property and sort order would take a gargantuan amount of space and time, and an app isn’t likely to need more than a fraction of those possibilities. In particular, the following queries require custom indexes: • A query with multiple sort orders • A query with an inequality filter on a property and filters on other properties A query that uses just equality filters on properties does not need a custom index in most cases thanks to a specialized query algorithm for this case, which we’ll look at in a moment. Also, filters on keys do not require custom indexes; they can operate on whatever indexes are used to fulfill the rest of the query. Let’s examine these queries and the indexes they require. Multiple Sort Orders The automatic single-property indexes provide enough information for one sort order. When two entities have the same value for the sorted property, the entities appear in the index in adjacent rows, ordered by their entity keys. If you want to order these entities using other criteria, you need an index with more information. The following query asks for all Player entities, sorted first by the level property in descending order, then, in the case of ties, sorted by the score property in descending order: SELECT * FROM Player ORDER BY level DESC, score DESC Custom Indexes and Complex Queries | 143
- 165. The index thisquery needs is straightforward: a table of Player entity keys, level values, and score values, sorted according to the query. This is not one of the indexes provided by the datastore automatically, so it is a custom index, and must be mentioned in the index configuration file. If you performed this query in the Python development web server, the server would add the following lines to the index.yaml file: - kind: Player properties: - name: level direction: desc - name: score direction: desc The order the properties appear in the configuration file matters. This is the order in which the rows are sorted: first by level descending, then by score descending. This configuration creates the index shown in Figure 5-9. The results appear in the table, and are returned for the query in the desired order. Figure 5-9. An index of the Player entity “level” and “score” properties, sorted by level descending, then score descending, then by key ascending Filters on Multiple Properties Consider the following query, which asks for every Player with a level less than the integer 10 and a charclass of the string 'mage': SELECT * FROM Player WHERE charclass='mage' AND level < 10 To be able to scan to a contiguous set of results meeting both filter criteria, the index must contain columns of values for these properties. The entities must be sorted first by charclass, then by level. 144 | Chapter 5: Datastore Queries
- 166. For Python, theindex configuration for this query would appear as follows in the index.yaml file: - kind: Player properties: - name: charclass direction: asc - name: level direction: asc This index is illustrated in Figure 5-10. Figure 5-10. An index of the Player entity “charclass” and “level” properties, sorted by charclass, then level, then key, with results for WHERE charclass = “mage” AND level < 10 The ordering sequence of these properties is important! Remember: the results for the query must all appear on adjacent rows in the index. If the index for this query were sorted first by level then by charclass, it would be possible for valid results to appear on nonadjacent rows. Figure 5-11 demonstrates this problem. The index ordering requirement for combining inequality and equality filters has sev- eral implications that may seem unusual when compared to the query engines of other databases. Heck, they’re downright weird. The first implication, illustrated previously, can be stated generally: The First Rule of Inequality Filters: If a query uses inequality filters on one property and equality filters on one or more other properties, the index must be ordered first by the properties used in equality filters, then by the property used in the inequality filters. Custom Indexes and Complex Queries | 145
- 167. Figure 5-11. Anindex of the Player entity “charclass” and “level” properties, sorted first by level then by charclass, which cannot satisfy WHERE charclass = “mage” AND level < 10 with consecutive rows This rule has a corollary regarding queries with both an inequality filter and sort orders. Consider the following possible query: SELECT * FROM Player WHERE level < 10 ORDER BY score DESC What would the index for this query look like? For starters, it would have a column for the level, so it can select the rows that match the filter. It would also have a column for the score, to determine the order of the results. But which column is ordered first? The First Rule implies that level must be ordered first. But the query requested that the results be returned sorted by score, descending. If the index were sorted by score, then by level, the rows may not be adjacent. To avoid confusion, App Engine requires that the correct sort order be stated explicitly in the query: SELECT * FROM Player WHERE level < 10 ORDER BY level, score DESC In general: The Second Rule of Inequality Filters: If a query uses inequality filters on one property and sort orders of one or more other properties, the index must be ordered first by the property used in the inequality filters (in either direction), then by the other desired sort orders. To avoid confusion, the query must state all sort orders explicitly. 146 | Chapter 5: Datastore Queries
- 168. There’s one lastimplication to consider with regard to inequality filters. The following possible query attempts to get all Player entities with a level less than 10 and a score less than 500: SELECT * FROM Player WHERE level < 10 AND score < 500 Consider an index ordered first by level, then by score, as shown in Figure 5-12. Figure 5-12. Neither possible index of the Player entity “level” and “score” properties can satisfy WHERE level < 10 AND score < 500 with consecutive rows In fact, there is no possible index that could satisfy this query completely using con- secutive rows. This is not a valid App Engine datastore query. The Third Rule of Inequality Filters: A query cannot use inequality filters on more than one property. A query can use multiple inequality filters on the same property, such as to test for a range of values. Multiple Equality Filters For queries using just equality filters, it’s easy to imagine custom indexes that satisfy them. For instance: SELECT * FROM Player WHERE charclass='mage' AND level=10 A custom index containing these properties, ordered in any sequence and direction, would meet the query’s requirements. But App Engine has another trick up its sleeve for this kind of query. For queries using just equality filters and no sort orders, instead of scanning a single table of all values, App Engine can scan the automatic single- property indexes for each property, and return the results as it finds them. App Engine can perform a “merge join” of the single-property indexes to satisfy this kind of query. Custom Indexes and Complex Queries | 147
- 169. In other words,the datastore doesn’t need a custom index to perform queries using just equality filters and no sort orders. If you add a suitable custom index to your configuration file, the datastore will use it. But a custom index is not required, and the development server’s automatic index configuration feature will not add one if it doesn’t exist. More specifically, the datastore usually doesn’t need a custom index for such queries. The algorithm that assembles the results performs well enough in typical cases, but in extreme cases it cannot fulfill the query in a reasonable amount of time. If an amount of time has passed and the algorithm has not assembled the requested number of results, the query will return an error. In Python, this raises a db.NeedIndexError exception. To understand the problematic extreme case, let’s consider how the algorithm would perform the following query using single-property indexes: SELECT * FROM Kind WHERE a=1 AND b=2 AND c=3 Recall that each of these tables contains a row for each entity with the property set, with fields for the entity’s key and the property’s value. The table is sorted first by the value, then by the key. The algorithm takes advantage of the fact that rows with the same value are consecutive, and within that consecutive block, rows are sorted by key. To perform the query, the datastore uses the following steps: 1. The datastore checks the a index for the first row with a value of 1. The entity whose key is on this row is a candidate, but not yet a confirmed result. 2. It then checks the b index for the first row whose value is 2 and whose key is greater than or equal to the candidate’s key. Other rows with a value of 2 may appear above this row in the b index, but the datastore knows those are not candidates because the first a scan determined the candidate with the smallest key. 3. If the datastore finds the candidate’s key in the matching region of b, that key is still a candidate, and the datastore proceeds with a similar check in the index for c. If the datastore does not find the candidate in the b index but does find another larger key with a matching value, that key becomes the new candidate, and it pro- ceeds to check for the new candidate in the c index. (It’ll eventually go back to check a with the new candidate before deciding it is a result.) If it finds neither the candidate nor a matching row with a larger key, the query is complete. 4. If a candidate is found to match all criteria in all indexes, the candidate is returned as a result. The datastore starts the search for a new candidate, using the previous candidate’s key as the minimum key. Figure 5-13 illustrates this zigzag search across the single-property indexes, first with a failed candidate, then two successful candidates. A key feature of this algorithm is that it finds results in the order in which they are to be returned: key order. The datastore does not need to compile a complete list of pos- sible results for the query—possibly millions of entities—then sort them to determine 148 | Chapter 5: Datastore Queries
- 170. Figure 5-13. Themerge join algorithm finding two entities WHERE a=1 AND b=2 AND c=3 which results ought to be first. Also, the datastore can stop scanning as soon as it has enough results to fulfill the query, which is always a limited number of entities. The problem case, the one that cannot be handled by this algorithm before the datastore times out the operation, arises when the algorithm has to reject too many candidates. Each scan takes time, and a rejected candidate represents several scans that do not produce a result. If too much time elapses before the algorithm has found enough results to satisfy the query, the datastore terminates the query. Using a custom index for a query with only equality filters and no sort orders speeds up the query, but it slows down updates to the properties mentioned in the query. If a query raises NeedIndexError, you must use a custom index. A zigzag-capable query using equality filters on properties can also use inequality filters on keys without needing a custom index. This is useful for fetching a large number of results in batches, using inequality filters on keys to define the range of the batch. Custom Indexes and Complex Queries | 149
- 171. Not-Equal and INFilters The Python API supports two operators we haven’t discussed yet: != (not-equal) and IN. These operators are not actually supported by the datastore itself. Instead, they are implemented by the datastore API as multiple queries in terms of the other operators. The filter prop != value matches every entity whose property does not equal the value. The datastore API determines the result set by performing two queries: one using prop < value in place of the not-equal filter, and one using prop > value in place of the filter. It returns both sets of results as one result set, which it can do reasonably quickly because the results are already in order. Because not-equal is actually implemented in terms of the inequality operators, it is subject to the three rules of inequality operators: 1. The query’s index must be ordered by the property used with the not-equal filter before other sort orders. 2. If the query uses other explicit sort orders, the not-equal filter’s property must be explicitly ordered first. 3. And finally, any query using a not-equal filter cannot also use inequality or not- equal filters on other properties. A not-equal filter will never return an entity that doesn’t have the filtered property. This is true for all filters, but can be especially counterintuitive in the case of not-equal. The filter prop IN (value1, value2, value3) matches every entity whose property equals any of the values. The datastore API implements this as a series of equality queries, one for each value to test. The more values that appear in the list, the longer the full set of queries will take to execute. If a single query includes multiple IN filters on multiple properties, the datastore API must perform equality queries for every combination of values in all of the filters. prop1 IN (value1, value2, value3, value4) AND prop2 IN (value5, value6, value7) is equivalent to 12 queries using equality filters. The != and IN operators are useful shortcuts. But because they actually perform multiple queries, they take longer to execute than the other operators. It’s worth understanding their performance implications before using them. Unset and Nonindexed Properties As you’ve seen, an index contains columns of property values. Typically, an app creates entities of the same kind with the same set of properties: every Player in our game has a name, a character class, a level, and a score. If every entity of a kind has a given property, then an index for that kind and property has a row corresponding to each entity of the kind. 150 | Chapter 5: Datastore Queries
- 172. But the datastoreneither requires nor enforces a common layout of properties across entities of a kind. It’s quite possible for an entity to not have a property that other entities of the same kind have. For instance, a Player entity might be created without a character class, and go without until the user chooses one. App Engine does not have a notion of an “empty field.” It is possible to set a property with a null value (Python’s None, Java’s null), but a property set to the null value is distinct from the property not being set at all. This is different from a tabular database, which requires a value (possibly null) for every cell in a row. If an entity does not have a property used in an index, the entity does not appear in the index. Stated conversely, an entity must have every property mentioned in an index to appear in the index. If a Player does not have a charclass property, it does not appear in any index with a charclass column. If an entity is not mentioned in an index, it cannot be returned as a result for a query that uses the index. Remember that queries use indexes for both filters and sort orders. A query that uses a property for any kind of filter or any sort order can never return an entity that doesn’t have that property. The charclass-less Player can never be a result for a Player query that sorts results by charclass. In “Strings, Text, and Blobs” on page 112, we mentioned that text and blob values are not indexed. Another way of saying this is that, for the purposes of indexes, a property with a text or blob value is treated as if it is unset. If an app performs a query using a filter or sort order on a property that is always set to a text or blob value, that query will always return no results. It is sometimes useful to store property values of other types, and exempt them from indexes. This saves space in index tables, and reduces the amount of time it takes to save the entity. In the Python API, the only way to declare a property as unindexed is with the Model API. There is currently no other way to set a specific property as unindexed in this API. We’ll look at this feature when we discuss the modeling API in Chapter 7. In the Java API, you can set a property as unindexed using the setUnindexedProperty() method of the Entity object, instead of the setProperty() method. An entity can only have one property of a given name, so an unindexed prop- erty overwrites an indexed one, and vice versa. You can also declare properties as unin- dexed in the JDO and JPA interfaces; see Chapter 8. If you need an entity to qualify as a result for a query, but it doesn’t make sense in your data model to give the entity every property used in the query, use the null value to represent the “no value” case, and always set it. The Python modeling API and the Java JDO and JPA interfaces make it easy to ensure that properties always have values. Unset and Nonindexed Properties | 151
- 173. Sort Orders andValue Types App Engine keeps the rows of an index sorted in an order that supports the corre- sponding query. Each type of property value has its own rules for comparing two values of the same type, and these rules are mostly intuitive: integers are sorted in numeric order, strings in Unicode order, and so forth. Two entities can have values of different types for the same property, so App Engine also has rules for comparing such values, though these rules are not so intuitive. Values are ordered first by type, then within their type. For instance, all integers are sorted above all strings. One effect of this that might be surprising is that all floats are sorted below all integers. The datastore treats floats and integers as separate value types, and so sorts them sep- arately. If your app relies on the correct ordering of numbers, make sure all numbers are stored using the same type of value. The datastore stores eight distinct types of values, not counting the nonindexed types (text and blob). The datastore supports several additional types by storing them as one of the eight types, then marshaling them between the internal representation and the value your app sees automatically. These additional types are sorted by their internal representation. For instance, a date-time value is actually stored as an integer, and will be sorted amongst other integer values in an index. Table 5-2 describes the eight indexable types supported by the datastore. The types are listed in their relative order, from first to last. Table 5-2. How the datastore value types are sorted Data type Python type Java type Ordering The null value None null - Integer and long, datetime.datetime, long (other integer types are widened), Numeric (datetime date-time db.Rating java.util.Date, is chronological) datastore.Rating Boolean bool (True or False) boolean (true or false) False, then true Byte string db.ByteString datastore.ShortBlob Byte order Unicode string unicode, db.Category, java.lang.String, Unicode character db.Email, db.IM, db.Link, datastore.Category, order db.PhoneNumber, datastore.Email, db.PostalAddress datastore.IMHandle, datastore.Link, datastore.PhoneNumber, datastore.PostalAddress Floating-point float double Numeric number 152 | Chapter 5: Datastore Queries
- 174. Data type Python type Java type Ordering Geographical db.GeoPt datastore.GeoPt By latitude, then point longitude (floating- point numbers) A Google users.User users.User By email address, account Unicode order Entity key db.Key datastore.Key Kind (byte string), then ID (numeric) or name (byte string) Queries and Multivalued Properties In a typical database, a field in a record stores a single value. A record represents a data object, and each field represents a single, simple aspect of the object. If a data object can have more than one of a particular thing, each of those things is typically repre- sented by a separate record of an appropriate kind, associated with the data object using the object’s key as a field value. App Engine supports both of these uses of fields: a property can contain a simple value, or the key of another entity. But the App Engine datastore can do something most other databases can’t: it can store more than one value for a single property. With multivalued properties (MVPs), you can represent a data object with more than one of something without resorting to cre- ating a separate entity for each of those things, if each thing could be represented by a simple value. One of the most useful features of multivalued properties is how they match an equality filter in a query. The datastore query engine considers a multivalued property equal to a filter value if any of the property’s values is equal to the filter value. This ability to test for membership means MVPs are useful for representing sets. Multivalued properties maintain the order of values, and can have repeated items. The values can be of any datastore type, and a single property can have values of different types. A Simple Example Consider the following example. The players of our online game can earn trophies for particular accomplishments. The app needs to display a list of all of the trophies a player has won, and the app needs to display a list of all of the players who have won a particular trophy. The app doesn’t need to maintain any data about the trophies them- selves; it’s sufficient to just store the name of the trophy. One option is to store the list of trophy names as a single delimited string value for each Player entity. This makes it easy to get the list of trophies for a particular player, but Queries and Multivalued Properties | 153
- 175. impossible to getthe list of players for a particular trophy. (A query filter can’t match patterns within string values.) Another option is to record each trophy win in a separate property named after the trophy. To get the list of players with a trophy, you just query for the existence of the corresponding property. However, getting the list of trophies for a given player would require either coding the names of all of the trophies in the display logic, or iterating over all of the Player entity’s properties looking for trophy names. With multivalued properties, we can store each trophy name as a separate value for the trophies property. To access a list of all trophies for a player, we simply access the property of the entity. To get a list of all players with a trophy, we use a query with an equality filter on the property. Here’s what this example looks like in Python: p = Player.get_by_key_name(user_id) p.trophies = ['Lava Polo Champion', 'World Building 2008, Bronze', 'Glarcon Fighter, 2nd class'] p.put() # List all trophies for a player. for trophy in p.trophies: # ... # Query all players that have a trophy. q = Player.gql("WHERE trophies = 'Lava Polo Champion'") for p in q: # ... MVPs in Python The Python API represents the values of a multivalued property as a Python list. Each member of the list must be of one of the supported datastore types. class Entity(db.Expando): pass e = Entity() e.prop = [ 'value1', 123, users.get_current_user() ] Remember that list is not a datastore type; it is only the mechanism for manipulating multivalued properties. A list cannot contain another list. A property must have at least one value, otherwise the property does not exist. To enforce this, the Python API does not allow you to assign an empty list to a property. Notice that the API can’t do otherwise: if a property doesn’t exist, then the API cannot know to represent the missing property as an empty list when the entity is retrieved from the datastore. (This being Python, the API could return the empty list whenever a nonexistent property is accessed, but that might be more trouble than it’s worth.) 154 | Chapter 5: Datastore Queries
- 176. Because it isoften useful for lists to behave like lists, including the ability to contain zero items, the Python data modeling API provides a mechanism that supports assign- ing the empty list to a property. We’ll look at this mechanism in Chapter 7. MVPs and Equality Filters As you’ve seen, when a multivalued property is the subject of an equality filter in a query, the entity matches if any of the property’s values are equal to the filter value. e1 = Entity() e1.prop = [ 3.14, 'a', 'b' ] e1.put() e2 = Entity() e2.prop = [ 'a', 1, 6 ] e2.put() # Returns e1 but not e2: q = Entity.gql('WHERE prop = 3.14') # Returns e2 but not e1: q = Entity.gql('WHERE prop = 6') # Returns both e1 and e2: q = Entity.gql("WHERE prop = 'a'") Recall that a query with a single equality filter uses an index that contains the keys of every entity of the given kind with the given property and the property values. If an entity has a single value for the property, the index contains one row that represents the entity and the value. If an entity has multiple values for the property, the index contains one row for each value. The index for this example is shown in Figure 5-14. This brings us to the first of several odd-looking queries that nonetheless make sense for multivalued properties. Since an equality filter is a membership test, it is possible for multiple equality filters to use the same property with different values and still return a result. An example in GQL: SELECT * FROM Entity WHERE prop = 'a' AND prop = 'b' App Engine uses the “merge join” algorithm described earlier for multiple equality filters to satisfy this query using the prop single-property index. This query returns the e1 entity because the entity key appears in two places in the index, once for each value requested by the filters. The way multivalued properties appear in an index gives us another way of thinking about multivalued properties: an entity has one or more properties, each with a name and a single value, and an entity can have multiple properties with the same name. The API represents the values of multiple properties with the same name as a list of values associated with that name. Queries and Multivalued Properties | 155
- 177. Figure 5-14. Anindex of two entities with multiple values for the “prop” property, with results for WHERE prop=‘a’ The datastore does not have a way to query for the exact set of values in a multivalued property. You can use multiple equality filters to test that each of several values belongs to the list, but there is no filter that ensures that those are the only values that belong to the list, or that each value appears only once. MVPs and Inequality Filters Just as an equality filter tests that any value of the property is equal to the filter value, an inequality filter tests that any value of the property meets the filter criterion. e1 = Entity() e1.prop = [ 1, 3, 5 ] e1.put() e2 = Entity() e2.prop = [ 4, 6, 8 ] e2.put() # Returns e1 but not e2: q = Entity.gql("WHERE prop < 2") # Returns e2 but not e1: q = Entity.gql("WHERE prop > 7") # Returns both e1 and e2: q = Entity.gql("WHERE prop > 3") Figure 5-15 shows the index for this example, with the results of prop > 3 highlighted. In the case of an inequality filter, it’s possible for the index scan to match rows for a single entity multiple times. When this happens, the first occurrence of each key in the 156 | Chapter 5: Datastore Queries
- 178. Figure 5-15. Anindex of two entities with multiple values for the “prop” property, with results for WHERE prop > 3 index determines the order of the results. If the index used for the query sorts the property in ascending order, the first occurrence is the smallest matching value. For descending, it’s the largest. In the example above, prop > 3 returns e2 before e1 because 4 appears before 5 in the index. MVPs and Sort Orders To summarize things we know about how multivalued properties are indexed: • A multivalued property appears in an index with one row per value. • All rows in an index are sorted by the values, possibly distributing property values for a single entity across the index. • The first occurrence of an entity in an index scan determines its place in the result set for a query. Together, these facts explain what happens when a query orders its results by a mul- tivalued property. When results are sorted by a multivalued property in ascending or- der, the smallest value for the property determines its location in the results. When results are sorted in descending order, the largest value for the property determines its location. This has a counterintuitive—but consistent—consequence: e1 = Entity() e1.prop = [ 1, 3, 5 ] e1.put() e2 = Entity() Queries and Multivalued Properties | 157
- 179. e2.prop = [2, 3, 4 ] e2.put() # Returns e1, e2: q = Entity.gql("ORDER BY prop ASC") # Also returns e1, e2: q = Entity.gql("ORDER BY prop DESC") Because e1 has both the smallest value and the largest value, it appears first in the result set in ascending order and in descending order. See Figure 5-16. Figure 5-16. Indexes of two entities with multiple values for the “prop” property, one ascending and one descending MVPs and the Query Planner The query planner tries to be smart by ignoring aspects of the query that are redundant or contradictory. For instance, a = 3 AND a = 4 would normally return no results, so the query planner catches those cases and doesn’t bother doing work it doesn’t need to do. However, most of these normalization techniques don’t apply to multivalued properties. In this case, the query could be asking, “Does this MVP have a value that is equal to 3 and another value equal to 4?” The datastore remembers which properties are MVPs (even those that end up with one or zero values), and never takes a shortcut that would produce incorrect results. But there is one exception. A query that has both an equality filter and a sort order will drop the sort order. If a query asks for a = 3 ORDER BY a DESC and a is a single-value property, the sort order has no effect because all of the values in the result are identical. For an MVP, however, a = 3 tests for membership, and two MVPs that meet that condition are not necessarily identical. 158 | Chapter 5: Datastore Queries
- 180. The datastore dropsthe sort order in this case anyway. To do otherwise would require too much index data and result in exploding indexes in cases that could otherwise survive. As always, the actual sort order is deterministic, but it won’t be the requested order. Exploding Indexes There’s one more thing to know about indexes when considering multivalued proper- ties for your data model. When an entity has multiple values for a property, each index that includes a column for the property must use multiple rows to represent the entity, one for each possible combination of values. In a single property index on the multivalued property, this is simply one row for each value, two columns each (the entity key and the property value). In an index of multiple properties where the entity has multiple values for one of the indexed properties and a single value for each of the others, the index includes one row for each value of the multivalued property. Each row has a column for each indexed property, plus the key. The values for the single-value properties are repeated in each row. Here’s the kicker: if an entity has more than one property with multiple values, and more than one multivalued property appears in an index, the index must contain one row for each combination of values to represent the entity completely. If you’re not careful, the number of index rows that need to be updated when the entity changes could grow very large. It may be so large that the datastore cannot complete an update of the entity before it reaches its safety limits, and returns an error. To help prevent “exploding indexes” from causing problems, App Engine limits the number of property values—that is, the number of rows times the number of columns—a single entity can occupy in an index. The limit is 5,000 property values, high enough for normal use, but low enough to prevent unusual index sizes from in- hibiting updates. If you do include a multivalued property in a custom index, be careful about the pos- sibility of exploding indexes. Configuring Indexes An application specifies the custom indexes it needs in a configuration file. Each index definition includes the kind, and the names and sort orders of the properties to include. A configuration file can contain zero or more index definitions. Configuring Indexes | 159
- 181. Most of thetime, you can leave the maintenance of this file to the development web server. The development server watches the queries the application makes, and if a query needs a custom index and that index is not defined in the configuration file, the server adds appropriate configuration automatically. The development server will not remove index configuration. If you are sure your app no longer needs an index, you can edit the file manually and remove it. You can disable the automatic index configuration feature. Doing so causes the devel- opment server to behave like App Engine: if a query doesn’t have an index and needs one, the query fails. How to do this is particular to the runtime environment, so we’ll get to that in a moment. Index configuration is global to all versions of your application. All versions of an app share the same datastore, including indexes. If you deploy a version of the app and the index configuration has changed, App Engine will use the new index configuration for all versions. Index Configuration for Python For Python apps, the index configuration file is named index.yaml, and is in the YAML format (similar to app.yaml). It appears in the application root directory. The structure is a single YAML list named indexes, with one element per index. Each index definition has a kind element (the kind name, a string) and a properties element. If the index supports queries with ancestor filters, it has an ancestor element with a value of yes. properties is a list, one element per column in the index, where each column has a name and an optional direction that is either asc (ascending order, the default) or desc (descending order). The order of the properties list is significant: the index is sorted by the first column first, then by the second column, and so on. Here’s an example index.yaml file using indexes from earlier in this chapter: indexes: - kind: Player properties: - name: charclass - name: level direction: desc - kind: Player properties: - name: level direction: desc - name: score direction: desc 160 | Chapter 5: Datastore Queries
- 182. By default, thedevelopment server adds index configuration to this file as needed. When it does, it does so beneath this line, adding it (and a descriptive comment) if it doesn’t find it: # AUTOGENERATED You can move index configuration above this line to take manual control over it. This isn’t strictly necessary, since the development server will never delete index configu- ration, not even that which was added automatically. To disable automatic index configuration in the development server, start the server with the --require_indexes command-line option. If you are using the Launcher, select the application, then go to the Edit menu and select Application Settings. Add the command-line option to the “Extra Flags” field, then click Update. Index Configuration for Java For Java apps, you add index configuration to a file named datastore-indexes.xml, in the directory WEB-INF/ in the WAR. This is an XML file with a root element named <datastore-indexes>. This contains zero or more <datastore-index> elements, one for each index. Each <datastore-index> specifies the kind using the kind attribute. It also has an ancestor attribute, which is true if the index supports queries with ancestor filters, and false otherwise. A <datastore-index> contains one or more <property> elements, one for each column in the index. Each <property> has a name attribute (the name of the property) and a direction attribute (asc for ascending, desc for descending). The order of the <property> elements is significant: the index is sorted by the first column first, then by the second column, and so on. An example: <datastore-indexes autoGenerate="true"> <datastore-index kind="Player" ancestor="false"> <property name="charclass" direction="asc" /> <property name="level" direction="desc" /> </datastore-index> <datastore-index kind="Player" ancestor="false"> <property name="level" direction="desc" /> <property name="score" direction="desc" /> </datastore-index> </datastore-indexes> The <datastore-indexes> root element has an attribute named autoGenerate. If true, or if the app does not have a datastore-indexes.xml file, the Java development server will generate new index configuration when needed for a query. If false, the develop- ment server behaves like App Engine: if a query needs an index that is not defined, the query fails. Configuring Indexes | 161
- 183. The development serverdoes not modify datastore-indexes.xml. Instead, it generates a separate file named datastore-indexes-auto.xml, in the directory WEB-INF/appengine- generated/. The complete index configuration is the total of the two configuration files. The Java server will never remove index configuration from the automatic file, so if you need to delete an index, you may need to remove it from the automatic file. You can move configuration from the automatic file to the manual file if that’s easier to manage, such as to check it into a revision control repository. 162 | Chapter 5: Datastore Queries
- 184. CHAPTER 6 Datastore Transactions With web applications, many users access and update data concurrently. Often, mul- tiple users need to read or write to the same unit of data at the same time. This requires a data system that can give some assurances that simultaneous operations will not corrupt any user’s view of the data. Most data systems guarantee that a single operation on a single unit of data maintains the integrity of the data, typically by scheduling operations that act on the same unit of data to be performed one at a time. Many applications need similar data integrity guarantees when performing a set of multiple operations, possibly over multiple units of data. Such a set of operations is called a transaction. A data system that supports transactions guarantees that if a trans- action succeeds, all of the operations in the transaction are executed completely. If any step of the transaction fails, then none of its effects are applied to the data. The data remains in a consistent and predictable state before and after the transaction, even if other processes are attempting to modify the data concurrently. For example, say you want to post a message to the bulletin board in the town square inviting other players to join your guild. The bulletin board maintains a count of how many messages have been posted to the board, so readers can see how many messages there are without reading every message object in the system. Posting a message requires three datastore operations: the app must read the old message count, update the mes- sage count with an incremented value, and create the new message object. Without transactions, these operations may succeed or fail independently. The count may be updated but the message object may not be created. Or, if you create the message object first, the object may be created, but the count not updated. In either case, the resulting count is inaccurate. By performing these operations in a single transaction, if any step fails, none of the effects are applied, and the application can try the entire transaction again. 163
- 185. Also consider whathappens when two players attempt to post to the message board at the same time. To increment the message count, each player process must read the old value, then update it with a new value calculated from the old one. Without trans- actions, these operations may be interleaved: the first process reads the original count (say, 10), the second process reads the count (also 10), the first process adds 1 and updates the count with the new value (11), then finally the second process adds 1 to its value and updates the count (11). Because the second process doesn’t know that the first process updated the value, the final count is 1 less than it ought to be (12). With transactions, the second process knows right away that the first process is updating the count and can do the right thing. A scalable web application has several requirements that are at odds with transactions. For one, the application needs access to data to be fast, and not be affected by how much data is in the system or how it is distributed across multiple servers. The longer it takes for a transaction to complete, the longer other processes have to wait to access the data reserved by the transaction. The combined effect on how many transactions can be completed in a period of time is called throughput. For web apps, high through- put is important. A web app also needs transactions to finish completely and consistently, so it knows that the effects of the transaction can be relied upon by other processes for further calculations. The promise that all processes can see the changes once a transaction is complete is known as strong consistency. (Alternatives such as eventual consistency trade this promise for greater flexibility in how changes are applied, but this makes life dif- ficult for an app that needs to present the user with a consistent view of the world with each request.) The App Engine datastore provides transactions with strong consistency and low over- head. It does this by limiting the scope of transactions: a single transaction can only read or write to entities that belong to a single entity group. Every entity belongs to an entity group, by default a group of its own. The app assigns an entity to a group when the entity is created, and the assignment is permanent. By having the app arrange entities into groups, the datastore can treat each group in- dependently when applying concurrent transactions. Two transactions that use differ- ent groups can occur simultaneously without harm. With a bit of thought, an app can ensure that entities are arranged to minimize the likelihood that two processes will need to access the same group, and thereby maximize throughput. In this chapter, we will discuss what happens when you update an entity group, how to create entities in entity groups, and how to perform multiple operations on an entity group using a transaction. We’ll also discuss batch operations, how query indexes are built, and the consistency guarantees of the datastore. 164 | Chapter 6: Datastore Transactions
- 186. Entities and EntityGroups When you create, update, or delete a single entity, the change occurs in a transaction: either all of your changes to the entity succeed, or none of them do. If you change two properties of an entity and save it, every request handler process that fetches the entity will see both changes. At no point during the save will a process see the new value for one property and the old value for the other. And if the update fails, the entity stays as it was before the save. In database terms, the act of updating an entity is atomic. It is often useful to update multiple entities atomically, such that any process’s view of the data is consistent across the entities. In the bulletin board example, the message count and each of the messages may be stored as separate entities, but the combined act of creating a new message entity and updating the count ought to be atomic. We need a way to combine multiple actions into a single transaction, so they all succeed or all fail. To do this in a scalable way, App Engine must know in advance which entities may be involved in a single transaction. These entities are stored and updated together, so the datastore can keep them consistent and still access them quickly. You tell App Engine which entities may be involved in the same transaction using entity groups. Every entity belongs to an entity group, possibly a group containing just itself. An entity can only belong to one group. You assign an entity to a group when the entity is created. Group membership is permanent; an entity cannot be moved to another group once it has been created. The datastore uses entity groups to determine what happens when two processes at- tempt to update data in the entity group at the same time. When this happens, the first update that completes “wins,” and the other update is canceled. App Engine notifies the process whose update is canceled by raising an exception. In most cases, the process can just try the update again and succeed. But the app must decide for itself how to go about retrying, since important data may have changed between attempts. This style of managing concurrent access is known as optimistic concurrency control. It’s “optimistic” in the sense that the database tries to perform the operations without checking whether another process is working with the same data (such as with a “lock- ing” mechanism), and only checks for collisions at the end, optimistic that the opera- tions will succeed. The update is not guaranteed to succeed, and the app must reattempt the operations or take some other course of action if the data changes during the update. Optimistic concurrency control is a good choice for web applications because reading data is fast—a reader never waits for updates—and almost always succeeds. If an up- date fails due to concurrency, it’s usually easy to try again, or return an error message to the user. Most web applications have only a small number of users updating the same piece of data, so concurrency failures are rare. Entities and Entity Groups | 165
- 187. Updating an entityin a group can potentially cancel updates to any other entity in the group by another process. You should design your data model so that entity groups do not need to be updated by many users simultaneously. Be especially careful if the number of simultaneous updates to a single group grows as your application gets more users. In this case, you usu- ally want to spread the load across multiple entity groups, and increase the number of entity groups automatically as the user base grows. Scal- able division of a data resource like this is known as sharding. Also be aware that some data modeling tasks may not be practical on a large scale. Incrementing a value in a single datastore entity every time any user visits the site’s home page is not likely to work well with a distributed strong consistency data system. Keys, Paths, and Ancestors To create an entity in a group with other entities, you associate it with the key of another entity from that group. The existing entity becomes the new entity’s parent, forming a path of ancestor relationships down to a root entity that does not have a parent. This path becomes part of the new entity’s key. Every entity whose key begins with the same root is in the same group, including the root entity itself. When you create an entity and do not specify a parent, the entity is created in a new group by itself. The new entity is the root of the new group. We alluded to paths earlier when we discussed keys, so let’s complete the picture. An entity’s key consists of the path of ancestors in the entity’s group, starting from the group’s root. Each entity in the path is represented by the entity’s kind followed by either the system-assigned numeric ID or the app-assigned string name. The full path is a sequence of kind and ID/name pairs. The following Python code creates four entities in the same group. The model class constructor accepts either an entity object or a Key object as the parent. # Creating a new entity group with a root entity: board = MessageBoard(key_name='The_Archonville_Times') board.title = 'The Archonville Times' board.put() # Using the object for the "parent" argument: msg1 = Message(parent=board, key_name='first!') msg1.put() # Using a Key for the "parent" argument: p_key = board.key() msg2 = Message(parent=p_key, key_name='pk_fest_aug_21') msg2.put() # Using an entity that isn't the root as the parent: 166 | Chapter 6: Datastore Transactions
- 188. msg3 = Message(parent=msg1,key_name='keep_clean') msg3.put() When you’re deriving the key of an entity with a parent, you must use the complete ancestor path. For example, in Python: k = db.Key.from_path('MessageBoard', 'The_Archonville_Times', 'Message', 'first!', 'Message', 'keep_clean') Similarly, GQL supports key literals with ancestors, as follows: SELECT * FROM MessageAttachment WHERE message = KEY(MessageBoard, 'The_Archonville_Times', Message, 'first!', Message, 'keep_clean') Notice that entities of different kinds can be in the same entity group. In the datastore, there is no relationship between kinds and entity groups. (You can enforce such a relationship in your app’s code, if you like.) Ancestors do not have to exist for a key to be valid. If you create an entity with a parent, then delete the parent, the key for the child is still valid and can still be assembled from its parts (such as with Python’s from_path() method). This is true even for a group’s root entity: the root can be deleted and other entities in the group remain in the group. You can even use a key for an entity that doesn’t exist as the parent for a new entity. Neither the kind nor the name/ID of an ancestor need to represent an actual entity. Group membership is defined by the first key part in the ancestor path, regardless of whether that part corresponds to an entity. Here is Python code that creates two entities in the same group without a root entity: root = db.Key.from_path('MessageBoard', 'The_Baskinville_Post') msg1 = Message(parent=root) msg1.put() msg2 = Message(parent=root) msg2.put() Ancestor Queries Though the full path on an entity key is not considered to determine group membership, ancestor paths are useful thanks to another datastore feature: ancestor queries. A datastore query can include a filter that limits the results to just those entities with a given ancestor. This can match any ancestor, not just the immediate parent. In other words, a query can match a sequence of key parts starting from the root. Continuing the town square bulletin board example, where each MessageBoard is the root of an entity group containing things attached to the board, the following GQL query returns the 10 most recent Message entities attached to a specific board: Entities and Entity Groups | 167
- 189. SELECT * FROMMessage WHERE ANCESTOR IS KEY(MessageBoard, 'The_Archonville_Times') ORDER BY post_date DESC LIMIT 10 Most queries that use an ancestor filter need custom indexes. There is one unusual exception: a query does not need a custom index if the query also contains equality filters on properties (and no inequality filters or sort orders). In this exceptional case, the “merge join” algorithm can use a built-in index of keys along with the built-in property indexes. In cases where the query would need a custom index anyway, the query can match the ancestor to the keys in the custom index. As we mentioned in Chapter 5, the datastore supports queries over entities of all kinds. Kindless queries are limited to key filters and ancestor filters. Since ancestors can have children of disparate kinds, kindless queries are useful for getting every child of a given ancestor, regardless of kind: SELECT * WHERE ANCESTOR IS KEY('MessageBoard', 'The_Archonville_Times') While ancestor queries can be useful, don’t get carried away building large ancestor trees. Remember that every entity with the same root belongs to the same entity group, and more simultaneous users that need to write to a group mean a greater likelihood of concurrency failures. If you want to model hierarchical relationships between entities without the consequences of entity groups, consider using multivalued proper- ties to store paths. For example, if there’s an entity whose path in your hierarchy can be represented as /A/B/C/D, you can store this path as: e.parents = ['/A', '/A/B', '/A/B/C'] Then you can perform a query similar to an ancestor query on this property: ... WHERE parents = '/ A/B'. What Can Happen in a Transaction Entity groups ensure that the operations performed within a transaction see a consistent view of the entities in a group. For this to work, a single transaction must limit its operations to entities in a single group. The entity group determines the scope of the transaction. Within a transaction, you can fetch, update, or delete an entity using the entity’s key. You can also create a new entity that either is a root entity of a new group that becomes the subject of the transaction, or that has a member of the transaction’s entity group as its parent. You can perform queries over the entities of a single entity group in a transaction. A query in a transaction must have an ancestor filter that matches the transaction’s entity group. The results of the query, including both the indexes that provide the results as 168 | Chapter 6: Datastore Transactions
- 190. well as theentities themselves, are guaranteed to be consistent with the rest of the transaction. In the Python and Java APIs, you do not need to declare the entity group for a trans- action explicitly. You simply perform datastore actions on entities of the same group. If you attempt to perform actions that involve different entity groups within a trans- action, the API raises an exception. The API also raises an exception if you attempt to perform a query in a transaction that does not have an ancestor filter. (We’ll discuss the specific exception that is raised in the next few sections.) Transactional Reads Sometimes it is useful to fetch entities in a transaction even if the transaction does not update any data. Reading multiple entities in a transaction ensures that the entities are consistent with one another. As with updates, entities fetched in a transaction must be members of the same entity group. A transaction that only reads entities never fails due to concurrency. As with reading a single entity, a read-only transaction sees the data as it appears at the beginning of the transaction, even if other processes make changes after the transaction starts and before it completes. The same is true for ancestor-only queries within a transaction. If the transaction does not create, update, or delete data from the entity group, it will not fail due to concurrency. The datastore can do this because it remembers previous versions of entities, using timestamps associated with the entity groups. The datastore notes the current time at the beginning of every operation and transaction, and this determines which version of the data the operation or transaction sees. This is known as multiversion concurrency control, a form of optimistic concurrency control. This mechanism is internal to the datastore; the application cannot access previous versions of data, nor can it see the timestamps. This timestamp mechanism has a minor implication for reading data within transac- tions. When you read an entity in a transaction, the datastore returns the version of the entity most recent to the beginning of the transaction. If you update an entity, then refetch the same entity within the same transaction, the datastore returns the entity as it appeared before the update. In most cases, you can just reuse the in-memory object you modified (which has your changes) instead of refetching the entity. Transactions in Python As we’ve seen, to create an entity in an entity group in Python, you provide either the Key object or the model object of the parent to the model class constructor, as the parent argument: Transactions in Python | 169
- 191. p = Entity() e = Entity(parent=p) The parent key does not have to represent an entity that exists. You can construct a fake key using db.Key.from_path() and use it as the parent: p_key = db.Key.from_path('Entity', 'fake_parent') e = Entity(parent=p_key) You can use methods on the model object to get the key for the entity’s parent, or fetch the entity’s parent from the datastore. These methods return None if the entity is a root entity for its entity group. p = e.parent() p_key = e.parent_key() Similarly, the Key object can return the Key of its parent, or None: e_key = e.key() p_key = e_key.parent() The Python API uses function objects to handle transactions. To perform multiple operations in a transaction, you define a function that executes the operations, then you pass it to the run_in_transaction() function (in the ...ext.db module). import datetime from google.appengine.ext import db class MessageBoard(db.Expando): pass class Message(db.Expando): pass def create_message_txn(board_name, message_name, message_title, message_text): board = db.get(db.Key.from_path('MessageBoard', board_name)) if not board: board = MessageBoard(key_name=board_name) board.count = 0 message = Message(key_name=message_name, parent=board) message.title = message_title message.text = message_text message.post_date = datetime.datetime.now() board.count += 1 db.put([board, message]) # ... try: db.run_in_transaction(create_message_txn, 170 | Chapter 6: Datastore Transactions
- 192. board_name=board_name, message_name=message_title, message_title=message_title, message_text=message_text) except db.TransactionFailedError, e: # Report an error to the user. # ... run_in_transaction() takes the transaction function as its first argument, and any po- sitional or keyword arguments intended for the transaction function as its remaining arguments. To perform the transaction, it starts a new transaction in the datastore, calls the function, then commits the transaction. If the transaction function returns a value, run_in_transaction() returns that value. All calls to the datastore to create, update, or delete entities within the transaction function take effect when the transaction is committed. Typically, you would update an entity by fetching or creating the model object, modifying it, then saving it, and continue to use the local object to represent the entity. In the rare case where the trans- action function fetches an entity after saving it, the fetch will see the entity as it was before the update, because the update has not yet been committed. If the transaction function raises an exception, the transaction is aborted, and run_in_transaction() reraises the exception. If you need to abort the transaction but do not want run_in_transaction() to raise an exception, raise the db.Rollback excep- tion. run_in_transaction() catches and does not reraise this exception. If the transaction cannot be completed due to a concurrency failure, run_in_transaction() retries the transaction again automatically up to three times. For each attempt, it calls the transaction function again with the same arguments. Make sure the transaction function can be called multiple times safely without undesirable side effects. If you don’t want the transaction to be retried in the event of the concurrency failure, or if you’d like it retried more than three times, you can use the alternate function run_in_transaction_custom_retries(). This function behaves identically to run_in_transaction() except for the custom number of retries. You pass it the number of retries as the first argument and the function as the second argument, followed by arguments for the function. If the transaction cannot be completed due to concurrency failures after all of the retries have been exhausted, run_in_transaction() (or ...custom_retries()) raises a db.TransactionFailedError. Since this condition can happen with any datastore up- date, you should catch it and provide an appropriate response to the user. Transactions in Python | 171
- 193. Transactions in Java TheJDO and JPA interfaces provide their own mechanisms for formulating transac- tions. Google’s online documentation describes JDO, and we’ll cover the JPA interface in Chapter 8. Using the low-level datastore interface, you create an entity in an entity group by pro- viding a Key object that represents an entity in that group to the Entity constructor. You can create this key using the KeyFactory class, or by calling the getKey() method on an Entity instance: Key parentKey = KeyFactory.createKey("MessageBoard", "The_Archonville_Times"); Entity message = new Entity("Message", parentKey); To create a key for a root entity (an entity without a parent), you call the KeyFactory.createKey() method with a kind (as a string) and the numeric ID or the string key name. To create a key with ancestors using createKey(), you also provide the Key of the parent. For convenience, the KeyFactory class provides an alternate mechanism for construct- ing keys with ancestors. The KeyFactory.Builder class represents a key path, which you can extend by calling its addChild() method. The method returns the object, so you can chain multiple calls on a single line. Once the path is complete, you call getKey() to get the Key object: Key k = new Builder("MessageBoard", "The_Archonville_times") .addChild("Message", 1427) .getKey(); To perform multiple operations within a single transaction, you call the beginTransaction() method of the DatastoreService instance. This method returns a Transaction object that represents the transaction. You perform the datastore operations as usual, calling the put(), get(), and delete() methods of the DatastoreService. Finally, you call the commit() method of the Transaction object to complete the transaction. Updates (put() and delete()) do not take effect until you commit the transaction. Fetching an entity using get() after it has been updated in the same transaction will return an Entity object that represents the state of the entity before the update. If the transaction cannot be committed due to a concurrency failure (or other datastore error), the commit throws a DatastoreFailureException: DatastoreService ds = DatastoreServiceFactory.getDatastoreService(); Key boardKey; Entity messageBoard; try { Transaction txn = ds.beginTransaction(); 172 | Chapter 6: Datastore Transactions
- 194. try { boardKey = KeyFactory.createKey("MessageBoard", boardName); messageBoard = ds.get(boardKey); } catch (EntityNotFoundException e) { messageBoard = new Entity("MessageBoard", boardName); messageBoard.setProperty("count", 0); boardKey = ds.put(messageBoard); } txn.commit(); } catch (DatastoreFailureException e) { // Report an error... } If you do not commit the transaction, the transaction is rolled back automatically after the servlet exits, and changes are not applied. You can roll back the transaction ex- plicitly by calling the rollback() method of the Transaction object. By default, each datastore operation is associated with the transaction started by the most recent call to beginTransaction(), known in the API as the “current” transaction. If you call beginTransaction() more than once, each Transaction is remembered in a stack. Calling commit() or rollback() removes the Transaction from the stack. If you commit or roll back the current transaction and there is another transaction on the stack, the next most recent transaction becomes the current transaction. You can associate a datastore operation with a specific transaction explicitly by passing the Transaction object to the operation method: Transaction txn = ds.beginTransaction(); ds.put(txn, messageBoard); txn.commit(); If there is no current transaction on the stack, calling an operation method performs the operation without a transaction. Updates occur immediately. If there is a current transaction but you would like the operation performed immediately and outside of the transaction, provide null as the first argument to the operation: Transaction txn = ds.beginTransaction(); // Add an update of entityOne to the transaction. ds.put(txn, entityOne); // Update entityTwo immediately, outside of the transaction. ds.put(null, entityTwo); // Commit the transaction, updating entityOne. txn.commit(); Transactions in Java | 173
- 195. With optimistic concurrencycontrol, it is usually appropriate for the application to try the transaction again in the event of a concurrency failure. The following example retries the transaction up to three times before reporting an error to the user: import com.google.appengine.api.datastore.DatastoreFailureException; import com.google.appengine.api.datastore.DatastoreService; import com.google.appengine.api.datastore.DatastoreServiceFactory; import com.google.appengine.api.datastore.Entity; import com.google.appengine.api.datastore.EntityNotFoundException; import com.google.appengine.api.datastore.Key; import com.google.appengine.api.datastore.KeyFactory; import com.google.appengine.api.datastore.PreparedQuery; import com.google.appengine.api.datastore.Query; import com.google.appengine.api.datastore.Transaction; // ... DatastoreService ds = DatastoreServiceFactory.getDatastoreService(); int retries = 3; boolean success = false; while (!success && retries > 0) { --retries; try { Transaction txn = ds.beginTransaction(); Key boardKey; Entity messageBoard; try { boardKey = KeyFactory.createKey("MessageBoard", boardName); messageBoard = ds.get(boardKey); } catch (EntityNotFoundException e) { messageBoard = new Entity("MessageBoard", boardName); messageBoard.setProperty("count", 0); boardKey = ds.put(messageBoard); } Entity message = new Entity("Message", boardKey); message.setProperty("message_title", messageTitle); message.setProperty("message_text", messageText); message.setProperty("post_date", postDate); ds.put(message); long count = (Long) messageBoard.getProperty("count"); ++count; messageBoard.setProperty("count", count); ds.put(messageBoard); log.info("Posting msg, updating count to " + count); txn.commit(); // Break out of retry loop. success = true; 174 | Chapter 6: Datastore Transactions
- 196. } catch (DatastoreFailureExceptione) { // Allow retry to occur. } } if (!success) { // Tell the user it didn't work out... resp.getWriter().println ("<p>A new message could not be posted. Try again later.</p>"); } // ... Key boardKey = KeyFactory.createKey("MessageBoard", boardName); try { Entity messageBoard = ds.get(boardKey); long count = (Long) messageBoard.getProperty("count"); resp.getWriter().println("<p>Latest messages posted to " + boardName + " (" + count + " total):</p>"); Query q = new Query("Message", boardKey); PreparedQuery pq = ds.prepare(q); for (Entity result : pq.asIterable()) { resp.getWriter().println("<h3>" + result.getProperty("message_title") + "</h3><p>" + result.getProperty("message_text") + "</p>"); } } catch (EntityNotFoundException e) { resp.getWriter().println("<p>No message board found.</p>"); } How Entities Are Updated To fully understand how the datastore guarantees that your data stays consistent, it’s worth discussing how transactions are performed behind the scenes. To do so, we must mention BigTable, Google’s distributed data system that is the basis of the App Engine datastore. We won’t go into the details of how entities, entity groups, and indexes are stored in BigTable, but we will refer to BigTable’s own notion of atomic transactions in our explanation. Figure 6-1 shows the phases of a successful transaction. The datastore uses a “journal” to keep track of changes that need to be applied to entities in an entity group. Each journal entry has a unique timestamp that indicates the order in which the changes were made. The datastore remembers the timestamp of the most recent change that has been committed, and guarantees that attempts to read the data will see all changes up to that point. When an app begins a transaction for an entity group, the datastore makes a note of the current last-committed timestamp for the group (point A in Figure 6-1). As the app How Entities Are Updated | 175
- 197. Figure 6-1. Thetimeline of a transaction: the operations, the commit phase, and the apply phase calls the datastore to update entities, the datastore writes the requested changes to the journal. Each change is marked as “uncommitted.” When the app finishes the transaction (point B), the datastore checks the group’s last- committed timestamp again. If the timestamp hasn’t changed since the transaction began, it marks all of the transaction’s changes as “committed,” then advances the group’s timestamp. Otherwise, the timestamp was advanced by another request han- dler since the beginning of the transaction, so the datastore aborts the current trans- action and reports a concurrency failure to the app. Verifying the timestamp, committing the journal entries, and updating the timestamp all occur in an atomic BigTable operation. If another process attempts to commit a transaction to the same entity group while the first transaction’s commit is in progress, the other process waits for the first commit to complete. This guarantees that if the first commit succeeds, the second process sees the updated timestamp and reports a con- currency failure. This is illustrated in Figure 6-2. Figure 6-2. A timeline of two concurrent transactions; the first to commit “wins” 176 | Chapter 6: Datastore Transactions
- 198. Once the journalentries have been committed (point C in Figure 6-1), the datastore applies each committed change to the appropriate entity and appropriate indexes, then marks the change as “applied.” If there are multiple unapplied changes for an entity in the journal, they are applied in the order they were performed. Here’s the sneaky bit. If the apply phase fails for whatever reason (hard drive failure, power outage, meteorite), the committed transaction is still considered successful. If there are committed but unapplied changes the next time someone performs an oper- ation on an entity in the entity group (within or without an explicit transaction), the datastore reruns the apply phase before performing the operation. This ensures that operations and transactions always see all changes that have been committed prior to the start of the operation or transaction. The datastore also uses background processes to roll forward unapplied operations, as well as purge old journal entries. The roll-forward mechanism also ensures that subsequent operations can see all com- mitted changes even if the apply phase is still in progress. At the beginning of an op- eration or transaction, the datastore notes the current time, then waits for all changes committed prior to that time to be applied before continuing. Notice that the apply phase does not need to occur inside a BigTable transaction. Be- cause of this, the datastore can spread multiple entities in the same group across mul- tiple machines, and can allow an entity group to get arbitrarily large. Only the group’s last-committed timestamp and journal need to be stored close enough together for a BigTable transaction. The datastore makes an effort to store entities of the same group “near” each other for performance reasons, but this does not limit the size or speed of entity groups. When an app updates an entity outside of a transaction, the datastore performs the update using the same transactional mechanism, as if it were a transaction of just one operation. The datastore assumes that an update performed outside of a transaction is safe to perform at any time, and will retry the update automatically in the event of a concurrency failure. If several attempts fail, the datastore throws the concurrency ex- ception. In contrast, an explicit transaction throws a concurrency exception on the first failure, because the datastore does not know if it is safe to commit the same changes. The app must retry the explicit transaction on its own. (The Python API’s run_in_transaction() function knows how to rerun the function that performs the transaction, but this occurs as part of the application code.) The App Engine datastore replicates all data to at least three places in each of at least two different data centers. An application uses one location as the “master” location, and replicates data to the others for use during failover or system maintenance. An update completes when the master location meets the consistency guarantees, and does not wait for replication to complete before returning. How Entities Are Updated | 177
- 199. When an appcalls the datastore to update data, the call does not return until the apply phase is complete. If an error occurs at any point in the process, the datastore call raises an exception in the application. This is true even if the error occurs during the apply phase, after the commit phase is complete and the update is guaranteed to be applied before the next transaction. Because the application can’t tell the dif- ference between an error during the commit phase and an error during the apply phase, the application should react as if the update has not taken place. In most cases, the app can simply retry the update. More care is needed if retrying the update relies on the previous attempt to have been un- successful, but these cases usually require testing the state of the data in a transaction, and the solution is to just retry the entire transaction. If the transaction creates a new entity, one way to avoid creating a du- plicate entity is to use a key name instead of a system-supplied numeric ID, precalculating and testing for the nonexistence of a global unique ID (GUID) if necessary. Failures during the apply phase are very rare, and most errors represent a failure to commit. One of the most important principles in scalable app design is to be tolerant of the most rare kinds of faults. How Entities Are Read The timestamp mechanism explains what happens when two processes attempt to write to the same entity group at the same time. When one process commits, it updates the timestamp for the group. When the other process tries to commit, it notices the timestamp has changed, and aborts. The app can retry the transaction with fresh data, or give up. The transaction is aborted only if the app attempted to update an entity during the transaction and another process has since committed changes. If the app only reads data in the transaction and does not make changes, the app simply sees the entities as they were at the beginning of the transaction. To support this, the datastore retains several old versions of each entity, marked with the timestamp of the most recently applied journal entry. Reading an entity in a transaction returns the version of the entity most recent to the timestamp at the beginning of the transaction. Reading an entity outside of a transaction does not roll forward unapplied but com- mitted changes. Instead, the read returns the entity as it appears on disk, as of the most recently applied changes. This is faster than waiting for pending changes to be applied, and usually not a concern for reads outside of transactions. But this means the entity may appear older or newer than other entities. If you need any consistency guarantees when reading multiple entities, use transactions. 178 | Chapter 6: Datastore Transactions
- 200. Batch Updates When youread, create, update, or delete an entity, the runtime environment makes a service call to the datastore. Each service call has some overhead, including serializing and deserializing parameters and transmitting them between machines in the data cen- ter. If you need to update multiple entities, you can save time by performing the updates together as a batch in one service call. We introduced batch calls in Chapter 4. Here’s a quick example of the Python batch API: # Creating multiple entities: e1 = Message(key_name='m1', text='...') e2 = Message(key_name='m2', text='...') e3 = Message(key_name='m3', text='...') message_keys = db.put([e1, e2, e3]) # Getting multiple entities using keys: message_keys = [db.Key('Message', 'm1'), db.Key('Message', 'm2'), db.Key('Message', 'm2')] messages = db.get(message_keys) for message in messages: # ... # Deleting multiple entities: db.delete(message_keys) When the datastore receives a batch call, it bundles the keys or entities by their entity groups, which it can determine from the keys. Then it dispatches calls to the datastore machines responsible for each entity group. The datastore returns results to the app when it has received all results from all machines. If the call includes changes for multiple entities in a single entity group, those changes are performed in a single transaction. There is no way to control this behavior, but there’s no reason to do it any other way. It’s faster to commit multiple changes to a group at once than to commit them individually, and no less likely to result in concur- rency failures. Each entity group involved in a batch update may fail to commit due to a concurrency failure. If a concurrency failure occurs for any update, the API raises the concurrency failure exception—even if updates to other groups were committed successfully. Batch updates in disparate entity groups are performed in separate threads, possibly by separate datastore machines, executed in parallel to one another. This can make batch updates especially fast compared to performing each update one at a time. Remember that if you use the batch API during a transaction, every entity or key in the batch must use the same entity group as the rest of the transaction. Also, be aware that App Engine puts a maximum size on service calls and responses, and the total size of the call or its responses must not exceed these limits. The number of items in a batch Batch Updates | 179
- 201. also has afixed limit, so a single batch operation cannot consume more than a fixed amount of memory. See Google’s official documentation for more information on these limits. How Indexes Are Updated As we saw in Chapter 5, datastore queries are powered by indexes. The datastore up- dates these indexes as entities change, so results for queries can be determined without examining the entity properties directly. This includes an index of keys, an index for each kind and property, and custom indexes described by your app’s configuration files that fulfill complex queries. When an entity is created, updated, or deleted, each rele- vant index is updated and sorted so subsequent queries match the new state of the data. The datastore updates indexes after changes have been committed, during the apply phase. Changes are applied to indexes and entities in parallel. Updates of indexes are themselves performed in parallel, so the number of indexes to update doesn’t neces- sarily affect how fast the update occurs. As with entities, the datastore retains multiple versions of index data, labeled with timestamps. When you perform a query, the datastore notes the current time, then uses the index data that is most current up to that time. However, unless the query has an ancestor filter, the datastore has no way to know which entity groups are involved in the result set and so cannot wait for changes in progress to be applied. This means that, for a brief period during an update, a query may return results that do not match the query criteria. While another process is updating an entity, the query may see the old version of its index but return the new version of the entity. And since changes to entities and changes to indexes are applied in parallel, it is possible for a query to see the new version of its index but return the old version of the entity. Figure 6-3 illustrates one possibility of what nontransactional reads and queries may see while changes are being applied. Figure 6-3. What nontransactional fetches and queries may see while changes are applied 180 | Chapter 6: Datastore Transactions
- 202. Even though bothfetches r-1 and r-2 occur after the commit, because they do not occur within transactions, they see different data: r-1 fetches the entity as it is before the update, and r-2 fetches the entity as it is after the update has been applied. Queries q-1 and q-2 may use the same (preupdate) index data to produce a list of results, but they return different entity data depending on whether changes have been applied to the entities. In rare cases, it’s also possible for changes to indexes to be applied prior to changes to entities, and for the apply phase to fail and leave committed changes unapplied until the next transaction on the entity group. If you need stronger guarantees, fetch or query entities within transactions to ensure all committed changes are applied before the data is used. A query with an ancestor filter knows its entity group, and can therefore offer the same strong consistency guarantees within transactions as fetches. But many useful queries span entity groups, and therefore cannot be performed in transactions. If it is important to your application that a result for a nontransactional query match the criteria exactly, verify the result in the application code before using it. How Indexes Are Updated | 181
- 204. CHAPTER 7 Data Modeling with Python Data modeling is the process of translating the data requirements of your application to the features of your data storage technology. While the application deals in players, towns, weapons, potions, and gold, the datastore knows only entities, entity groups, keys, properties, and indexes. The data model describes how the data is stored and how it is manipulated. Entities represent players and game objects; properties describe the status of objects and the relationships between them. When an object changes location, the data is updated in a transaction, so the object cannot be in two places at once. When a player wants to know about the weapons in her inventory, the application performs a query for all weapon objects whose location is the player, possibly requiring an index. In the last few chapters, we’ve been using the Python class db.Expando to create and manipulate entities and their properties. As we’ve been doing it, this class illustrates the flexible nature of the datastore. The datastore itself does not impose or enforce a structure on entities or their properties, giving the application control over how indi- vidual entities represent data objects. This flexibility is also an essential feature for scalability: changing the structure of millions of records is a large task, and the proper strategy for doing this is specific to the task and the application. But structure is needed. Every player has a number of health points, and a Player entity without a health property, or with a health property whose value is not an integer, is likely to confuse the battle system. The data ought to conform to a structure, or schema, to meet the expectations of the code. Because the datastore does not enforce this schema itself—the datastore is schemaless—it is up to the application to ensure that entities are created and updated properly. App Engine includes a data modeling library for defining and enforcing data schemas in Python. This library resides in the google.appengine.ext.db package. It includes several related classes for representing data objects, including db.Model, db.Expando and db.PolyModel. To give structure to entities of a given kind, you create a subclass of one of these classes. The definition of the class specifies the properties for those objects, their allowed value types, and other requirements. 183
- 205. In this chapter,we’ll introduce the Python data modeling library and discuss how to use it to enforce a schema for the otherwise schemaless datastore. We’ll also discuss how the library works and how to extend it. Models and Properties The db.Model superclass lets you specify a structure for every entity of a kind. This structure can include the names of the properties, the types of the values allowed for those properties, whether the property is required or optional, and a default value. Here is a definition of a Book class similar to the one we created in Chapter 4: from google.appengine.ext import db import datetime class Book(db.Model): title = db.StringProperty(required=True) author = db.StringProperty(required=True) copyright_year = db.IntegerProperty() author_birthdate = db.DateProperty() obj = Book(title='The Grapes of Wrath', author='John Steinbeck') obj.copyright_year = 1939 obj.author_birthdate = datetime.date(1902, 2, 27) obj.put() This Book class inherits from db.Model. In the class definition, we declare that all Book entities have four properties, and we declare their value types: title and author are strings, copyright_year is an integer, and author_birthdate is a date-time. If someone tries to assign a value of the wrong type to one of these properties, the assignment raises a db.BadValueError. We also declare that title and author are required properties. If someone tries to create a Book without these properties set as arguments to the Book constructor, the attempt raises a db.BadValueError. copyright_year and author_birthdate are optional, so we can leave them unset on construction, and assign values to the properties later. If these properties are not set by the time the object is saved, the resulting entity will not have these properties—and that’s allowed by this model. A property declaration ensures that the entity created from the object has a value for the property, possibly None. As we’ll see in the next section, you can further specify what values are considered valid using arguments to the property declaration. A model class that inherits from db.Model ignores all attributes that are not declared as properties when it comes time to save the object to the datastore. In the resulting entity, all declared properties are set, and no others. This is the sole difference between db.Model and db.Expando. A db.Model class ignores undeclared properties. A db.Expando class saves all attributes of the object as properties 184 | Chapter 7: Data Modeling with Python
- 206. of the correspondingentity. That is, a model using a db.Expando class “expands” to accommodate assignments to undeclared properties. You can use property declarations with db.Expando just as with db.Model. The result is a data object that validates the values of the declared properties, and accepts any values for additional undeclared properties. The official documentation refers to properties with declarations as static properties and properties on a db.Expando without declarations as dynamic properties. These terms have a nice correspondence with the notions of static and dynamic typing in programming languages. Prop- erty declarations implement a sort of runtime validated static typing for model classes, on top of Python’s own dynamic typing. As we’ll see, property declarations are even more powerful than static typing, because they can validate more than just the type of the value. For both db.Model and db.Expando, object attributes whose names begin with an un- derscore (_) are always ignored. You can use these private attributes to attach transient data or functions to model objects. (It’s possible to create an entity with a property whose name starts with an underscore; this convention only applies to object attributes in the modeling API.) Because model objects also have attributes that are methods and other features, you cannot use certain names for properties in the Python model API. Some of the more pernicious reserved names are key, kind, and parent. The official documentation has a complete list of reserved names. In the next section, we’ll see a way to use these reserved names for datastore properties even though they aren’t allowed as attribute names in the API. Beyond the model definition, db.Model and db.Expando have the same interface for sav- ing, fetching, and deleting entities, and for performing queries and transactions. db.Expando is a subclass of db.Model. Property Declarations You declare a property for a model by assigning a property declaration object to an attribute of the model class. The name of the attribute is the name of the datastore property. The value is an object that describes the terms of the declaration. As discussed earlier, the db.StringProperty object assigned to the title class attribute says that the entity that an instance of the class represents can only have a string value for its title property. The required=True argument to the db.StringProperty constructor says that the object is not valid unless it has a value for the title property. This can look a little confusing if you’re expecting the class attribute to shine through as an attribute of an instance of the class, as it normally does in Python. Instead, the Property Declarations | 185
- 207. db.Model class hooksinto the attribute assignment mechanism so it can use the property declaration to validate a value assigned to an attribute of the object. In Python terms, the model uses property descriptors to enhance the behavior of attribute assignment. Property declarations act as intermediaries between the application and the datastore. They can ensure that only values that meet certain criteria are assigned to properties. They can assign default values when constructing an object. They can even convert values between a data type used by the application and one of the datastore’s native value types, or otherwise customize how values are stored. Property Value Types db.StringProperty is an example of a property declaration class. There are several property declaration classes included with the Python SDK, one for each native data- store type. Each one ensures that the property can only be assigned a value of the corresponding type: class Book(db.Model): title = db.StringProperty() b = Book() b.title = 99 # db.BadValueError, title must be a string b.title = 'The Grapes of Wrath' # OK Table 7-1 lists the datastore native value types and their corresponding property dec- laration classes. Table 7-1. Datastore property value types and the corresponding property declaration classes Data type Python type Property class Unicode text string (up to unicode or str (converted to unicode as ASCII) db.StringProperty 500 bytes, indexed) Long Unicode text string db.Text db.TextProperty (not indexed) Short byte string (up to 500 db.ByteString db.ByteStringProperty bytes, indexed) Long byte string (not in- db.Blob db.BlobProperty dexed) Boolean bool db.BooleanProperty Integer (64-bit) int or long (converted to 64-bit long) db.IntegerProperty Float (double precision) float db.FloatProperty Date-time datetime.date db.DateProperty datetime.datetime db.DateTimeProperty datetime.time db.TimeProperty 186 | Chapter 7: Data Modeling with Python
- 208. Data type Python type Property class Entity key db.Key or a model instance db.ReferenceProperty, db.SelfReferenceProperty A Google account users.User db.UserProperty Property Validation You can customize the behavior of a property declaration by passing arguments to the declaration’s constructor. We’ve already seen one example: the required argument. All property declaration classes support the required argument. If True, the property is required and must not be None. You must provide an initial value for each required property to the constructor when creating a new object. (You can provide an initial value for any property this way.) class Book(db.Model): title = db.StringProperty(required=True) b = Book() # db.BadValueError, title is required b = Book(title='The Grapes of Wrath') # OK The datastore makes a distinction between a property that is not set and a property that is set to the null value (None). Property declarations do not make this distinction, because all declared properties must be set (possibly to None). Unless you say otherwise, the default value for declared properties is None, so the required validator treats the None value as an unspecified property. You can change the default value with the default argument. When you create an object without a value for a property that has a default value, the constructor assigns the default value to the property. A property that is required and has a default value uses the default if constructed with- out an explicit value. The value can never be None. class Book(db.Model): rating = db.IntegerProperty(default=1) b = Book() # b.rating == 1 b = Book(rating=5) # b.rating == 5 By default, the name of the class attribute is used as the name of the datastore property. If you wish to use a different name for the datastore property than is used for the attribute, specify a name argument. This allows you to use names already taken by the API for class or instance attributes as datastore properties. class Song(db.Model): song_key = db.StringProperty(name='key') s = Song() Property Declarations | 187
- 209. s.song_key = 'C#min' # The song_key attribute is stored as the # datastore property named 'key'. s.put() You can declare that a property should contain only one of a fixed set of values by providing a list of possible values as the choices argument. If None is not one of the choices, this acts as a more restrictive form of required: the property must be set to one of the valid choices using a keyword argument to the constructor. _KEYS = ['C', 'C min', 'C 7', 'C#', 'C# min', 'C# 7', # ... ] class Song(db.Model): song_key = db.StringProperty(choices=_KEYS) s = Song(song_key='H min') # db.BadValueError s = Song() # db.BadValueError, None is not an option s = Song(song_key='C# min') # OK All of these features validate the value assigned to a property, and raise a db.BadValueError if the value does not meet the appropriate conditions. For even greater control over value validation, you can define your own validation function and assign it to a property declaration as the validator argument. The function should take the value as an argument, and raise a db.BadValueError (or an exception of your choos- ing) if the value should not be allowed. def is_recent_year(val): if val < 1923: raise db.BadValueError class Book(db.Model): copyright_year = db.IntegerProperty(validator=is_recent_year) b = Book(copyright_year=1922) # db.BadValueError b = Book(copyright_year=1924) # OK Nonindexed Properties In Chapter 5, we mentioned that you can set properties of an entity in such a way that they are available on the entity, but are considered unset for the purposes of indexes. In the Python API, you establish a property as nonindexed using a property declaration. If the property declaration is given an indexed argument of False, entities created with that model class will set that property as nonindexed. 188 | Chapter 7: Data Modeling with Python
- 210. class Book(db.Model): first_sentence = db.StringProperty(indexed=False) b = Book() b.first_sentence = "On the Internet, popularity is swift and fleeting." b.put() # Count the number of Book entities with # an indexed first_sentence property... c = Book.all().order('first_sentence').count(1000) # c = 0 Automatic Values Several property declaration classes include features for setting values automatically. The db.DateProperty, db.DateTimeProperty, and db.TimeProperty classes can populate the value automatically with the current date and time. To enable this behavior, you provide the auto_now or auto_now_add arguments to the property declaration. If you set auto_now=True, the declaration class overwrites the property value with the current date and time when you save the object. This is useful when you want to keep track of the last time an object was saved. class Book(db.Model): last_updated = db.DateTimeProperty(auto_now=True) b = Book() b.put() # last_updated is set to the current time # ... b.put() # last_updated is set to the current time again If you set auto_now_add=True, the property is set to the current time only when the object is saved for the first time. Subsequent saves do not overwrite the value. class Book(db.Model): create_time = db.DateTimeProperty(auto_now_add=True) b = Book() b.put() # create_time is set to the current time # ... b.put() # create_time stays the same The db.UserProperty declaration class also includes an automatic value feature. If you provide the argument auto_current_user=True, the value is set to the user accessing the current request handler if the user is signed in. If you provide auto_current_user_add=True, the value is only set to the current user when the entity Property Declarations | 189
- 211. is saved forthe first time, and left untouched thereafter. If the current user is not signed in, the value is set to None. class BookReview(db.Model): created_by_user = db.UserProperty(auto_current_user_add=True) last_edited_by_user = db.UserProperty(auto_current_user=True) br = BookReview() br.put() # created_by_user and last_edited_by_user set # ... br.put() # last_edited_by_user set again At first glance, it might seem reasonable to set a default for a db.UserProperty this way: from google.appengine.api import users class BookReview(db.Model): created_by_user = db.UserProperty( default=users.get_current_user()) # WRONG This would set the default value to be the user who is signed in when the class is imported. Requests that use a cached instance of the appli- cation may end up using a previous user instead of the current user as the default. To guard against this mistake, db.UserProperty does not accept the default argument. You can use only auto_current_user or auto_current_user_add to set an automatic value. List Properties The data modeling API provides a property declaration class for multivalued properties, called db.ListProperty. This class ensures that every value for the property is of the same type. You pass this type to the property declaration, like so: class Book(db.Model): tags = db.ListProperty(basestring) b = Book() b.tags = ['python', 'app engine', 'data'] The type argument to the db.ListProperty constructor must be the Python represen- tation of one of the native datastore types. Refer back to Table 4-1 for a complete list. The datastore does not distinguish between a multivalued property with no elements and no property at all. As such, an undeclared property on a db.Expando object can’t store the empty list. If it did, when the entity is loaded back into an object, the property simply wouldn’t be there, potentially confusing code that’s expecting to find an empty 190 | Chapter 7: Data Modeling with Python
- 212. list. To avoidconfusion, db.Expando disallows assigning an empty list to an undeclared property. The db.ListProperty declaration makes it possible to keep an empty list value on a multivalued property. The declaration interprets the state of an entity that doesn’t have the declared property as the property being set to the empty list, and maintains that distinction on the object. This also means that you cannot assign None to a declared list property—but this isn’t of the expected type for the property anyway. The datastore does distinguish between a property with a single value and a multivalued property with a single value. An undeclared property on a db.Expando object can store a list with one element, and represent it as a list value the next time the entity is loaded. The example above declares a list of string values. (basestring is the Python base type for str and unicode.) This case is so common that the API also provides db.StringListProperty. You can provide a default value to db.ListProperty using the default argument. If you specify a nonempty list as the default, a shallow copy of the list value is made for each new object that doesn’t have an initial value for the property. db.ListProperty does not support the required validator, since every list property tech- nically has a list value (possibly empty). If you wish to disallow the empty list, you can provide your own validator function that does so: def is_not_empty(lst): if len(lst) == 0: raise db.BadValueError class Book(db.Model): tags = db.ListProperty(basestring, validator=is_not_empty) b = Book(tags=[]) # db.BadValueError b = Book() # db.BadValueError, default "tags" is empty b = Book(tags=['awesome']) # OK db.ListProperty does not allow None as an element in the list because it doesn’t match the required value type. It is possible to store None as an element in a list for an unde- clared property. Models and Schema Migration Property declarations prevent the application from creating an invalid data object, or assigning an invalid value to a property. If the application always uses the same model classes to create and manipulate entities, then all entities in the datastore will be con- sistent with the rules you establish using property declarations. Property Declarations | 191
- 213. In real life,it is possible for an entity that does not fit a model to exist in the datastore. When you change a model class—and you will change model classes in the lifetime of your application—you are making a change to your application code, not the datastore. Entities created from a previous version of a model stay the way they are. If an existing entity does not comply with the validity requirements of a model class, you’ll get a db.BadValueError when you try to fetch the entity from the datastore. Fetching an entity gets the entity’s data, then calls the model class constructor with its values. This executes each property’s validation routines on the data. Some model changes are “backward compatible” such that old entities can be loaded into the new model class and be considered valid. Whether it is sufficient to make a backward-compatible change without updating existing entities depends on your ap- plication. Changing the type of a property declaration or adding a required property are almost always incompatible changes. Adding an optional property will not cause a db.BadValueError when an old entity is loaded, but if you have indexes on the new property, old entities will not appear in those indexes (and therefore won’t be results for those queries) until the entities are loaded and then saved with the new property’s default value. The most straightforward way to migrate old entities to new schemas is to write a script that queries all of the entities and applies the changes. We’ll discuss how to implement this kind of batch operation in Chapter 12. Modeling Relationships You can model relationships between entities by storing entity keys as property values. The Python data modeling interface includes several powerful features for managing relationships. The db.ReferenceProperty declaration describes a relationship between one model class and another. It stores the key of an entity as the property value. The first argument to the db.ReferenceProperty constructor is the model class of the kind of entity refer- enced by the property. If someone creates a relationship to an entity that is not of the appropriate kind, the assignment raises a db.BadValueError. You can assign a data object directly to the property. The property declaration stores the key of the object as the property’s value to create the relationship. You can also assign a db.Key directly. class Book(db.Model): title = db.StringProperty() author = db.StringProperty() class BookReview(db.Model): book = db.ReferenceProperty(Book, collection_name='reviews') b = Book() 192 | Chapter 7: Data Modeling with Python
- 214. b.put() br = BookReview() br.book = b # sets br's 'book' property to b's key br.book = b.key() # same thing We’ll explain what collection_name does in a moment. The referenced object must have a “complete” key before it can be assigned to a refer- ence property. A key is complete when it has all of its parts, including the string name or the system-assigned numeric ID. If you create a new object without a key name, the key is not complete until you save the object. When you save the object, the system completes the key with a numeric ID. If you create the object (or a db.Key) with a key name, the key is already complete, and you can use it for a reference without saving it first. b = Book() br = BookReview() br.book = b # db.BadValueError, b's key is not complete b.put() br.book = b # OK, b's key has system ID b = Book(key_name='The_Grapes_of_Wrath') br = BookReview() br.book = b # OK, b's key has a name db.put([b, br]) A model class must be defined before it can be the subject of a db.ReferenceProperty. To declare a reference property that can refer to another instance of the same class, you use a different declaration, db.SelfReferenceProperty. class Book(db.Model): previous_edition = db.SelfReferenceProperty() b1 = Book() b2 = Book() b2.previous_edition = b1 Reference properties have a powerful and intuitive syntax for accessing referenced ob- jects. When you access the value of a reference property, the property fetches the entity from the datastore using the stored key, then returns it as an instance of its model class. A referenced entity is loaded “lazily”: it is not fetched from the datastore until the property is dereferenced. br = db.get(book_review_key) # br is a BookReview instance title = br.book.title # fetches book, gets its title property Modeling Relationships | 193
- 215. This automatic dereferencingof reference properties occurs the first time you access the reference property. Subsequent uses of the property use the in-memory instance of the data object. This caching of the referenced entity is specific to the object with the property. If another object has a reference to the same entity, accessing its reference fetches the entity anew. db.ReferenceProperty does another clever thing: it creates automatic back-references from a referenced object to the objects that refer to it. If a BookReview class has a reference property that refers to the Book class, the Book class gets a special property whose name is specified by the collection_name argument to the declaration (e.g., reviews). This property is special because it isn’t actually a property stored on the entity. Instead, when you access the back-reference property, the API performs a datastore query for all BookReview entities whose reference property equals the key of the Book. Since this is a single-property query, it uses the built-in indexes, and never requires a custom index. b = db.get(book_key) # b is a Book instance for review in b.reviews: # review is a BookReview instance # ... If you don’t specify a collection_name, the name of the back-reference property is the name of the referring class followed by _set. If a class has multiple reference properties that refer to the same class, you must provide a collection_name to disambiguate the back-reference properties. class BookReview(db.Model): # Book gets a BookReview_set special property. book = db.ReferenceProperty(Book) # Book gets a recommended_book_set special property. recommended_book = db.ReferenceProperty(Book, collection_name='recommended_book_set') Because the back-reference property is implemented as a query, it incurs no overhead if you don’t use it. As with storing db.Key values as properties, neither the datastore nor the property dec- laration requires that a reference property refer to an entity that exists. Dereferencing a reference property that points to an entity that does not exist raises an error. (As of this writing, there is no specific exception for this case, just an Error with a descriptive message.) Keys cannot change, so a relationship is only severed when the referenced entity is deleted from the datastore. 194 | Chapter 7: Data Modeling with Python
- 216. One-to-Many Relationships A referenceproperty and its corresponding back-reference represent a one-to-many relationship between classes in your data model. The reference property establishes a one-way relationship from one entity to another, and the declaration sets up the back- reference mechanism on the referenced class. The back-reference uses the built-in query index, so determining which objects refer to the referenced object is reasonably fast. It’s not quite as fast as storing a list of keys on a property, but it’s easier to maintain. A common use of one-to-many relationships is to model ownership. In the previous example, each BookReview was related to a single Book, and a Book could have many BookReviews. The BookReviews belong to the Book. One-to-One Relationships You can also use a reference property to model a one-to-one relationship. The property declaration doesn’t enforce that only one entity can refer to a given entity, but this is easy to maintain in the application code. Because the performance of queries scales with the size of the result set and not the size of the data set, it’s usually sufficient to use the back-reference query to follow a one-to-one relationship back to the object with the reference. If you’d prefer not to use a query to traverse the back-reference, you could also store a reference on the second object back to the first, at the expense of having to maintain the relationship in two places. This is tricky, because the class has to be defined before it can be the subject of a ReferenceProperty. One option is to use db.Expando and an undeclared property for one of the classes. A one-to-one relationship can be used to model partnership. A good use of one-to-one relationships in App Engine is to split a large object into multiple entities to provide selective access to its properties. A player might have an avatar image up to 64 kilobytes in size, but the application probably doesn’t need the 64 KB of image data every time it fetches the Player entity. You can create a separate PlayerAvatarImage entity to con- tain the image, and establish a one-to-one relationship by creating a reference property from the Player to the PlayerAvatarImage. The application must know to delete the related objects when deleting a Player: class PlayerAvatarImage(db.Model): image_data = db.BlobProperty() mime_type = db.StringProperty() class Player(db.Model): name = db.StringProperty() avatar = db.ReferenceProperty(PlayerAvatarImage) # Fetch the name of the player (a string) a # reference to the avatar image (a key). p = db.get(player_key) Modeling Relationships | 195
- 217. # Fetch theavatar image entity and access its # image_data property. image_data = p.avatar.image_data Many-to-Many Relationships A many-to-many relationship is a type of relationship between entities of two kinds where entities of either kind can have that relationship with many entities of the other kind, and vice versa. For instance, a player may be a member of one or more guilds, and a guild can have many members. There are at least two ways to implement many-to-many relationships using the data- store. Let’s consider two of these. The first method we’ll call “the key list method,” and the second we’ll call “the link model method.” The key list method With the key list method, you store a list of entity keys on one side of the relationship using a db.ListProperty. Such a declaration does not have any of the features of a db.ReferenceProperty such as back-references or automatic dereferencing, because it does not involve that class. To model the relationship in the other direction, you can implement the back-reference feature using a method and the Python annotation @property: class Player(db.Model): name = db.StringProperty() guilds = db.ListProperty(db.Key) class Guild(db.Model): name = db.StringProperty() @property def members(self): return Player.all().filter('guilds', self.key()) # Guilds to which a player belongs: p = db.get(player_key) guilds = db.get(p.guilds) # batch get using list of keys for guild in guilds: # ... # Players that belong to a guild: g = db.get(guild_key) for player in g.members: # ... Instead of manipulating the list of keys, you could implement automatic dereferencing using advanced Python techniques to extend how the values in the list property are accessed. A good way to do this is with a custom property declaration. We’ll consider this in a later section. 196 | Chapter 7: Data Modeling with Python
- 218. The key listmethod is best suited for situations where there are fewer objects on one side of the relationship than on the other, and the short list is small enough to store directly on an entity. In this example, many players each belong to a few guilds; each player has a short list of guilds, while each guild may have a long list of players. We put the list property on the Player side of the relationship to keep the entity small, and use queries to produce the long list when it is needed. The link model method The link model method represents each relationship as an entity. The relationship entity has reference properties pointing to the related classes. You traverse the relationship by going through the relationship entity via the back-references. class Player(db.Model): name = db.StringProperty() class Guild(db.Model): name = db.StringProperty() class GuildMembership(db.Model): player = db.ReferenceProperty(Player, collection_name='guild_memberships') guild = db.ReferenceProperty(Guild, collection_name='player_memberships') p = Player() g = Guild() db.put([p, g]) gm = GuildMembership(player=p, guild=g) db.put(gm) # Guilds to which a player belongs: for gm in p.guild_memberships: guild_name = gm.guild.name # ... # Players that belong to a guild: for gm in g.player_memberships: player_name = gm.player.name # ... This technique is similar to how you’d use “join tables” in a SQL database. It’s a good choice if either side of the relationship may get too large to store on the entity itself. You can also use the relationship entity to store metadata about the relationship (such as when the player joined the guild), or model more complex relationships between multiple classes. The link model method is more expensive than the key list method. It requires fetching the relationship entity to access the related object. Modeling Relationships | 197
- 219. Remember that AppEngine doesn’t support SQL-style join queries on these objects. You can achieve a limited sort of join by repeating infor- mation from the data objects on the link model objects, using code on the model classes to keep the values in sync. To do this with strong consistency, the link model object and the two related objects would need to be in the same entity group, which is not always possible or practical. If eventual consistency would suffice, you could use task queues to propagate the information. See Chapter 13. Model Inheritance In data modeling, it’s often useful to derive new kinds of objects from other kinds. The game world may contain many different kinds of carryable objects, with shared prop- erties and features common to all objects you can carry. Since you implement classes from the data model as Python classes, you’d expect to be able to use inheritance in the implementation to represent inheritance in the model. And you can, sort of. If you define a class based on either db.Model or db.Expando, you can create other classes that inherit from that data class, like so: class CarryableObject(db.Model): weight = db.IntegerProperty() location = db.ReferenceProperty(Location) class Bottle(CarryableObject): contents = db.StringProperty() amount = db.IntegerProperty() is_closed = db.BooleanProperty() The subclass inherits the property declarations of the parent class. A Bottle has five property declarations: weight, location, contents, amount, and is_closed. Objects based on the child class will be stored as entities whose kind is the name of the child class. The datastore has no notion of inheritance, and so by default will not treat Bottle entities as if they are CarryableObject entities. This is mostly significant for queries, and we have a solution for that in the next section. If a child class declares a property already declared by a parent class, the class definition raises a db.DuplicatePropertyError. The data modeling API does not support overrid- ing property declarations in subclasses. A model class can inherit from multiple classes, using Python’s own support for mul- tiple inheritance: class PourableObject(GameObject): contents = db.StringProperty() amount = db.IntegerProperty() 198 | Chapter 7: Data Modeling with Python
- 220. class Bottle(CarryableObject, PourableObject): is_closed = db.BooleanProperty() Each parent class must not declare a property with the same name as declarations in the other parent classes, or the class definition raises a db.DuplicatePropertyError. However, the modeling API does the work to support “diamond inheritance,” where two parent classes themselves share a parent class: class GameObject(db.Model): name = db.StringProperty() location = db.ReferenceProperty(Location) class CarryableObject(GameObject): weight = db.IntegerProperty() class PourableObject(GameObject): contents = db.StringProperty() amount = db.IntegerProperty() class Bottle(CarryableObject, PourableObject): is_closed = db.BooleanProperty() In this example, both CarryableObject and PourableObject inherit two property dec- larations from GameObject, and are both used as parent classes to Bottle. The model API allows this because the two properties are defined in the same class, so there is no conflict. Bottle gets its name and location declarations from GameObject. Queries and PolyModels The datastore knows nothing of our modeling classes and inheritance. Instances of the Bottle class are stored as entities of the kind 'Bottle', with no inherent knowledge of the parent classes. It’d be nice to be able to perform a query for CarryableObject entities and get back Bottle entities and others. That is, it’d be nice if a query could treat Bottle entities as if they were instances of the parent classes, as Python does in our application code. We want polymorphism in our queries. For this, the data modeling API provides a special base class: db.PolyModel. Model classes using this base class support polymorphic queries. Consider the Bottle class defined previously. Let’s change the base class of GameObject to db.PolyModel, like so: from google.appengine.ext.db import PolyModel class GameObject(db.PolyModel): # ... We can now perform queries for any kind in the hierarchy, and get the expected results: here = db.get(location_key) q = CarryableObject.all() q.filter('location', here) q.filter('weight >', 100) Queries and PolyModels | 199
- 221. for obj inq: # obj is a carryable object that is here # and weighs more than 100 kilos. # ... This query can return any CarryableObject, including Bottle entities. The query can use filters on any property of the specified class (such as weight from CarryableObject) or parent classes (such as location from GameObject). Behind the scenes, db.PolyModel does three clever things differently from its cousins: • Objects of the class GameObject or any of its child classes are all stored as entities of the kind 'GameObject'. • All such objects are given a property named class that represents the inheritance hierarchy starting from the root class. This is a multivalued property, where each value is the name of an ancestor class, in order. • Queries for objects of any kind in the hierarchy are translated by the db.PolyModel class into queries for the base class, with additional equality filters that compare the class being queried to the class property’s values. In short, db.PolyModel stores information about the inheritance hierarchy on the enti- ties, then uses it for queries to support polymorphism. Each model class that inherits directly from db.PolyModel is the root of a class hierarchy. All objects from the hierarchy are stored as entities whose kind is the name of the root class. As such, your data will be easier to maintain if you use many root classes to form many class hierarchies, as opposed to putting all classes in a single hierarchy. That way, the datastore viewer and bulk loading tools can still use the datastore’s built-in notion of entity kinds to distinguish between kinds of objects. Creating Your Own Property Classes The property declaration classes serve several functions in your data model: Value validation The model calls the class when a value is assigned to the property, and the class can raise an exception if the value does not meet its conditions. Type conversion The model calls the class to convert from the value type used by the app to one of the core datastore types for storage, and back again. Default behavior The model calls the class if no value was assigned to determine an appropriate default value. Every property declaration class inherits from the db.Property base class. This class implements features common to all property declarations, including support for the 200 | Chapter 7: Data Modeling with Python
- 222. common constructor arguments(such as required, name, and indexed). Declaration classes override methods and members to specialize the validation and type conversion routines. Validating Property Values Here is a very simple property declaration class. It accepts any string value, and stores it as a datastore short string (the default behavior for Python string values). from google.appengine.ext import db class PlayerNameProperty(db.Property): data_type = basestring def validate(self, value): value = super(PlayerNameProperty, self).validate(value) if value is not None and not isinstance(value, self.data_type): raise db.BadValueError('Property %s must be a %s.' % (self.name, self.data_type.__name__)) return value And here is how you would use the new property declaration: class Player(db.Model): player_name = PlayerNameProperty() p = Player() p.player_name = 'Ned Nederlander' p.player_name = 12345 # db.BadValueError The validate() method takes the value as an argument, and either returns the value, returns a different value, or raises an exception. The value returned by the method becomes the application-facing value for the attribute, so you can use the validate() method for things like type coercion. In this example, the method raises a db.BadValueError if the value is not a string or None. The exception message can refer to the name of the property using self.name. The data_type member is used by the base class. It represents the core datastore type the property uses to store the value. For string values, this is basestring. The validate() method should call the superclass’s implementation before checking its own conditions. The base class’s validator supports the required, choices, and validator arguments of the declaration constructor. If the app does not provide a value for a property when it constructs the data object, the property starts out with a default value. This default value is passed to the validate() method during the object constructor. If it is appropriate for your property declaration to allow a default value of None, make sure your validate() method allows it. Creating Your Own Property Classes | 201
- 223. So far, thisexample doesn’t do much beyond db.StringProperty. This by itself can be useful to give the property type a class for future expansion. Let’s add a requirement that player names be between 6 and 30 characters in length by extending the validate() method: class PlayerNameProperty(db.Property): data_type = basestring def validate(self, value): value = super(PlayerNameProperty, self).validate(value) if value is not None: if not isinstance(value, self.data_type): raise db.BadValueError('Property %s must be a %s.' % (self.name, self.data_type.__name__)) if (len(value) < 6 or len(value) > 30): raise db.BadValueError(('Property %s must be between 6 and ' + '30 characters.') % self.name) return value The new validation logic disallows strings with an inappropriate length: p = Player() p.player_name = 'Ned' # db.BadValueError p.player_name = 'Ned Nederlander' # OK p = Player(player_name = 'Ned') # db.BadValueError Marshaling Value Types The datastore supports a fixed set of core value types for properties, listed in Ta- ble 4-1. A property declaration can support the use of other types of values in the attributes of model instances by marshaling between the desired type and one of the core datastore types. For example, the db.ListProperty class converts between the empty list of the app side and the condition of being unset on the datastore side. The get_value_for_datastore() method converts the application value to the datastore value. Its argument is the complete model object, so you can access other aspects of the model when doing the conversion. The make_value_from_datastore() method takes the datastore value and converts it to the type to be used in the application. It takes the datastore value and returns the desired object attribute value. Say we wanted to represent player name values within the application using a PlayerName class instead of a simple string. Each player name has a surname and an optional first name. We can store this value as a single property, using the property declaration to convert between the application type (PlayerName) and a core datastore type (such as unicode). class PlayerName(object): def __init__(self, first_name, surname): 202 | Chapter 7: Data Modeling with Python
- 224. self.first_name = first_name self.surname = surname def is_valid(self): return (isinstance(self.first_name, unicode) and isinstance(self.surname, unicode) and len(self.surname) >= 6) class PlayerNameProperty(db.Property): data_type = basestring def validate(self, value): value = super(PlayerNameProperty, self).validate(value) if value is not None: if not isinstance(value, PlayerName): raise db.BadValueError('Property %s must be a PlayerName.' % (self.name)) # Let the data class have a say in validity. if not value.is_valid(): raise db.BadValueError('Property %s must be a valid PlayerName.' % self.name) # Disallow the serialization delimiter in the first field. if value.surname.find('|') != -1: raise db.BadValueError(('PlayerName surname in property %s cannot ' + 'contain a "|".') % self.name) return value def get_value_for_datastore(self, model_instance): # Convert the data object's PlayerName to a unicode. return (getattr(model_instance, self.name).surname + u'|' + getattr(model_instance, self.name).first_name) def make_value_for_datastore(self, value): # Convert a unicode to a PlayerName. i = value.find(u'|') return PlayerName(first_name=value[i+1:], surname=value[:i]) And here’s how you’d use it: p = Player() p.player_name = PlayerName('Ned', 'Nederlander') p.player_name = PlayerName('Ned', 'Neder|lander') # db.BadValueError, surname contains serialization delimiter p.player_name = PlayerName('Ned', 'Neder') # db.BadValueError, PlayerName.is_valid() == False Here, the application value type is a PlayerName instance, and the datastore value type is that value encoded as a Unicode string. The encoding format is the surname field, followed by a delimiter, followed by the first_name field. We disallow the delimiter character in the surname using the validate() method. (Instead of disallowing Creating Your Own Property Classes | 203
- 225. it, we couldalso escape it in get_value_for_datastore() and unescape it in make_value_for_datastore().) In this example, PlayerName('Ned', 'Nederlander') is stored as this Unicode string: Nederlander|Ned The datastore value puts the surname first so that the datastore will sort PlayerName values first by surname, then by first name. In general, you choose a serialization format that has the desired ordering characteristics for your custom property type. (The core type you choose also impacts how your values are ordered when mixed with other types, though if you’re modeling consistently this isn’t usually an issue.) If the conversion from the application type to the datastore type may fail, put a check for the conversion failure in the validate() method. This way, the error is caught when the bad value is assigned, instead of when the object is saved. Customizing Default Values When the app constructs a data object and does not provide a value for a declared property, the model calls the property declaration class to determine a default value. The base class implementation sets the default value to None, and allows the app to customize the default value in the model using the default argument to the declaration. A few of the built-in declaration classes provide more sophisticated default values. For instance, if a db.DateTimeProperty was set with auto_now_add=True, the default value is the current system date and time. (db.DateTimeProperty uses get_value_for_datastore() to implement auto_now=True, so the value is updated whether or not it has a value.) The default value passes through the validation logic after it is set. This allows the app to customize the validation logic and disallow the default value. This is what happens when required=True: the base class’s validation logic disallows the None value, which is the base class’s default value. To specify custom default behavior, override the default_value() method. This method takes no arguments and returns the desired default value. Here’s a simple implementation of default_value() for PlayerNameProperty: class PlayerNameProperty(db.Property): # ... def default_value(self): default = super(PlayerNameProperty, self).default() if default is not None: return default return PlayerName('', 'Anonymous') 204 | Chapter 7: Data Modeling with Python
- 226. In this example,we call the superclass default() method to support the default argu- ment to the constructor, which allows the app to override the default value in the model. If that returns None, we create a new PlayerName instance to be the default value. Without further changes, this implementation breaks the required feature of the base class, because the value of the property is never None (unless the app explicitly assigns a None value). We can fix this by amending our validation logic to check self.required and disallow the anonymous PlayerName value if it’s True. Accepting Arguments If you want the application to be able to control the behavior of your custom property declaration class using arguments, you override the __init__() method. The method should call the superclass __init__() method to enable the features of the superclass that use arguments (like required). The Property API requires that the verbose_name property come first, but after that all __init__() arguments are keyword values. class PlayerNameProperty(db.Property): # ... def __init__(self, verbose_name=None, require_first_name=False, **kwds): super(PlayerNameProperty, self).__init__(verbose_name, **kwds) self.require_first_name = require_first_name def validate(self, value): value = super(PlayerNameProperty, self).validate(value) if value is not None: # ... if self.require_first_name and not value.first_name: raise db.BadValueError('Property %s PlayerName needs a first_name.' % self.name) # ... You’d use this feature like this: class Player(db.Model): player_name = PlayerNameProperty(require_first_name=True) p = Player(player_name=PlayerName('Ned', 'Nederlander')) p.player_name = PlayerName(u'', u'Charo') # db.BadValueError, first name required p = Player() # db.BadValueError, default value PlayerName('', 'Anonymous') has empty first_name Creating Your Own Property Classes | 205
- 228. CHAPTER 8 The Java Persistence API The App Engine Java SDK includes implementations of two data access interface standards: the Java Persistence API (JPA) and Java Data Objects (JDO). These interfaces provide two essential features. For one, these interfaces define a mechanism for describing the structure of data objects in terms of Java classes. You can use them to define and enforce consistent data schemas on top of App Engine’s schemaless datastore, and take advantage of type safety in the Java language. These interfaces serve as a data modeling layer. Because each interface is a standard supported by other popular data storage solutions, using a standard interface makes it easier to port an application to and from these other solutions. Different databases have varying degrees of support for these standards. Since the standards were developed with SQL-based relational databases in mind, the App Engine datastore can only be said to support a portion of the standard, and it is often easier to port away from App Engine than to it. But this alone adds value, as you can reserve the right to move your app to your company’s own servers at any time. These interfaces are a portability layer. The App Engine SDK uses an open source product called DataNucleus Access Platform as the basis for its implementations of JDO and JPA. Access Platform uses an adapter layer that translates both standards to an underlying implementation. The App Engine SDK includes an Access Platform adapter based on its low-level datastore API. Access Platform version 1.1 supports version 1.0 of the JPA specification and version 2.3 of JDO. The JDO and JPA standards are similar, and share similar roots. The concepts that apply to the App Engine datastore have similar interfaces in both standards but with different terminology and minor behavioral differences. Which one you choose may depend on how familiar you are with it, or how well it is implemented for your most likely porting target, if you have one in mind. 207
- 229. In this chapter,we’ll look at how to use JPA with App Engine. If you’d prefer to use JDO, check out the official documentation for App Engine, which includes a JDO tutorial. A quick note on terminology: JPA refers to data objects as “entities.” This similarity to datastore entities is convenient in some ways, and not so convenient in others. For this chapter, we’ll refer to JPA entities as “data objects” (or just “objects”) to avoid confu- sion with datastore entities. Setting Up JPA Using JPA requires a few steps common to all App Engine applications. We performed these steps back in Chapter 2. JPA needs a configuration file that specifies that you want to use the App Engine im- plementation of the interface. This file is named persistence.xml, and should appear in your WAR’s WEB-INF/classes/META-INF/ directory. If you’re using Eclipse, you can create this file in the src/META-INF/ directory, and Eclipse will copy it to the final location automatically. We saw the contents of this file in Example 2-14, but here it is again for reference: <?xml version="1.0" encoding="UTF-8" ?> <persistence xmlns="http://java.sun.com/xml/ns/persistence" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://java.sun.com/xml/ns/persistence http://java.sun.com/xml/ns/persistence/persistence_1_0.xsd" version="1.0"> <persistence-unit name="transactions-optional"> <provider> org.datanucleus.store.appengine.jpa.DatastorePersistenceProvider </provider> <properties> <property name="datanucleus.NontransactionalRead" value="true"/> <property name="datanucleus.NontransactionalWrite" value="true"/> <property name="datanucleus.ConnectionURL" value="appengine"/> </properties> </persistence-unit> </persistence> This configuration tells Access Platform to use the "appengine" adapter. It also says to allow reads and writes outside of transactions (NontransactionalRead and NontransactionalWrite are true), which fits the semantics of the datastore that we de- scribed earlier. We named this configuration set "transactions-optional" to match. Your application uses an EntityManager object to perform a set of datastore operations. The application creates an EntityManager using an EntityManagerFactory. The factory loads the configuration file and uses it for subsequent datastore interactions. You get an instance of the factory by calling a static method and passing it the name of the configuration set ("transactions-optional"): 208 | Chapter 8: The Java Persistence API
- 230. import javax.persistence.EntityManagerFactory; import javax.persistence.Persistence; // ... EntityManagerFactory emfInstance = Persistence.createEntityManagerFactory("transactions-optional"); The createEntityManagerFactory() static method performs a nontrivial amount of work. You can think of the factory as a connection pool, and each EntityManager as an individual connection. Since you only need one factory for the entire existence of the application, a best practice is to call the method only once, store the factory in a static member, and reuse it for multiple web requests. We did this in Chapter 2 with the EMF class, repeated here for reference: package clock; import javax.persistence.EntityManagerFactory; import javax.persistence.Persistence; public final class EMF { private static final EntityManagerFactory emfInstance = Persistence.createEntityManagerFactory("transactions-optional"); private EMF() {} public static EntityManagerFactory get() { return emfInstance; } } Access Platform hooks up the persistence plumbing to your JPA data classes in a post- compilation process that it calls “enhancement.” If you are using Eclipse with the Goo- gle Plugin, the plug-in performs this step automatically. If you are not using Eclipse, you must add the enhancement step to your build process. See the official documen- tation for information about performing this build step using Apache Ant. Entities and Keys In JPA, you define data classes as plain old Java objects (POJOs). You use annotations to tell JPA which classes to persist to the datastore, and how to store its members. Defining your data exclusively in terms of the Java classes your application uses makes it easy to manipulate your persistent data. It also makes it easy to test your application, since you can create mock data objects directly from the classes. JPA also lets you use an XML file instead of annotations to describe how to persist data classes. We’ll only cover the annotation style here, but if you are familiar with the XML file mechanism, you can use it with Access Platform. Entities and Keys | 209
- 231. Here’s a simpleexample of a data class: import java.util.Date; import javax.persistence.Entity; import javax.persistence.Id; @Entity(name = "Book") public class Book { @Id private String isbn; private String title; private String author; private int copyrightYear; private Date authorBirthdate; // ... constructors, accessors ... } JPA knows instances of the Book class can be made persistent (saved to the datastore) because of the @Entity annotation. This annotation takes a name argument that specifies the name to be used in JPA queries for objects of this class. The name must be unique across all data classes in the application. By default, the name of the datastore kind is derived from the name of the class. Spe- cifically, this is the simple name of the class, without the package path (everything after the last ., e.g., "Book"). If you have two data classes with the same simple name in different packages, you can specify an alternate kind name using the @Table annotation. (JPA was designed with tabular databases in mind, but the concept is equivalent.) import javax.persistence.Entity; import javax.persistence.Table; @Entity(name = "Book") @Table(name = "BookItem") public class Book { // ... } The Book class has five members. Four of these members are stored as properties on the datastore entity: title, author, copyrightYear, and authorBirthdate. The fifth member, isbn, represents the key name for the entity. JPA knows this because the member has the @Id annotation, and because the type of the member is String. Every data class needs a member that represents the object’s primary key, annotated with @Id. If the type of this member is String and it has no other annotations, then the key has no ancestors, and the value of the member is the string key name. The appli- cation must set this field before saving the object for the first time. To tell JPA to let the datastore assign a unique numeric ID instead of using an app- provided key name string, you declare the member with a type of Long and give it the annotation @GeneratedValue(strategy = GenerationType.IDENTITY), like so: 210 | Chapter 8: The Java Persistence API
- 232. import javax.persistence.Entity; import javax.persistence.GeneratedValue; import javax.persistence.GenerationType; import javax.persistence.Id; @Entity(name = "Book") public class Book { @Id @GeneratedValue(strategy = GenerationType.IDENTITY) private Long id; // ... } The member is set with the system-assigned ID when the object is saved to the datastore for the first time. These simple key member types are sufficient for entities without ancestors. Together with the entity kind ("Book"), the member represents the complete key of a root entity. If an instance of the class may represent an entity with ancestors, the key member must be able to represent the full key path. There are two ways to do this. One way is to declare the type of the key member to be the com.google.appen gine.api.datastore.Key class: import javax.persistence.Entity; import javax.persistence.Id; import com.google.appengine.api.datastore.Key; @Entity(name = "Book") public class Book { @Id private Key id; // ... } You can use this key member type to create a complete Key with a string name. You can also use system-assigned numeric IDs with ancestors by using the @Generated Value annotation, then assigning a Key value with neither the name nor the ID set. If you’d prefer not to create a dependency on an App Engine–specific class, there is another way to implement a key with ancestors: declare the ID field’s type as String and use a DataNucleus JPA extension that encodes the complete key as a string value, like so: import javax.persistence.Entity; import javax.persistence.Id; import org.datanucleus.jpa.annotations.Extension; import com.google.appengine.api.datastore.Key; @Entity(name = "Book") public class Book { @Id Entities and Keys | 211
- 233. @Extension(vendorName = "datanucleus", key = "gae.encoded-pk", value = "true") private Key id; // ... } You can convert between a Key and a string-encoded key using the KeyFactory class’s keyToString() and stringToKey() methods. (Note that the Key class’s toString() method returns something else.) You can use a Key ID field or a string-encoded ID field in combination with the @GeneratedValue annotation to produce keys with ancestors and system-assigned nu- meric IDs. Entity Properties The fields of the object become the properties of the corresponding entity. The name of a field is used as the name of the property. The @Id field is not stored as a property value, only as the key. JPA and App Engine support many types of fields. Any of the types mentioned in Table 4-1 can be used as a field type. A field can contain a serializable object, stored as a single property. A field can also be a collection of one of the core datastore types or a serializable class, to be stored as a multivalued property. Additionally, App Engine supports JPA embedded data objects and relationships between entities using fields. In some cases, JPA must be told which fields to save to the datastore. For the Java standard types (such as Long or String or Date), JPA assumes that fields of those types should be saved. For other types, especially the datastore-specific classes such as datastore.ShortBlob, you must tell JPA to save the field by giving it the @Basic anno- tation. If you have a field that should not be saved to the datastore, give it the @Transient annotation. import java.util.List; import javax.persistence.Basic; import javax.persistence.Id; import javax.persistence.Transient; import com.google.appengine.api.datastore.ShortBlob; @Entity(name = "Book") public class Book { // ... private String title; // saved @Basic // saved private ShortBlob coverIcon; @Basic // saved 212 | Chapter 8: The Java Persistence API
- 234. private List<String> tags; @Transient // not saved private int debugAccessCount; } As with the low-level API, some types are widened before being stored. int and Integer are converted to Long, and float and Float become Double. With the JPA in- terface, these values are converted back to the declared field types when loaded into an object. A Serializable class can be used as a field type, using the @Lob annotation. These values are stored in serialized form as datastore.Blob values. As such, these values are not indexed, and cannot be used in queries. Collection types are stored as multivalued properties in iteration order. When loaded into the data class, multivalued properties are converted back into the specified col- lection type. By default, the name of a field is used as the name of the corresponding property. You can override this using the @Column annotation: import javax.persistence.Column; import javax.persistence.Entity; @Entity(name = "Book") public class Book { // ... @Column(name = "long_description") private String longDescription; } You can declare that the datastore property of a field should not be mentioned in indexes—the property of each entity should be created as a nonindexed property— using an @Extension annotation: import org.datanucleus.jpa.annotations.Extension; @Entity(name = "Book") public class Book { // ... @Extension(vendorName = "datanucleus", key = "gae.unindexed", value = "true") private String firstSentence; } Embedded Objects App Engine supports JPA embedded classes by storing the fields of the embedded class as properties on the same datastore entity as the fields of the primary class. You must Embedded Objects | 213
- 235. declare the classto embed using the @Embeddable annotation. To embed the class, simply use it as a field type: import javax.persistence.Embeddable; import javax.persistence.Entity; @Embeddable public class Publisher { private String name; private String address; private String city; private String stateOrProvince; private String postalCode; // ... } @Entity(name = "Book") public class Book { // ... private Publisher publisher; } Because fields of embedded classes are stored as separate properties, they are queryable just like other properties. You can refer to an embedded field in a property with the name of the outer field with a dot-notation, such as publisher.name. The actual property name is just the name of the inner field, and you can change this if needed using an @Column annotation. Saving, Fetching, and Deleting Objects To start a session with the datastore, you use the EntityManagerFactory to create an EntityManager. You must create a new EntityManager for each request handler, and close it when you’re done. import javax.persistence.EntityManager; // ... // emf is the EntityManagerFactory EntityManager em; try { em = emf.createEntityManager(); // ... do datastore stuff ... } finally { em.close(); } To create a new data object, you construct the data class, then call the EntityManager’s persist() method with the object: EntityManager em; try { em = emf.createEntityManager(); 214 | Chapter 8: The Java Persistence API
- 236. book = newBook(); book.title = "The Grapes of Wrath"; // ... em.persist(book); } finally { em.close(); } If you create an object with a complete key, and an entity with that key already exists in the datastore, saving the new object will overwrite the old one. In App Engine’s implementation, JPA’s merge() method is equivalent to persist() in this way. (Other implementations may do something different in this case.) To fetch an entity with a known key, you use the find() method. This method takes the class of the object in which to load the entity, and the key of the object. The key can be any appropriate type: a string key name, a numeric ID, a datastore.Key object, or a string-encoded complete key. The method returns an object of the given class, or null if no object with that key is found: Book book = em.find(Book.class, "9780596156732"); if (book == null) { // not found } The ability of find() to accept all four key types is nonstandard. To make your code more portable, only call find() using the type of key you used in the data class. When you create or fetch an entity (or get an entity back from a query), the data object becomes “attached” to (or managed by) the entity manager. If you make changes to an attached object and do not save them by calling the persist() method explicitly, the object is saved automatically when you close the entity manager. As we’ll see in the next section, if you need the entity to be updated in a transaction, you pass the updated object to the persist() method at the moment it needs to be saved. To delete an entity, you call the remove() method. This method takes the data object as its sole argument. The object still exists in memory after it is removed from the datastore. em.remove(book); The remove() method requires a loaded data object. There is no way to delete an entity with this method without fetching its object first. (You can delete entities without fetching them using a JPQL delete query. See the section “Queries and JPQL” on page 217.) Saving, Fetching, and Deleting Objects | 215
- 237. Remember to closethe EntityManager by calling its close() method. If you don’t, changes to objects will not be saved to the datastore. The best way to do this is in a finally block, as shown above, so the manager still gets closed in the event of an uncaught exception. Transactions in JPA The API for performing transactions in JPA is similar to the low-level datastore API. You call a method on the entity manager to create a Transaction object, then call meth- ods on the object to begin and commit or roll back the transaction. import javax.persistence.EntityTransaction; // ... EntityTransaction txn = em.getTransaction(); txn.begin(); try { Book book = em.find(Book.class, "9780596156732"); BookReview bookReview = new BookReview(); bookReview.setRating(5); book.getBookReviews().add(bookReview); // Persist all updates and commit. txn.commit(); } finally { if (txn.isActive()) { txn.rollback(); } } The JPA transaction interface was designed for databases that support global transac- tions, so it knows nothing of App Engine’s local transactions and entity groups. It’s up to the application to know which operations are appropriate to perform in a single transaction. You can manage entity groups and ancestors using App Engine’s exten- sions to JPA. One way to set up a data class that can represent entities with parents is to use either a datastore.Key or a string-encoded key for the @Id field. When you create a new object, you can construct the complete key, including ancestors, and assign it to this field. Alternatively, you can establish a second field to contain the parent key as either a Key or string-encoded key, using an extension: import javax.persistence.Basic; import javax.persistence.Entity; import javax.persistence.GeneratedValue; import javax.persistence.GenerationType; import javax.persistence.Id; import org.datanucleus.jpa.annotations.Extension; 216 | Chapter 8: The Java Persistence API
- 238. @Entity public class BookReview { @Id @GeneratedValue(strategy = GenerationType.IDENTITY) @Extension(vendorName = "datanucleus", key = "gae.encoded-pk", value = "true") private String keyString; @Basic @Extension(vendorName = "datanucleus", key = "gae.parent-pk", value = "true") private String bookKeyString; } The parent key field makes it easier to port your application to another database at a later time. It declares a slot in the data class for the ancestor relationship that is separate from the entity’s key name. The parent key field is required if you want to perform queries using ancestor filters. As we’ll see in the next section, JPA queries must refer to fields on the data class. The App Engine implementation of JPA includes features for managing entity groups automatically using object relationships. We’ll discuss relationships later in this chapter. Queries and JPQL JPA includes a SQL-like query language called JPQL. JPQL provides access to the un- derlying database’s query functionality at the level of abstraction of JPA data objects. You form queries for data objects in terms of the data classes, and get objects as results. To perform a query, you call the entity manager’s createQuery() method with the text of the JPQL query. This returns a Query object. To get the results, you call getResultList() on the Query object. import java.util.List; import javax.persistence.Query; // ... Query query = em.createQuery("SELECT b FROM Book b"); @SuppressWarnings("unchecked") List<Book> results = (List<Book>) query.getResultList(); In this example, the cast to List<Book> generates a compiler warning, so we suppress this warning using an @SuppressWarnings annotation. JPA knows which class to use for each result from the @Entity(name = "...") annotation on the class. You can also use the full package path of the class in the query. Queries and JPQL | 217
- 239. You can useparameters in your JPQL query, and replace the parameters with values by calling setParameter(): Query query = em.createQuery( "SELECT b FROM Book b WHERE copyrightYear >= :earliestYear"); query.setParameter("earliestYear", 1923); getResultList() returns a special App Engine–specific implementation of List that knows how to fetch results in batches. If you iterate over the entire list, the List im- plementation may make multiple calls to the datastore to fetch results. If you are only expecting one result, you can call getSingleResult() instead. This gets the first result if any, or null if there are no results. Book book = (Book) query.getSingleResult(); You can fetch a range of results by setting an offset and a maximum number of results, using the setFirstResult() and setMaxResults() methods before calling getResultList(). // Get results 5-15. query.setFirstResult(4); query.setMaxResults(10); @SuppressWarnings("unchecked") List<Book> results = (List<Book>) query.getResultList(); The syntax of JPQL is straightforward, and similar to SQL or the Python API’s GQL. JPQL keywords can be all uppercase or all lowercase, and are shown as uppercase here, as is tradition. Class and field names are case-sensitive. The query begins by identifying the simple name of the class of objects to query, corresponding to the kind of the entities: SELECT b FROM Book b This query returns all Book data objects, where Book is the value of the name argument to the @Entity annotation on the data class (which happens to also be named Book). The class name is followed by an identifier (b); stating that identifier after the word SELECT tells JPA to return objects of that class as results. To perform a keys-only query, give the name of the key field instead of the class iden- tifier. The methods that return results return values of the type used for the @Id field in the data class. Query query = em.createQuery("SELECT isbn FROM Book"); @SuppressWarnings("unchecked") List<String> results = (List<String>) query.getResultList(); The App Engine implementation of JPQL supports queries for specific fields, though perhaps not in the way you’d expect. For a query for specific fields, the datastore returns the complete data for each entity to the application, and the interface implementation 218 | Chapter 8: The Java Persistence API
- 240. selects the requestedfields and assembles the results. This is only true if one of the requested fields is a datastore property, and is not true if the only field is a key field (@Id). If the query is for one field, each result is a value of the type of that field. If the query is for multiple fields, each result is an Object[] whose elements are the field values in the order specified in the query. Query query = em.createQuery("SELECT isbn, title, author FROM Book"); // Fetch complete Book objects, then // produce result objects from 3 fields // of each result @SuppressWarnings("unchecked") List<Object[]> results = (List<Object[]>) query.getResultList(); for (Object[] result : results) { String isbn = result[0]; String title = result[1]; String author = result[2]; // ... } You specify filters on fields using a WHERE clause and one or more conditions: SELECT b FROM Book b WHERE author = "John Steinbeck" AND copyrightYear >= 1940 To filter on the entity key, refer to the field that represents the key in the data class (the @Id field): SELECT b FROM Book b WHERE author = "John Steinbeck" AND isbn > :firstKeyToFetch You can perform an ancestor filter by establishing a parent key field (as we did in the previous section) and referring to that field in the query: SELECT br FROM BookReview br WHERE bookKey = :pk As with find(), you can use any of the four key types with parameterized queries, but the most portable way is to use the type used in the class. You specify sort orders using an ORDER BY clause. Multiple sort orders are comma- delimited. Each sort order can have a direction of ASC (the default) or DESC. SELECT b FROM Book b ORDER BY rating DESC title The App Engine implementation of JPQL includes a couple of additional tricks that the datastore can support natively. One such trick is the string prefix trick: SELECT b FROM Book b WHERE title LIKE 'The Grape%' The implementation translates this to WHERE title >= 'The Grape', which does the same thing: it returns all books with a title that begins with the string The Grape, in- cluding "The Grape", "The Grapefruit", and "The Grapes of Wrath". Queries and JPQL | 219
- 241. This trick onlysupports a single wildcard at the end of a string. It does not support a wildcard at the beginning of the string. Another trick App Engine’s JPQL implementation knows how to do is to translate queries on key fields into batch gets. For example: SELECT b FROM Book b WHERE isbn IN (:i1, :i2, :i3) This becomes a batch get using three keys, and does not perform a query at all. In addition to these SELECT queries, App Engine’s JPA implementation supports delet- ing entities that meet criteria with JPQL. A delete query can include filters on keys and properties to specify the entities to delete. DELETE FROM Book b WHERE isbn >= "TEST_000" AND isbn <= "TEST_999" As with other mechanisms for modifying data, if you perform a delete query outside of a transaction, it is possible for a delete of one entity to fail while the others succeed. If you perform it inside a transaction, it’ll be all or nothing, but all entities must be in the same entity group, and the delete query must use an ancestor filter. The JPA specification supports many features of queries that are com- mon to SQL databases, but are not supported natively in the App Engine datastore. With a SQL database, using one of these features calls the database directly, with all of the performance implications (good and bad) of the datastore’s implementation. When an app uses a feature of JPQL that the underlying database does not support, DataNucleus Access Platform tries to make up the differ- ence using its own in-memory query evaluator. It attempts to load all of the information it needs to perform the query into memory, execute the nonnative operations itself, then return the result. Because such features are potential scalability hazards—an AVG() query would require fetching every entity of the kind, for example—the App Engine implementation disables the Access Platform in-memory query evaluator. Relationships Most useful data models involve relationships between classes of data objects. Players are members of guilds, book reviews are about books, messages are posted to message boards, customers place orders, and orders have multiple line items. For logical reasons or architectural reasons, two concepts may be modeled as separate but related classes. Those relationships are as much a part of the data model as the data fields of the objects. In the App Engine datastore (and most databases), one easy way to model a relationship between two objects is to store the entity key of one object as a property of the other, and (if needed) vice versa. The datastore supports Key values as a native property value 220 | Chapter 8: The Java Persistence API
- 242. type, and alsoprovides a way to encode key values as strings. You don’t need any help from JPA to model relationships this way. But relationships are so important to data modeling that JPA has a family of features to support them. In JPA, you can define owned relationships in the data model that enforce constraints by managing changes. With owned relationships, you can say that a book has zero or more book reviews, and JPA ensures that you can’t have a book review without a book. If you delete a Book object, JPA knows to also delete all of its BookReview objects. In the Java code, the relationship is represented by a field whose type is of the related data class, ensuring that only the appropriate classes are used on either side of the relationship. The App Engine implementation of JPA supports one-to-one and one-to-many rela- tionships. It does not yet support JPA’s notion of many-to-many relationships. An unowned relationship is a relationship without these constraints. App Engine sup- ports unowned relationships through the storing of literal key values, but does not yet support them through JPA. You can use multivalued properties of Key values to model unowned one-to-one, one-to-many and many-to-many relationships. To completely support the semantics of JPA owned relationships, App Engine stores objects with owned relationships in the same entity group. It’s easy to see why this has to be the case. If one object is deleted within a transaction, the relationship says the related object must also be deleted. But to do that in the same transaction requires that both objects be in the same entity group. If one object is deleted outside of a transaction, then the other object must be deleted in a separate operation, and if one delete or the other fails, an object remains that doesn’t meet the relationship requirement. While the use of entity groups may sound constraining, it’s also a powerful feature. You can use JPA owned relationships to perform transactions on related entities, and the JPA implementation will manage entity groups for you automatically. You specify an owned one-to-one relationship by creating a field whose type is of the related class, and giving the field an @OneToOne annotation. For example, you could associate each book with a cover image, like so: import javax.persistence.Entity; import javax.persistence.OneToOne; import bookstore.BookCoverImage; @Entity(name = "Book") public class Book { // ... @OneToOne(cascade=CascadeType.ALL) private BookCoverImage bookCoverImage; } This annotation declares a one-to-one relationship between the Book and BookCoverImage classes. Relationships | 221
- 243. In every relationship,one class “owns” the relationship. The owner of a relationship is responsible for propagating changes to related objects. In this example, the Book class is the “owner” of the relationship. The cascade=CascadeType.ALL argument annotation says that all kinds of changes should propagate to related objects (including PERSIST, REFRESH, REMOVE, and MERGE). For example: // EntityManager em; // ... Book book = new Book(); book.setBookCoverImage(new BookCoverImage()); book.setTitle("The Grapes of Wrath"); book.bookCoverImage.setType("image/jpg"); EntityTransaction txn = em.getTransaction(); txn.begin(); try { em.persist(book); txn.commit(); } finally { if (txn.isActive()) { txn.rollback(); } } em.close(); This code creates a Book and a related BookCoverImage. When it makes the Book persis- tent, the BookCoverImage is made persistent automatically (the PERSIST action cascades). Similarly, if we were to delete the Book, the BookCoverImage would also be deleted (the DELETE action cascades). Cascading actions follow all ownership paths from “owner” to “owned,” and do the right thing if the objects they find have changed since they were loaded from the datastore. You can have JPA populate a field on the “owned” class that points back to the owner automatically, like so: import javax.persistence.Entity; import javax.persistence.OneToOne; import bookstore.Book; @Entity(name = "BookCoverImage") public class BookCoverImage { // ... @OneToOne(mappedBy="bookCoverImage") private Book book; } 222 | Chapter 8: The Java Persistence API
- 244. The mappedBy argumenttells JPA that the book field refers to the Book object that is related to this object. This is managed from the “owner” side of the relationship: when the BookCoverImage is assigned to the Book’s field, JPA knows that the back-reference refers to the Book object. To specify a one-to-many relationship, you use a field type that is a List or Set of the related class, and use the @OneToMany annotation on the “one” class: import java.util.List; import javax.persistence.CascadeType; import javax.persistence.Entity; import javax.persistence.OneToMany; import bookstore.BookReview; @Entity(name = "Book") public class Book { // ... @OneToMany(cascade=CascadeType.ALL) private List<BookReview> bookReviews; } To create a back-reference from the “many” class to the “one” class, you use a @ManyToOne annotation, then specify a mappedBy on the “one” class’s @OneToMany: // BookReview.java @Entity(name = "BookReview") public class BookReview { // ... @ManyToOne(mappedBy="bookReviews") private Book book; } // Book.java @Entity(name = "Book") public class Book { // ... @OneToMany(cascade=CascadeType.ALL, mappedBy="book") private List<BookReview> bookReviews; } In a one-to-many relationship, the “one” is always the owner class, and the “many” is the owned class. In a one-to-one relationship, JPA knows which is the “owned” class by the absence of a back-reference field, or a back-reference field mentioned by a mappedBy annotation argument. When you fetch a data object that has a relationship field, JPA does not fetch the related objects right away. Instead, it waits until you access the field to fetch the object (or objects). This is called “lazy” fetching, and it saves your app from unnecessary datastore operations. The App Engine implementation of JPA only supports lazy fetching Relationships | 223
- 245. (FetchType.LAZY), and doesnot yet support its opposite, “eager” fetching (FetchType.EAGER). Note that you must access related objects prior to closing the EntityManager, so they are fetched into memory. // Fetch a Book, but not its BookCoverImage. Book book = em.find(Book.class, "9780596156732"); // ... // The BookCoverImage is fetched when it is first accessed. resp.setContentType(book.bookCoverImage.type); In the datastore, the relationship is represented using ancestors. The owner object is the parent, and all owned objects are children. When you access the relationship field on the owner object, the interface uses an ancestor query to get the owned objects. When you access a back-reference to the owner from the owned object, the interface parses the owned object’s key to get the parent. Related objects are created in the same entity group, so they can all be updated within the same transaction if necessary. The owner’s entity is created first (if necessary) and becomes the parent of the owned objects’ entities. If you declare a back-reference using mappedBy, no property is stored on the owned object’s entity. Instead, when the field is dereferenced, the implementation uses the owned object’s key path to determine the owner’s key and fetches it. The App Engine implementation does not support many-to-one relationships where the “many” is the owner. That is, it does not support a one-to-many relationship where actions cascade from the many to the one. Creating new relationships between existing data classes can be tricky, because the entity group requirements must be met in the migrated data. Adding a relationship to the owner class is like adding a field: the entities that represent instances of the owner class must be updated to have appropriate key properties. The “owned” side is trickier: since an owned object’s entity must have the owner as its parent, if the owned object already exists in the datastore, it must be deleted and re-created with the new parent. You can’t change an entity’s parent after the entity has been created. This use of datastore ancestors means you cannot reassign an owned object to another owner after it has been saved to the datastore. This also means that one object cannot be on the “owned” side of more than one relationship, since the entity can have only one parent. Relationships and cascading actions imply that an operation on a data object can translate to multiple datastore operations on multiple entities, all in the same entity group. If you want these operations to occur in a single transaction, you must perform the initial operation (such as em.merge(...)) within an explicit transaction. 224 | Chapter 8: The Java Persistence API
- 246. If you performa cascading action outside of an explicit transaction, each of the data- store operations performed by JPA occurs in a separate operation. Some of these op- erations may fail while others succeed. As such, it’s a best practice to perform all JPA updates within explicit transactions, so there is no confusion as to what succeeded and what failed. You can perform queries on relationship fields in JPQL using key values. For a query on the owner class, the query is a simple key property query: SELECT FROM Book b WHERE bookCoverImage = :bci AND publishDate > :pdate For a query on an owned class, a query on the back-reference field becomes an ancestor filter: SELECT FROM BookCoverImage bci WHERE book = :b You cannot refer to properties of the related entity in the query filter. App Engine does not support join queries. For More Information The JDO and JPA interfaces have many useful features that work with the App Engine implementation. One excellent source of documentation on these interfaces is the DataNucleus Access Platform website: http://www.datanucleus.org/products/accessplatform/ The App Engine implementation is an open source project hosted on Google Code. The source includes many unit tests that exercise and demonstrate many features of JDO and JPA. You can browse the source code at the project page: http://code.google.com/p/datanucleus-appengine/ To read the unit tests, click the Source tab, then click Browse in the navigation bar. The path is svn/trunk/tests/org/datanucleus/store/appengine/. For More Information | 225
- 248. CHAPTER 9 The Memory Cache Durable data storage requires a storage medium that retains data through power loss and system restarts. Today’s medium of choice is the hard drive, a storage device com- posed of circular platters coated with magnetic material on which data is encoded. The platters spin at a constant rate while a sensor moves along the radius, reading and writing bits on the platters as they travel past. Reading or writing a specific piece of data requires a disk seek to position the sensor at the proper radius and wait for the platter to rotate until the desired data is underneath. Hard drives are astonishingly fast, all things considered, but for web applications, disk seeks can be costly. Fetching an entity from the datastore by key can take time on the order of tens of milliseconds. Most high-performance web applications use a memory cache. A memory cache uses a volatile storage medium, usually the RAM of the cache machines, for very fast read and write access to values. A distributed memory cache provides scalable, consistent tem- porary storage for distributed systems, so many processes can access the same data. Because memory is volatile—it gets erased during an outage—the cache is not useful for long-term storage, or even short-term primary storage for important data. But it’s excellent as a secondary system for fast access to data also kept elsewhere, such as the datastore. It’s also sufficient as global high-speed memory for some uses. The App Engine distributed memory cache service, known as memcache in honor of the original memcached system that it resembles, stores key-value pairs. You can set a value with a key, and get the value given the key. A value can be up to a megabyte in size. A key is up to 250 bytes, and the API accepts larger keys and uses a hash algorithm to convert them to 250 bytes. Setting a single value in the cache is atomic: the key either gets the new value or retains the old one (or remains unset). The memcache does not support transactions like the datastore does. If you get a value, then try to set a value based on what you got, the first value may have changed since it was fetched. The App Engine memcache includes the ability to increment and decrement numeric values as an atomic operation. 227
- 249. A common wayto use the memcache with the datastore is to cache datastore entities by their keys. When you want to fetch an entity by key, you first check the memcache for a value with that key, and use it if found (known as a cache hit). If it’s not in the memcache (a cache miss), you fetch it from the datastore, then put it in the memcache so future attempts to access it will find it there. At the expense of a small amount of overhead during the first fetch, subsequent fetches become much faster. If the entity changes in the datastore, you can attempt to update the memcache when the entity is updated in the datastore, so subsequent requests can continue to go to the cache but see fresh data. This mostly works, but it has two minor problems. For one, it is possible that the memcache update will fail even if the datastore update succeeds, leaving old data in the cache. Also, if two processes update the same datastore entity, then update the memcache, the datastore will have correct data (thanks to datastore transactions), but the memcache update will have the value of whichever update occurs last. Because of this possibility, it’s somewhat better to just delete the memcache key when the datastore changes, and let the next read attempt populate the cache with a current value. Naturally, the delete could also fail. Because there is no way to update both the datastore and the memcache in a single transaction, there is no way to avoid the possibility that the cache may contain old data. To minimize the duration that the memcache will have a stale value, you can give the value an expiration time when you set it. When the expiration time elapses, the cache unsets the key, and a subsequent read results in a cache miss and triggers a fresh fetch from the datastore. Of course, this caching pattern works for more than just datastore entities. You can use it for datastore queries, web service calls made with URL Fetch, expensive calcu- lations, or any other data that can be replaced with a slow operation, where the benefits of fast access outweigh the possibility of staleness. This is so often the case with web applications that a best practice is to cache aggres- sively. Look through your application for opportunities to make this trade-off, and implement caching whenever the same value is needed an arbitrary number of times, especially if that number increases with traffic. Site content such as an article on a news website often falls into this category. Caching speeds up requests and saves CPU time, and reduces the likelihood of datastore read failures impacting your application. The APIs for the memcache service are straightforward. We’ll look at the Python in- terface first, then the Java interface, with an emphasis on caching datastore entities. The Python Memcache API The Python API to the memcache service is designed to be compatible with the Python interface to memcached. It includes a Client class with static methods you can call to interact with the service. The App Engine API also provides these methods as straight functions in the google.appengine.api.memcache package. 228 | Chapter 9: The Memory Cache
- 250. Keys are strings.The API also accepts a memcached-style tuple of a hash code and a string as a key for compatibility, but it ignores the hash code. The string can be any size; strings larger than 250 bytes are hashed to 250 bytes. Values can be any Python value that can be serialized using the pickle module. This includes most simple values, and data structures and objects whose members are pickleable values. Pretty much any db.Model instance falls into this category. You can check the documentation for the pickle module for a more complete description. Setting and Getting Values in Python To store a value in memcache, you call the set() function (or method, if you’re using the Client class) with the key and the value: from google.appengine.api import memcache # headlines is a pickle-able object headlines = ['...', '...', '...'] success = memcache.set("headlines", headlines) if not success: # Problem accessing memcache... set() returns True on success, or False if the value could not be stored. If there already exists a value with the given key, set() overwrites it. If there is no value for the key, set() sets it. To set a value only if there is no value for the key, you call the add() function with similar arguments: success = memcache.add("headlines", headlines) if not success: # Key already set, or another problem setting... add() works like set(), but does not update the value and returns False if a value is currently set for the given key. To update a value only if it’s already set, you call the replace() function, also with similar arguments: success = memcache.replace("headlines", headlines) if not success: # Key not set, or another problem setting... To get a value, you call the get() method with the key: headlines = memcache.get("headlines") if headlines is None: # Key not set, or another problem getting... get() returns the value, or None if there is no value set for the given key. The Python Memcache API | 229
- 251. Setting and GettingMultiple Values You can set and get multiple values at a time using the batch interface. A batch call to memcache is performed as a single call to the service, which is faster than multiple individual calls for many items. Like the datastore batch operations, the total amount of data sent to or received from memcache in a single call must not exceed one megabyte. set_multi(), add_multi(), and replace_multi() are similar to their singular counter- parts, except instead of taking a key and a value as arguments, they take a Python mapping (such as a dict) of keys to values. article_summaries = {'article00174': '...', 'article05234': '...', 'article15820': '...', } failed_keys = memcache.set_multi(article_summaries) if failed_keys: # One or more keys failed to update... These methods return a list of keys that failed to be set. A completely successful batch update returns the empty list. To get multiple values in a single batch, you call the get_multi() method with a list of keys: article_summary_keys = ['article00174', 'article05234', 'article15820', ] article_summaries = memcache.get_multi(article_summary_keys) get_multi() returns a dict of keys and values that were present and retrieved successfully. All of the batch operations accept a key_prefix keyword argument that prepends every key with the given prefix string. When used with get_multi(), the keys on the returned values do not include the prefix. article_summaries = {'00174': '...', '05234': '...', '15820': '...', } failed_keys = memcache.set_multi(article_summaries, key_prefix='article') if failed_keys: # ... # ... article_summary_keys = ['00174', 230 | Chapter 9: The Memory Cache
- 252. '05234', '15820', ] article_summaries = memcache.get_multi(article_summary_keys, key_prefix='article') Batch calls to the memcache service are not atomic actions. set_multi() may be successful for some items and not others. get_multi() will return the data seen at the time of each individual get, and does not guarantee that all retrieved data is read at the same time. Memcache Namespaces You can specify an optional namespace for a memcache key with any setter or getter call. Namespaces let you segment the key space by the kind of thing you are storing, so you don’t have to worry about key collisions between disparate uses of the cache. All of the setter and getter functions accept a namespace argument, a string that declares the namespace for all keys involved in the call. The batch functions use one namespace for all keys in the batch. article_summaries = {'article00174': '...', 'article05234': '...', 'article15820': '...', } failed_keys = memcache.set_multi(article_summaries, namespace='News') if failed_keys: # ... # ... summary = memcache.get('article05234', namespace='News') Namespaces are different from key prefixes in batch calls. Key prefixes merely prepend keys in a batch with a common string, as a convenience for batch processing. The namespace is a separate part of the key, and must be specified with an explicit namespace argument in all calls. Cache Expiration You can tell the memcache to delete (or evict) a value automatically after a period of time when you set or update the value. All of the setter methods accept a time argument. Its value is either a number of seconds from the current time up to one month (up to 2,678,400 seconds), or it is a Unix epoch time (seconds since January 1, 1970, such as returned by time.time()) in the future. The Python Memcache API | 231
- 253. (The API knowsthe difference by the size of the value, since it’s been longer than one month since the epoch.) # Refresh the headlines cache once every 5 minutes. success = memcache.set('headlines', headlines, time=300) A key’s expiration timer is reset every time it is updated using set() or replace() (or the _multi() equivalents). If no time argument is provided to a setter, the default behavior is to keep the values in the memcache as long as possible. Deleting Keys You can delete keys immediately by calling the delete() or delete_multi() functions. delete() takes a key as its argument, and delete_multi() takes a list of keys. # Force the headlines cache to refresh. result = memcache.delete('headlines') if not result: # Delete failed... For compatibility with the Python memcached interface, the delete() function returns one of three success values, represented by constants in the API: DELETE_NETWORK_FAILURE means there was a problem caused by a network failure; DELETE_ITEM_MISSING means there was no value with the given key to delete; and DELETE_SUCCESSFUL means the key was deleted. If you don’t care about the distinction between a successfully deleted value and a value that was never there, you can treat the return value as a true-or-false value, since the network failure case is equivalent to zero and the others are nonzero. The delete_multi() method returns True if all keys were successfully deleted or did not exist, and False if at least one key was not deleted successfully. delete() and delete_multi() accept the namespace parameter. delete_multi() also ac- cepts the key_prefix parameter, similar to the other _multi() functions. You can specify an amount of time during which processes are not allowed to readd the key. This is useful for ensuring that multiple simultaneous worker processes work- ing with the same data do not accidentally undo cache deletes with an update. To specify a readd lock duration, you provide the seconds argument to either delete() or delete_multi(). This argument is either a relative number of seconds up to one month, or a date in the future as a Unix epoch time. result = memcache.delete('tempnode91512', seconds=5) You can delete every memcache value for your application by calling the flush_all() function. It takes no arguments, and returns True on success or False on failure. memcache.flush_all() 232 | Chapter 9: The Memory Cache
- 254. Memcache Counters The AppEngine memcache includes a feature that lets you increment and decrement numeric values in the cache using atomic operations. The memcache is not transac- tional otherwise, so this is the only way to modify an existing value atomically. In Python, you call the incr() method to increment a value, and the decr() method to decrement a value. These methods accept a key, and an optional amount to increment or decrement as the delta argument (the default is 1). As with the other methods, these methods accept the namespace argument. work_done = memcache.incr('work_done') if not work_done: # ... work_done = memcache.incr('work_done', delta=5) if not work_done: # ... As shown, the methods return the new value, or None if the key was not set, if it was not set to an integer, or if there was another problem. You can specify a starting value to use if the key is unset by providing the initial_value argument. If the key is found, its value is used as the starting value. Otherwise, the provided initial value is used as the starting value for the operation. If you don’t provide an initial value and the key is not set, the call does nothing and returns None. (This is the default.) work_done = memcache.incr('work_done', initial_value=0) Cache Statistics The memcache maintains a set of statistics about cache hits and misses, items in the cache, and data transferred out of the cache. You can inspect this data programmatically by calling the get_stats() method. This method returns a dictionary containing the statistics. Note that these statistics are just a best effort, and may be reset by the system at any time. import logging stats = memcache.get_stats() logging.info('Memcache statistics:') for stat in stats.iteritems(): logging.info('%s = %d' % stat) Available statistics include: hits The number of cache hits counted. The Python Memcache API | 233
- 255. misses The number of cache misses counted. items The number of items currently in the cache. bytes The total size of items currently in the cache. byte_hits The total of bytes returned in response to cache hits, including keys and values. oldest_item_age The age of the least recently accessed item in the cache, in milliseconds. The Python development server includes a memcache viewer in its Console. The viewer lets you query, set, and delete values, and also displays statistics. Refer back to “The Development Console” on page 54 for more information. The Java Memcache API As with other services, App Engine includes two Java interfaces to the memcache service. One is a proprietary interface that provides direct access to the features of the service. The other is an implementation of JCache, an interface standard proposed by JSR 107. As of this writing, JSR 107 is not a ratified standard, and the recommended way to use it on App Engine is in flux. So for now, we’ll just consider the proprietary interface. For more information on using the JCache interface with App Engine, consult the App Engine documentation. The memcache service Java API allows for the use of any serializable object as the key or the value of a key-value pair stored in the cache. Keys whose serialized form is larger than 250 bytes are hashed to 250 bytes. Values can be up to one megabyte in size. In many cases, you can make a JPA data class serializable simply by declaring that the class implements Serializable, with no other changes. If you’re using the low-level datastore API, the Entity class and all of the property value classes implement the Serializable interface. With App Engine’s memcache service Java API, you interact with the service using a MemcacheService object, which you get from MemcacheServiceFactory.getMemcacheService(). The basic operations are available as methods of MemcacheService: put(), get(), and delete(). import com.google.appengine.api.memcache.MemcacheService; import com.google.appengine.api.memcache.MemcacheServiceFactory; // ... List<String> headlines; 234 | Chapter 9: The Memory Cache
- 256. MemcacheService memcache =MemcacheServiceFactory.getMemcacheService(); memcache.put("headlines", headlines); headlines = (List<String>) memcache.get("headlines"); memcache.delete("headlines"); The put() method takes a key and a value, either of which can be serializable objects. Called with just these two arguments, the new value is stored with no expiration time, and will either create a new value with the given key or replace an existing one, if any. To set a value with an expiration time, you pass put() an Expiration object as its third argument. You produce this value with a static method in the Expiration class: import com.google.appengine.api.memcache.Expiration; // ... memcache.put("headlines", headlines, Expiration.byDeltaSeconds(300)); You can use the alternate methods byDeltaMillis(int) for an expiration time in milli- seconds, or onDate(java.util.Date) for an expiration time at an absolute time in the future. To change how put() deals with an existing value with the given key, provide a MemcacheService.SetPolicy enumeration value as the fourth argument. You can use null for the third argument if you don’t want to set an expiration. import com.google.appengine.api.memcache.MemcacheService.SetPolicy; // ... memcache.put("headlines", headlines, null, SetPolicy.ADD_ONLY_IF_NOT_PRESENT); The three set-policy options are SET_ALWAYS (the default), ADD_ONLY_IF_NOT_PRESENT (only create, do not overwrite), and REPLACE_ONLY_IF_PRESENT (only overwrite, do not create). You can test whether the cache contains a value with a given key without fetching the value using the contains() method. This method takes a key as its argument, and re- turns true or false. boolean headlinesAreCached = memcache.contains("headlines"); To delete a value for a given key, you call the delete() method with the key as its argument. You can specify an amount of time during which processes are not allowed to readd the key, as a long integer number of milliseconds in the second argument. This is useful for ensuring that multiple simultaneous worker processes working with the same data do not accidentally undo cache deletes with an update. memcache.delete("tempnode91512", 5); The Java Memcache API | 235
- 257. The low-level APIincludes methods for setting, fetching, or deleting multiple items in a single batch call. This is faster than making one service call per item. However, the total size of the call and its response is limited to one megabyte. The putAll() method stores multiple values in a batch. It accepts a Map<Object, Object> of keys and values as its first argument. It optionally accepts Expiration and SetPolicy values, like its single-item counterpart. import java.util.HashMap; import java.util.Map; // ... Map<Object, Object> articleSummaries = new HashMap<Object, Object>(); articleSummaries.put("article00174", "..."); articleSummaries.put("article05234", "..."); articleSummaries.put("article15820", "..."); memcache.putAll(articleSummaries); The getAll() method fetches multiple items in a batch. It accepts a Collection<Object> of keys as its argument, and returns a Map<Object, Object> of keys and values for all of the provided keys that were found in the cache. import java.util.List; // ... List<Object> articleSummaryKeys = Arrays.<Object>asList( "article00174", "article05234", "article15820"); Map<Object, Object> articleSummaries = memcache.getAll(articleSummaryKeys); The deleteAll() method deletes multiple items in a batch. It accepts a Collection of keys as its argument. It also optionally accepts a readd lock time in milliseconds as its second argument. memcache.deleteAll(articleSummaryKeys); Any memcached value can be stored in a namespace. Namespaces let you segment the key space by the kind of thing you are storing, so you don’t have to worry about key collisions between disparate uses of the cache. To use namespaces in the low-level Java API, you set the namespace on the MemcacheService by calling its setNamespace() method. All subsequent operations use the given namespace when manipulating values. memcache.setNamespace("News"); memcache.put("headlines", headlines); memcache.setNamespace("User"); memcache.put("headlines", userHeadlines); // Get User:"headlines". userHeadlines = (List<String>) memcache.get("headlines"); 236 | Chapter 9: The Memory Cache
- 258. // Get News:"headlines". memcache.setNamespace("News"); headlines = (List<String>) memcache.get("headlines"); If a memcache value is an integer (int or long), you can increase or decrease the amount atomically—that is, without colliding with other processes doing the same thing— using the increment() method. The method takes the key and the amount to increment (which can be negative). It returns the new value (a Long), or null if there is no value already stored with that key. memcache.put("work_done", 0); // ... Long workDone = memcache.increment("work_done", 1); As of this writing, the Java interface does not support supplying an initial value for an unset cache counter in the same atomic action. The low-level Java API provides access to more statistics than the JCache API. The getStatistics() method returns a Stats object, with a method for each of the available stats. import com.google.appengine.api.memcache.Stats; // ... Stats stats = memcache.getStatistics(); int ageOfOldestItemMillis = stats.getMaxTimeWithoutAccess(); Available statistics include: getHitCount() The number of cache hits counted. getMissCount() The number of cache misses counted. getItemCount() The number of items currently in the cache. getTotalItemBytes() The total size of items currently in the cache. getBytesReturnedForHits() The total of bytes returned in response to cache hits, including keys and values. getMaxTimeWithoutAccess() The age of the least recently accessed item in the cache, in seconds. The Java Memcache API | 237
- 259. There is noway to explicitly reset the memcache statistics. They are accumulated over the uptime of the service, and so are mostly useful for relative comparisons over brief periods of time. By default, if the memcache service is unavailable or there is an error accessing the service, the low-level API and the JCache API behave as if keys do not exist. Attempts to put new values fail silently; the four-argument put() method will return false in this case as if the put failed due to the set policy. Attempts to get values will behave as cache misses. In the low-level Java API, you can change this behavior by installing an alternate error handler. The setErrorHandler() method of MemcacheService takes an object that implements the ErrorHandler interface. Two such implementations are provided: LogAndContinueErrorHandler and StrictErrorHandler. The default is LogAndContinueErrorHandler with its log level set to FINE (the “debug” level in the Ad- ministration Console). StrictErrorHandler throws MemcacheServiceException for all transient service errors. import com.google.appengine.api.memcache.StrictErrorHandler; // ... memcache.setErrorHandler(new StrictErrorHandler()); Error handlers can have custom responses for invalid values and service errors. Other kinds of exceptions thrown by the API behave as usual. 238 | Chapter 9: The Memory Cache
- 260. CHAPTER 10 Fetching URLs and Web Resources An App Engine application can connect to other sites on the Internet to retrieve data and communicate with web services. It does this not by opening a connection to the remote host from the application server, but through a scalable service called the URL Fetch service. This takes the burden of maintaining connections away from the app servers, and ensures that resource fetching performs well regardless of how many re- quest handlers are fetching resources simultaneously. As with other parts of the App Engine infrastructure, the URL Fetch service is used by other Google applications to fetch web pages. The service supports fetching URLs using the HTTP protocol, and using HTTP with SSL (HTTPS). Other methods sometimes associated with URLs (such as FTP) are not supported. Note that with HTTPS connections, the service can’t authenticate the des- tination of the connection, because there is no certificate trust chain. The service accepts all certificates. This means the connection protocol itself can’t protect against “man-in-the-middle” attacks, though the traffic is still encrypted. Because the URL Fetch service is based on Google infrastructure, the service inherits a few restrictions that were put in place in the original design of the underlying HTTP proxy. The service supports the five most common HTTP actions (GET, POST, PUT, HEAD, and DELETE) but does not allow for others or for using a nonstandard action. Also, it can only connect to the standard TCP ports for each method (80 for HTTP, and 443 for HTTPS). The proxy uses HTTP 1.1 to connect to the remote host. As a safety measure against accidental request loops in an application, the URL Fetch service cannot fetch the URL that maps to the request handler doing the fetching. An app can make connections to other URLs of its own, so request loops are still possible, but this restriction provides a simple sanity check. The outgoing request can contain URL parameters, a request body, and HTTP headers. A few headers cannot be modified for security reasons, which mostly means that an app cannot issue a malformed request, such as a request whose Content-Length header does not accurately reflect the actual content length of the request body. In these cases, the service uses the correct values, or does not include the header. 239
- 261. Both the Pythonand Java runtime environments include interfaces to the URL Fetch service that emulate standard library routines. The Python environment replaces the urllib and urllib2 modules with workalike equivalents that use the URL Fetch service instead of making direct connections. Similarly, the Java environment replaces java.net.URLConnection with a service-based implementation. These standard interfa- ces allow for existing code and third-party components to function in an App Engine app and still meet App Engine’s scaling requirements. These interfaces (and their low-level counterparts) implement synchronous URL fetch- ing, where the application waits for the remote host to respond before continuing. The URL Fetch service also supports an asynchronous interface that allows the app to trigger fetch requests, do other work, then retrieve results when the requests are ready. This can significantly reduce the amount of clock time spent in a request that relies on a remote host, and allows an app to make multiple URL Fetch connections simultaneously. As of this writing, only the Python environment has an asynchronous URL Fetch interface. A similar interface is being developed for Java apps. Check the official App Engine website for updates. In this chapter, we will introduce the standard and low-level interfaces for fetching URLs synchronously in Python and Java, as well as the asynchronous interface for Python apps. Fetching URLs in Python The Python runtime environment overrides portions of the urllib, urllib2, and httplib modules in the Python standard library so that HTTP and HTTPS connections made with these modules use the URL Fetch service. This allows existing software that depends on these libraries to function on App Engine, as long as the requests function within the limitations of the service. In most cases, if you are writing Python code that fetches URLs, you will want to use the urllib2 module. This module provides rich extensible support for features of re- mote web servers such as HTTP authentication and cookies. We won’t go into the details of this module here, but Example 10-1 shows a brief example using the module’s urlopen() convenience function. Example 10-1. A simple example of using the urllib2 module to access the URL Fetch service import urllib2 from google.appengine.api import urlfetch # ... try: 240 | Chapter 10: Fetching URLs and Web Resources
- 262. newsfeed = urllib2.urlopen('http://ae-book.appspot.com/blog/atom.xml/') newsfeed_xml = newsfeed.read() except urllib2.URLError, e: # Handle urllib2 error... except urlfetch.Error, e: # Handle urlfetch error... In this example, we catch both exceptions raised by urllib2 and exceptions raised from the URL Fetch Python API, google.appengine.api.urlfetch. The service may throw one of its own exceptions for conditions that urllib2 doesn’t catch, such as a request exceeding its deadline. Naturally, not all of the URL Fetch service’s features are available in these libraries, and vice versa. When using the standard libraries, the service has the following default behaviors: • If the remote host doesn’t respond within five seconds, the request is canceled and a service exception is raised. • The service follows HTTP redirects up to five times before returning the response to the application. (A urllib2 redirect handler will not see most redirects.) • Responses from remote hosts that exceed one megabyte in size are truncated to one megabyte. The application is not told whether the response is truncated. If you use the service API directly, you can customize these behaviors. Example 10-2 shows a similar example using the urlfetch module, with several options changed. Example 10-2. Customizing URL Fetch behaviors using the urlfetch module from google.appengine.api import urlfetch # ... try: newsfeed = urlfetch.fetch('http://ae-book.appspot.com/blog/atom.xml/', allow_truncated=False, follow_redirects=False, deadline=10) newsfeed_xml = newsfeed.content except urlfetch.Error, e: # Handle urlfetch error... The fetch() function in the google.appengine.api.urlfetch module calls the service to fetch a URL, waits for the response, then returns an object that represents the result. The first and only required argument to this function is the full URL, including the http:// or https:// and the full host name. (You cannot use a “relative” URL path to request a URL of the application doing the requesting. You can derive an appropriate hostname from the handler’s own request data, if you need to.) Fetching URLs in Python | 241
- 263. The remaining argumentsto the function are all optional. They specify the payload (the request body, empty by default), the method (default is GET), and any additional headers to include (as a dict-like value). The allow_truncated argument specifies what to do if the response from the remote host is larger than the maximum supported size of one megabyte. If set to True, the service truncates the response to one megabyte before returning it to the application. If False, the service raises a urlfetch.ResponseTooLargeError when this happens. The default when calling fetch() is False. Using the standard libraries (urllib2 et al.) sets this to True, so there is one fewer custom exception you need to catch. The follow_redirects argument tells the service how to handle HTTP redirect codes (HTTP response codes 301 and 302, accompanied by a Location header specifying a new URL). If True, the service will follow HTTP redirects and return the final page to the application. If the service encounters more than five consecutive redirects, it stops at the fifth and returns the last redirect response to the app. If this argument is False, the service returns whatever the remote host returned for the requested URL, which may be an HTTP redirect. The argument’s default value is True. Lastly, the deadline argument tells the service how long to wait for the remote host to respond. The default is 5 seconds, and you can set this to anywhere from 1 to 10 sec- onds. If the fetch exceeds the deadline, the service raises a urlfetch.Error. (As of this writing, there is no exception class for this specific condition.) The Python development server simulates the URL Fetch service by making HTTP connections directly from your computer. If the remote host might behave differently when your app connects from your computer rather than from Google’s proxy servers, be sure to test your URL Fetch calls on App Engine. Fetching URLs in Java The Java runtime includes a custom implementation of the URLConnection class in the JRE-standard java.net package that calls the URL Fetch service instead of making a direct socket connection. As with the other standard interfaces, you can use this inter- face and rest assured that you can port your app to another platform easily. Example 10-3 shows a simple example of using a convenience method in the URL class, which in turn uses the URLConnection class, to fetch the contents of a web page. The openStream() method of the URL object returns an input stream of bytes. As shown, you can use an InputStreamReader (from java.io) to process the byte stream as a character stream. The BufferedReader class makes it easy to read lines of text from the InputStreamReader. 242 | Chapter 10: Fetching URLs and Web Resources
- 264. Example 10-3. Usingjava.net.URL to call the URL Fetch service import java.net.URL; import java.net.MalformedURLException; import java.io.IOException; import java.io.InputStream; import java.io.InputStreamReader; import java.io.BufferedReader; // ... try { URL url = new URL("http://ae-book.appspot.com/blog/atom.xml/"); InputStream inStream = url.openStream(); InputStreamReader inStreamReader = new InputStreamReader(inStream); BufferedReader reader = new BufferedReader(inStreamReader); // ... read characters or lines with reader ... reader.close(); } catch (MalformedURLException e) { // ... } catch (IOException e) { // ... } Note that the URL Fetch service has already buffered the entire response into the ap- plication’s memory by the time the app begins to read. The app reads the response data from memory, not from a network stream from the socket or the service. You can use other features of the URLConnection interface, as long as they operate within the functionality of the service API. Notably, the URL Fetch service does not maintain a persistent connection with the remote host, so features that require such a connection will not work. By default, the URL Fetch service waits up to five seconds for a response from the remote server. If the server does not respond by the deadline, the service throws an IOException. You can adjust the amount of time to wait using the setConnectTimeout() method of the URLConnection. (The setReadTimeout() method has the same effect; the service uses the greater of the two values.) The deadline can be between 1 and 10 seconds. When using the URLConnection interface, the URL Fetch service follows HTTP redirects automatically, up to five consecutive redirects. The low-level API for the URL Fetch service lets you customize several behaviors of the service. Example 10-4 demonstrates how to fetch a URL with this API with options specified. As shown, the FetchOptions object tells the service not to follow any redirects, and to throw a ResponseTooLargeException if the response exceeds the maximum size of one megabyte instead of truncating the data. Fetching URLs in Java | 243
- 265. Example 10-4. Usingthe low-level API to call the URL Fetch service, with options import java.net.URL; import java.net.MalformedURLException; import com.google.appengine.api.urlfetch.FetchOptions; import com.google.appengine.api.urlfetch.HTTPResponse; import com.google.appengine.api.urlfetch.ResponseTooLargeException; import com.google.appengine.api.urlfetch.URLFetchService; import com.google.appengine.api.urlfetch.URLFetchServiceFactory; // ... try { URL url = new URL("http://ae-book.appspot.com/blog/atom.xml/"); FetchOptions options = FetchOptions.Builder .doNotFollowRedirects() .disallowTruncate(); HTTPRequest request = new HTTPRequest(url, HTTPMethod.GET, options); URLFetchService service = URLFetchServiceFactory.getURLFetchService(); HTTPResponse response = service.fetch(request); // ... process response.content ... } catch (ResponseTooLargeException e) { // ... } catch (MalformedURLException e) { // ... } catch (IOException e) { // ... } The Java development server simulates the URL Fetch service by making HTTP con- nections directly from your computer. Be sure to test that your app can access the URLs it needs when running on App Engine. Asynchronous Requests in Python All of the calls to the URL Fetch service we’ve seen so far in this chapter have been synchronous calls: the app asks the service to fetch a URL, then waits for the service to return the result. But time is precious within a web request, and waiting for a remote host to return data is likely to be the most time-consuming part of a request handler. If the handler must fetch more than one URL, these wait times add up as the app must wait for the first host to respond before starting the next fetch. But while the service is busy fetching URLs, the app server is doing nothing. The app could be using this val- uable clock time to perform other tasks. To support this, the URL Fetch service supports asychronous calls. With an asynchro- nous call, the app starts a URL fetch by calling the service with the request data as usual. But unlike a synchronous call, this call returns immediately, and the app can continue doing other things. When the app needs the results, it makes another call to the service, 244 | Chapter 10: Fetching URLs and Web Resources
- 266. and the servicereturns the results. The app can fetch multiple URLs simultaneously, and the URL Fetch service performs the fetches in parallel. Figure 10-1 shows the sequence of events for a synchronous call and an asynchronous call. As shown, if the fetch is not complete when the app asks for results, the app waits for the remainder of the duration. If the fetch is complete by the time the app asks for results, the service returns the results (or an error message) immediately. Figure 10-1. Sequence diagrams of a synchronous URL fetch and an asynchronous URL fetch Example 10-5 demonstrates a simple asynchronous call to the URL Fetch service. The create_rpc() function returns an object that represents the asynchronous call (“remote procedure call,” or RPC). The make_fetch_call() function initiates the fetch for the given URL and associates it with the given RPC object, then returns immediately. Fi- nally, calling the get_result() method of the RPC object waits for the fetch to finish, and returns the result. Asynchronous Requests in Python | 245
- 267. Example 10-5. FetchingURLs in the background using the asynchronous URL Fetch API from google.appengine.api import urlfetch # ... try: rpc = urlfetch.create_rpc() urlfetch.make_fetch_call(rpc, 'http://ae-book.appspot.com/blog/atom.xml/') # Do other things. newsfeed = rpc.get_result() newsfeed_xml = newsfeed.content except urlfetch.Error, e: # Handle urlfetch error... If the app does nothing between make_fetch_call() and rpc.get_result(), it behaves as if it made a synchronous call. In fact, this is precisely how the synchronous fetch() function is implemented. You can specify a fetch deadline of up to 10 seconds by passing a deadline argument to create_rpc(). The default deadline is five seconds. If the remote host does not re- spond before the deadline elapses, the fetch is canceled, and the API raises a DownloadError when you attempt to access the results. The make_fetch_call() function accepts the same request parameters as the fetch() function, including the url, payload, method, headers, allow_truncated, and follow_redirects. The development server simulates asynchronous calls to the URL Fetch service by per- forming them synchronously. If your app would respond significantly faster when using asynchronous fetches, especially if the app makes multiple simultaneous fetches, you won’t notice the time savings until the app is running on App Engine. Note that the Python standard libraries (urllib2 et al.) only support synchronous calls. To take advantage of asynchronous calls, you must use the urlfetch interface. RPC Objects The RPC object represents the call throughout the process. An RPC object can only represent one call. If you are fetching multiple URLs, you must create an RPC object for each fetch by calling the create_rpc() function. The RPC object advances through four states during its lifetime: 1. Created. The object has been created, but has not been associated with a fetch call. 2. In progress. A fetch has been initiated and associated with the object. 3. Ready. The fetch has completed, either with results or with an error. 246 | Chapter 10: Fetching URLs and Web Resources
- 268. 4. Checked. Thestatus of the fetch has been reported to the application, such as by having raised an exception to represent an error. In the simple example just given, create_rpc() created the object in its initial “created” state, and make_fetch_call() advanced it to its “in progress” state. Calling the object’s get_result() method advanced it to the “ready” state by waiting for the fetch to finish, then advanced it to the “checked” state by verifying the results and raising an exception if necessary. You can advance the object through the last two states manually using methods. The wait() method waits for the fetch to finish (“in progress” to “ready”). The check_result() method verifies and reports the final status (“ready” to “checked”). Calling any of these methods advances the object to the appropriate state, performing the tasks along the way. If the starting state for the method has already passed, such as calling wait() when in the “ready” state, the method does nothing. Figure 10-2 illustrates the RPC object states and transitions. Figure 10-2. The URL Fetch RPC object states and transitions Processing Results with Callbacks To make the most of the parallel execution of asynchronous URL fetching, the request handler should initiate the fetch as soon as possible in the handler’s lifetime. This can be as straightforward as creating the RPC objects and calling make_fetch_call() at the top of a routine, and calling the get_results() method at the point in the routine where the results are needed. If your handler uses multiple diverse components to perform tasks, and each component may require that URLs be fetched, you could have the main routine ask each component to initiate its fetches, then allow the components to get their own results as control reaches the appropriate points in the code. The URL Fetch RPC object offers another way to organize the code that handles the results of fetches: callbacks. A callback is a function associated with the RPC object that is called when the RPC’s fetch has completed, during the transition from the “in progress” state to the “ready” state. In other words, an RPC’s callback function is called if the RPC is in the “in progress” state and the app calls the wait(), check_results(), or get_results() method. Asynchronous Requests in Python | 247
- 269. The callback functionis called without arguments. But a common use for a callback function is to process the results of the fetch, so the function needs access to the RPC object. There are several ways to give the callback function access to the object. One way is to use a bound method, a feature of Python that lets you refer to a method of an instance of a class as a callable object. Define a class with a method that processes the results of the fetch, using an RPC object stored as a member of the class. Create an instance of the class, then create the RPC object, assigning the bound method as the callback by passing it as the callback argument to the create_rpc() function. Exam- ple 10-6 demonstrates this technique. Example 10-6. Using an object method as a callback to access the RPC object from google.appengine.api import urlfetch # ... class CatalogUpdater(object): def prepare_urlfetch_rpc(self): self.rpc = urlfetch.create_rpc(callback=self.process_results) urlfetch.make_fetch_call(self.rpc, 'http://api.example.com/catalog_feed') return self.rpc def process_results(self): try: results = self.rpc.get_result() # Process results.content... except urlfetch.Error, e: # Handle urlfetch errors... class MainHandler(webapp.RequestHandler): def get(self): rpcs = [] catalog_updater = CatalogUpdater(self.response) rpcs.append(catalog_updater.prepare_urlfetch_rpc()) # ... for rpc in rpcs: rpc.wait() Another way to give the callback access to the RPC object is to use a nested function (sometimes called a closure). If the callback function is defined in the same scope as a variable whose value is the RPC object, the function can access the variable when it is called. To allow for this style of calling, the URL Fetch API lets you assign the callback function to the RPC object’s callback property after the object has been created. Example 10-7 demonstrates the use of a nested function as a callback. The create_callback() function creates a function object, a lambda expression, that calls another function with the RPC object as an argument. This function object is assigned to the callback property of the RPC object. 248 | Chapter 10: Fetching URLs and Web Resources
- 270. Example 10-7. Usinga nested function as a callback to access the RPC object from google.appengine.api import urlfetch def process_results(rpc): try: results = self.rpc.get_result() # Process results.content... except urlfetch.Error, e: # Handle urlfetch errors... def create_callback(rpc): # Use a function to define the scope for the lambda. return lambda: process_results(rpc) # ... rpc = urlfetch.create_rpc() rpc.callback = create_callback(rpc) urlfetch.make_fetch_call(self.rpc, 'http://api.example.com/catalog_feed') # ... rpc.wait() If you’ve used other programming languages that support function objects, the create_callback() function may seem unnecessary. Why not create the function object directly where it is used? In Python, the scope of an inner function is the outer function, including its variables. If the outer function redefines the variable containing the RPC object (rpc), when the inner function is called it will use that value. By wrapping the creation of the inner function in a dedicated outer function, the value of rpc in the scope of the callback is always set to the intended object. Someone still needs to call the wait() method on the RPC object so the callback can be called. But herein lies the value of callbacks: the component that calls wait() does not have to know anything about what needs to be done with the results. The main routine can query its subcomponents to prepare and return RPC objects, then later it can call wait() on each of the objects. The callbacks assigned by the subcomponents are called to process each result. If you have multiple asynchronous fetches in progress simultaneously, the callback for a fetch is called if the fetch finishes during any call to wait()—even if the wait() is for another fetch’s RPC object. Of course, the wait() doesn’t return until the fetch for its own RPC object finishes and its callbacks are called. A callback is only called once: if you call wait() for a fetch whose callback has already been called, it does nothing and returns immediately. If your code makes multiple simultaneous fetches, be sure not to rely on a fetch’s callback being called only during its RPC object’s wait(). Asynchronous Requests in Python | 249
- 272. CHAPTER 11 Sending and Receiving Mail and Instant Messages App Engine apps can communicate with the outside world in three ways. The first method we covered in Chapter 10: an app can receive and respond to HTTP requests, and an app can initiate HTTP requests and get responses with the URL Fetch service. The second method of networked communication available to apps is email. An app can send email messages by calling the Mail service with message data and a list of recipients. An app can also receive email messages at any of several addresses monitored by App Engine. The third method is instant messages, specifically the XMPP protocol. An app can participate in a chat dialog with a user of any XMPP-compatible chat service, including Google Talk and any Jabber server. An app can also use XMPP to communicate with custom clients via such services. App Engine does not act as an XMPP service itself; instead, it connects to Google Talk’s infrastructure to participate as a chat user. An app might send email to notify users of system events or the actions of other users (such as to send social networking invitations), confirm user actions (such as to confirm an order), follow up on long-term user actions (such as to send a shipping notice for an order), or send system notifications to administrators. The app can send email on behalf of itself, the app’s administrators, or the currently signed-in user (during the request handler that’s sending the email). An app can receive email sent to specific addresses, such as to provide an email interface to the application, or to moderate or monitor email discussions. An app can reply to the email immediately, or set up work that causes a reply to be sent later. The XMPP service is useful for chat interfaces, such as a chat-based query engine. With a custom client (like a mobile phone client, or Flash in a browser) and an XMPP server or chat service (like Google Talk), the app can use XMPP to push real-time updates to clients. App Engine doesn’t run its own XMPP server, so clients would connect to the chat service (using an account on that service) to communicate with the app. 251
- 273. Sending email andchat messages is similar to initiating HTTP requests: the app calls a service using an API, and the service takes care of making remote connections and managing the appropriate protocols. Unlike the URL Fetch service, the Mail and XMPP services do not return a response immediately. Receiving email and chat messages is also similar to receiving HTTP requests. In fact, they use the same mechanism: request handlers. When a service receives an email or XMPP message intended for your app, the service sends an HTTP request to the app using a specified URL with the message in the HTTP payload. The app processes in- coming messages using request handlers mapped to the specified URLs. The service ignores the response for the request; if the app needs to reply to the user, it can send a message using the API. Figure 11-1 illustrates the flow of incoming email and XMPP messages. Figure 11-1. Architecture of incoming mail and XMPP, calling web hooks in response to incoming message events 252 | Chapter 11: Sending and Receiving Mail and Instant Messages
- 274. Every app hasits own set of email and XMPP addresses, based on its application ID. App Engine does not currently support email and XMPP addresses with custom domains. An app can also receive messages at arbitrary addresses with an app-specific domain name. For email, the app can receive messages at addresses of these forms: app-id@appspotmail.com anything@app-id.appspotmail.com For XMPP chat, the app can receive messages at addresses of these forms (note the differences in the domain names): app-id@appspot.com anything@app-id.appspotchat.com In this chapter, we’ll discuss the APIs for sending and receiving email and XMPP mes- sages, and language-specific tools for creating and processing those messages. Enabling Inbound Services Incoming email and XMPP messages are disabled by default. If an app has an inbound service disabled, then messages sent to an app’s address are ignored, and the service will never attempt to access the app at a web hook URL. To enable inbound services, you add a section to the app’s configuration file. In Python, you add a section similar to the following in the app.yaml file: inbound_services: - mail - xmpp_message In Java, you add a similar section to the appengine-web.xml file, anywhere inside the root element: <inbound-services> <service>mail</service> <service>xmpp_message</service> </inbound-services> These examples enable both services. You can omit one service to keep it disabled while enabling the other. Once your app is deployed, you can confirm that these services are enabled from the Administration Console, under Application Settings. If your app does not appear to be receiving HTTP requests for incoming messages, check the Console and update the configuration if necessary. Enabling Inbound Services | 253
- 275. Sending Email Messages Tosend an email message, you call the API of the Mail service. The outgoing message has a sender address (“From”), one or more recipients (“To”, “Cc”, or “Bcc”), a subject, a message body, and optional file attachments. A outgoing message can have only these fields, and cannot use other email message headers. The Mail service attaches additional headers to the message for tracking pur- poses, such as the date and time the message is sent. You can specify a multipart message body, such as to include both plain text and HTML versions of the message, and to include attachments. The total size of the message, including all headers, cannot exceed one megabyte. The call to the Mail service is asynchronous. When your application calls the Mail service to send a message, the message is enqueued for delivery, and the service call returns. If there is a problem delivering the message, such as if the remote mail server cannot be contacted or the remote server says the address is invalid, an error message is sent via email to the sender address. The app is not notified of the failure by the service directly. (You can use an incoming email address for the app as the sender address. The app will have to parse the message sent by the remote server for an error.) When your app runs in the development server, sending a message causes the server to print information about the message to the logs, and no message is sent. In the Python development server only, you can configure the server to actually send email messages using either Sendmail (if it’s set up on your machine) or an SMTP server. To configure the Python development server to use Sendmail to send email, give the server the --enable_sendmail flag: dev_appserver.py --enable_sendmail appdir To configure the Python development server to use an SMTP server to send email, use the --smtp_host=... (with optional --smtp_port=...), --smtp_user=..., and --smtp_password=... arguments: dev_appserver.py --smtp_host=smtp.example.com --smtp_user=exmail --smtp_password="t3!!t43w0r!d" appdir The Java development server does not support sending email; it just logs the message at the “info” log level. You can adjust the level at which outgoing mail messages are logged by running the development server with the mail.log_mail_level property set to a value such as WARNING. You can also tell the development server to log the body of the message by setting the mail.log_mail_body system property to true. From the com- mand line: dev_appserver.sh --jvm_flag=-Dmail.log_mail_level=WARNING 254 | Chapter 11: Sending and Receiving Mail and Instant Messages
- 276. --jvm_flag=-Dmail.log_message_body=true appdir From Eclipse, select the Run menu, Debug Configurations..., and then select your app’s configuration. Select the Arguments tab, then in the “VM arguments” section, set VM properties like this: -Dmail.log_mail_level=WARNING -Dmail.log_mail_body=true Sender Addresses The sender (“From”) address on an outgoing email message must be one of the allowed addresses: • The Google Account address of one of the application administrators • The address of the user currently signed in to the app with Google Accounts (during the request handler that is sending the message) • A valid incoming email address for the application Replies to messages sent by the app go to the sender address, as do error messages sent by the outgoing mail server (such as “Could not connect to remote host”) or the remote mail server (such as “User not found”). You can use an application developer’s Google Account address as the sender address. To add accounts as application administrators, go to the Developers section of the Administration Console. If you do not want to use the account of a specific developer as the sender address, you can create a new Google Account for a general purpose address, then add it as a developer for the app. Make sure you protect the password of the address’s Google account, since anyone who can sign in to that account can modify your application. You can use Gmail to monitor the account for replies, and you can set up automatic email forwarding in Gmail to relay replies to specific administrators or a mailing list (or Google Group) automatically. You can create a Gmail account using the application ID, and add app- id@gmail.com as a developer. Note that if you create the Gmail account before you register the application ID, you must be signed in using the Gmail account when you register the application ID. App Engine won’t let you register an app ID that matches a Gmail account name unless you are signed in with that account. Technically it’s possible for someone else to register a Gmail account with your app ID after you have registered the app ID, though it’s un- likely. If you’d like to claim your ID as a Gmail address, do it soon. Of course, the owner of the Gmail account can’t administer your app unless you add the account as a developer. Sending Email Messages | 255
- 277. You can usethe email address of a user as the sender address if and only if the address is of a registered Google Account, and the user is signed in and initiated the request whose handler is sending the email. That is, you can send email on behalf of the “cur- rent” user. This is useful if the email is triggered by the user’s action and if replies to the message ought to go to the user’s email address. The Google Accounts API does not expose the user’s human-readable name, so you won’t be able to provide that unless you get it from the user yourself. As we mentioned earlier, an application can receive email messages at addresses of the form app-id@appspotmail.com or anything@app-id.appspotmail.com, where app-id is your application ID and anything can be any string that’s valid on the left side of the email address (it can’t contain an @ symbol). You can use an incoming email address as the sender of an email message to have replies routed to a request handler. The “anything” lets you create custom sender addresses on the fly. For example, a customer support app could start an email conversation with a unique ID and include the ID in the email address (support+ID@app-id.appspotmail.com), and save replies for that conversation in the datastore so the entire thread can be viewed by customer service personnel. Note that the sender address will also receive error (“bounce”) messages. If you use an incoming mail address as the sender, you could have the app process error messages to remove invalid email addresses automatically. Note that different remote email servers may use different formatting for error messages. Any email address can also have a human-friendly name, such as "The Example Team <admin@example.com>". How you do this is specific to the interface; we’ll look at the interfaces in a moment. You can include a separate “Reply-to” address in addition to the sender (“From”) ad- dress. Most mail readers and servers will use this address instead of the sender address for replies and error messages. The “Reply-to” address must meet the same require- ments as the sender address. The development server does not check that the sender address meets these conditions because it doesn’t know who the app’s developers are. Be sure to test features that send email while running on App Engine. Recipients An outgoing email message can use any address for a recipient, and can have multiple recipients. A recipient can be a primary recipient (the “To” field), a secondary or “carbon-copied” recipient (the “Cc” field), or a “blind carbon-copied” recipient (“Bcc”). The “To” and “Cc” recipients are included in the content of the message, so a reply intended for all recipients can be sent to the visible addresses. The “Bcc” recipients receive the message, 256 | Chapter 11: Sending and Receiving Mail and Instant Messages
- 278. but their addressesare not included in the content of the message, and so are not included in replies. The “Bcc” recipient type is especially useful if you want a single message to go to mul- tiple recipients, but you do not want any recipient to know who received the message. You can use this technique to send an email newsletter to users without exposing the users’ email addresses. A common technique for newsletters is to use the sender address as the sole “To” recipient, and make everyone else a “Bcc” recipient. The number of recipients for an email message counts toward an email recipient quota. This quota is initially small to prevent unsolicited email advertisers from abusing the system. You can raise this quota by allocating part of your budget toward email recipients. Attachments An app can attach files to an email message. For security reasons (mostly having to do with insecure email clients), only certain types of files are allowed as attachments. Filenames must end in an extension that reflects the file’s type. Table 11-1 lists the allowed types and the corresponding allowed filename extensions. Table 11-1. The file types allowed for email attachments, and their required filename extensions MIME content type Filename extension Description application/pdf .pdf Portable Document Format (PDF) application/rss+xml .rss RSS feed image/gif .gif GIF image image/jpeg .jpeg , .jpg, .jpe JPEG image image/png .png PNG image image/tiff .tiff, .tif TIFF image image/vnd.wap.wbmp .wbmp WAP bitmap image image/x-ms-bmp .bmp Microsoft Windows bitmap image text/calendar .ics iCalendar event invitation text/comma-separated-values .csv Comma-separated values spreadsheet text/css .css Cascading stylesheet for HTML text/html .html , .htm HTML document text/plain .text, .txt, .asc, .diff, .pot Plain text text/x-vcard .vcf vCard electronic business card These types are usually enough for rich email content, but not necessarily for business documents or desktop applications. If you want to deliver other kinds of files to users, one option is to send a link to a request handler that delivers the file through the Sending Email Messages | 257
- 279. browser. The linkcan be personalized with a temporary unique ID, or restricted using Google Accounts authentication. Sending Email in Python The Python API includes two ways of preparing and sending messages. One way is to call a function with the fields of the message as keyword arguments. Another is to prepare the message in an object, then call a method on the object to send the message. The Mail service API is provided by the google.appengine.api.mail package. The send_mail() method takes the fields of the message as parameters: from google.appengine.api import mail from google.appengine.api import users message_body = ''' Welcome to Example! Your account has been created. You can edit your user profile by clicking the following link: http://www.example.com/profile/ Let us know if you have any questions. The Example Team ''' # (admin@example.com is a Google Account that has # been added as a developer for the app.) mail.send_mail( sender='The Example Team <admin@example.com>', to=users.get_current_user().email(), subject='Welcome to Example.com!', body=message_body) Alternatively, you can prepare the message using an EmailMessage object, then call its send() method. The EmailMessage constructor accepts the same arguments as the send_mail() function: message = mail.EmailMessage( sender='The Example Team <admin@example.com>', to=users.get_current_user().email(), subject='Welcome to Example.com!', body=message_body) message.send() You can also set the fields of an EmailMessage using attributes of the object. This allows you to reuse the same object to send multiple messages with modified values. The possible fields of a message are listed in Table 11-2. 258 | Chapter 11: Sending and Receiving Mail and Instant Messages
- 280. Table 11-2. Fieldsof an email message in the Python interface Field Value Required? sender The sender’s email address. A string. Required to A “To” recipient address as a string, or multiple “To” recipient addresses as a list of strings. Required cc A “Cc” recipient address as a string, or multiple “Cc” recipient addresses as a list of strings. Optional bcc A “Bcc” recipient address as a string, or multiple “Bcc” recipient addresses as a list of strings. Optional “Bcc” recipients receive the message, but are not included in the content of the message. reply_to An alternate address to which clients should send replies instead of the sender address. A Optional string. subject The subject of the message. A string. Required body The plain-text body of the message. A string. Required html An alternate HTML representation of the body of the message, displayed instead of body by Optional HTML-capable email readers. A string. attachments File attachments for the message. A list of tuples, one per attachment, each containing the Optional filename and the file data. The value of an email address field ( sender, to, cc, bcc, reply_to) can be a plain email address: 'juliet@example.com' It can also be an address with a human-readable name, in the standard format (RFC 822): 'Juliet <juliet@example.com>' When you call the send() method or the send_mail() function, the API checks the message to make sure it is valid. This includes testing the email addresses for validity, and making sure the message has all of the required fields. You can call functions to perform these checks separately. The is_email_valid(address) function returns True if it considers an email address valid. The is_initialized() method of an EmailAddress object returns True if the object has all of the fields necessary for sending. The API includes a shortcut method that sends an email message to all administrators (developers) for the application. The send_mail_to_admins() function accepts the same arguments as send_mail(), but without the recipient fields. There is also an AdminEmailMessage class that is similar to the EmailMessage class, but with recipients set to be the app administrators automatically. When calling this function, the message size is limited to 16 kilobytes. (This is a safety limit that ensures delivery of important administrative messages.) Example 11-1 shows a larger example using EmailMessage, with both plain-text and HTML parts, and an attachment. Sending Email Messages | 259
- 281. Example 11-1. Anexample of sending an email message in Python, using several features from google.appengine.api import mail def send_registration_key(user_addr, software_key_data): message_body = ''' Thank you for purchasing The Example App, the best example on the market! Your registration key is attached to this email. To install your key, download the attachment, then select "Register..." from the Help menu. Select the key file, then click "Register". You can download the app at any time from: http://www.example.com/downloads/ Thanks again! The Example Team ''' html_message_body = ''' <p>Thank you for purchasing The Example App, the best example on the market! Your registration key is attached to this email.</p> <p>To install your key, download the attachment, then select <b>Register...</b> from the <b>Help</b> menu. Select the key file, then click <b>Register</b>.</p> <p>You can download the app at any time from:</p> <p> <a href="http://www.example.com/downloads/"> http://www.example.com/downloads/ </a> </p> <p>Thanks again!</p> <p>The Example Team<br /> <img src="http://www.example.com/images/logo_email.gif" /></p> ''' message = mail.EmailMessage( sender='The Example Team <admin@example.com>', to=user_addr, subject='Your Example Registration Key', body=message_body, html=html_message_body, attachments=[('example_key.txt', software_key_data)]) message.send() 260 | Chapter 11: Sending and Receiving Mail and Instant Messages
- 282. Sending Email inJava The Java interface to the Mail service is the JavaMail standard interface (javax.mail.*). There is also a low-level interface, though you can access every feature of the service through the JavaMail implementation. (As such, we’ll only discuss the JavaMail interface here.) To use JavaMail, you first create a JavaMail “session.” The Session object usually con- tains information needed to connect to a mail server, but with App Engine, no config- uration is needed. You prepare the message as a MimeMessage object, then send it using the send() static method of the Transport class. The Transport class uses the most recently created session to send the message. import java.util.Properties; import javax.mail.Message; import javax.mail.MessagingException; import javax.mail.Session; import javax.mail.Transport; import javax.mail.internet.AddressException; import javax.mail.internet.InternetAddress; import javax.mail.internet.MimeMessage; import com.google.appengine.api.users.User; import com.google.appengine.api.users.UserServiceFactory; // ... User user = UserServiceFactory.getUserService().getCurrentUser(); String recipientAddress = user.getEmail(); Properties props = new Properties(); Session session = Session.getDefaultInstance(props, null); String messageBody = "Welcome to Example! Your account has been created." + "You can edit your user profile by clicking the" + "following link:nn" + "http://www.example.com/profile/nn" + "Let us know if you have any questions.nn" + "The Example Teamn"; try { Message message = new MimeMessage(session); message.setFrom(new InternetAddress("admin@example.com", "The Example Team")); message.addRecipient(Message.RecipientType.TO, new InternetAddress(recipientAddress)); message.setSubject("Welcome to Example.com!"); message.setText(messageBody); Transport.send(message); } catch (AddressException e) { // An email address was invalid. // ... } catch (MessagingException e) { Sending Email Messages | 261
- 283. // There wasan error contacting the Mail service. // ... } As shown here, you call methods on the MimeMessage to set fields and to add recipients and content. The simplest message has a sender (setFrom()), one “To” recipient (addRecipient()), a subject (setSubject()), and a plain-text message body (setText()). The setFrom() method takes an InternetAddress. You can create an InternetAddress with just the email address (a String) or the address and a human-readable name as arguments to the constructor. The email address of the sender must meet the require- ments described earlier. You can use any string for the human-readable name. The addRecipient() method takes a recipient type and an InternetAddress. The allowed recipient types are Message.RecipientType.TO (“To,” a primary recipient), Message.RecipientType.CC (“Cc” or “carbon-copy,” a secondary recipient), and Message.RecipientType.BCC (“Bcc” or “blind carbon-copy,” where the recipient is sent the message but the address does not appear in the message content). You can call addRecipient() multiple times to add multiple recipients of any type. The setText() method sets the plain-text body for the message. To include an HTML version of the message body for mail readers that support HTML, you create a MimeMultipart object, then create a MimeBodyPart for the plain-text body and another for the HTML body and add them to the MimeMultipart. You then make the MimeMultipart the content of the MimeMessage: import javax.mail.Multipart; import javax.mail.internet.MimeBodyPart; import javax.mail.internet.MimeMultipart; // ... String textBody = "...text..."; String htmlBody = "...HTML..."; Multipart multipart = new MimeMultipart(); MimeBodyPart textPart = new MimeBodyPart(); textPart.setContent(textBody, "text/plain"); multipart.addBodyPart(htmlPart); MimeBodyPart htmlPart = new MimeBodyPart(); htmlPart.setContent(htmlBody, "text/html"); multipart.addBodyPart(htmlPart); message.setContent(multipart); You attach files to the email message in a similar way: Multipart multipart = new MimeMultipart(); // ... byte[] fileData = getBrochureData(); String fileName = "brochure.pdf"; 262 | Chapter 11: Sending and Receiving Mail and Instant Messages
- 284. String fileType ="application/pdf"; MimeBodyPart attachmentPart = new MimeBodyPart(); attachmentPart.setContent(fileData, fileType); attachmentPart.setFileName(fileName); multipart.addBodyPart(attachmentPart); // ... message.setContent(multipart); Remember that an attachment must be of one of the approved types, and must have a filename that ends with a corresponding filename extension, as listed in Table 11-1. You can add multiple MimeBodyPart objects to a single MimeMultipart. The plain-text body, the HTML body, and the file attachments are each part of a MIME multipart message. When using a MimeMultipart, you must include a text/plain part to be the plain-text body of the message. The multipart object overrides any plain-text content set on the MimeMessage with setText(). App Engine’s implementation of the JavaMail interface includes a shortcut for sending an email message to all of the app’s administrators. To send a message to all adminis- trators, use a recipient address of "admins", with no @ symbol or domain name. Receiving Email Messages With the mail inbound service enabled in configuration, an application can receive email messages at any of several addresses. An incoming mail message is routed to the app in the form of an HTTP request. Email sent to addresses of the following forms are routed to the default version of the app: app-id@appspotmail.com anything@app-id.appspotmail.com The HTTP request uses the POST action, and is sent to the following URL path: /_ah/mail/to-address The recipient email address of the message is included at the end of the URL path, so the app can distinguish between different values of “anything.” The body content of the HTTP POST request is the complete MIME email message, including the mail headers and body. It can be parsed by any library capable of parsing MIME email messages. We’ll look at examples in Python and Java in the next two sections. Receiving Email Messages | 263
- 285. The development serverconsole (http://localhost:8080/_ah/admin/) includes a fea- ture for simulating incoming email by submitting a web form. The development server cannot receive actual email messages. If the app has the incoming mail service enabled but does not have a request handler for the appropriate URL, or if the request handler re- turns an HTTP response code other than 200 for the request, the mes- sage gets “bounced” and the sender receives an error email message. Receiving Email in Python To receive email in Python, you map the incoming email URL path to a script handler in app.yaml file: handlers: - url: /_ah/mail/.+ script: handle_email.py The app address used for the message is included in the URL path, so you can set up separate handlers for different addresses directly in the configuration: handlers: - url: /_ah/mail/support%40.*app-id.appspotmail.com script: support_contact.py - url: /_ah/mail/.+ script: handle_email.py Email addresses are URL-encoded in the final URL, so this pattern uses %40 to represent an @ symbol. Also notice you must include a .* before the application ID when using this technique, so the pattern works for messages sent to version-specific addresses (such as support@dev.latest.app-id.appspotmail.com). The Python SDK includes a class for parsing the POST content into a convenient object, called InboundEmailMessage (in the google.appengine.api.mail package). It takes the multipart MIME data (the POST body) as an argument to its constructor. Here’s an example using the webapp framework: from google.appengine.api import mail from google.appengine.ext import webapp from google.appengine.ext.webapp.util import run_wsgi_app class IncomingMailHandler(webapp.RequestHandler): def post(self): message = mail.InboundEmailMessage(self.request.body) # sender = message.sender # recipients = message.to # body = list(message.bodies(content_type='text/plain'))[0] # ... application = webapp.WSGIApplication([('/_ah/mail/.+', IncomingMailHandler)], debug=True) 264 | Chapter 11: Sending and Receiving Mail and Instant Messages
- 286. def main(): run_wsgi_app(application) if __name__ == '__main__': main() The InboundEmailMessage object includes attributes for the fields of the message, similar to EmailMessage. sender is the sender’s email address, possibly with a displayable name in the standard format (Mr. Sender <sender@example.com>). to is a list of primary recipient addresses, and cc is a list of secondary recipients. (There is no bcc on an incoming message, because blind-carbon-copied recipients are not included in the message content.) subject is the message’s subject. The InboundEmailMessage object may have more than one message body: an HTML body and a plain-text body. You can iterate over the MIME multipart parts of the types text/html and text/plain using the bodies() method. Without arguments, this method returns an iterator that returns the HTML parts first, then the plain-text parts. You can limit the parts returned to just the HTML or plain-text parts by setting the content_type parameter. For example, to get just the plain-text bodies: for text_body in message.bodies(content_type='text/plain'): # ... In the example earlier, we extracted the first plain-text body by passing the iterator to the list() type, then indexing its first argument (which assumes one exists): text = list(message.bodies(content_type='text/plain'))[0] If the incoming message has file attachments, then these are accessible on the attachments attribute. As with using EmailMessage for sending, this attribute is a list of tuples whose first element is the filename and whose second element is the data byte string. InboundEmailMessage allows all file types for incoming attachments, and does not require that the filename accurately represent the file type. Be careful when using files sent to the app by users, since they may not be what they say they are, and have not been scanned for viruses. The Python SDK includes a convenient webapp handler base class for processing incoming email, called InboundMailHandler in the google.appengine.ext.webapp.mail_handlers package. You use the handler by creating a subclass that overrides the receive() method, then installing it like any other handler. When the handler receives an email message, the receive() method is called with an InboundEmailMessage object as its argument. from google.appengine.ext.webapp import mail_handlers class MyMailHandler(mail_handlers.InboundMailHandler): def receive(self, message): # ... application = webapp.WSGIApplication([('/_ah/mail/.+', MyMailHandler)], Receiving Email Messages | 265
- 287. debug=True) def main(): run_wsgi_app(application) if __name__ == '__main__': main() Receiving Email in Java In Java, you map the incoming email URL path to a servlet with an entry in the de- ployment descriptor (web.xml): <servlet> <servlet-name>mailreceiver</servlet-name> <servlet-class>myapp.MailReceiverServlet</servlet-class> </servlet> <servlet-mapping> <servlet-name>mailreceiver</servlet-name> <url-pattern>/_ah/mail/*</url-pattern> </servlet-mapping> The JavaMail and servlet APIs provide everything we need to parse the MIME multipart message in the HTTP POST request. The MimeMessage class (in the javax.mail.internet package) has a constructor that accepts a java.io.InputStream, which we can get from the HttpServletRequest using its getInputStream() method. The MimeMessage constructor also needs a JavaMail Session, which, as with sending email, can use the default empty configuration. import java.io.IOException; import java.util.Properties; import javax.mail.Session; import javax.mail.MessagingException; import javax.mail.Multipart; import javax.mail.Part; import javax.mail.internet.MimeMessage; import javax.servlet.http.*; public class MailReceiverServlet extends HttpServlet { public void doPost(HttpServletRequest req, HttpServletResponse resp) throws IOException { Properties props = new Properties(); Session session = Session.getDefaultInstance(props, null); try { MimeMessage message = new MimeMessage(session, req.getInputStream()); String contentType = message.getContentType(); Object content = message.getContent(); if (content instanceof String) { // A plain text body. // ... } else if (content instanceof Multipart) { 266 | Chapter 11: Sending and Receiving Mail and Instant Messages
- 288. // A multipartbody. for (int i = 0; i < ((Multipart) content).getCount(); i++) { Part part = ((Multipart) content).getBodyPart(i); // ... } } } catch (MessagingException e) { // Problem parsing the message data. // ... } } } If the incoming message is a MIME multipart message (such as a message with an HTML body, or attachments), the getContent() method of the MimeMessage returns an object that implements the Multipart interface. You can use this interface to get a count of the parts (getCount()) and select parts by index (getBodyPart(int index), which returns a BodyPart). Sending XMPP Messages An App Engine application can send instant messages to users of XMPP-compliant chat services, including Google Talk and Jabber servers, using the XMPP service. Each participant in an XMPP communication has an address similar to an email ad- dress, known as a JID. (JID is short for “Jabber ID,” named after the Jabber project, where XMPP originated.) A JID consists of a username, an “at” symbol (@), and the domain name of the XMPP server. A JID can also have an optional “resource” string, which is sometimes used to identify specific clients connected to the service with the username; a message sent to the ID without the resource goes to all connected clients. username @ domain / resource To send a message, a chat participant sends an XMPP message to its own XMPP server. The participant’s chat service contacts the recipient service’s host using the domain name of the JID and a standard port, then delivers the message. If the remote service accepts messages for the JID and someone is connected to the service with a chat client for that JID, the service delivers the message to the client. Most services, including Google Talk, require that the user has established a previous relationship with a sender before it will route chat messages to the user. The user can establish this relationship with an App Engine app either by sending a chat invitation to the app, or by accepting a chat invitation sent by the app. App Engine applications can send chat invitations as well as messages. Sending XMPP Messages | 267
- 289. In most cases,App Engine applications should only send a chat invita- tion to a user at the request of the user. Sending unsolicited chat invi- tations could be considered abuse and a violation of the terms of service. App Engine’s XMPP service uses the same infrastructure as Google Talk, Google’s Jabber-compatible instant messaging service. You can use Google Talk directly via the Google Talk client for Windows, from Gmail’s web interface, from a Google Talk web gadget, or any Jabber-compatible chat client. You can also develop a custom client that connects to Google Talk via XMPP with a Google Account. This could be an application that runs on a desktop computer or an Internet-capable mobile phone, or it could be a Flash applet that runs in a browser window. (Web browsers do not support making an XMPP connection directly. Gmail’s interface to Google Talk uses an HTTP relay hosted by Google, which does not have a public protocol.) A custom client connected in this way can have messages pushed to it from your App Engine app. For more information on developing custom Google Talk clients, see the Google Talk developer documentation: http://code.google.com/apis/talk/ Any user with a Google Account can access the Google Talk service for free with any XMPP-compatible chat client. This includes users with Google Accounts on Google Apps domains. An XMPP message includes a sender address, one or more recipient addresses, a mes- sage type, and a message body. The sender address must be one of the app’s incoming XMPP addresses. These are of the form app-id@appspot.com or anything@app-id.appspotchat.com, where app-id is your application ID and anything can be any string that’s valid on the left side of a JID (it can’t contain an @ symbol). Unlike incoming email addresses, it’s not as convenient to use the “anything” form for creating IDs on the fly, since the recipient needs to accept an invitation from that ID before receiving messages, but it can still be useful for sessions that begin with an invitation, or addresses that represent specific purposes or users of the app (support@app-id.appspotchat.com). If the version of the app that is sending an XMPP message is not the default version, App Engine modifies the sender address to a version-specific address, so replies go directly to the correct version: anything@version.latest.app-id.appspotchat.com (app- id@appspot.com becomes app-id@version.latest.app-id.appspotchat.com.) App Engine adds a “resource” to the end of the sender JID (after the domain name) that looks like this: /bot. This is mostly just to comply with the best practice of sending 268 | Chapter 11: Sending and Receiving Mail and Instant Messages
- 290. messages using JIDswith resources. It isn’t noticed by chat users, and is not needed when a user wishes to send a message to the app. You’ll see it in log messages. The message type can be any of the types in the XMPP standard, including “chat,” “error,” “groupchat,” “headline,” and “normal.” An app can only receive messages of the types “chat” and “normal,” and so cannot participate in group chats. For straight- forward communication between an app and a chat user, you usually want to send “chat” messages. For an app and a custom client, you can do what you like. When an app is running in the development server, sending an XMPP chat message or invitation causes the server to print the message to the console. The development server does not contact the XMPP service or send messages. The APIs for calling the XMPP service are straightforward. In Python, each major action is a function in the google.appengine.api.xmpp module. In Java, each action is a method of an XMPPService object, which you get from XMPPServiceFactory.getXMPPService(). (There is no Java standard API for this one.) Sending a Chat Invitation To invite a user to chat in Python, you call the send_invite() method. It takes the recipient JID as its first argument, and an optional sender JID (from_jid) as its second argument. By default, it uses app-id@appspot.com as the sender JID. from google.appengine.api import xmpp jid = 'juliet@example.com' xmpp.send_invite(jid) # from app-id@appspot.com xmpp.send_invite(jid, from_jid='support@app-id.appspotchat.com') # from a custom JID In Java, each JID is represented by a JID object, which you create by passing the address as a string to the JID constructor. To send an invitation, you call the sendInvitation() method with either one or two arguments: import com.google.appengine.api.xmpp.JID; import com.google.appengine.api.xmpp.XMPPService; import com.google.appengine.api.xmpp.XMPPServiceFactory; // ... XMPPService xmpp = XMPPServiceFactory.getXMPPService(); // From app-id@appspot.com: xmpp.sendInvitation(JID("juliet@example.com")); // From a custom JID: xmpp.sendInvitation(JID("juliet@example.com"), JID("support@app-id.appspotchat.com")); Sending XMPP Messages | 269
- 291. Sending a ChatMessage To send a chat message in Python, you call the send_message() function. The function takes a JID or list of JIDs, the body of the message, and an optional sender JID (from_jid). It returns a success code, or a list of success codes, one for each recipient JID. result = xmpp.send_message( 'juliet@example.com', 'Your dog has reached level 12!') if result != xmpp.NO_ERROR: # ... By default, this sends a message of the “chat” type. You can send a message of a different type by setting the message_type parameter. Acceptable values include xmpp.MESSAGE_TYPE_CHAT (the default), xmpp.MESSAGE_TYPE_ERROR, xmpp.MESSAGE_TYPE_GROUPCHAT, xmpp.MESSAGE_TYPE_HEADLINE, and xmpp.MESSAGE_TYPE_NORMAL. Complete XMPP messages are sent over the network as XML data. By default, send_message() treats the text of the message as plain text, and knows to escape XML characters. Instead of a text message, you can send an XML stanza. This is included verbatim (assuming the stanza is well formed) in the XMPP message, so you can send structured data to XMPP clients. To tell send_message() that the content is an XML stanza so it doesn’t escape XML characters, provide the raw_xml=True parameter. The send_message() function returns a status code for each recipient JID, as a single value if called with a single JID, or as a list of codes if called with a list of JIDs. The possible status values are xmpp.NO_ERROR, xmpp.INVALID_JID, and xmpp.OTHER_ERROR. In Java, you send a message by calling the sendMessage() method of the XMPPService. The method takes a Message object, which you build with a MessageBuilder object. sendMessage() returns a SendResponse object, which contains status codes for each in- tended recipient of the message. import com.google.appengine.api.xmpp.JID; import com.google.appengine.api.xmpp.Message; import com.google.appengine.api.xmpp.MessageBuilder; import com.google.appengine.api.xmpp.SendResponse; import com.google.appengine.api.xmpp.XMPPService; import com.google.appengine.api.xmpp.XMPPServiceFactory; // ... XMPPService xmpp = XMPPServiceFactory.getXMPPService(); JID recipient = new JID("juliet@example.com"); Message message = new MessageBuilder() .withRecipientJids(recipient) .withBody("Your dog has reached level 12!") .build(); SendResponse success = xmpp.sendMessage(message); 270 | Chapter 11: Sending and Receiving Mail and Instant Messages
- 292. if (success.getStatusMap().get(recipient) !=SendResponse.Status.SUCCESS) { // ... } You use the MessageBuilder class to assemble the (immutable) Message object. You can chain its methods to construct a complete message in a single statement. Relevant methods include: withBody(String body) Sets the message body. asXml(boolean asXml) Declares that the body contains a well-formed XML stanza (and not plain text). withFromJid(JID jid) Sets the sender JID. withRecipientJids(JID jid1, ...) Adds one or more recipient JIDs. withMessageType(MessageType type) Sets the message type. build() Returns the finished Message. Message types are represented by the MessageType enum: MessageType.CHAT, MessageType.ERROR, MessageType.GROUPCHAT, MessageType.HEADLINE, and MessageType.NORMAL. The sendMessage() method returns a SendResponse object. Calling this object’s getStatusMap() method returns a Map<JID, SendResponseStatus>, a map of recipient JIDs to status codes. The possible status codes are SendResponse.Status.SUCCESS, SendResponse.Status.INVALID_ID, and SendResponse.Status.OTHER_ERROR. Checking a Google Talk User’s Status App Engine’s XMPP service supports inquiring about the availability (or presence) of a Google Talk user, whether the user is online and accepting chat messages. Querying for presence is specific to the Google Talk service and not XMPP, so this feature only works with Talk users. (XMPP chat clients use presence notification messages, which are not yet supported by App Engine.) A Google Talk user will only appear as available to an App Engine app if the user has a prior chat relationship with the app (the user accepted an invitation from the app, or has sent the app an invitation). All other users appear as “not present.” A user’s presence value has no relationship with the user’s “available” status. A con- nected user appears as present even if her status is set to “away.” Also, there is no way to read the user’s status message. Sending XMPP Messages | 271
- 293. To get thestatus of a Google Talk user in Python, you call the get_presence() function with the user’s JID and an optional custom sender JID (from_jid). The function returns True if the user is connected and can receive chat messages from the app. if xmpp.get_presence('juliet@example.com'): # User can receive chat messages from the app. # ... In Java, you call the getPresence() method of the XMPPService instance with the user’s JID and an optional sender JID. The method returns a Presence object, whose isAvailable() method returns true if the user is connected. import com.google.appengine.api.xmpp.Presence; // ... JID jid = new JID("juliet@example.com"); Presence presence = xmpp.getPresence(jid); if (presence.isAvailable()) { // ... } When running in the development server, all user addresses are present. Receiving XMPP Messages With the xmpp_message inbound service enabled, an app can receive XMPP messages sent by users and clients of XMPP chat services, such as Google Talk. Incoming mes- sages are routed to the application in the form of HTTP requests to a specified URL. This is similar to how an app can receive email. An app can receive two kinds of XMPP messages: “chat” and “normal.” Chat messages are sent by user-operated instant messaging clients, and usually represent a conversa- tion. Normal messages are general purpose messages, which you might use for custom clients. Notice that this means an app cannot be a participant in a group chat. An app receives XMPP messages at several addresses. Messages sent to addresses of these forms are routed to the default version of the app: app-id@appspot.com anything@app-id.appspotchat.com Messages sent to addresses of this form are routed to the specified version of the app, useful for testing: anything@version.latest.app-id.appspotmail.com Each message is delivered to the app as an HTTP POST request to the following URL path: /_ah/xmpp/message/chat/ (Unlike incoming email, the sender JID is not included in the URL path.) 272 | Chapter 11: Sending and Receiving Mail and Instant Messages
- 294. The body contentof the HTTP POST request is a MIME multipart message, with a part for each field of the message: from The sender’s JID. to The app JID to which this message was sent. body The message body content (with characters as they were originally typed). stanza The full XML stanza of the message, including the previous fields (with XML spe- cial characters escaped); useful for communicating with a custom client using XML. The Python and Java SDKs include classes for parsing the request data into objects. When the service is enabled, all incoming messages are accepted and routed to the app. All incoming chat invitations are accepted automatically. The app always appears as connected and available when someone inquires for the app’s presence. The development server console (http://localhost:8080/_ah/admin/) includes a fea- ture for simulating incoming XMPP messages by submitting a web form. The devel- opment server cannot receive actual XMPP messages. Receiving XMPP Messages in Python In Python, you map the URL path to a script handler in app.yaml file, as usual: handlers: - url: /_ah/xmpp/message/chat/ script: handle_xmpp.py You can parse the incoming message into a Message object by passing a mapping of the POST parameters to its constructor. With the webapp framework, this is a simple matter of passing the parsed POST data (a mapping of the POST parameter names to values) in directly: from google.appengine.api import xmpp from google.appengine.ext import webapp from google.appengine.ext.webapp.util import run_wsgi_app class IncomingXMPPHandler(webapp.RequestHandler): def post(self): message = xmpp.Message(self.request.POST) message.reply('I got your message!') application = webapp.WSGIApplication([('/_ah/xmpp/message/chat/', IncomingXMPPHandler)], debug=True) Receiving XMPP Messages | 273
- 295. def main(): run_wsgi_app(application) if __name__ == '__main__': main() The Message object has attributes for each message field: from, to, and body. It also includes a convenience method for replying to the message: reply(), which takes the body of the reply as its first argument. The Python SDK includes an extension for webapp that makes it easy to implement chat interfaces that perform user-issued commands. Here’s an example that responds to the commands /stats and /score username, and ignores all other messages: from google.appengine.api import xmpp from google.appengine.ext import webapp from google.appengine.ext.webapp import xmpp_handlers from google.appengine.ext.webapp.util import run_wsgi_app def get_stats(): # ... def get_score_for_user(username): # ... class UnknownUserError(Exception): pass class ScoreBotHandler(xmpp_handlers.CommandHandler): def stats_command(self, message): stats = get_stats() if stats: message.reply('The latest stats: %s' % stats) else: message.reply('Stats are not available right now.') def score_command(self, message): try: score = get_score_for_user(message.arg) message.reply('Score for user %s: %d' % (message.arg, score)) except UnknownUserError, e: message.reply('Unknown user %s' % message.arg) application = webapp.WSGIApplication([('/_ah/xmpp/message/chat/', ScoreBotHandler)], debug=True) def main(): run_wsgi_app(application) if __name__ == '__main__': main() 274 | Chapter 11: Sending and Receiving Mail and Instant Messages
- 296. The CommandHandler base class, provided by the google.appengine.ext.webapp.xmpp_handlers package, parses the incoming message for a command of the following syntax: /commandname argument If the message contains such a command, the handler attempts to call a method named after the command, such as commandname_command(). If no such method exists, it calls the unhandled_command() method, whose default implementation replies to the message with “Unknown command.” You can override this method to customize its behavior. The base handler calls the command method with the Message object as an argument. The handler annotates the object with two additional attributes: command, the name of the command, and arg, the text following the command, if any. If the incoming message does not start with a /commandname-style command, the base handler calls the text_message() method with the Message as its argument. The default implementation does nothing, and you can override it to specify behavior in this case. This package also contains a simpler handler class named BaseHandler, with several useful features. It parses incoming messages, and logs and ignores malformed messages. If a message is valid, it calls its message_received() method, which you override with the intended behavior. The class also overrides webapp.RequestHandler’s handle_exception() method to send an XMPP reply with a generic error message when an uncaught exception occurs, so the user isn’t left to wonder whether the message was received. (CommandHandler extends BaseHandler, and so also has these features.) Receiving XMPP Messages in Java In Java, you process incoming XMPP messages by mapping a servlet to the URL path called by the XMPP service, in the deployment descriptor: <servlet> <servlet-name>xmppreceiver</servlet-name> <servlet-class>myapp.XMPPReceiverServlet</servlet-class> </servlet> <servlet-mapping> <servlet-name>xmppreceiver</servlet-name> <url-pattern>/_ah/xmpp/message/chat/</url-pattern> </servlet-mapping> The XMPPService object includes a parseMessage() method that knows how to parse the incoming request data into a Message object. You access the data using the Message’s methods. import java.io.IOException; import javax.servlet.http.*; import com.google.appengine.api.xmpp.Message; import com.google.appengine.api.xmpp.XMPPService; import com.google.appengine.api.xmpp.XMPPServiceFactory; Receiving XMPP Messages | 275
- 297. public class XMPPReceiverServletextends HttpServlet { public void doPost(HttpServletRequest req, HttpServletResponse resp) throws IOException { XMPPService xmpp = XMPPServiceFactory.getXMPPService(); Message message = xmpp.parseMessage(req); // ... } } You access the fields of a Message using methods: getFromJid() The sender JID. getMessageType() The MessageType of the message (MessageType.CHAT or MessageType.NORMAL). getRecipientJids() The app JID used, as a single-element JID[]. getBody() The String content of the message. getStanza() The String raw XML of the message. 276 | Chapter 11: Sending and Receiving Mail and Instant Messages
- 298. CHAPTER 12 Bulk Data Operations and Remote Access The datastore is designed for retrieving and processing data at a rate suitable for inter- active web applications. A single request may need to update a few entities, or query for and retrieve several dozen. Larger amounts of work triggered by interactive actions are not typically expected to be performed within a single web request. (App Engine’s facility for such work, task queues, is discussed in Chapter 13.) User-initiated actions tend to fit into this model of short requests. Administrative actions, such as one-off chores performed by the application’s main- tainers and staff, sometimes follow a different model. You may need to create or update thousands of entities from a datafile, or download the complete contents of the data- store for archival or offline testing. A change to a data model in a new version of the software may require modifying millions of entities to match before the software can be deployed. You could build features into the application to do these things, and that may be ap- propriate for routine maintenance tasks. But App Engine has a feature that makes it easy to do ad hoc work with a live application, driven from your own computer: a remote access API. With the remote access API, utilities running on your computer can call the live appli- cation’s services, such as the datastore, using the same APIs and libraries that the ap- plication uses. Each service call becomes a web request that is processed by a request handler. Each interaction uses a secure connection and is restricted to developer ac- counts. Figure 12-1 shows how the remote API request handler gives transparent access to the live application’s services. 277
- 299. Figure 12-1. Architectureof remote API, with remote calls performed by a web hook App Engine includes a tool for uploading and downloading data via the remote API. The tool can create new datastore entities using data from a comma-separated values (CSV) datafile (which can be exported and imported from most spreadsheet applica- tions), and can create CSV files with data from the app’s datastore. It can use the app’s data models to validate data before storing it, and can be extended to perform custom transformations on the data. The tool also makes it easy to do a straight download of the entire datastore, and can upload that data back to the app, or to another app. App Engine also has a tool for running a Python command-line session with the remote API. With this remote shell tool, you can execute arbitrary Python statements using the Python API, and service calls will operate on the live application. These remote access features come in two parts: the remote API request handler, and the tools and libraries that call the handler. The remote API handler is available for both Python and Java applications, and is included with each runtime environment. Both versions of the remote API use the same network protocol, so any tool that works with the Python version also works with the Java version. Currently, the remote access tools and libraries are only available for Python. You can use these tools and libraries with a Java application via the Java remote API request handler. You will need to download and install the Python SDK to get the tools and libraries. For some features, you may also need to reimplement your data models using the Python data modeling library (described in Chapter 7). See the section “Installing the Python SDK” on page 16. In this chapter, we’ll describe how to set up the remote API request handler for Python and Java apps, and how to use the bulk upload and download tools from the Python SDK. We’ll also look at the Python remote shell utility. Finally, we’ll look at how to use the remote API directly in your own Python scripts. Setting Up the Remote API for Python The Python remote API request handler is included in the runtime environment. You can set it up at a URL path of your choosing in your app.yaml file. Be sure to restrict this URL path to administrators using login: admin. 278 | Chapter 12: Bulk Data Operations and Remote Access
- 300. Here’s an excerptthat sets up the remote API URL using the path /remote_api: handlers: - url: /remote_api script: $PYTHON_LIB/google/appengine/ext/remote_api/handler.py login: admin The $PYTHON_LIB portion of the script path gets replaced by the system with the actual path to the App Engine libraries. You can test this URL in a browser using the development server. Visit the URL (such as http://localhost:8080/remote_api), and make sure it redirects to a fake authenti- cation form. Check the box to sign in as an administrator, and click Submit. You should see this message: This request did not contain a necessary header. The remote API expects an HTTP header identifying the remote API protocol version to use, which the browser does not provide. But this is sufficient to test that the handler is configured correctly. Setting Up the Remote API for Java To use the remote API tools with a Java application, you set up a URL path with a servlet provided by the SDK, namely com.google.apphosting.utils.remoteapi.RemoteApiServlet. You can choose any URL path; you will give this path to the remote API tools in a command-line argument. Be sure to restrict access to the URL path to administrators. The following excerpt for your deployment descriptor (web.xml) associates the remote API servlet with the URL path /remote_api, and restricts it to administrator accounts: <servlet> <servlet-name>remoteapi</servlet-name> <servlet-class> com.google.apphosting.utils.remoteapi.RemoteApiServlet </servlet-class> </servlet> <servlet-mapping> <servlet-name>remoteapi</servlet-name> <url-pattern>/remote_api</url-pattern> </servlet-mapping> <security-constraint> <web-resource-collection> <web-resource-name>remoteapi</web-resource-name> <url-pattern>/remote_api</url-pattern> </web-resource-collection> <auth-constraint> <role-name>admin</role-name> </auth-constraint> </security-constraint> Setting Up the Remote API for Java | 279
- 301. Using the BulkLoader Tool The bulk loader tool is part of the Python SDK, and requires that Python be installed on your computer. The main tool is called bulkloader.py. On Windows, the Python SDK installer puts this command in the command path. On Mac OS X, the Launcher creates a symbolic link to the tool at /usr/local/bin/bulkloader.py when you first run it, which may or may not be in your command path. If you installed the SDK using the ZIP archive, the command is in the SDK’s root directory. You can verify you can run the bulkloader.py command by running it without arguments: bulkloader.py If you have a Python application, you can use the appcfg.py upload_data and appcfg.py download_data commands to do data uploads and downloads. These are equivalent to bulkloader.py, but can also derive the application ID and the remote API URL from your app.yaml file. (Future releases may make more bulk loader features available via appcfg.py; check the App Engine website for the latest.) If you have a Java application, you just run the bulkloader.py tool, specifying the ap- plication ID and remote API URL as command-line arguments: bulkloader.py --app-id=appid --url=http://appid.appspot.com/remote_api ... You can use the bulk loader tool with the development server by spec- ifying the equivalent remote API URL for it using the --url=... argu- ment. This is useful for testing the remote API URL configuration, for verifying data uploads before doing them with the live application, and for managing test data during development. You must still specify an accurate --app_id=... argument when using the development server. bulkloader.py --app_id=app-id --url=http://localhost:8080/remote_api ... The tool will prompt for an email address and password. You can enter any values for these when using the development server. If you get an “Authentication Failed” error when using the loader with the development server, then your remote API URL is not working. Check your configuration and the development server output. Installing SQLite The loader tool keeps track of its progress during an upload or download in a datafile, so you can interrupt it and start it again from where it left off instead of starting the job from the beginning. To do this, the tool requires a Python module called sqlite3. You may already have this module installed in your version of Python. Verify that you have 280 | Chapter 12: Bulk Data Operations and Remote Access
- 302. the sqlite3 moduleinstalled by starting a Python shell and importing it. If nothing happens, the module is installed: % python Python 2.6.1 (r261:67515, Jul 7 2009, 23:51:51) [GCC 4.2.1 (Apple Inc. build 5646)] on darwin Type "help", "copyright", "credits" or "license" for more information. >>> import sqlite3 >>> If you do not have the module, you can get it from the SQLite website. Note that if you compiled your Python installation from source, you will need to recompile it with SQLite support. http://www.sqlite.org/download.html If you’d prefer not to install SQLite, you can disable the use of the progress datafile in the bulk loader tool by providing this command-line argument when running the tool: --db_filename=skip. Backup and Restore One of the easiest features of the bulk loader to use is the backup and restore feature. The bulk loader can download every entity of a given kind to a datafile, and can upload that datafile back to the app to restore those entities. It can also upload the datafile to another app, effectively moving all of the entities of a kind from one app to another. You can also use these features to download data from the live app, and “upload” it to the development server for testing (using the --url=... argument mentioned earlier). These features require no additional configuration or code beyond the installation of the remote API handler. To download every entity of a given kind, run the bulkloader.py tool with the --dump, --kind=..., and --filename=... arguments: bulkloader.py --dump --app_id=app-id --url=http://app-id.appspot.com/remote_api --kind=kind --filename=backup.dat (This is a single command line. The characters are line continuation markers, sup- ported by some command shells. You can remove these characters and type all of the arguments on one line.) To upload a datafile created by --dump, thereby re-creating every entity in the file, use the --restore argument instead of --dump: bulkloader.py --restore --app_id=app-id --url=http://app-id.appspot.com/remote_api --kind=kind --filename=backup.dat Using the Bulk Loader Tool | 281
- 303. When using --dumpand --restore, entities are not validated using the app’s model classes. They are dumped and restored directly as they appear in the datastore. Entity keys are preserved, including system-assigned numeric IDs. Properties with key values will remain valid as long as the entities they are referring to are in the app or are part of the dump being restored. Restoring an entity with the same key as an existing entity will replace the existing entity. Keys are preserved even if you upload to an app other than the one that originally contained the dumped entities. If you restore an entity with a system-assigned numeric ID, the ID counter for the given key path and kind is advanced past the restored ID, so future new entities will not overwrite the restored entity. Uploading Data The bulk loader uploads data from your computer to your application. Out of the box, the loader supports reading data from a datafile, and the loader can be customized extensively. This makes it easy to produce a set of fixed information, such as informa- tion about products in an online store, and upload it to the app as entities in the data- store. The translation is obvious: each row represents one entity, and each column represents a property of the entity. The tool reads datafiles in the CSV file format, which can be exported by most spreadsheet applications, including Microsoft Excel and Google Spreadsheets. To perform the translation, the bulk loader needs to know the intended data types and property names for each column. You describe this in a file of Python code that defines a “loader” class for each kind to be imported. The class can define additional code to use to translate each string value from the datafile to an appropriate datastore value, and can postprocess entities before they are saved to add or delete properties and create additional entities. The loader class can even replace the CSV file reading routine to read data in other formats. Here is a simple loader definition file. Let’s call it loaders.py: from google.appengine.tools import bulkloader import datetime # Import the app's data models directly into # this namespace. We must add the app # directory to the path explicitly. import sys import os.path sys.path.append( os.path.abspath( os.path.dirname( os.path.realpath(__file__)))) from models import * def get_date_from_str(s): if s: 282 | Chapter 12: Bulk Data Operations and Remote Access
- 304. return datetime.datetime.strptime(s, '%m/%d/%Y').date() else: return None class BookLoader(bulkloader.Loader): def __init__(self): bulkloader.Loader.__init__(self, 'Book', [('title', str), ('author', str), ('copyright_year', int), ('author_birthdate', get_date_from_str), ]) loaders = [BookLoader] The loader definition file must do several things: 1. It must import or define a data model class for each kind to import, using the Python data modeling API. The loader file does not have the app directory in the module load path, so the loader file must add it. The modules should be imported directly into the local namespace. 2. It must define (or import) a loader class for each kind to import, a subclass of the Loader class in the package google.appengine.tools.bulkloader. This class must have a constructor (an __init__() method) that calls the parent class’s constructor with arguments that describe the CSV file’s columns. 3. It must define a module-global variable named loaders, whose value is a list of loader classes. The loader class’s __init__() method calls its parent class’s __init__() method with three arguments: self (the object being initialized), the kind of the entity to create as a string, and a list of tuples. Each tuple represents a column from the CSV file, in order from left to right. The first element of each tuple is the property name to use. The second element is a Python function that takes the CSV file value as a string and returns the desired datastore value in the appropriate Python value type. (Refer to Table 4-1 for a list of the Python value types for the datastore.) For simple data types, you can use the constructor for the type as the type conversion function. In the example above, we use str to leave the value from the CSV file as a string, and int to parse the CSV string value as an integer. Depending on the output of your spreadsheet program, you can use this technique for long, float, db.Text, and other types whose constructors accept strings as initial arguments. For instance, users.User takes an email address as a string argument. The bool type constructor returns False for the empty string, and True in all other cases. If your CSV file contains Boolean values in other forms, you’ll need to write a converter function, like this: def bool_from_string(s): return s == 'TRUE' Using the Bulk Loader Tool | 283
- 305. Dates, times, anddate-times usually require a custom converter function to make an appropriate datetime.datetime value from the string. As shown previously, you can use the datetime.datetime.strptime() function to parse the string in a particular format. Tailor this to the output of your spreadsheet program. To create a multivalued property, you can use a type conversion function that returns a list value. However, the function must get all of its data from a single cell (though you can work around this by tweaking the file reader routine, described below). It cannot reference other cells in the row, or other rows in the table. Storing a key value (a reference to another entity) is usually a straightforward matter of defining a type conversion function that uses db.Key.from_path() with the desired kind and a key name derived from a CSV file value. By default, the bulk loader creates each entity using a key with no ancestors, the given kind, and a system-assigned numeric ID. You can tell the bulk loader to use a custom key name instead of a numeric ID by defining a method of your loader class named generate_key(). This method takes as arguments a loop counter (i) and the row’s col- umn values as a list of strings (values), and returns either a string key name or a db.Key object. The db.Key object can include an entity group parent key. If the function returns None, a system-assigned numeric ID will be used. class BookLoader(bulkloader.Loader): # ... def generate_key(self, i, values): # Use the first column as the key name. return values[0] If an entity is uploaded with the same key as an entity that already exists in the datastore, the new entity will replace the existing entity. You can use this effect to download, modify, then reupload data to update ex- isting entities instead of creating new ones. You can “skip” an unused column in the CSV file by giving it a property name that begins with an underscore. The Python data modeling API does not save attributes beginning with an underscore as properties, so the final entity will not have a property with this name. You can reuse the same dummy property name for all columns you wish to skip: class BookLoader(bulkloader.Loader): def __init__(self): bulkloader.Loader.__init__(self, 'Book', [('_UNUSED', lambda x: None), ('title', str), ('author', str), ('_UNUSED', lambda x: None), ('_UNUSED', lambda x: None), ('copyright_year', int), 284 | Chapter 12: Bulk Data Operations and Remote Access
- 306. # ... ]) If you have a column that is to be the entity key name (via generate_key()) and you do not want this value to also become a property, you must “skip” the column in the column list using this technique. The bulk loader will complain if it finds more columns than are declared in the list. The bulk loader uses the Python data modeling API to construct, validate, and save entities. If your app is written in Python, you just use the model classes of your app for this purpose, and get the validation for free. If your app is written in Java, you have a choice to make. You can either implement Python data model classes that resemble your application’s data model, or you can define Python data models that allow arbitrary properties to be set by the loader without validation. For information on defining data models in Python, see Chapter 7. To define a model class that allows all properties and does no validation, simply define a class like the following for each kind in the loader file, above the definition of the loader classes: from google.appengine.ext import db class Book(db.Expando): pass The loader class can do postprocessing on an entity after it is created but before it is saved to the datastore. This gives you the opportunity to inspect the entity, add or remove properties, and even create additional entities based on each row of the datafile. To perform postprocessing, implement a method named handle_entity() on the loader class. This takes an instance of the model class as its only argument (entity). It should return the instance to create, a list of model instances if more than one entity should be created, or None if no entity should be created. You can extend the loader class to read file formats other than CSV files. To do this, you override the generate_records() method. This method takes a filename (filename), and returns a Python generator. Each iteration of the generator should yield a list of strings for each entity to create, where each string corresponds to a property defined in the loader’s constructor. You can use a generate_records() routine to mas- sage the columns processed by the loader so that all the information needed for a single property appears in the column’s value, such as to assemble what will eventually be- come a multivalued property from multiple columns in the datafile. You can override the initialize() method of the loader to perform actions at the be- ginning of an upload. This method takes the input file’s name (filename) and a string given on the command line as the --loader_opts=... argument (loader_opts), which you can use to pass options to this method. Similarly, you can override the finalize() method to do something after the upload is complete. Using the Bulk Loader Tool | 285
- 307. To start theupload for a Java app or a Python app, run the bulkloader.py command as follows: bulkloader.py --app_id=app-id --url=http://app-id.appspot.com/remote_api --config_file=loaders.py --kind=Book --filename=bookdata.csv If you have a Python app, you can use appcfg.py upload_data to do the same thing, with fewer arguments to type. (The app ID and remote API URL are derived from the app’s configuration file.) appcfg.py upload_data --config_file=loaders.py --kind=Book --filename=bookdata.csv app-dir The --config_file=... argument specifies the file of Python code that defines the loader classes. The --kind=... argument states the kind of the entities to import from the datafile; the loader figures out which Loader class and model class to use using the kind. --filename=... is the name of the CSV file. If your CSV file’s first row is a header row and not actual data, add the --has_header argument to skip the first row. If any row of the CSV file fails to translate to an entity, either because a type conversion function raised an exception or because the corresponding model class does not con- sider a converted value valid, the bulk loader stops the upload and reports the error. If you have sqlite3 installed, you can start the upload from where it left off (after fixing the offending value in the CSV file) by running the command again with the same arguments. You can do a test run of an upload without actually uploading any data by adding the --dry_run argument. This causes the loader to process the CSV file and construct and validate each entity object, and report any errors it finds. It doesn’t upload any data or otherwise access the app. Downloading Data You can use the bulk loader to download data from the app’s datastore as well as upload it. The download feature is symmetrical with the upload feature: it supports down- loading an app’s entities and adding them as rows to a CSV datafile. You provide Python code that describes how entities should be saved, and you can customize the download process to perform data transformations or output other file formats. Similar to the loader class for uploads, the download feature uses a file of Python code that defines an exporter class for each kind of entity to download. The exporter class 286 | Chapter 12: Bulk Data Operations and Remote Access
- 308. specifies which propertygoes in which column of the resulting CSV file, and can include value transformation functions to get the entity data into a format accepted by your spreadsheet program. Here is a simple exporter definition file that is symmetric with the loader definition example in the previous section. Data downloaded with this exporter can be loaded back in with the loader. Let’s call this file exporters.py: from google.appengine.tools import bulkloader import datetime # Import the app's data models directly into # this namespace. We must add the app # directory to the path explicitly. import sys import os.path sys.path.append( os.path.abspath( os.path.dirname( os.path.realpath(__file__)))) from models import * def make_str_for_date(d): if d: return d.strftime('%m/%d/%Y') else: return '' class BookExporter(bulkloader.Exporter): def __init__(self): bulkloader.Exporter.__init__(self, 'Book', [('title', str, None), ('author', str, ''), ('copyright_year', int, ''), ('author_birthdate', make_str_for_date, ''), ]) exporters = [BookExporter] The exporter definition file must do several things: 1. It must import or define a data model class for each kind to export, using the Python data modeling API. The exporter file does not have the app directory in the module load path, so the exporter file must add it. The modules should be imported directly into the local namespace. 2. It must define (or import) an exporter class for each kind to export, a subclass of the Exporter class in the package google.appengine.tools.bulkloader. This class must have a constructor (an __init__() method) that calls the parent class’s con- structor with arguments that describe the CSV file’s columns. 3. It must define a module-global variable named exporters, whose value is a list of exporter classes. Using the Bulk Loader Tool | 287
- 309. You can safelyput your exporter classes in the same file as the loader classes, if you like. You can even put them in the same file as your models if that makes sense for your application. You provide the name of the file that defines the exporters using the --filename=... command-line argument. The loader class’s __init__() method calls its parent class’s __init__() method with three arguments: self (the object being initialized), the kind of the entity to create as a string, and a list of tuples. Each tuple represents a column from the CSV file, in order from left to right. The first element of each tuple is the name of a property on each entity. The second element is a Python function that takes the datastore value (as a Python representation of a datastore value type) and returns the string value to add to the CSV file. For the exporter, each tuple also requires a third element that specifies the default value to use if a given entity does not have this property (a string). If the default value is None, then the bulk loader will consider it an error if it finds an entity without this property, and it will abort. The desired behavior of the type conversion function depends on your spreadsheet application. In most cases, the result of passing the datastore value to the str con- structor is sufficient. For Boolean values, you may want a converter function that emits 'TRUE' instead of 'True': def make_str_for_bool(b): return 'TRUE' if b else 'FALSE' For datetime value types (such as datetime.date), you can call the strftime() method of the value in a converter function to produce a string form of the date in the desired format, as in the previous example. If you pass a Key value (such as a reference property) to str, you get an encoded key string. This value contains all of the information about the key, and you can reconstruct the Key object by passing this string to the Key constructor. The default exporter behavior does not save the key of each entity to the CSV file. You can change the behavior of the exporter by overriding the output_entities() method. This method takes a generator that yields model instances in key order (entity_generator), and does all the work to produce a file. The output filename is accessible using the output_filename member of the exporter (self.output_filename). You can also override the initialize() method to do work before the download begins. This method takes the name given by --filename=... and the string given by --exporter_opts=..., as the filename and exporter_opts arguments, respectively. There’s also a finalize() method you can override, which is called after the download is complete (with no arguments). To download data with the bulk loader and exporter classes, run the bulkloader.py command as follows (notice the --download option that wasn’t there in the upload command): 288 | Chapter 12: Bulk Data Operations and Remote Access
- 310. bulkloader.py --download --app_id=app-id --url=http://app-id.appspot.com/remote_api --config_file=exporters.py --kind=Book --filename=bookdata.csv For a Python application, you can also start a download using appcfg.py download_data, like so: appcfg.py download_data --config_file=exporters.py --kind=Book --filename=bookdata.csv app-dir Controlling the Bulk Loader You can control the rate that the bulk loader transfers data using command-line arguments: --num_threads=... The maximum number of loader threads to spawn. The default is 10. --batch_size=... The maximum number of entities to create or download with each remote API call. The loader combines puts and gets into batches to reduce the number of remote API calls needed. If you have large entities, set this number low so that the total size of a batch call does not exceed one megabyte. The default is 10. --bandwidth_limit=... The maximum number of bytes per second to send or receive on average. The default is 250 kilobytes per second. --rps_limit=... The maximum number of records per second to send or receive on average. The default is 20 records per second. --http_limit=... The maximum number of HTTP requests to make per second. The default is 15 requests per 2 seconds (7.5 per second). If you restrict your app’s Google Accounts authentication to a Google Apps domain, the bulk loader needs to know the domain for authentication purposes. You provide this with the --auth_domain=... argument. The bulk loader uses a temporary database file to keep track of where in a loading job it is in case it is interrupted. Rerunning the interrupted tool with the same arguments restarts the job from where it left off. By default, the name of this file is bulkloader- progress-..., where ... is the timestamp of the current run. You can specify a custom name and location for this file with the --db_filename=... argument. If you specify a Using the Bulk Loader Tool | 289
- 311. value of skipfor this argument, the bulk loader will not use a file to record progress. (If you do not have sqlite3 installed, you must specify --db_filename=skip.) The loader also writes a log of its actions to a file named bulkloader-log-... (also with a timestamp in the name). You can change this name with --log_filename=.... For downloads, the loader stashes downloaded entities in a temporary datafile named bulkloader-results-.... You can change the name of this file with --result_db_filename=.... Using the Remote Shell Tool With the remote API handler installed, you can use a tool included with the Python SDK to manipulate a live application’s services from an interactive Python shell. You interact with the shell using Python statements and the Python service APIs. This tool works with both Java and Python applications using the remote API handler. To start a shell session, run the remote_api_shell.py command. As with the other Py- thon SDK commands, this command may already be in your command path. remote_api_shell.py app-id The tool prompts for your developer account email address and password. (Only reg- istered developers for the app can run this tool, or any of the remote API tools.) By default, the tool connects to the application using the domain name app-id.appspot.com, and assumes the remote API handler is installed with the URL path /remote_api. To use a different URL path, provide the path as an argument after the application ID: remote_api_shell.py app-id /admin/util/remote_api To use a different domain name, such as to use a specific application version, or to test the tool with the development server, give the domain name with the -s ... argument: remote_api_shell.py -s dev.latest.app-id.appspot.com app-id The shell can use any service API that is supported by the remote API handler. This includes URL Fetch, memcache, Images, Mail, Google Accounts, and of course the datastore. (As of this writing, XMPP is not supported by the remote API handler.) Several of the API modules are imported by default for easy access. The tool does not add the current working directory to the module load path by default, nor does it know about your application directory. You may need to adjust the load path (sys.path) to import your app’s classes, such as your data models. Here is an example of a short shell session: % remote_api_shell.py clock Email: juliet@example.com Password: App Engine remote_api shell 290 | Chapter 12: Bulk Data Operations and Remote Access
- 312. Python 2.5.1 (r251:54863,Feb 6 2009, 19:02:12) [GCC 4.0.1 (Apple Inc. build 5465)] The db, users, urlfetch, and memcache modules are imported. clock> import os.path clock> import sys clock> sys.path.append(os.path.realpath('.')) clock> import models clock> books = models.Book.all().fetch(6) clock> books [<models.Book object at 0x7a2c30>, <models.Book object at 0x7a2bf0>, <models.Book object at 0x7a2cd0>, <models.Book object at 0x7a2cb0>, <models.Book object at 0x7a2d30>, <models.Book object at 0x7a2c90>] clock> books[0].title u'The Grapes of Wrath' clock> from google.appengine.api import mail clock> mail.send_mail('juliet@example.com', 'test@example.com', 'Test email', 'This is a test message.') clock> To exit the shell, press Ctrl-D. Using the Remote API from a Script You can call the remote API directly from your own Python scripts using a library from the Python SDK. This configures the Python API to use the remote API handler for your application for all service calls, so you can use the service APIs as you would from a request handler directly in your scripts. Here’s a simple example script that prompts for a developer account email address and password, then accesses the datastore of a live application: #!/usr/bin/python import getpass import sys # Add the Python SDK to the package path. # Adjust these paths accordingly. sys.path.append('~/google_appengine') sys.path.append('~/google_appengine/lib/yaml/lib') from google.appengine.ext.remote_api import remote_api_stub from google.appengine.ext import db import models # Your app ID and remote API URL path go here. APP-ID = 'app-id' REMOTE_API_PATH = '/remote_api' def auth_func(): email_address = raw_input('Email address: ') password = getpass.getpass('Password: ') Using the Remote API from a Script | 291
- 313. return email_address, password def initialize_remote_api(app-id=APP-ID, path=REMOTE_API_PATH): remote_api_stub.ConfigureRemoteApi( app-id, path, auth_func) remote_api_stub.MaybeInvokeAuthentication() def main(args): initialize_remote_api() books = models.Book.all().fetch(10) for book in books: print book.title return 0 if __name__ == '__main__': sys.exit(main(sys.argv[1:])) The ConfigureRemoteApi() function (yes, it has a TitleCase name) sets up the remote API access. It takes as arguments the application ID, the remote API handler URL path, and a callable that returns a tuple containing the email address and password to use when connecting. In this example, we define a function that prompts for the email address and password, and pass the function to ConfigureRemoteApi(). The function also accepts an optional fourth argument specifying an alternate domain name for the connection. By default, it uses app-id.appspot.com, where app-id is the application ID in the first argument. The MaybeInvokeAuthentication() function sends an empty request to verify that the email address and password are correct, and raises an exception if they are not. (With- out this, the script would wait until the first remote call to verify the authentication.) Remember that every call to an App Engine library that performs a service call does so over the network via an HTTP request to the application. This is inevitably slower than running within the live application. It also consumes application resources like web requests do, including bandwidth and request counts, which are not normally con- sumed by service calls in the live app. On the plus side, since your code runs on your local computer, it is not constrained by the App Engine runtime sandbox or the 30-second request deadline. You can run long jobs and interactive applications on your computer without restriction, using any Py- thon modules you like—at the expense of consuming app resources to marshal service calls over HTTP. 292 | Chapter 12: Bulk Data Operations and Remote Access
- 314. CHAPTER 13 Task Queues and Scheduled Tasks The App Engine architecture is optimized for handling web requests, small amounts of work that run in a stateless environment with the intent of returning a response to the user as fast as possible. But many web applications have other kinds of work that need to get done, work that doesn’t fit in the fast response model. Instead of doing the work while the user waits, it’s often acceptable to record what work needs to get done, respond to the user right away, then do the work later, within seconds or minutes. The ability to make this trade-off is especially useful with scalable web applications that use a read-optimized datastore, since updating an element of data may require several re- lated but time-consuming updates, and it may not be essential for those updates to happen right away. Then there’s the work that needs to get done independently of user actions: fetching and caching fresh data from remote sources, sending daily status email messages, re- spawning health packs and enemies throughout the island. For this, it would be useful to configure a schedule of work that needs to be done on a regular basis, and that gets performed regardless of web traffic from users. App Engine provides facilities for these categories of work: task queues and scheduled tasks. With task queues, a request handler invoked by a user request can enqueue a unit of work—a task—and return a response to the user and exit. App Engine executes enqueued tasks at a configurable rate. If a task fails, App Engine retries the task, en- suring that tasks that fail due to transient conditions (such as datastore contention) eventually succeed. An app can also provide a schedule of tasks to be performed on a regular basis at particular times of an hour, day, or week, and App Engine executes the tasks according to the schedule. Scheduled tasks are not retried, but a scheduled task can use a task queue to take advantage of the retry behavior. So, what is a task? A task is a request to a request handler. The task specifies a URL path of the application to be called, and can also include data that App Engine passes to the request handler as HTTP parameters. A URL of a networked application called by a system to perform a task is known as a web hook. 293
- 315. App Engine invokesa request handler for a task in the same environment as it does a handler for a user request. This means the task handler runs with the same sandbox and resource restrictions, including the 30-second deadline. A task can enqueue an- other task, allowing a large unit of work to be initiated by a single task and performed in batches. By treating task handlers as request handlers, App Engine can parallelize and scale tasks using the same infrastructure it uses to scale user requests. Fig- ure 13-1 illustrates how task queues and scheduled tasks take advantage of the request handler infrastructure. Figure 13-1. Architecture of task queues and scheduled tasks (“cron”), calling web hooks In this chapter, we’ll describe the concepts of task queues and scheduled tasks, and how to use them in Python and Java applications. Google initially released the task queue feature as an “experimental” feature, reserving the right to change the design, API, and resource limits before declaring the feature stable. As of this writing, the feature is still under this “experimental” status. Please refer to the official App Engine documentation for the latest information on the task queue API. Task Queues An app tells App Engine to perform a task by adding the task to a task queue; it enqueues the task. The task queue keeps track of the tasks to perform, and processes them (by issuing HTTP requests) at a particular rate. You can configure the rate for a queue to control how quickly tasks can consume resources. An app can have up to 10 different queues configured at different rates. Once a task has been enqueued success- fully, the task queue ensures that the task will be performed at least once, and at least 294 | Chapter 13: Task Queues and Scheduled Tasks
- 316. until the taskreports success. Queues ensure the eventual success of tasks by retrying them until they succeed. Enqueueing a task is fast, about three times faster than writing to the datastore. This makes tasks useful for pretty much anything whose success or failure doesn’t need to be reported to the user in the response to the HTTP request that initiates the work. For example, an app can write a value to the memcache, then enqueue a task to persist that value to the datastore. This saves time during the user request, and allows the task to do bookkeeping or make other time-consuming updates based on the change (assuming it meets the application’s needs that the bookkeeping happens later than the initial update). Because tasks are just calls to application URLs, tasks are executed in parallel, just like user requests. This makes tasks useful for performing lots of work quickly. Consider a user request that needs to send 100 email messages. Without tasks, the only option would be to make 100 calls to the Mail service, one at a time, while the user waits for a response. With tasks, the user request can enqueue tasks and respond to the user right away. And if each email message is sent in a separate task, App Engine can fire up multiple application instances to send multiple messages simultaneously. With a bit of effort in the browser to check on task progress, task queues can provide a seamless and fast real-time user experience for performing large amounts of work. Task queues are also useful for batch processing, such as converting datastore entities from one data schema to another without the use of an offline driving mechanism. A task can process a chunk of the batch (such as a range of entity IDs), then enqueue another task to process the next chunk. If processing a chunk fails, the queue can retry the chunk without restarting the entire batch job. You can combine task chaining with parallelization for fine-grained control over how quickly batch jobs execute, and how quickly they consume resources. Processing Rates and Token Buckets The processing rate for a queue is controlled using a “token bucket” algorithm. In this algorithm, a queue has a number of “tokens,” and it spends a token for each task it executes. Tokens are replenished at a steady rate up to a maximum number of tokens (the “bucket size”). Both the replenishment rate and the bucket size are configurable for a queue. If a queue contains a task and has a token, it usually executes the task immediately. If a queue has many tasks and many available tokens, it executes as many tasks as it can afford, immediately and in parallel. If there are tasks remaining, the queue must wait until a token is replenished before executing the next task. The token bucket algorithm gives a queue the flexibility to handle bursts of new tasks, while still remaining within acceptable limits. The larger the bucket, the more tasks an idle queue will execute immediately when the tasks are enqueued all at once. Task Queues | 295
- 317. I say itusually executes the tasks immediately because App Engine may adjust the method and rate of how it executes tasks based on the performance of the system. In general, task queue schedules are approximate, and may vary as App Engine balances resources. A queue does not wait for one task to finish before executing the next task. Instead, it initiates the next task as soon as a token is available, in parallel with any currently running tasks. Tasks are not strictly ordered, but App Engine makes an effort to perform tasks in the order they are enqueued. Tasks must not rely on being executed serially or in a specific order. Each task queue has a name and processing rate (token replenishment rate and bucket size). Every app has a queue named “default” that processes 5 tasks per second, with a bucket size of 5. If you don’t specify a queue name when enqueueing a task, the task is added to the default queue. You can adjust the rate and bucket size of the default queue, and can set the rate to 0 to turn it off. Tasks enqueued to a paused queue remain on the queue until you upload the new configuration with a positive rate. Task queues and token buckets help you control how tasks are executed so you can plan for maximizing throughput, making the most efficient use of system resources to execute tasks in parallel. Tasks inevitably share resources, even if the resource is just the pool of warmed-up application servers. Executing a bunch of tasks simultaneously may not be the fastest way to complete all of the tasks, since App Engine may need to start up new instances of the application to handle the sudden load. If multiple tasks operate on the same entity groups in the datastore, it may be faster to perform only a few tasks at a time and let datastore retries sort out contention, instead of relying on task retries to drive in all of the changes. Limiting the execution rate with token buckets can actually result in faster completion of multiple tasks. Queue processing rates are configured using a configuration file. This file can be up- loaded separately from the rest of the application. If a given queue is processing tasks too quickly, you can upload a new temporary configuration for the queue that tells it to run at a slower rate, and the change will take effect immediately. You can experiment with different rates and token bucket sizes to improve task throughput. The maximum rate of queue processing for an application is 10 tasks per second. That’s the total rate across all queues. If your queue configuration adds up to a collective rate of more than 10 tasks per second, you will get an error when you try to upload the app or the configuration file. Elements of a Task You define a task by specifying the aspects of the HTTP request to be made when the task is executed. This can include a URL path, a method (such as POST), and a message body or form parameters. 296 | Chapter 13: Task Queues and Scheduled Tasks
- 318. If the taskdoes not specify a URL path, it assumes a URL path of the following form, based on the queue name: /_ah/queue/queue-name This is convenient for using the same handler for every task in a queue, or otherwise selecting the handler based on the queue name. The default URL for tasks on the default queue is /_ah/queue/default. A task can include a delay, either a number of seconds from when the task is enqueued or a specific date and time in the future. The queue will not execute the task until after the delay has elapsed. The queue may not execute the task right away after the delay depending on token availability, but the queue will execute the task at the next oppor- tunity after its delay has elapsed. Every task has a name that identifies it uniquely. You can supply the name, or you can let the system assign a name. Task names allow you to ensure that a given unit of work is only enqueued once in situations where the app may call the task queue API more than once for the same work. If a task with a given name is already in a queue, or has been enqueued in the past seven days, an attempt to enqueue another task with the same name will fail. If you do not specify a name for a task, the system assigns a name guaranteed to be unique across all pending and completed tasks currently in the system. Completed tasks that remain in the system to prevent duplicates are called tombstones, or tombstoned tasks. A common case where custom names and tombstones are useful is when handling a web form. Suppose clicking a button or submitting a form causes the application to perform some expensive work that ought to only be performed once. If the user acci- dentally clicks the button twice, the browser may send two web requests for the same action. You can prevent the work from being performed twice by rendering the form with a unique ID (such as the button’s “value”) and performing the work in a task with that name. Two requests to perform the work with the same name will result in only one task being created and performed. Currently, tombstones remain in the system for about seven days. This may be con- figurable in a future release. Task Handlers and Retries When a queue executes a task, it issues an HTTP request to the app via the frontend, which routes it to a request handler according to the app’s URL configuration. The frontend recognizes a request coming from a queue as if it were coming from a developer account, so you can restrict access to task URLs in the handler configuration. The request handler for the task URL should perform the work, then return an HTTP response code of 200, indicating the work was completed successfully. If the handler returns a different response code, App Engine assumes the task failed completely, and puts it back on the queue to retry it again later. A failed task is reenqueued with a delay Task Queues | 297
- 319. that gets exponentiallylonger the more times it has been retried so that problems caused by resource consumption or contention can be relieved automatically, up to a maxi- mum delay of one hour. Retries guarantee the eventual success of the task, assuming it is possible for the task to succeed. If a task enqueues successfully, it stays on the queue until it is successful. Tasks are retried indefinitely, and never expire. (Currently, there is no way to flush queues, or identify tasks that fail repeatedly, but these are being considered as future features.) Under very rare circumstances, such as after a system failure, a task may be retried even if it completed successfully. This is a design trade-off that favors fast task creation over built-in once-only fault tolerance. A task that can be executed multiple times without changing the end result is called idempotent. Whether a task’s code must be strictly idempotent depends on what the task is doing and how important it is that the calcu- lation it is performing be accurate. For instance, a task that deletes a datastore entity can be retried because the second delete fails harmlessly. Since a task is always run with its task name (assigned when the task is created, either by the application or by the system), a task can guard against reruns by storing and recognizing its task name. App Engine provides the name of the task in an HTTP header with the request, along with three other useful headers: X-AppEngine-TaskName The name of the task, or the system-generated unique ID. X-AppEngine-QueueName The name of the queue executing the task. X-AppEngine-TaskRetryCount The number of times the task has been retried (0 for the first attempt). The body of a response from a task’s request handler is ignored. If the task needs to store or communicate information, it must do so by using the appropriate services or by logging messages. Task queues have special behavior with regard to app versions. If the version of the app that enqueued a task was the default version, then the task uses the default version of the app when it executes—even if the default version has changed since the task was enqueued. If the version of the app that enqueued the task was not the default version at the time, then the task uses that version specifically when it executes. This allows you to test nondefault versions that use tasks before making them the default. A call to a task handler appears in the request log, just like a user-initiated web request. You can monitor and analyze the performance of tasks just as you would user requests. 298 | Chapter 13: Task Queues and Scheduled Tasks
- 320. Testing and ManagingTasks The development web server does not process task queues. Instead, it remembers tasks that are added to queues, and lets you execute them manually from the developer console. Recall that the development server console URL looks like this: http://localhost:8080/_ah/admin/ Select Task Queues from the sidebar. Each named queue is listed with its rate and bucket size, the age of its oldest task, and the number of tasks in the queue. You can delete all tasks from the queue by clicking the Flush Queue button. To execute tasks on a queue, click the queue’s name. The Console displays a list of the tasks on the queue. For any task, click the Run button to run the task, or the Delete button to delete it without running it. If the task runs successfully, it disappears from the list. Be sure to have a window open showing the development server’s log output when testing tasks. When you run a task, it appears as a logged request, along with its response code. If the response code is not 200, the task is considered to have failed, and it remains in the list of tasks in the Console. You can tell that something actually happened by watching the log messages. You can inspect task queues for an app running on App Engine from the Administration Console. As of this writing, the App Engine team is still building features for managing tasks, such as the ability to flush task queues manually from the Administration Con- sole, or the ability to delete tasks using the API. Currently, once a task is enqueued, it cannot be removed from the queue for the app or from the Administration Console. You can add logic to your task to watch the retry count (the X-AppEngine-TaskRetryCount header) and return success without performing work to bail if it is retried too many times. If your current code doesn’t do this bail out trick and you have tasks enqueued and failing, just upload new code using the same app version number, and the tasks will use it when they are retried. Using Task Queues in Python A Python application enqueues tasks using the API provided by the google.appengine.api.labs.taskqueue module. The simplest way to enqueue a task is to call the add() function. Without arguments, the add() function enqueues a task to the default queue using the default URL for that queue: from google.appengine.api.labs import taskqueue taskqueue.add() Task Queues | 299
- 321. When the taskqueue feature is no longer considered “experimental,” Google will change the package name of the Python API to google.appengine.api.taskqueue (moving it out of the labs package). Recall that the default queue’s name is default, and if a task does not specify its own handler URL path, it uses a URL path derived from the queue name, such as: /_ah/queue/default You map the URL to a request handler that performs the task in the app.yaml file. You can restrict the URL so that it can only be called by task queues and by the app’s developers (for testing) using login: admin, like so: handlers: - url: /_ah/queue/default script: default_task.py login: admin With this configuration, the no-argument call to the add() function enqueues the task on the default queue with the default URL and no arguments, and returns immediately. When the task queue processes this task, it issues an HTTP POST request with default arguments to the URL path /_ah/queue/default, which invokes the default_task.py re- quest handler. You can pass parameters to the task, as a set of named parameters (like a web form) or an HTTP POST message body. The add() function accepts either a Python dict of named parameters as the params argument, or a bytestring (str) message body as the payload argument. taskqueue.add(params={'address': 'juliet@example.com'}) # ... class DefaultQueueHandler(webapp.RequestHandler): def post(self): address = self.request.get('address') # ... You can specify a custom URL for a task to use instead of the default URL for the queue using the url parameter. In fact, you can choose to ignore the default URLs entirely and just use per-task URLs. taskqueue.add( url='/send_invitation_task', params={'address': 'juliet@example.com'} ) You can control other aspects of the HTTP request as well. You can change the HTTP method used with the method parameter; the default is POST. (You can only use a payload or params for HTTP methods that support them.) You can also add arbitrary HTTP headers with the headers argument, which takes a Python dict. 300 | Chapter 13: Task Queues and Scheduled Tasks
- 322. You can seta delay for a task, such that the task does not become eligible for execution until the delay has elapsed. You can specify the delay as a number of seconds from the time the task is enqueued using the countdown argument, or as a datetime.datetime in the future using the eta (as in “estimated time of arrival”) argument. To set the name for a task, you specify the name argument. As described earlier, the name ensures that only one task with that name can be enqueued for a period of time, at least seven days with the current implementation. If you do not specify a name, the system assigns a unique name that has not been used in the past seven days. The Python API also includes an object interface for enqueueing tasks. The add() func- tion is equivalent to creating a Queue object that represents the default queue, creating a Task object with the given parameters, and adding it to the queue with the add() method of the Queue. q = taskqueue.Queue() t = taskqueue.Task( url='/send_invitation_task', params={'address': 'juliet@example.com'} ) q.add(t) An app can have up to 10 named queues, each with its own task execution rate (token replenishment rate and bucket size). In a Python app, you configure queues using a configuration file named queue.yaml, which looks like this: queue: - name: fastest rate: 10/s bucket_size: 10 - name: limitedemail rate: 1000/d The file is in the YAML format, just like other Python configuration files. It contains one element named queue whose value is a list of mappings, one mapping for each queue. Each queue has a name, a token replenishment rate, and a bucket_size. The name and rate elements are required. If you omit bucket_size, it defaults to 5. The rate is specified as a number, a slash, and a unit of time: s for second, m for minute, h for hour, and d for day. Tokens are replenished continuously over the interval, so 1/s and 60/m are equivalent. The default queue uses a rate of 5/s and a bucket size of 5. You can configure the default queue using queue.yaml by defining a queue named default. queue: - name: default rate: 1/s Task Queues | 301
- 323. To enqueue tasksto named queues, you use the object interface, passing the name of the queue to the Queue constructor: q = taskqueue.Queue('limitedemail') t = taskqueue.Task( url='/send_invitation_task', params={'address': 'juliet@example.com'} ) q.add(t) Remember that each named queue has a default task URL, which you can use by omit- ting the url parameter for the task. In this example, the default URL for the limitedemail queue is: /_ah/queue/limitedemail Similar to index.yaml, the queue.yaml file applies to all versions of the app. It is uploaded when the app is uploaded with appcfg.py update. You can update just the queue con- figuration for the app using appcfg.py update_queues: appcfg.py update_queues app-dir Deferring work with tasks The Python SDK includes a handy utility that makes it easy to throw work into a task without writing a custom task handler. The utility uses a prepackaged general purpose task handler to process deferred work. To use this utility, you create a Python function or other callable object that performs the work to be executed outside of the current request handler, then pass that callable object to the defer() function from the google.appengine.ext.deferred package. The defer() function takes the object to call within the task and arguments to pass to the callable object. Here’s a simple example that spans two Python modules, one containing the deferred function, and one containing the request handler that defers it. First, here’s the func- tion, to appear in a module named invitation.py: from google.appengine.api import mail import logging _INVITATION_MESSAGE_BODY = ''' You have been invited to join our community... ''' def send_invitation(recipient): mail.send_mail('support@example.com', recipient, 'You're invited!', _INVITATION_MESSAGE_BODY) logging.info('Sent invitation to %s' % recipient) 302 | Chapter 13: Task Queues and Scheduled Tasks
- 324. And here’s therequest handler script: from google.appengine.ext import webapp from google.appengine.ext.webapp.util import run_wsgi_app from google.appengine.ext import deferred import invitation class SendInvitationHandler(webapp.RequestHandler): def get(self): # recipient = ... deferred.defer(invitation.send_invitation, recipient) # ... application = webapp.WSGIApplication([ ('/sendinvite', SendInvitationHandler), ], debug=True) def main(): run_wsgi_app(application) if __name__ == '__main__': main() You set up the request handler for deferred work using the following configuration in your app.yaml file: handlers: - url: /_ah/queue/deferred script: $PYTHON_LIB/google/appengine/ext/deferred/ login: admin This maps the URL path /_ah/queue/deferred to the deferred module in the API. ($PYTHON_LIB is replaced with the path to the Python directory containing the SDK on App Engine.) Tasks created with defer() use this URL path by default. login: admin restricts this handler to task queues and application developers. The defer() function enqueues a task on the default queue that calls the given callable object with the given arguments. (Despite its use of the URL /_ah/queue/deferred for deferred tasks, this utility uses the default queue, not a queue named deferred.) The arguments are serialized and deserialized using Python’s pickle module; all argument values must be pickle-able. Most Python callable objects can be used with defer(), including functions and classes defined at the top level of a module, methods of objects, class methods, instances of classes that implement __call__(), and built-in functions and methods. defer() does not work with lambda functions, nested functions, nested classes, or instances of nested classes. The task handler must be able to access the callable object by name, possibly via a serializable object, since it does not preserve the scope of the call to defer(). You also can’t use a function or class in the same module as the request handler class from which you call defer(). This is because pickle believes the module of the request handler class to be __main__ while it is running, and so it doesn’t save the correct Task Queues | 303
- 325. package name. Thisis why the previous example keeps the deferred function in a sep- arate module. You can control the parameters of the task, such as the delay, by passing additional arguments to defer(). These are the same arguments you would pass to Task(), but with the argument names prepended with an underscore so they are not confused with arguments for the callable. deferred.defer(invitation.send_invitation, 'juliet@example.com', _countdown=86400) To call the callable, the task handler determines the module location of the callable from the description saved by the defer() function, imports the required module, re- creates any required objects from their serialized forms, then calls the callable. If the module containing the callable imports other modules, those imports will occur during the task. If the deferred callable requires any additional setup, such as changes to the module import path, make sure this happens in the callable’s module, or within the callable itself. The task handler determines the success or failure of the task based on exceptions raised by the callable. If the callable raises a special exception called deferred.PermanentTaskFailure, the task handler logs the error, but returns a success code to the task queue so the task is not retried. If the callable raises any other exception, the exception is propagated to the Python runtime and the handler returns an error code, which causes the task queue to retry the task. If the callable does not raise an exception, the task is considered successful. The deferred library is careful to raise deferred.PermanentTaskFailure for errors it knows will prevent the task from ever succeeding. Such errors log messages, then return success to flush the task from the queue. Using Task Queues in Java The Java task queue API consists of classes in the com.google.appengine.api.labs.taskqueue package. To enqueue a task, you use static methods of the QueueFactory class to get a Queue object that represents the queue, then you call the add() method on the Queue. This example enqueues a task on the default queue, using the default URL (/_ah/queue) and no data payload: import com.google.appengine.api.labs.taskqueue.Queue; import com.google.appengine.api.labs.taskqueue.QueueFactory; // ... Queue queue = QueueFactory.getDefaultQueue(); queue.add(); 304 | Chapter 13: Task Queues and Scheduled Tasks
- 326. When the taskqueue feature is no longer considered “experimental,” Google will change the package name of the Java API to com.google.appengine.api.taskqueue (moving it out of the labs package). As usual, you map the URL to a request handler that performs the task in the deploy- ment descriptor (web.xml). You can restrict the URL so that it can only be called by task queues and by the app’s developers (for testing), like so: <servlet> <servlet-name>defaulttaskurl</servlet-name> <servlet-class>myapp.DefaultQueueServlet</servlet-class> </servlet> <servlet-mapping> <servlet-name>defaulttaskurl</servlet-name> <url-pattern>/_ah/queue</url-pattern> </servlet-mapping> <security-constraint> <web-resource-collection> <web-resource-name>defaulttaskurl</web-resource-name> <url-pattern>/_ah/queue</url-pattern> </web-resource-collection> <auth-constraint> <role-name>admin</role-name> </auth-constraint> </security-constraint> This example restricts the default task URL (/_ah/queue) to be accessible only by ad- ministrators and task queues. With this configuration, the no-argument call to the add() method enqueues the task on the default queue with the default URL and no arguments, and returns immediately. When the task queue processes this task, it issues an HTTP POST request with default arguments to the URL path /_ah/queue, which invokes DefaultQueueServlet. You can pass parameters to the task as a set of named parameters (like a web form) or an HTTP POST message body. The add() method accepts a TaskOptions object that describes data and options for the task. You create a TaskOptions object using the TaskOptions.Builder static methods. Each builder method returns a TaskOptions ob- ject, and the object itself has similar methods for chaining purposes. import com.google.appengine.api.labs.taskqueue.TaskOptions; // ... TaskOptions taskOptions = TaskOptions.Builder.url("/send_invitation_task") .param("address", "juliet@example.com") .param("firstname", "Juliet"); Queue queue = QueueFactory.getDefaultQueue(); queue.add(taskOptions); Task Queues | 305
- 327. The url() methodsets a custom URL for the task. If no custom URL is set, the task calls the default URL for the queue. The param() method adds a request parameter to be included with the task’s HTTP request. The method takes the name of the parameter, and either a String or a byte[] as the value of the parameter. Parameters are added to the HTTP request body URL-encoded, similar to a web form. The removeParam() method removes a parameter from the TaskOptions object; it takes the parameter name as its argument. You can also set the HTTP request body with the payload() method. This method takes the request body as a String. You can also include a character set specification (such as "utf-8") as the second argument. You can’t set both a payload and parameters for the same task. You can control other aspects of the HTTP request as well. The method() method sets the HTTP method; the default is "POST". You can also add arbitrary HTTP headers by calling the header() method (which takes a String key and String value as arguments) for each header, or by calling the headers() method with a Map<String, String>. You can set a delay for the task. The countdownMillis() method sets the delay as a number of milliseconds (a long) from the time the task is enqueued. The etaMillis() method sets the delay to be an absolute date and time, as a number of milliseconds from midnight, January 1, 1970, UTC. To set the name for a task, you call the name() method. As described earlier, the name ensures that only one task with that name can be enqueued for a period of time, at least seven days with the current implementation. If you do not specify a name, the system assigns a unique name that has not been used in the past seven days. An app can have up to 10 named queues, each with its own task execution rate (token replenishment rate and bucket size). In a Java app, you configure queues using a con- figuration file named queue.xml, in the WEB-INF/ directory, which looks like this: <queue-entries> <queue> <name>fastest</name> <rate>10/s</rate> <bucket-size>10</bucket-size> </queue> <queue> <name>limitedemail</name> <rate>1000/d</rate> <bucket-size>5</bucket-size> </queue> </queue-entries> The file includes one <queue> element per queue to be configured. Each queue has a <name>, a token replenishment <rate>, and a <bucket-size>. 306 | Chapter 13: Task Queues and Scheduled Tasks
- 328. The <rate> isspecified as a number, a slash, and a unit of time: s for second, m for minute, h for hour, and d for day. Tokens are replenished continuously over the interval, so 1/s and 60/m are equivalent. The default queue uses a rate of 5/s and a bucket size of 5. You can configure the default queue using queue.xml by defining a queue named default. <queue-entries> <queue> <name>default</name> <rate>1/s</rate> <bucket-size>1</bucket-size> </queue> </queue-entries> To enqueue tasks to named queues, you pass the queue name to the QueueFactory method getQueue(). Tasks enqueued with the Queue object it returns will use the named queue. Queue queue = QueueFactory.getQueue("limitedemail"); queue.add(taskOptions); Remember that each named queue has a default task URL, which you can use by not setting a URL for the task. In this example, the default URL for the limitedemail queue is: /_ah/queue/limitedemail Similar to datastore-indexes.xml, the queue.xml file applies to all versions of the app. It is uploaded when the app is uploaded with appcfg update (appcfg.sh update on Mac OS X or Linux). You can update just the queue configuration for the app using appcfg update_queues: appcfg update_queues war Transactional Task Enqueueing There’s an exciting feature of tasks that had not yet been released by the time this book went to press, but is likely to be available by the time you read this: the ability to enqueue a task as part of a datastore transaction. A task enqueued within a transaction is only enqueued if the transaction succeeds. If the transaction fails, the task is not enqueued. This opens up a world of possibilities for the datastore. Specifically, it enables easy transactions that operate across multiple entity groups. Consider the message board example from Chapter 6. To maintain an accurate count of every message in each conversation, we have to update the count each time a message is posted. To do this with strong consistency, the count and the message have to be updated in the same transaction, which means they have to be in the same entity group—and therefore every message in the thread has to be in the same entity group. This might be acceptable for a count of messages per conversation, since it’s unlikely Task Queues | 307
- 329. that many userswill be posting to the same conversation simultaneously, and even so, the delay for resolving concurrency failures might not be noticed. But what if we want a count of every message on the website? Putting every message in a single entity group would be impractical, as it would effectively serialize all updates to the entire site. We need a way to update the count reliably without keeping every- thing in one entity group. Transactional task enqueueing lets us update the count reliably without concern for entity groups. To post a message, we use a transaction to create the message entity and enqueue a task that will update the count. If the transaction fails, the task is not en- queued, so the count remains accurate. The task is performed outside of the transac- tion, so the count does not need to be in the same entity group as the message, but transactional enqueueing and task retries ensure that the count is updated, but only under the proper circumstances. Of course, this comes with a trade-off: it must be acceptable for the count to be inac- curate between the time the message entity is created and the time the count is updated. In other words, we must trade strong consistency for eventual consistency. Transac- tional task enqueueing gives us a simple way to implement eventually consistent global transactions. You might think that eventual consistency is suitable for the global message count, because who cares if the message count is accurate? But eventual consistency is useful for important data as well. Say Alicandria wants to give TheBarbarousBobmcgrath 1,000 gold pieces. Since any player can trade gold with any other player, it is impractical to put all players in the same entity group. So we use transactional task enqueueing: the app deducts 1,000 gold pieces from Alicandria’s inventory, then enqueues a task to give 1,000 gold pieces to TheBarbarousBobmcgrath, all in a transaction. We use task names and memcache locks to ensure the system doesn’t accidentally create new gold pieces if it retries the task. Also, since Bob might get angry if he doesn’t get his money quickly, we configure the gold transfer queue to execute at a fast rate and with a large token bucket. We’re not able to include a tested code sample here because this feature wasn’t available when this book was written. Check the App Engine website for information on the release of this feature and how to use it. Scheduled Tasks Applications do work in response to external stimuli: user requests, incoming email and XMPP messages, HTTP requests sent by a script on your computer. And while task queues can be used to trigger events across a period of time, a task must be enqueued by application code before anything happens. 308 | Chapter 13: Task Queues and Scheduled Tasks
- 330. Sometimes you wantan application to do something “on its own.” For instance, an app may need to send nightly email reports of the day’s activity, or fetch new headlines from a news service. For this purpose, App Engine lets you specify a schedule of tasks to perform on a regular basis. In the App Engine API, scheduled tasks are also known as “cron jobs,” named after a similar feature in the Unix operating system. A scheduled task consists of a URL path to call and a description of the recurring times of the day, week, or month at which to call it. It can also include a description, which is displayed in the Administration Console and other reports of the schedule. To execute a scheduled task, App Engine calls the URL path using an empty GET request. A scheduled task cannot be configured with parameters, headers, or a different HTTP method. If you need something more complicated, you can do it in the code for the request handler mapped to the scheduled task’s URL path. As with task queue handlers, you can secure the URL path by restricting it to application developers in the frontend configuration. The system can call such URL paths to exe- cute scheduled tasks. The HTTP request includes the header X-AppEngine-Cron: true to differentiate it from other App Engine–initiated requests. Of course, this isn’t enough to ensure that the request isn’t from a malicious user. To block external users, you must secure the URL in the frontend configuration. In Python, the schedule is a configuration file named cron.yaml. It contains a value named cron, which is a list of task schedules. Each task schedule has a description, a url, and a schedule. You can also specify a timezone for the schedule. cron: - description: Send nightly reports. url: /cron/reports schedule: every day 23:59 timezone: America/Los_Angeles - description: Refresh news. url: /cron/getnews schedule: every 1 hours In Java, the corresponding file in the WEB-INF/ directoryis named cron.xml. A <cronentries> element contains zero or more <cron> elements, one for each schedule. The <url>, <description>, <schedule>, and <timezone> elements define the scheduled task. <cronentries> <cron> <url>/cron/reports</url> <description>Send nightly reports.</description> <schedule>every day 23:59</schedule> <timezone>America/Los_Angeles</timezone> </cron> <cron> <url>/cron/getnews</url> <description>Refresh news.</description> Scheduled Tasks | 309
- 331. <schedule>every 1 hours</schedule> </cron> </cronentries> The value for the schedule element (in either Python or Java) uses a simplified English- like format for describing the recurrence of the task. It accepts simple recurrences, such as every 30 minutes or every 3 hours, with a minimum interval of every 1 minutes. The parser’s English isn’t that good: it doesn’t understand every 1 minute or every minute. It does understand every day, as an exception. The interval every day also accepts an optional time of day, as a 24-hour hh:mm time; every day 23:59 runs every day at 11:59 p.m. You can have a task recur weekly using the name of a weekday, as in every tuesday, and can also include a time: every tuesday 23:59. In another English parsing foible, day names must use all lowercase letters. You can abbreviate day names using just the first three letters, such as every tue 23:59. You can have a task recur monthly or on several days of a given month by specifying a comma-delimited list of ordinals (such as 2nd, or first,third) and a comma-delimited list of weekday names (monday,wednesday,friday or sat,sun). For instance, 2nd,4th sunday occurs on the second and fourth Sunday of each month. You can also include a time of day, as earlier. Finally, you can have a task recur yearly by including the word “of” and a comma- delimited list of lowercase month names (january,july, or oct,nov,dec). Here’s a schedule that uses every feature to execute at 6 p.m. on six specific days of the year: 3rd,4th tue,wed,thu of march 18:00 You can specify a timezone element to set the time zone for the time of day in a task schedule. This element is optional: if omitted, it defaults to Coordinated Universal Time (UTC). The value of timezone is a string from the “zoneinfo” standard list of time zone identifiers, such as America/Los_Angeles (also known as Pacific Time in the United States). See the App Engine documentation for a link to a list of time zone identifiers. If you choose a time zone identifier where Daylight Saving Time (DST) is used and have a task scheduled during the DST hour, your task will be skipped when DST advances forward an hour, and run twice when DST retreats back an hour. Unless this is desired, pick a time zone that does not use DST, or do not schedule tasks during the DST hour. Scheduled tasks are always invoked using the default version of the app. There is no way to schedule a task to execute using a specific nondefault version of the app. The development console does not execute scheduled tasks automatically. If you need to test a scheduled task, you can visit the task URL path while signed in as an admin- istrator. The Python version of the development console includes a Cron Jobs section that lists the URL paths in the configuration file for easy access. 310 | Chapter 13: Task Queues and Scheduled Tasks
- 332. As with otherservice configuration files, the scheduled task configuration file applies to the entire app, and is uploaded along with the application. You can also upload it separately. With the Python SDK: appcfg.py update_cron app-dir With the Java SDK (appcfg.sh update_cron on Mac OS X or Linux): appcfg update_cron war You can validate your task schedule and get a human-readable description of it using appcfg.py cron_info app-dir (Python) or appcfg cron_info war (Java). The report in- cludes the exact days and times of the next few runs, so you can make sure that the schedule is what you want. Scheduled Tasks | 311
- 334. CHAPTER 14 The Django Web Application Framework This chapter discusses how to use a major web application framework with the Python runtime environment. Java developers may be interested in the general discussion of frameworks that follows, but the rest of this chapter is specific to Python. Several frameworks for Java are known to work well with App Engine, and you can find in- formation on these by searching the Web and asking the App Engine Java forum. As with all major categories of software, web applications have a common set of prob- lems that need to be solved in code. Most web apps need software to interface with the server’s networking layer, communicate using the HTTP protocol, define the resources and actions of the application, describe and implement the persistent data objects, enforce site-wide policies such as access control, and describe the browser interface in a way that makes it easily built and modified by designers. Many of these components involve complex and detailed best practices for interoperating with remote clients and protecting against a variety of security vulnerabilities. A web application framework is a collection of solutions and best practices that you assemble and extend to make an app. A framework provides the structure for an app, and most frameworks can be run without changes to demonstrate that the initial skel- eton is functional. You use the toolkit provided by the framework to build the data model, business logic, and user interface for your app, and the framework takes care of the details. Frameworks are so useful that selecting one is often the first step when starting a new web app project. Notice that App Engine isn’t a web application framework, exactly. App Engine pro- vides scaling infrastructure, services, and interfaces that solve many common problems, but these operate at a level of abstraction just below most web app frameworks. A better example of a framework is webapp, a framework for Python included with the App Engine Python SDK. webapp lets you implement request handlers as Python classes, and takes care of the details of interfacing with the Python runtime environment and routing requests to handler classes. 313
- 335. Several major frameworksfor Python work well with App Engine. Django, Pylons, web2py, and CherryPy work as is or with the help of an adapter component, and some frameworks (such as web2py) have added explicit support for App Engine. These frameworks are mature, robust, and widely used, and have large thriving support com- munities and substantial online documentation. You can buy books about these frameworks. Not every feature of every framework works with App Engine. Notably, many frame- works include a mechanism for defining data models, but these are usually implemen- ted for relational databases, and don’t work with the App Engine datastore. In some cases, you can just replace the framework’s data modeling library with App Engine’s. Some features of frameworks also have issues running within App Engine’s sandbox restrictions, such as by depending upon unsupported libraries. Developers have written adapter components that work around many of these issues. In general, to use a framework, you add the framework’s libraries to your application directory, then map all dynamic URLs (all URLs except those for static files) to a script that invokes the framework. Because the interface between the runtime environment and the app is CGI, you can use a WSGI adapter in the script to call a WSGI framework, just as we did with webapp. Most frameworks have their own mechanism for associ- ating URL paths with request handlers, and it’s often easiest to send all dynamic re- quests to the framework and let it route them. You may still want to use app.yaml to institute Google Accounts–based access control for some URLs. Since many web developers will want to use a framework more substantial than webapp, App Engine’s Python runtime environment includes the Django web appli- cation framework. You don’t need to add Django’s libraries to your application direc- tory, though you may still want to use a “helper” component that simplifies setting it up. In this chapter, we’ll look at how to create an App Engine app using the Django web application framework and the Python runtime environment. We’ll discuss how to set up Django for development, how to select the Django version, and how to install and use the Django App Engine Helper, an open source tool, to create an app. For more information about Django, see the Django project website: http://www.djangoproject.com/ Installing Django When App Engine first debuted in early 2008, the latest version of Django was 0.96. The Django project has since released version 1.0 and a significantly updated version 1.1. The 1.1 release includes many new features and changes to the API that are not backward compatible with 0.96, but the new features are substantial enough that most new apps should prefer the latest version. 314 | Chapter 14: The Django Web Application Framework
- 336. Because App Enginepromises not to make backward-incompatible changes to the runtime environment, the default version of Django on App Engine (in version 1 of the Python runtime) is still 0.96. However, App Engine includes a mechanism for explicitly selecting a newer version, such that all subsequent attempts to import Django modules will use the new version. You can use this mechanism in your main request handler script to use Django 1.1 for your app. Before you do that, you must download and install Django 1.1 on your local computer. You won’t need to add it to your application directory (it’s available when running on App Engine), but you will need it locally while building and testing your application. Go to the Django project website and download Django 1.1: http://www.djangoproject.com/download/1.1/tarball/ Follow the instructions to install the Python modules in your local Python environment by running the following commands. (The sudo command will require your computer’s administrator password.) tar xzvf Django-1.1.tar.gz cd Django-1.1 sudo python setup.py install Django 1.1 was the latest release when this book was published. If you see a newer version available on the Django website, be sure to check the App Engine documentation to confirm that the new version has been added to the Python runtime environment before using it locally. If you’d prefer to use your own copy of Django instead of the one pro- vided by the Python runtime environment (such as a newer version not yet in the environment), you can add it directly to your application di- rectory. Doing so requires an additional step to remove the version of Django that’s part of the runtime environment from the Python module search path. To do this, put this in your main request handler script prior to importing django modules: for k in [k for k in sys.modules if k.startswith('django')]: del sys.modules[k] Creating a Django Project In Django, the complete set of code, configuration, and static files for a website is called a project. You create a new project by running a command that was installed with Django, called django-admin.py, like so: django-admin.py startproject myproject This command creates in your current working directory a new directory named after the project, in this case myproject/, with several starter files: Creating a Django Project | 315
- 337. __init__.py A file that tells Python that code files in this directory can be imported as modules (this directory is a Python package). manage.py A command-line utility you will use to build and manage this project, with many features. settings.py Configuration for this project, in the form of a Python source file. urls.py Mappings of URL paths to Python code, as a Python source file. If you’re following along with a Django tutorial or book, the next step is usually to start the Django development server using the manage.py command. If you did so now, you would be running the Django server, but it would know nothing of App Engine. To run the project in the App Engine development server, we must connect the Django project to the App Engine runtime environment using a request handler script and configuration in the app.yaml file. The Request Handler Script All that’s left to do is to invoke the Django project from a handler script. Exam- ple 14-1 is a minimal script that hands the request off to Django 1.1 using the WSGI runner function from the webapp library. Create a file named main.py in the Django project directory you created in the previous step (e.g., myproject/). Example 14-1. A simple request handler for invoking the Django framework from google.appengine.dist import use_library use_library('django', '1.1') import os os.environ['DJANGO_SETTINGS_MODULE'] = 'settings' import django.core.handlers.wsgi from google.appengine.ext.webapp import util def main(): # Run Django via WSGI. application = django.core.handlers.wsgi.WSGIHandler() util.run_wsgi_app(application) if __name__ == '__main__': main() The use_library() function is an App Engine utility that tells the runtime environment that version 1.1 of Django should be used when importing libraries from the django package. Without this call, importing from django would use Django 0.96. 316 | Chapter 14: The Django Web Application Framework
- 338. use_library() does nothave any effect when run in the development server, so make sure your Python installation has the correct version of Django in its module load path. The script sets the DJANGO_SETTINGS_MODULE environment variable before importing anything from Django so imported modules know where to find your application’s Django configuration. The value 'settings' says this is in a file named settings.py in the application root directory. (This file was created by Django when you started the project.) The rest of the handler script uses a main() routine (which is cached with app caching) to create the Django WSGI handler, then run it with webapp’s WSGI runner utility. You could also use a generic WSGI-to-CGI adapter, but this one includes features such as smart handling of uncaught exceptions. A typical setup routes all dynamic URLs for the site to Django via the handler script. To do this, create an app.yaml file in the project directory such as the following: application: myproject version: 1 runtime: python api_version: 1 handlers: - url: /static static_dir: static - url: /.* script: main.py Though you will use Django’s own URL routing feature for most of the application, you must still use app.yaml to configure the App Engine frontend, and tell the App Master to serve static files using the static file servers. There’s one small tweak we must make before we can try this out. Edit settings.py, and look for the line that defines ROOT_URLCONF. By default, this is set to a module path that includes the project directory as a package name, such as myproject.urls. Since the project directory is our app’s root directory, this path needs to be adjusted. Change this to read just urls: ROOT_URLCONF = 'urls' Run the App Engine development server (using the Launcher or the dev_appserver.py command), then visit the server URL in your browser. If Django says “It worked!” you’re ready to develop your Django app on App Engine. The Django App Engine Helper Django has many useful features that work well with App Engine, but some features need a little help to work as described in the Django documentation. A few features are not supported, especially those that depend on Django’s database interface, which expects to be backed by a relational database. If you develop with Django using just The Django App Engine Helper | 317
- 339. the minimal requesthandler, you’ll need to be careful not to use an unsupported fea- ture, or a feature that depends on something that is unsupported. Notably, Django’s admin application relies heavily on the database interface, and does not work with App Engine. Thankfully, there are several open source projects that restore and extend support for features of Django for App Engine. One such project is the Django App Engine Helper. The Helper makes it easy to use Django’s maintenance tools with App Engine, such as using manage.py to start the App Engine development server. It provides an adapter layer so you can use App Engine data models with some Django components expecting Django’s own data classes. The Helper also makes it easy to use Django’s unit testing framework, including fixtures that install test data in the development server’s data- store. Other features of Django are adapted to use App Engine services and infrastruc- ture, and otherwise behave as they do in other Django apps. To install the Helper, download it from the project website: http://code.google.com/p/google-app-engine-django/ Click the Downloads tab, then select the revision with the highest release number. The download is a ZIP archive. Unpack the archive, then copy all of its files to the application root directory. The Helper includes custom versions of settings.py and manage.py, its own main.py request handler and app.yaml configuration, and everything else you need to start a new project. (You can omit the documentation and auxiliary files, such as README and Makefile.) Be sure to edit app.yaml to use your application ID. The Django App Engine Helper distribution is itself a complete skeleton project. To start a new project, simply copy the Helper files to a new application directory and edit app.yaml to add your application ID. You don’t need a separate handler script; the Helper provides one for you. The Helper looks in several places for Django. First, it checks the application directory for a version of Django in either a django package directory or a django.zip archive containing the contents of the package. This lets you override the version of Django included with the App Engine runtime environment using a custom version. If it doesn’t find either of those, it calls use_library() to select an appropriate version. When run in the development server, the Helper uses whatever version is installed with Python on your computer. On App Engine, it uses the most recent version of Django supported by that specific version of the Helper. As of the “r94” release, that’s Django 1.1. You can request a specific version of Django that’s supported by App Engine by editing main.py and manage.py in your project, finding the line that calls 318 | Chapter 14: The Django Web Application Framework
- 340. InstallAppengineHelperForDjango(), and addingthe version identifier as an argument to that function call. For instance, to select Django 1.0: InstallAppengineHelperForDjango('1.0') Now that the Helper is installed, you can start the App Engine development server using manage.py as you would with any other Django project: python manage.py runserver Thanks to the Helper, this is no different than running the dev_appserver.py command or using the Launcher to start the server. If you’d prefer to use those methods, you may, without losing any functionality of the Helper. Test that the default project is working locally by visiting the server’s URL in a web browser. By default, Django uses port 8000, not 8080, so this URL is: http://localhost:8000/ The Helper adds several features to manage.py that work with the App Engine devel- opment server. One especially useful feature is the shell command, which starts a Python shell in the context of the Django project—and the development server. You can use this shell to import App Engine modules and interact with the development server, the simulated sandbox, and the simulated services. python manage.py shell Here is a simple shell session that demonstrates calling the development server’s simulated Mail service (several harmless warning messages have been elided): % python manage.py shell Python 2.6.1 (r261:67515, Jul 7 2009, 23:51:51) [GCC 4.2.1 (Apple Inc. build 5646)] on darwin Type "help", "copyright", "credits" or "license" for more information. (InteractiveConsole) >>> from google.appengine.api import mail >>> mail.send_mail('sender@example.com', 'to@example.com', 'my subject', 'message body') INFO:root:MailService.Send INFO:root: From: sender@example.com INFO:root: To: to@example.com INFO:root: Subject: my subject INFO:root: Body: INFO:root: Content-type: text/plain INFO:root: Data length: 12 INFO:root:You are not currently sending out real email. If you have sendmail installed you can use it by using the server with --enable_sendmail >>> You can open a Python shell that connects to the live application via remote_api using the console command. This requires that you set up remote_api as described in Chap- ter 12 using a URL path of /remote_api. python manage.py console The Django App Engine Helper | 319
- 341. With console, theprevious command session would actually attempt to send an email message using the Mail service. (It would fail, unless sender@example.com happens to be a developer of your application.) You can deploy and manage your App Engine application using manage.py. The update, rollback, and vacuum_indexes commands all work as manage.py targets, and they as- sume that the Django project in the current directory is the intended application. python manage.py update You can erase the development server datastore using the flush (or reset) command. Unlike dev_appserver.py, the Helper is careful to maintain a separate development datastore for each Django project. python manage.py flush For more information on features of the Helper, see the Helper’s project website and the Articles section of the App Engine documentation. Creating a Django Application In Django terms, a Django project consists of one or more “applications.” Each appli- cation has a self-contained purpose, such as a user management application or a shop- ping cart application. A complete website may use multiple applications together, each of which can be reusable for other sites. You create a Django application by running the project’s manage.py command with the startapp action: python manage.py startapp bookstore This creates a subdirectory in the project’s root directory named after the new appli- cation, in this case bookstore. The directory contains three files: an empty __init__.py, models.py, and views.py. Running manage.py uses code paths that run through the development server and that may emit warning messages about the datastore and images API, even for actions that don’t start a development server. You can safely ignore these warnings. The new app must be added to the project’s configuration. Edit settings.py in the project root directory, and find the INSTALLED_APPS variable (a Python tuple). Add the module path of the app as an element of the tuple, retaining 'appengine_django'. INSTALLED_APPS = ( 'appengine_django', 'bookstore', ) 320 | Chapter 14: The Django Web Application Framework
- 342. A simple Djangoapplication defines request handlers (“views”) and data models for performing the app’s functions. The handlers are (typically) functions in the views.py file, and data models are classes in models.py. The Django project hooks up the app’s handlers to the site’s URLs in its urls.py file, and describes the appearance of the site and its applications using templates. To see how views and URLs work, let’s define a simple view that returns an HTTP response. Edit views.py in the bookstore/ directory so it looks like this: from django.http import HttpResponse def home(request): return HttpResponse('The Book Store home page.') Now edit the urls.py file for the project (in the project root directory) so that it looks like this: from django.conf.urls.defaults import * urlpatterns = patterns('', (r'^books/', 'bookstore.views.home'), ) Start the server (python manage.py runserver) then visit the view’s URL in a browser: http://localhost:8000/books/. The message from the view is displayed. In this simple example, URLs that match the pattern '^books/' are routed to the home() function defined in the bookstore.views module. Django’s URL routing facility includes many features for defining URL patterns in a modular way, including the ability to define a set of partial URL paths within each application. URL patterns can also match parts of URL paths and pass them to view functions as arguments. A view is a function that takes a django.http.HttpRequest object as an argument (along with any other arguments passed in by the URL pattern) and returns a django.http.HttpResponse. Django has many handy ways for creating and returning response objects, including convenient ways to use its templating system. Let’s update this simple example to use a template, which we’ll expand upon later. Edit views.py to look like this: from django.shortcuts import render_to_response import datetime def home(request): return render_to_response( 'bookstore/index.html', { 'clock': datetime.datetime.now() }, ) Next, create a directory named templates/ in the project root directory. Note that this is the project directory, not the bookstore/ application directory. The project (the web- site) owns the appearance of the site. The project can declare multiple template direc- tories in settings.py; the templates/ directory is declared in the default settings. Creating a Django Application | 321
- 343. Create a subdirectoryin templates/ named bookstore/, so the full path from the project root is templates/bookstore/. Inside, create the file index.html with the page contents to display: <html> <body> <p>Welcome to The Book Store! {{ clock }}</p> </body> </html> Reload the page to see the template displayed by the new view. The render_to_response() shortcut function takes as its first argument the path to the template file, relative to one of the template directories specified in the TEMPLATE_DIRS variable of the settings.py file for the project. The second argument is a Python mapping that defines variables to be used within the template. In this example, we set a template variable named clock to be the current datetime.datetime. Within the template, {{ code }} interpolates this value as a string. There are too many cool features of Django URLs, views, and templates to cover here. For more information, see the Django website. Using App Engine Models With Django Django includes a data modeling library that provides an object interface to relational databases. Since App Engine’s datastore is not a relational database, there is no obvious way to translate between Django’s modeling interface and the datastore. The Django data interface cannot be used with App Engine, nor can any component that relies on it, including the Django data administration application. Instead, you can use App Engine’s data modeling library, which we discussed in Chap- ter 7. While it’s not a drop-in replacement for Django data modeling, it uses a similar API design. For instance, you can pass entities and query objects to Django templates for displaying data on web pages. Unlike with Django data models, you do not need to run python manage.py syncdb or related commands when changing App Engine models. The datastore is schemaless, so there are no table definitions to update. You can use the App Engine modeling interface directly, or you can use a feature pro- vided by the Django App Engine Helper to define your models. The Helper includes a BaseModel class, a subclass of App Engine’s db.Model that adds the ability to use an App Engine model class in a Django unit test fixture. (More on testing in a moment.) Let’s use BaseModel to implement some data classes for the message board. Edit models.py in the bookstore/ directory to look like the following. (The Helper’s version of python manage.py startapp has already added the import statements for the BaseModel class and db package.) 322 | Chapter 14: The Django Web Application Framework
- 344. from appengine_django.models importBaseModel from google.appengine.ext import db class Book(BaseModel): title = db.StringProperty() author = db.StringProperty() copyright_year = db.IntegerProperty() author_birthdate = db.DateProperty() class BookReview(BaseModel): book = db.ReferenceProperty(Book, collection_name='reviews') review_author = db.UserProperty() review_text = db.TextProperty() rating = db.StringProperty(choices=['Poor', 'OK', 'Good', 'Very Good', 'Great'], default='Great') create_date = db.DateTimeProperty(auto_now_add=True) Edit views.py to query the database for books and pass the query object to the template, like this: from django.shortcuts import render_to_response from google.appengine.ext import db from bookstore import models def home(request): q = models.Book.all().order('title') return render_to_response('bookstore/index.html', { 'books': q }) Finally, edit templates/bookstore/index.html to display the list of books: <html> <body> <p>Welcome to The Book Store!</p> <p>Books in our catalog:</p> <ul> {% for book in books %} <li>{{ book.title }}, by {{ book.author }} ({{ book.copyright_year }})</li> {% endfor %} </ul> </body> </html> Notice that we’re passing the db.Query object directly to the template. The template uses the query object’s iterator interface to fetch the Book entities from the datastore. If you load the page right now (and feel free to do so to test the code), you’ll see an empty list of books. Let’s create some books from the Python shell. Stop the server (hit Ctrl-C), then start the shell: python manage.py shell You can create objects in the development server’s datastore by typing Python code directly into the shell. Here is a complete shell session, including output, that creates several Book entities: Using App Engine Models With Django | 323
- 345. % python manage.pyshell Python 2.6.1 (r261:67515, Jul 7 2009, 23:51:51) [GCC 4.2.1 (Apple Inc. build 5646)] on darwin Type "help", "copyright", "credits" or "license" for more information. (InteractiveConsole) >>> from bookstore import models >>> import datetime >>> b = models.Book(title='Grapes of Wrath, The', author='Steinbeck, John', copyright_year=1939, author_birthdate=datetime.date(1902, 2, 27)) >>> b.put() datastore_types.Key.from_path(u'Book', 1, _app_id_namespace=u'google-app-engine-django') >>> b = models.Book(title='Cat in the Hat, The', author='Dr. Seuss', copyright_year=1957, author_birthdate=datetime.date(1904, 3, 2)) >>> b.put() datastore_types.Key.from_path(u'Book', 2, _app_id_namespace=u'google-app-engine-django') >>> Exit the shell (press Ctrl-D), then restart the server and load the page. Using Django Unit Tests and Fixtures The Django App Engine Helper includes support for unit tests using Django’s unit test runner and the App Engine development server. The unit test runner runs tests inside the development server’s environment, so tests can access the simulated services, in- cluding the simulated datastore. The test runner supports tests defined with the standard Python libraries unittest or doctest, or other test frameworks with test runner customizations. Django provides its own class for defining unittest test cases, called django.test.TestCase, which adds several Django-specific features. For unittest tests, the runner looks in files in the Django application directory (e.g., bookstore/) named models.py and tests.py for classes that inherit from unittest.TestCase or django.test.TestCase. Each method of a test case class whose name begins with test is a separate test. For each test, the runner initializes the test environment, instantiates the TestCase class, and calls the method for the test. If the class defines a setUp() method, it calls it before each test. If the class defines a tear Down() method, it calls it after each test. You can use setUp() and tearDown() to create and clean up test data and preparation for every test in a test case. The TestCase class provides methods that cause the test to report failure if the condi- tions tested are not met. self.assertEqual() takes two arguments, and causes the test to fail if the arguments are not equal. self.assert_() takes one argument, and fails the test if the argument is a false value (False, None, 0, the empty string, and so forth). self.assertRaises() takes an exception class, a callable object, and arguments for the callable object, calls the callable, then fails the test if the callable does not raise the given exception. There are several other test methods, as well. Each of these methods takes an optional message argument that is printed in the test report if the test fails. 324 | Chapter 14: The Django Web Application Framework
- 346. Here are afew trivial tests that illustrate the mechanism. Add this to models.py: from django import test import datetime # ... class TestBook(test.TestCase): def setUp(self): self.book = Book(title='test title', author='test author', copyright_year=1922, author_birthdate=datetime.date(1900, 1, 1)) def testCreateDate(self): self.assert_(self.book.author_birthdate and isinstance(self.book.author_birthdate, datetime.date)) def testAuthorMustBeAString(self): def badAssignment(): self.book.author = 999 self.assertRaises(db.BadValueError, badAssignment) def testAuthorIsAString(self): self.book.author = 'Steinbeck, John' The last test, testAuthorIsAString(), tests that assigning a str to the author property of a Book does not raise an exception. If it did, the test would fail. To run all of the tests for the bookstore app, run the following command (from the project root directory, as usual): python manage.py test bookstore The tests run, and the runner prints a summary of successful and failed tests. Real tests can get somewhat complicated, and may require that test data be present in the datastore. Django and the Helper ensure that the datastore starts out empty for each test. You can load data into the datastore in the setUp() routine, or in the test method itself. But for large amounts of test data, there is a better way: fixtures. A fixture is a set of test data that is imported into the datastore at the beginning of a test. The easiest way to create test data is with the Django command python manage.py dumpdata. With the help of the Helper and the BaseModel base class, the dumpdata com- mand knows how to dump the contents of the development server’s simulated data- store in a format that can be used as a Django test fixture. First, clear the datastore with the reset command: python manage.py reset Next, start up a shell: python manage.py shell Using Django Unit Tests and Fixtures | 325
- 347. Enter Python statementsto create test data in the datastore. Here is a shell session that creates a few Book and BookReview entities: % python manage.py shell Python 2.6.1 (r261:67515, Jul 7 2009, 23:51:51) [GCC 4.2.1 (Apple Inc. build 5646)] on darwin Type "help", "copyright", "credits" or "license" for more information. (InteractiveConsole) >>> from google.appengine.ext import db >>> from bookstore import models >>> b1 = models.Book(title="one", author="") >>> b1.put() datastore_types.Key.from_path(u'Book', 1, _app_id_namespace=u'google-app-engine-django') >>> br1 = models.BookReview(book=b1, review_text="review1") >>> br2 = models.BookReview(book=b1, review_text="review2") >>> br3 = models.BookReview(book=b1, review_text="review3") >>> db.put(br1) datastore_types.Key.from_path(u'BookReview', 2, _app_id_namespace=u'google-app-engine-django') >>> db.put(br2) datastore_types.Key.from_path(u'BookReview', 3, _app_id_namespace=u'google-app-engine-django') >>> db.put(br3) datastore_types.Key.from_path(u'BookReview', 4, _app_id_namespace=u'google-app-engine-django') >>> b2 = models.Book(title="two", author="") >>> b2.put() datastore_types.Key.from_path(u'Book', 5, _app_id_namespace=u'google-app-engine-django') >>> b2br1 = models.BookReview(book=b2, review_text="review1") >>> b2br2 = models.BookReview(book=b2, review_text="review2") >>> db.put(b2br1) datastore_types.Key.from_path(u'BookReview', 6, _app_id_namespace=u'google-app-engine-django') >>> db.put(b2br2) datastore_types.Key.from_path(u'BookReview', 7, _app_id_namespace=u'google-app-engine-django') >>> Exit the shell (Ctrl-D). Create a directory named fixtures/ in the bookstore/ app directory. Now run the dumpdata command, and redirect its output to a file in this new directory: python manage.py dumpdata bookstore >bookstore/fixtures/bookstore.json The resulting file contains representations of all of the datastore entities for the kinds whose models are defined in the bookstore app, in the JSON format. To load this fixture for every test in a test case, you create a class attribute for the django.test.TestCase named fixtures, whose value is a list of the names of the fixture files to load from the fixtures/ directory. Here’s another (trivial) test case that uses fixtures, which you can put in models.py: 326 | Chapter 14: The Django Web Application Framework
- 348. # ... class TestBooksAndReviews(test.TestCase): fixtures = ['bookstore.json'] def testBookReviewOrder(self): d = Book.all().filter('title =', 'one').get() self.assert_(d, 'Book "one" must be in test fixture') q = BookReview.all().filter('book =', d) q.order('-create_date') reviews = q.fetch(3) self.assertEqual('review3', reviews[0].review_text) self.assertEqual('review2', reviews[1].review_text) self.assertEqual('review1', reviews[2].review_text) Run the new test using the test command, as before: python manage.py test bookstore For each test in the TestBooksAndReviews test case, Django flushes the simulated data- store, loads the fixtures, calls the test case’s setUp() method (if any), then calls the test method, and finally calls the tearDown() method (if any). You can also load fixtures into the project’s development datastore manually using the python manage.py loaddata command: python manage.py loaddata bookstore/fixtures/bookstore.json See the Django documentation for more information about unit testing and fixtures. Using Django Forms Django includes a powerful feature for building web forms based on data model defi- nitions. The Django forms library can generate HTML for forms, validate that submit- ted data meets the requirements of the model, and redisplay the form to the user with error messages. The default appearance is useful, and you can customize the appearance extensively. Django’s version of the form library only works with the Django data modeling API, and does not work directly with App Engine models. Thankfully, the App Engine Py- thon runtime environment includes an App Engine–compatible implementation of Django forms, in the package google.appengine.ext.db.djangoforms. We won’t go into the details of how Django forms work—see the Django documen- tation for a complete explanation—but let’s walk through a quick example to see how the pieces fit together. Our example will use the following behavior for creating and editing Book entities: • An HTTP GET request to /books/book/ displays an empty form for creating a new Book. Using Django Forms | 327
- 349. • An HTTPPOST request to /books/book/ processes the book creation form, and either creates the book and redirects to /books (the book listing page) or redisplays the form with errors, if any. • An HTTP GET request to /books/book/1234 displays the form to edit the Book en- tity, with the fields filled out with the current values. • An HTTP POST request to /books/book/1234 updates the book with that ID, with the same error-handling behavior as the book creation form. Edit urls.py to use a new view function named book_form() to handle these URLs: from django.conf.urls.defaults import * urlpatterns = patterns('', (r'^books/book/(d*)', 'bookstore.views.book_form'), (r'^books/', 'bookstore.views.home'), ) The regular expression '^books/book/(d*)' captures the book ID in the URL, if any, and passes it to the view function as an argument. Edit views.py to create the BookForm class and the book_form() view function: # ... from google.appengine.ext.db import djangoforms from django.http import HttpResponseRedirect # ... class BookForm(djangoforms.ModelForm): class Meta: model = models.Book def book_form(request, book_id=None): if request.method == 'POST': # The form was submitted. if book_id: # Fetch the existing Book and update it from the form. book = models.Book.get_by_id(int(book_id)) form = BookForm(request.POST, instance=book) else: # Create a new Book based on the form. form = BookForm(request.POST) if form.is_valid(): book = form.save(commit=False) book.put() return HttpResponseRedirect('/books/') # else fall through to redisplay the form with error messages else: # The user wants to see the form. if book_id: # Show the form to edit an existing Book. book = models.Book.get_by_id(int(book_id)) form = BookForm(instance=book) 328 | Chapter 14: The Django Web Application Framework
- 350. else: # Show the form to create a new Book. form = BookForm() return render_to_response('bookstore/bookform.html', { 'book_id': book_id, 'form': form, }) # ... The BookForm class is a subclass of google.appengine.ext.db.djangoforms.ModelForm. It can examine an App Engine model class (a subclass of db.Model) and can render and process forms with fields based on the model’s property declarations. It knows which model class to use from the Meta inner class, whose model class attribute is set to our Book class. The ModelForm has useful default rendering and processing behavior for each of the default property declaration types, and you can customize this extensively. For now, we’ll use the defaults. The book_form() view function takes the HTTP request object and the book_id captured by the regular expression in urls.py as arguments. If the request method is 'POST', then it processes the submitted form; otherwise it assumes the method is 'GET' and just displays the form. In either case, the form is represented by an instance of the BookForm class. If constructed without arguments, the BookForm represents an empty form for creating a new Book entity. If constructed with the instance argument set to a Book object, the form’s fields are prepopulated with the object’s property values. To process a submitted form, you pass the dictionary of POST parameters (request.POST) to the BookForm constructor as its first positional argument. If you also provide the instance argument, the instance sets the initial values—including the entity key—and the form data overwrites everything else, as provided. The BookForm object knows how to render the form based on the model class and the provided model instance (if any). It also knows how to validate data submitted by the user, and render the form with the user’s input and any appropriate error messages included. The is_valid() method tells you if the submitted data is acceptable for saving to the datastore. If it isn’t, you send the BookForm to the template just as you would when displaying the form for the first time. If the data submitted by the user is valid, the BookForm knows how to produce the final entity object. The save() method saves the entity and returns it; if you set the commit=False argument, it just returns the entity and does not save it, so you can make further changes and save it yourself. In this example, a successful create or update redirects the user to /books/ (which we’ve hardcoded in the view for simplicity) instead of rendering a template. Using Django Forms | 329
- 351. To display theform, we simply pass the BookForm object to a template. There are several methods on the object for rendering it in different ways; we’ll use the as_p() method to display the form fields in <p> elements. Create that template now, named templates/bookstore/bookform.html: <html> <body> {% if book_id %} <p>Edit book {{ book_id }}:</p> <form action="/books/book/{{ book_id }}" method="POST"> {% else %} <p>Create book:</p> <form action="/books/book/" method="POST"> {% endif %} {{ form.as_p }} <input type="submit" /> </form> </body> </html> The template is responsible for outputting the <form> tag and the Submit button. The BookForm does the rest. Before we test this, let’s make one more change to the book-listing template to add some “create” and “edit” links. Edit templates/bookstore/index.html: <html> <body> <p>Welcome to The Book Store!</p> <p>Books in our catalog:</p> <ul> {% for book in books %} <li>{{ book.title }}, by {{ book.author }} ({{ book.copyright_year }}) [<a href="/books/book/{{ book.key.id }}">edit</a>]</li> {% endfor %} </ul> <p>[<a href="/books/book/">add a book</a>]</p> </body> </html> Restart the development server, then load the book list URL (/books/) in the browser. Click “add a book” to show the book creation form. Enter some data for the book, then submit it to create it. The default form widget for a date field is just a text field, and it’s finicky about the format. In this case, the “author birthdate” field expects input in the form YYYY-MM-DD, such as 1902-02-27. 330 | Chapter 14: The Django Web Application Framework
- 352. Continue the testby clicking the “edit” link next to one of the books listed. The form displays with that book’s data. Edit some of the data, then submit the form to update the entity. Also try entering invalid data for a field, such as nonnumeric data for the “copyright year” field, or a date that doesn’t use the expected format. Notice that the form redis- plays with your original input, and with error messages. The main thing to notice about this example is that the data model class itself (in this case Book) completely describes the default form, including the display of its form fields and the validation logic. The default field names are based on the names of the prop- erties. You can change a field’s name by specifying a verbose_name argument to the property declaration on the model class: class Book(BaseModel): title = db.StringProperty(verbose_name="Book title") # ... See the Django documentation for more information about customizing the display and handling of forms, and other best practices regarding form handling. Using Django Forms | 331
- 354. CHAPTER 15 Deploying andManaging Applications Uploading your application to App Engine is easy: just click a button or run a command, then enter your developer account email address and password. All of your applica- tion’s code, configuration, and static files are sent to App Engine, and seconds later your new app is running. Easy deployment is one of App Engine’s most useful features. You don’t have to worry about which servers have which software, how servers connect to services, or whether machines need to be rebooted or processes restarted. Other than your developer ac- count, there are no database passwords to remember, no secret keys to generate, and no need to administer and troubleshoot individual machines. Your application exists as a single logical entity, and running it on large-scale infrastructure is as easy as running it on your local computer. App Engine includes features for testing a new version of an app before making it public, reverting to a previous version quickly in case of a problem, and migrating the datastore and service configuration for the app from one version to another. These features let you control how quickly changes are applied to your application: you can make a new version public immediately to fix small bugs, or you can test a new version for a while before making it the version everyone sees. Service configuration is shared across all versions of an app, including datastore in- dexes, task queues, and scheduled tasks. When you upload the app, the service con- figuration files on your computer are uploaded as well, and take effect immediately for all app versions. You can also upload each configuration file separately. This is espe- cially useful for datastore indexes, since new indexes based on existing entities take time to build before they are ready to serve datastore queries. Notice that service configuration is separate from the application configuration, which includes URL mappings, runtime environment selection, and inbound service activa- tion. Application configuration is bound to a specific app version. 333
- 355. App Engine providesrich facilities for inspecting the performance of an app while it is serving live traffic. Most of these features are available in the Administration Console, including analytic graphs of traffic and resource usage, and browsable request and message logs. You can also use the Console to inspect and make one-off changes to datastore entities. You also use the Administration Console to perform maintenance tasks, such as giving other developers access to the Console, changing settings, and setting up a billing account. In this chapter, we’ll discuss how to upload a new app, how to update an existing app, and how to use App Engine’s versioning feature to test a new version of an app on App Engine while your users continue to use the previous version. We’ll look at how to migrate the datastore and service configuration from one version to another. We’ll also look at features of the SDK and the Administration Console for inspecting, trouble- shooting, and analyzing how your live application is running. And finally, we’ll discuss other application maintenance tasks, billing, and where to get more information. Uploading an Application We introduced uploading an application way back in Chapter 2, but let’s begin with a brief review. If you’re developing a Python app using the Launcher for Windows or Mac OS X, you can deploy your app by selecting it in the app list then clicking the Deploy button. The Launcher prompts for your developer account email address and password. The Mac version can remember your account credentials on your Keychain, but it always prompts in case you want to specify a different account. The upload begins, and the Launcher opens a window to display status messages emitted by the uploader tool. If you’re developing a Java app using Eclipse and the Google Plugin, you can deploy your app by clicking on the Deploy App Engine Project button in the Eclipse toolbar (the one that looks like the App Engine logo, a gray airplane engine with blue wings). It prompts for your developer account email address and password. The upload begins, and an Eclipse progress window reports on the status. You can also upload a Python or Java app using the AppCfg command-line tool from the SDK. For Python, the tool takes the update action and a path to your application root directory (the directory containing the app.yaml file): appcfg.py update clock For Java, the tool takes the update action and a path to the WAR directory: appcfg update clock/war (For Mac OS X or Linux and a Java app, use appcfg.sh.) 334 | Chapter 15: Deploying and Managing Applications
- 356. We said thisearlier, but it’s worth repeating: there is no way to down- load the app’s files once they have been uploaded. We strongly advise that you keep backups and use a revision control system. Using Versions The upload tool determines the application ID from the appropriate configuration file. For Python, this is the application element in the app.yaml file. For Java, this is the <application> element in the appengine-web.xml file. The tool also uses this file to determine the version ID to use for this upload, from the version element. If App Engine does not yet have a version of this app with this ID, then it creates a new version with the uploaded files. If App Engine does have such a version, then it replaces the existing version. The replacement is total: no remnants of the previous version’s code or static files remain. The new app has only the files present in the project directory on your computer at the time of the upload. Of course, data stored by the services for the app remain, including the datastore, memcache, log data, and enqueued tasks. The version of the app that is visible on your app’s primary domain name—either app- id.appspot.com or your Google Apps domain—is known as the default version. When you upload your app for the first time, the initial version becomes the default version automatically. If you subsequently upload a version with a different version ID, the original version remains the default until you change the default using the Adminis- tration Console. Recall from Chapter 3 that each version has its own appspot.com URL that includes the version ID as well as the application ID: version-id.latest.app-id.appspot.com Remember that there are no special protections on the version URLs. If the app does not restrict access using code or configuration, then anyone who knows an unrestricted URL can access it. If you don’t want a user to be able to access a version other than the default, you can check the Host header in the app and respond accordingly. You can also upload the nondefault version with configuration that restricts all URLs to ad- ministrators. Be sure to upload it again with the real configuration be- fore making it the default version. When you replace an existing version by uploading the app with that version ID, App Engine starts using the uploaded app for requests for that version within seconds of the upload. It is not guaranteed that every request after a particular time will use the new code and static files, but it usually doesn’t take more than a few seconds for the App Using Versions | 335
- 357. Master to updateall of the frontend servers. The App Master ensures that all of the files are in place on a frontend server before using the new files to handle requests. Internally, App Engine maintains a “minor” version number for each version ID. When you upload the app, App Engine creates a new minor version for the app ID and version ID mentioned in the app’s configuration file, and associates all code, static files, and frontend configuration with that minor version. When it is done creating the new minor version, it declares the new minor version the latest for the version ID, and the frontend starts to use it. Minor version numbers are an internal implementation detail, and are not exposed to the app or the Administration Console. You cannot access previous minor versions or the latest minor version number, and you cannot revert to previous minor versions. (This explains the word latest in the version URL, but you can’t ac- tually replace the word with a minor version number to access other minor versions.) If you upload the app with the same version ID as that of the version that’s currently the default, your users will start seeing the updated app within a few seconds of up- loading. This is fine for small, low-risk changes that don’t depend on changes to your data schemas or datastore indexes. For larger changes, it’s better to upload the app with a new version ID (in the application configuration file), test the app using the version URL, then switch the default version to direct traffic to the new version. To switch the default version, sign in to the Ad- ministration Console, then select Versions from the sidebar. Select the radio button next to the desired version, then click the Make Default button. This is shown in Figure 15-1. Figure 15-1. The Administration Console Versions panel App Engine can host up to 10 different version IDs per app at one time. You can delete unused versions from the Administration Console by clicking the Delete button on the appropriate row. 336 | Chapter 15: Deploying and Managing Applications
- 358. Many actions inthe Administration Console refer to a specific version of the app, in- cluding usage statistics and the log viewer. You can control which version you’re look- ing at by selecting it from the drop-down menu in the top-left corner of the screen, next to the application ID. The version ID only appears as a drop-down menu if you have more than one version of the app. Similarly, the application ID appears as a drop-down menu if you have more than one app associated with your developer account. Managing Service Configuration All versions of an application use the same services. Service configuration and appli- cation data are shared across all versions of the app. An app can have several service-related configuration files: index.yaml or datastore-indexes.xml A description of all of the required datastore indexes. queue.yaml or queue.xml Configuration for task queues. cron.yaml or cron.xml The schedule for scheduled tasks (cron jobs). Whenever you upload the app, these configuration files are uploaded from your com- puter to the services and take effect for the entire app, replacing any configuration that was once there. This is true regardless of whether the app version ID is new or already exists, or whether the version ID is the default. You can update the configuration for each service without uploading the entire app using the AppCfg tool. To update just the index configuration, use the update_indexes action, with the project directory (e.g., app-dir): appcfg.py update_indexes app-dir To update just the task queue configuration, use update_queues: appcfg.py update_queues app-dir And to update just the scheduled task schedule, use update_cron: appcfg.py update_cron app-dir Managing Indexes When you upload the datastore index configuration for an app, the datastore begins building indexes that appear in the configuration but do not yet exist. This process is not instantaneous, and may take many minutes for new indexes that contain many rows. The datastore needs time to crawl all of the entities to build the new indexes. Managing Indexes | 337
- 359. You can checkon the build status of new indexes using the Administration Console, in the Indexes section. An index being built appears with a status of “Building.” When it is ready, the status changes to “Serving.” Figure 15-2 shows a simple example of the “Indexes” section. Figure 15-2. The Administration Console Indexes panel, with two indexes in the “Serving” status If an index’s build status is “Error,” the index has exploded (as described earlier in “Exploding Indexes” on page 159). To clear this condition, you must first remove the index from your configuration, then upload the new configuration. You can change the definition of the index so that it is no longer an exploding index, or you can delete the entities that are causing the problem. Once that is done, you can add the index configuration back and upload it again. If your application performs a query while the index for the query is building, the query will fail. You can avoid this by uploading the index configuration, waiting until the index is built, then making the app that uses that query available. The most convenient way to do this depends on whether you upload the new application in a new version: • If you are uploading the new application using the version identifier that is cur- rently the “default” version, upload the index configuration alone using the appcfg.py update_indexes command. When the indexes are built, upload the app. • If you are uploading the application as a new version, or as a version that isn’t the default and that nobody is actively using, you can safely upload the application and index configuration together (appcfg.py update). Wait until the indexes are built before making the new version the default. 338 | Chapter 15: Deploying and Managing Applications
- 360. If you uploadindex configuration that does not mention an index that has already been built, the datastore does not delete the unused index, since it might still be in use by an older version of the app. You must tell App Engine to purge unused indexes. To do this, run the AppCfg command with the vacuum_indexes option. For instance, in Python: appcfg.py vacuum_indexes app-dir App Engine will purge all custom indexes not mentioned in the index configuration uploaded most recently. This reclaims the storage space used by those indexes. Browsing and Downloading Logs Once your app is up and running on App Engine, you’ll want to monitor, analyze, and manage its behavior. The Administration Console and the AppCfg tool include several important features for doing this. One of the most important is the application’s request and message log. App Engine maintains a log of every request for the app, including the date and time of the request; the IP address of the client; the URL requested; the HTTP status the server returned; and information provided by the client, such as the user agent (the browser identification string) and the referrer (if the user followed a link, which page contained the link). App Engine also records resource usage for each request. An application can also write its own messages to the log, to associate diagnostic in- formation with the request. We looked at the interfaces for doing this back in Chapter 3. You can browse your application’s logs using the Administration Console. To do so, visit the Console, select your app, then choose Logs from the sidebar menu. Separate logs are maintained for each version of the app, so make sure the version whose logs you want to browse is selected. The log browser can show all recent requests for the app, or just recent log messages emitted by the app at a given log level. Each entry can be expanded to view the request details. Figure 15-3 shows an example of the “Info”-level messages for an app, with one request expanded. You can perform advanced filtering in the log viewer by clicking the Options widget. Here you can limit the view to messages before a specific date and time, and can filter for textual patterns in the app’s log messages or in specific fields in the request logs. You can download your application’s log data for offline analysis using the AppCfg tool from the SDK. This command works similarly with both the Python SDK (appcfg.py) and the Java SDK (appcfg or appcfg.sh). Browsing and Downloading Logs | 339
- 361. Figure 15-3. TheAdministration Console log browser The following command downloads request logs for the app in the development di- rectory clock, and saves them to a file named logs.txt: appcfg.py request_logs clock logs.txt Request data appears in the file in a common format known as the Apache Combined (or “NCSA Combined”) logfile format, one request per line (shown here as two lines to fit on the page): 127.0.0.1 - - [05/Jul/2009:06:46:30 -0700] "GET /blog/ HTTP/1.1" 200 14598 - "Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10_5_8; en-us)...,gzip(gfe)" From left to right, the fields are: • The IP address of the client • A - (an unused field retained for backward compatibility) • The email address of the user who made the request, if the user is signed in using Google Accounts; - otherwise • The date and time of the request • The HTTP command string in double quotes, including the method and URL path • The HTTP response code returned by the server • The size of the response, as a number of bytes • The “Referrer” header provided by the client, usually the URL of the page that linked to this URL • The “User-Agent” header provided by the client, usually identifying the browser and its capabilities 340 | Chapter 15: Deploying and Managing Applications
- 362. By default, thisdata does not include the domain name (the “host”) of the request. You can add it by giving the appcfg.py request_logs command the --include_vhost option. The host field is added at the end of the line: appcfg.py request_logs clock logs.txt --include_vhost You can download just the requests for a given host using the --vhost option. This is useful if you are using multiple appspot.com domain names for different purposes, or if you want to see log data for requests directed at specific versions of the app. appcfg.py request_logs clock logs.txt --vhost msgs.example.appspot.com The command above downloads request data only. To download log messages emitted by the application, include a minimum severity level specified as a number, where 0 is all log messages (“debug” level and up) and 5 is only “critical” messages, using the --severity argument: appcfg.py request_logs clock logs.txt --severity=1 Application messages appear in the file on separate lines immediately following the corresponding request. The format for this line is a tab, the severity of the message as a number, a colon, a numeric timestamp for the message, then the message: 1:1246801590.938119 get_published_entries cache HIT Log data is ordered chronologically by request, from earliest to latest. Application messages are ordered within each request by their timestamps. The downloaded log data is the combined log data for all versions of the app. For requests to specific versions using the versioned appspot.com URL, you can use the --include_vhost or --vhost options to differentiate between requests for different ver- sions. There is no built-in way to tell the difference between a request for a previous default version and a newer default version, though the app could add such an indica- tion to its log messages if desired. By default, appcfg request_logs downloads log data for the current calendar day, back to midnight, Pacific Time. You can use the --num_days option to request more than one day’s worth of logs, or specify 0 to download all of the log data available. You can also specify an alternate end date with the --end_date=... option, whose value is of the form YYYY-MM-DD (such as 2009-11-04). For the purposes of appcfg request_logs, App Engine retains up to 90 days of request log data, and about 1,000 application log messages for each log level. You can keep a running archive of all messages using the --append option. If this option is provided, appcfg request_logs will check your local logfile for the latest message it has, then download all available log messages since that message and append them to the file. How often you need to do this for a complete data set depends on how fast the log data grows, based on how often the application writes messages to the log. Log data does not count toward your storage limits, and there is (currently) no way to increase the amount of log data retained by App Engine. Browsing and Downloading Logs | 341
- 363. Inspecting the Datastore Youcan use the Administration Console to inspect and modify entities in the datastore of your live application. You can view all entities of a kind, perform queries using GQL, edit the properties of existing entities, and create new entities based on existing ones. These features are in the Data Viewer section of the Console. The editing facilities of this interface are somewhat limited. You can’t create new prop- erties, delete properties, or change value types. New entities created through the Con- sole have properties and value types derived from other examples of the given kind. You cannot create new kinds. You also can’t view or edit binary (“blob”) data. In most cases, these abilities go beyond the needs of inspecting a live app’s data—most apps can’t use kinds they don’t already know about—but you might notice if the app relies on the existence or nonexistence of properties. You can perform real-time queries on entities from the Console using GQL, even for Java applications. However, you can only perform queries for which the app already has indexes (built-in or custom). If you try to perform a query that needs an index that the app doesn’t have, the Console displays a helpful error message. Refer to the section “GQL” on page 123 for more information about the Administration Console’s data viewer. Application Settings The Application Settings panel of the Administration Console lets you change or view several miscellaneous aspects of your application. This panel lets you view the application ID and the authentication method (Google accounts or Google Apps) that you set when you registered the application ID. These cannot be changed. If you need different values, you must register a new application ID. The application title is the user-visible name of your application. It appears on the Google Accounts sign-in screen. You set this when you created the application, and you can change it at any time from this panel. The “cookie expiration” time is the amount of time a user signed in with a Google account will remain signed in. If the user signs in to your app, he will not have to sign in again from the computer he is using until the expiration time elapses. Also on this screen, you can view which optional services are enabled for this applica- tion. Optional services are those that have passive behavior that affect the app, specif- ically the inbound services of incoming email and incoming XMPP. You can enable these for an app in the app’s configuration file, as discussed in previous chapters. This screen offers an alternate path for setting up a domain name for an app using Google Apps. Instead of adding the app as a service within Google Apps, you can use this setup procedure to associate the app with an Apps domain with default settings. 342 | Chapter 15: Deploying and Managing Applications
- 364. Either setup procedurehas the same effect, and you must still register the domain and create the Google Apps account separately. (There is a link to set up Google Apps on this screen.) Managing Developers When you register an application ID, you become a developer for the application au- tomatically. You can invite other people to be developers for the application from the Developers section of the Administration Console. To invite a developer, you enter the person’s email address in the appropriate field and click Invite. App Engine sends an email to the developer inviting her to set up an ac- count. If the email address you invited is for a Google account, the developer can use the existing account to access App Engine, though she must still accept the invitation by clicking on a link in the invitation email message. If the email address does not have a corresponding Google account, the developer can create a Google account for that address by following the instructions in the message. The developer cannot accept the invitation from a Google account with a different address; you must invite the alternate address explicitly. An invited developer who has not yet accepted the invitation appears in the list with a status of “Pending.” After the developer accepts the invitation, she appears with a status of “Active.” You can remove any developer from the list by clicking the Remove button for the developer. The developer loses all access immediately. All developers of an application have the same rights and access. They can access the Administration Console, upload new application files, create and delete versions, change the default version, access logs, and inspect and tweak the datastore. They can even remove your administrator access, take over billing, and disable or delete the app. When a developer signs in to the application itself via Google Accounts, the app rec- ognizes her as an administrator, allowing her to access URLs configured as administrator-only and telling the app about the administrator status when the app inquires via the API. While you can only register up to 10 application IDs yourself, you can be a developer on as many applications as you like, assuming you can get invited. Of course, it’s against the Terms of Service to create fake accounts just to be able to create more apps and invite your main account to them, so don’t do that. Every administrative action performed by a developer adds a corresponding entry in a special log. Any developer can view this log in the Admin Logs section of the Console. This log cannot be edited or erased. Activities that get logged include app updates, version changes, developer invites and removals, index changes, settings changes, and changes to datastore entities made from the Console. Managing Developers | 343
- 365. Quotas and Billing Thefirst thing you see when you visit the Administration Console and select an app is the “dashboard.” This handy screen provides a visual overview of your app’s traffic, resource usage, and errors. For an example of the dashboard for a new app, refer back to Figure 2-11 in Chapter 2. The topmost chart displays time-based data over the past 24 hours. You can select from several data sets to view via the drop-down menu, including requests per second, clock time or CPU time per request, bandwidth, errors, and quota denials. You can adjust the period for this chart by clicking on the buttons (such as “6 hr”). Below the chart is a graph showing how much of the billable quotas have been con- sumed for the calendar day, and how much of your daily budget has been spent for each quota. A message at the upper-right of the chart indicates how much of the cal- endar day is remaining. If any of the bars look like they might fill up before the next reset, you may need to increase your budget for that quota to avoid quota denials. Near the bottom of the dashboard are lists of popular URL paths and URLs that are returning errors. You can click on a URL to view the detailed request logs for that URL path. The dashboard’s time-based chart and URL traffic lists show data for the version of the app selected by the drop-down menu in the upper-left corner of the screen. When you first sign in to the Console, the default version is selected. To view data for a different version of the app, select it from the drop-down menu. You can view a more comprehensive chart of how the app is consuming resources with quotas from the Quota Details section of the Administration Console. This chart shows billable quotas as well as several fixed quotas, such as API calls and service bandwidth. If your app is having quota-denial errors, check this screen for information on how the app is consuming resources. The resource usage chart on the dashboard and the quota details screen show the total of all resource usage for all versions of the app. All versions of an app share the same budget and quotas. When your app is ready to outgrow the free quotas, you can set up billing and set a budget for additional resources. App Engine allocates more resources as needed ac- cording to the budget you establish, and you are only billed for the resources actually consumed. To set up billing for an application, you select the Billing Settings panel in the Administration Console and click the Enable Billing button. The developer account used to enable billing becomes the “billing account,” and the owner of that account is solely responsible for setting the budget and paying for resources consumed. When you enable billing, the Console prompts you for billing information and au- thorization to charge via Google Checkout, Google’s payment service. The process looks as though you’re making a purchase via Google Checkout, but no money is 344 | Chapter 15: Deploying and Managing Applications
- 366. charged right away.You set the budget for the app, a maximum amount of money that can be charged to your account, then authorize Google Checkout to charge up to that amount. If you increase the budget at a later date, you must reauthorize the amount with Google Checkout. Once you have set the budget amount, you can allocate that budget across the billable quotas. You can select from several presets, or select Custom to set the values directly. Figure 15-4 is an example of the budget settings screen. Figure 15-4. Setting the budget in the Administration Console (rates shown may not be current) You can view a history of charges made to the billing account, including a breakdown of the billed resource usage for each calendar day, in the Billing History panel. Getting Help If you have questions not answered by this book, you may find your answers in the official documentation on Google’s website: http://code.google.com/appengine/ The documentation includes complete references for the APIs and tools for both the Python and Java runtime environments; a list of frequently asked questions and answers (the FAQ); and a large collection of articles describing best practices, interesting fea- tures, and complete working examples. You may also want to browse the contents of the SDK, as installed by either the Eclipse plug-in or the Python Launcher, and also available as a ZIP archive from the website. Getting Help | 345
- 367. The source codefor the Python SDK serves as supplementary documentation, and includes several undocumented (and unsupported) features and extensions. Both SDKs also include a set of functional example applications. All App Engine developers should subscribe to Google’s App Engine downtime mailing list. This low-traffic, announcement-only list is used by the App Engine team to an- nounce scheduled times when services are taken down for maintenance, and also to report the status of unexpected problems: http://groups.google.com/group/google-appengine-downtime-notify You can check the current and past health of App Engine and its services by consulting the system status site: http://code.google.com/status/appengine If you believe you have found a bug in App Engine, the SDK, or the Administration Console, or if you have a feature you’d like to request, you can post to the App Engine issue tracker. You can also see features others have requested, and vote for your favor- ites. (The App Engine team does indeed consider the highly requested features in this list when determining the product road map.) http://code.google.com/p/googleappengine/issues/list Google hosts several mailing lists where the large and vibrant App Engine community gathers to post questions and discuss techniques. The App Engine team is also sub- scribed to these lists, and they often answer questions directly. This page has more information on the mailing lists and other community resources: http://code.google.com/appengine/community.html 346 | Chapter 15: Deploying and Managing Applications
- 368. Index Symbols __init__ () method, property declaration class, 205 != (not equal to) operator, 125, 127 not-equal filter in queries, 150 ( ) (parentheses), grouping in regular A expressions, 76 Access Platform (DataNucleus), 207 . (dot notation) for embedded fields in property, website information on JDO and JPA, 225 214 addresses, email and XMPP, 253 < (less than) operator, 124, 127 AdminEmailMessage class (Python), 259 query with less-than filter, 138 Administration Console, 11 <= (less than or equal to) operator, 124, 127 accessing with Google Apps account, 60 = (equality) operator, 124, 127 application IDs and versions, 70 query with one equality filter, 137 Application Settings, 253 > (greater than) operator, 124, 127 Billing History, 345 query with greater-than filter, 138 Billing Settings, 100, 344 >= (greater than or equal to) operator, 124, Dashboard, 61, 344 127 Data Viewer panel, 123, 342 @Basic annotation, 48, 212 Developers panel, 84, 255, 343 @Column annotation, 213 Indexes panel, 338 @Embeddable annotation, 214 introduction to, 61 @Entity annotation, 48, 210 log browser, 339 name argument, 217 Logs panel, 93 @Extension annotation, 213 Quota Details panel, 96 @GeneratedValue annotation, 210, 212 registering application, 56 @Id annotation, 48, 210 scheduled tasks, 309 @Lob annotation, 213 task queues and, 299 @OneToMany annotation, 223 Versions panel, 336 @OneToOne annotation, 221 ancestor queries, 167 mappedBy argument, 223 ancestors @property annotation (Python), 196 entities and entity groups, 166 @Table annotation, 210 entities with, implementing data class key in @Transient annotation, 212 JPA, 211 _ (underscore) in data object attribute names, entities without, data class keys, 210 185, 284 filtering on, in JPQL query, 219 key with, creating in Java, 172 managing in JPA, 216 We’d like to hear your suggestions for improving our indexes. Send email to index@oreilly.com. 347
- 369. owned objects inrelationships and, 224 vacuum_indexes, 339 Python code creating entity group with root appcfg.sh command entity, 166 update, 61, 334 representing relationships, 224 update_cron, 310 annotations in JPA, 46, 209 update_queues, 307 @Basic, 48, 212 appengine-web.xml file @Column, 213 application IDs and versions, 41, 57, 71 @Embeddable, 214 enabling inbound services, 253 @Entity, 48, 210 example, 68 name argument, 217 system properties, 95 @Extension, 213 appengine_django.models module, 322 @GeneratedValue, 210, 212 application development (see developing @Id, 48, 210 applications) @Lob, 213 application ID, 57, 70, 335 @OneToMany, 223 getting using ApiProxy class, 88 @OneToOne, 221 setting in Java applications, 71 mappedBy argument, 223 setting in Python applications, 71 @Table, 210 application servers, 3, 65 @Transient, 212 CGI communication with Python code, 86 Apache Combined logfile format, 340 application settings, 342 ApiProxy class (Java), 88 application title, 57 ApiProxy.OverQuotaException class (Java), applications, registering (see registering 98 applications) app caching, 89–93 appspot.com domain, 56 in Java, 92 secure connections, 81 in Python, 90 asynchronous calls, 245 app master, 66 asynchronous URL Fetch requests in Python, app.yaml file 240, 244–249 configuring frontend, 67 processing results with callbacks, 247 configuring static files in Python, 76 RPC objects, 246 configuring URL paths, 72 atomic transactions, 165 enabling inbound services, 253, 264 attachments to email restricting access to URLs, 59 email received in Java, 267 specifying app ID, 57 email received in Python, 265 specifying version ID, 71 file types allowed and filename extensions, appcfg command 257 cron_info, 311 sending in Java, 262 update, 61, 334 sending in Python, 259 update_cron, 310 authentication update_queues, 307 with Google Accounts, configuring frontend appcfg.py command for, 83–85 cron_info, 311 restricting access to your application, 59 download_data, 280, 289 viewing method for an application, 342 request_logs, 340, 341 automatic property values, 189 update, 60, 334 auto_now or auto_now_add arguments in with update_cron, 310, 337 date-time property declarations update_indexes, 337 (Python), 189 update_queues, 302, 337 upload_data, 280, 286 348 | Index
- 370. B C back-reference properties, 194 cache expiration for static files back-reference query following one-to-one in Java, 80 relationship, 195 in Python, 77 backup and restore (bulkloader), 281 cache expiration in memcache BadValueError class (Python), 184, 188 setting in Java API, 235 entities not complying with data model, setting in Python API, 231 192 cache hits and misses, 228 bandwidth limits, 100 cache statistics from memcache BaseHandler class (Python), 275 getting from Java API, 237 BaseModel class (appengine_django.models, getting from Python API, 233 Python), 322 caching (see app caching; memcache) batch API calls to services callbacks, Python URL Fetch API, 247–249 datastore, from Python API, 179 CascadeType (javax.persistence), 223 memcache cascading actions in relationships in JPA, 222, using Java API, 236 224 using Python API, 230–231 CGI (Common Gateway Interface), 86 batch processing, 12 changes, propagating to related objects in JPA, using task queues, 295 222 BigTable, 175 chat (see XMPP messages) billable quotas, 100 choices argument, property validator, 188 billing for an application, 344 classes Blob class (Java), 112, 213 datastore entity, creating in Python, 105 Blob class (Python), 112 empty class definition in Python, 106 blobs Client class (Python), 228 storing in datastore, Java API, 213 closures in Python, 248 storing in datastore, Python and Java APIs, Collection class (java.util), 114 112 collection types in data object fields, 213 bound methods, 248 collection_name argument, property BufferedReader class (java.io), 242 declarations, 194 bulk data operations and remote access, 277– Column class (javax.persistence), 213 292 com.google.appengine.api.datastore package, backup and restore with bulkloader, 281 109, 110, 132 bulk loader tool, 280 com.google.appengine.api.labs.taskqueue downloading data, 286–290 package, 304 installing SQLite, 281 com.google.appengine.api.memcache setting up remote API for Java, 279 package, 234 setting up remote API for Python, 278 com.google.appengine.api.urlfetch package, uploading data, 282–286 243 using remote API from a script, 291–292 com.google.appengine.api.users package, 43, using remote shell tool, 290–291 44 bulkloader tool, 280–290 com.google.appengine.api.xmpp package, backup and restore, 281 270, 275 controlling with command-line arguments, com.google.apphosting.api package, 88, 97 289 com.google.apphosting.utils.remoteapi downloading data, 288 package, 279 exporter definition, 287 comma-separated-values files (see CSV files) starting upload for Java or Python app, 286 CommandHandler class (Python), 275 bulkloader.Loader class, 282 comparison operators Index | 349
- 371. supported by datastorequeries and GQL, Python data modeling interface, 34 124 using App Engine models with Django, 322– supported by db.Query class, filter( ) 324 method, 127 web application frameworks, 314 compiling Java project from command line, 42 data modeling with Java (see JPA) concurrency data modeling with Python, 183–205 failure during Java transaction commit, creating custom property classes, 200 172 accepting arguments, 205 failures in Python transactions, 171 customizing default values, 204 multiversion concurrency control, 169 marshaling value types, 202 optimistic concurrency control, 165 validating property values, 201 retrying transaction in Java after model inheritance, 198 concurrency failure, 174 modeling relationships, 192 configuration files many-to-many relationships, 196–198 app.yaml, 67 one-to-many relationships, 195 appengine-web.xml file example, 68 one-to-one relationships, 195 application ID in, 335 models and properties, 184 cron.xml file in Java, 309 private object attributes beginning with _ cron.yaml in Python, 309 (underscore), 185 Django, in app.yaml file, 317 property declarations, 185–192 enabling inbound services, 253 automatic values, 189 index, 135, 159–162 list properties, 190 datastore-indexes.xml file in Java, 161 models and schema migration, 191 index.yaml file in Python, 160 nonindexed properties, 188 persistence.xml, 45, 208 property value types and property queue.xml, 306 declaration classes, 186 request handler for deferred work in Python, validator arguments, 187 303 queries and PolyModels, 199 scheduled task, uploading, 311 data objects, 208 service-related, 337 Java, 46 web.xml, 68 saving, fetching, and deleting in JPA, 214 YAML format, 26 data schemas, 183 consistency in datastore transactions, 164 data storage limits, 100 cookie expiration time, 342 data types counters in memcache, 233 datastore property value types and CPU limits, 97 equivalent language types, 110 CPython interpreter, 3, 86 datastore value types and sort orders, 152 cron jobs (see scheduled tasks) property value types and property cron.xml file, 309 declaration classes, 186 cron.yaml file, 309 property value types in JPA, 212 CSV (comma-separated-values) files, 282, 283 property value types, in Python, 283 skipping unused column, 284 used to store property values, 201 custom domain names, 12 value types for datstore entity properties, 184 value types for properties, 202 D DataNucleus Access Platform, 207 data classes defined in JPA, 209 website information on JDO and JPA, 225 data modeling datastore, 4 bulkloader, use of Python API, 285 ModelForm class in Django, 329 350 | Index
- 372. accessing, with JavaData Objects (JDO) or automatic indexes and simple queries, 136– Java Persistence API (JPA), 45 143 for Django projects, 320 kindless queries in Python, 142 entities and properties, 5 queries on keys in Python, 141 inspecting and modifying entities in, 342 query for all entities of a kind, 137 Python application storing user preferences query with greater-than and less-than in, 33–37 filters, 138 service limits and, 98 query with one equality filter, 137 storing UserPrefs objects in Java application, query with one sort order, 139 45–51 back-reference query following one-to-one transactions, 6 relationship, 195 types, GQL value literals for, 125 comparison operators, 124 updates to, memcache and, 228 configuring in Django application, 323 user preferences stored in, 25 configuring indexes, 159–162 using memcache with, 227 in Java, 161 what it doesn't do, 13 in Python, 160 datastore entities, 5, 103–119 custom indexes and complex queries, 143– deleting in Java, 119 149 deleting in Python, 119 query with filters on multiple properties, entities and entity groups, transactions with, 144 165–168 query with multiple equality filters, 147 entities and keys in JPA, 209–212 query with multiple sort orders, 143 getting entities using keys, 116 GQL, 123–126 in Java, 119 in python, 123 in Python API, 116 during index updates, 180 inspecting entity objects, 117 introduction to indexes, 134 Java datastore API, 108–110 Java API, 132–134 keys and key objects in Java API, 114 keys-only queries, 133 keys and key objects in Python API, 114– JPQL queries in JPA, 217–220, 225 116 kindless queries, 168 keys and properties, 104 kinds and, 122 postprocessing by bulkloader loader class, multivalued properties and, 153–159 285 exploding indexes, 159 property values, 110–114 MVPs and equality filters, 155 multivalued properties in Java API, 114 MVPs and inequality filters, 156 multivalued properties in Python API, MVPs and sort orders, 157 113 MVPs in Python, 154 Python code setting values, 111 simple example in Python, 153 strings, text and blobs in Java API, 112 not-equal (!=) and IN filters, 150 strings, text and blobs in Python API, PolyModels and, 199 112 Python API, 126–131 unset versus null value in Java API, 113 GqlQuery class, 128 unset versus null value in Python API, keys-only queries, 131 112 Query class, 127 Python API for datastore, 105–108 retrieving results, 129 saving, 118 results and keys, 122 datastore entity properties (see properties) sort orders and value types, 152 datastore queries, 6, 121–162 unset and nonindexed properties, 150 ancestor, 167 datastore transactions, 163–181 Index | 351
- 373. allowing reads andwrites outside of using versions, 335–337 transactions, 208 deployment descriptors batch updates, 179 example web.xml file, 68 entities and entity groups, 165–168 JSP, setting explicit URL mapping, 74 ancestor queries, 167 web.xml file for ClockServlet (example), 41 keys, paths, and ancestors, 166 deployment limits, 98 entity updates, 175–178 dereferencing of reference properties, 194 in Java, 172–175 list properties, 196 in JPA, 216 developer tools, 10 in Python, 169–171 developers, managing, 84, 343 reading entities, 178 developing applications, 24–55 task enqueueing, 307 Java application, 39–55 updating indexes, 180 caching with memcache, 51–54 what can happen in, 168 web forms and the datastore, 45–51 datastore-indexes.xml file, 161 Python application, 25–38 DatastoreFailureException class (Java), 172 storing user preferences in datastore, 33 DatastoreService class (Java), 117, 132 users and Google Accounts, 30 beginTransaction( ) method, 172 webapp framework, 28 DatastoreServiceFacoty class (Java) user preferences pattern, 24 getDatastoreService( ) method, 117 development console, 54 Date class (java.util), 235 execution of scheduled tasks, 310 DateProperty class (Python), 189 task execution, 299 dates, times, and date-times as property value development servers types, 284 configuring Python server to use Sendmail, datetime module, 26 254 DateTimeProperty class (Python) configuring Python server to use SMTP automatic property values, 189 server for email, 254 default values, 204 datastore for Django projects, 320 DeadlineExceededError class (Python), 97 logging of email messages by Java server, DeadlineExceededException class (Java), 97 254 default values for properties simulation of receiving email, 264 customizing, 204 dev_appserver command, 42 validation, 187 dev_appserver.py command, 27 default version, 335 --help option, 17 deferred module, 302 dev_appserver.sh command, 42 defer( ) function, 303 disk seeks, 227 mapping URL path to, 303 distributed memory cache, 227 PermanentTaskFailure class, 304 Django web application framework, 28, 313– DELETE action cascades, 222 331 DELETE statements in JPQL, 220 App Engine Helper, 318–320 deploying and managing applications, 333– creating a Django application, 320–322 346 creating a Django project, 315 application settings, 342 installing, 314 browsing and downloading logs, 339–341 request handler script, 316 developers, 343 unit tests and fixtures, 324 inspecting the datastore, 342 using App Engine data models with, 322– quotas and billing, 344 324 service configuration, 337 using Django forms, 327–331 uploading application, 334 django-admin.py command, 315 352 | Index
- 374. django.core.handlers.wsgi module, 316 EmailAddress class (Python), 259 django.http module, 321 EmailMessage class (Python), 258 django.http.HttpRequest object, 321 example of sending email in Python, 259 django.shortcuts module, 321 Embeddable class (javax.persistence), 214 domain names, 69 embedded objects, JPA, 214 domains enhancement process (JPA), 48, 209 App Engine applications, 2 entities, 5, 6, 104 setting up app domain name using Google (see also datastore entities) Apps, 343 Entity class (Java), 113 setting up custom domain name for getKey( ) method, 114, 172 application, 58 methods to inspect the entity, 117 site identification by, secure connections results from keys-only queries, 134 and, 81 setUnindexedProperty( ) method, 151 using custom domain names, 12 entity groups, 7, 164 using custom domain other than batch updates and, 179 appspot.com when registering keys, paths, and ancestors, 166 application, 56 kinds of entities and, 167 dot notation for embedded fields in property, managing in JPA, 216 214 related objects in, 224 DownloadError class (Python), 246 transactions with, 169 DuplicatePropertyError class (Python), 198 entity ID, 104 dynamic properties, 185 entity properties (see properties) EntityManager class (javax.persistence), 45, 208, 214 E find( ) method, 215 Eclipse IDE, 10 persist( ) method, 214 installing Java SDK with Google Plugin, 21 remove( ) method, 215 Packagae Explorer, 39 EntityManagerFactory class uploading Java app, 61 (javax.persistence), 45, 208 versions for Mac, 32-bit and 64-bit, 21 createEntityManager( ) method, 214 editors, 16 EntityTransaction class (javax.persistence), email and instant messages, 251–276 216 checking email addresses for validity, 259 environment variables, 85 email and XMPP addresses for applications, set by App Engine at beginning of requests, 253 87 enabling inbound services, 253 equality filters flow of incoming mail and XMPP messages, combining with inequality filters, 145 252 multivalued properties and, 155 receiving email messages, 263–267 query using multiple equality filters, 147 in Java, 266–267 Error class (Python), 242 in Python, 264–266 error handler for memcache, 238 receiving XMPP messages, 272–276 ErrorHandler class (Java), 238 sending email messages, 254–263 exceptions, property validation errors, 188 attachments, 257 Expando class (Python), 106, 183 in Java, 261–263 classes based on, inheritance from, 198 in Python, 258–261 difference from Model class, 185 recipients, 256 Expiration class (Java), 235 sender addresses, 255 exploding indexes, 159 sending email using remote shell tool, 290 Exporter class (Python), 287 sending XMPP messages, 267–272 Index | 353
- 375. exporter class fordownloads, 287–289 G GeneratedValue class (javax.persistence), 211 F GenerationType class (javax.persistence), 210 fetching URLs, 239–249 Gmail accounts for application developers, asynchronous requests in Python, 244–249 255 in Java, 242–244 Gmail user names as application IDs, 57 in Python, 240–242 Google Accounts, 9 FetchOptions class (Java), 132, 243 application developer address as sender fields of data objects in JPA, 212 address for email, 255 queries for specific fields, 219 authorization with, 83–85 fields of email message in Python interface, Java servlet displaying information and 258 links, 43 fields of embedded classes in JPA, 214 Python API for, 32 file count limit for applications, 98 Python application displaying account file servers for static files, 4 information and links, 30–33 file size limits for application files, 98 using for authentication, 25 file types for email attachments, 257 Google App Engine filename extensions, MIME types and, 77 things it doesn’t do, 12 filters, 122 website documentation for, 345 ancestor queries, 167 website, SDK downloads, 13 comparison operators for, 124 Google App Engine Launcher (see Launcher) not-equal (!=) and IN filters, 150 Google Apps queries on keys, 141 app using Google Apps domain name, 61 queries using multiple equality filters, 147 application access policy, 84 Query class (Python), filter( ) method, 127 authentication, 59 query with filters on multiple properties, Google BigTable distributed data system, 175 144 Google Checkout, 345 query with greater-than and less-than filters, billing management, 100 138 Google Data (GD) protocol, 111 query with one equality filter, 137 Google Plugin for Eclipse ranges in query filters, 139 installing Java SDK with, 21 specifying in JPQL queries, 219 using to develop Java application, 39 WHERE clause in GQL, 124 Google Talk, 251 fixtures in Django, 325–327 checking user’s status, 271 frontends, 64 use by XMPP service, 268 configuring, 66–85 google.appengine.api.labs.taskqueue module, app IDs and versions, 70 299 authorization with Google Accounts, google.appengine.api.mail module, 258, 264 83–85 google.appengine.api.memcache module, 38, domain names, 69 228 Java application, 68 google.appengine.api.quota module, 98 Python application, 67 google.appengine.api.urlfetch module, 240, request handlers, 72–75 241 secure connections, 81–83 google.appengine.api.users module, 32 static files and resource files, 75–80 google.appengine.api.xmpp module, 269 function objects in Python google.appengine.dist module, 316 create_callback( ) function and, 249 google.appengine.ext.db module, 34, 105, 110 handling transactions, 170 data modeling library, 183 354 | Index
- 376. google.appengine.ext.db.djangoforms module, HTTP methods supported by URL Fetch 327 service, 239 google.appengine.ext.db.polymodel module, HTTP POST requests, 36, 263 199 email received in Java, 266 google.appengine.ext.deferred module, 302 XMPP messages received, 272 google.appengine.ext.remote_api.remote_api_ XMPP messages received in Python, 273 stub module, 291 httplib module (Python), 240 google.appengine.ext.webapp module, 29 HttpRequest object (django.http), 321 google.appengine.ext.webapp.mail_handlers HttpResponseRedirect class (django.http, module, 265 Python), 328 google.appengine.ext.webapp.util module, HTTPS connections, 81, 82 264 (see also secure connections) google.appengine.ext.webapp.xmpp_handlers over TCP port 443, 82 module, 275 URL Fetch service, 239 google.appengine.runtime module, 96 HttpServlet class (javax.servlet.http), 41 google.appengine.runtime.apiproxy_errors HttpServletRequest class (javax.servlet.http), module, 98 41, 266 google.appengine.tools.bulkloader module, getUserPrincipal( ) method, 44 282, 286 HttpServletResponse class (javax.servlet.http), GQL, 123–126 41 comparison operators, 124 HTTTP POST requests in Python, 128 email received in Python, 264 key literals with ancestors, 167 keywords, 126 LIMIT and OFFSET clause, 126 I query for all entities of a kind, 137 Id class (javax.persistence), 212 query with multivalued properties, 155 idempotent, 298 SELECT * and SELECT __key__, 126 image processing, 9 SELECT statements, 124 imports value literals for datastore types, 125 multiple requests handling by same WHERE clause, 124 application instance, 90 GqlQuery class (Python), 128 Python code files stored in ZIP archives, 99 fetch( ) method result limit and offset, 130 IN operator, 125 iteration with, 131 filters in queries, 150 keys-only queries, 131 InboundEmailMessage class (Python), 264 kindless query, 143 InboundMailHandler class (Python), 265 grouping in regular expressions, 76 incrementing and decrementing values in memcache, 233 in Java API, 237 H index.yaml file, 160 HashMap class (java.util), 236 configuration for query with multiple sort help, 345 orders, 144 HTTP GET requests, 309 for query with filters on multiple properties, HTTP headers 145 cache instructions for client, 75 indexed argument in property declarations, checking Host header using Python and 188 webapp, 69 indexes, 6 domain name in, 81 ancestor queries, 168 MIME type information in, 77 automatic, and simple queries, 136–143 resource usage, 101 kindless queries, 142 Index | 355
- 377. queries on keys,141 iterable query objects in Python, 131 query for all entities of a kind, 137 query with greater-than and less-than filters, 138 J query with one equality filter, 137 Jabber, 251 query with one sort order, 139 Jabber ID (see JID) configuration files, 135 JAR (Java Archive) files, 42 configuring, 159–162 using to reduce file count for application in Java, 161 code, 99 in Python, 160 Java custom, and complex queries, 143–149 App Engine runtime environment, 3 query with filters on multiple properties, application IDs and versions, 71, 335 144 bulkloader loader class definition, 285 query with multiple equality filters, 147 bulkloader tool, running, 280 query with multiple sort orders, 143 configuring frontend, 68 database and App Engine datastore, 121 datastore API, 108–110 introduction to, 134 datastore entities maintenance of, 136 deleting, 119 managing for an application, 337 getting using keys, 117 multivalued properties and, 155 inspecting, 117 exploding indexes, 159 saving, 119 MVPs and inequality filters, 156 datastore entity keys and Key objects, 114 MVPs and sort orders, 158 datastore entity property values, 110–114 nonindexed properties and datastore Entity class, setting and unsetting queries, 150 properties, 113 nonindexed properties in JPA, 213 listing of equivalent types, 110 updating, 180 multivalued properties, 114 updating index configuration, 337 strings and text, and blobs, 112 inequality filters deploying an application, 334 combining with equality filters, index order deployment limits on files, 99 and, 145 developing applications (see developing multivalued properties and, 156 applications) query cannot use inequality filters on more development console, 54 than one property, 147 downloading logs, 339 query on keys in Python, 141 enabling inbound services, 253 query using with sort orders, index order Enterprise Edition (Java EE), 20 and, 146 fetching URLs, 242–244 inheritance of data model, 198 using low-level API to call URL Fetch multiple inheritance, 198 service, 243 InputStream class (java.io), 266 frontend configuration for authentication, InputStreamReader class (java.io), 242 85 instant messages, 8, 251 index configuration, 161 (see also XMPP messages) installing App Engine SDK, 20–24 InternetAddress class (javax.mail.internet), on Mac OS X, 20 262 with Google Plugin for Eclipse, 21 interpreter without Eclipse, 23 App Engine Python interpreter, 87 Java 6 JVM, 20 in runtime environment, 85 logging in, 94 IOException class (java.io), 243 memcache API, 234–238 query API, 132–134 356 | Index
- 378. keys-only queries, 133 java.util.Date class, 235 receiving email, 266–267 java.util.HashMap class, 236 parsing MIME multipart messages in java.util.List class, 114, 236 HTTP POST, 266 java.util.logging library, 94 receiving XMPP messages, 275 controlling log level of messages, 95 remote API, setting up, 279 java.util.logging.Logger class, 94 request handlers, 73 java.util.Map class, 236 request limits, 97 javac command, 42 resource files, 79 --version, 20 runtime environment, 87 javax.mail package, 261 app caching, 92 javax.mail.internet package, 261 scheduled tasks configuration file, 309 javax.mail.internet.InternetAddress class, 262 SDK, plug-in for Eclipse IDE, 10 javax.mail.internet.MimeBodyPart class, 262 secure connections, 83 javax.mail.internet.MimeMessage class, 261, sending email, 261–263 262, 266 attachments, 262 javax.mail.internet.MimeMultipart class, 262 creating JavaMail session, 261 javax.mail.Session class, 261 setting MIME type for message body, javax.mail.Transport class, 261 262 javax.persistence.CascadeType, 223 sending XMPP messages, 270 javax.persistence.Column class, 213 checking Google Talk user’s status, 272 javax.persistence.Embeddable class, 214 service limits, 98 javax.persistence.EntityManager class, 214 static files in, 79–80 javax.persistence.EntityManagerFactory class, task queues in, 304–307 45, 208 transactions in, 172–175 javax.persistence.EntityTransaction class, 216 committing and rolling back javax.persistence.GeneratedValue class, 211 transactions, 173 javax.persistence.GenerationType class, 210 concurrency failure during commit, 172 javax.persistence.Id class, 212 retrying transaction after concurrency javax.persistence.OneToMany class, 223 failure, 174 javax.persistence.OneToOne class, 221 unindexed properties, 151 javax.persistence.Persistence class, 208 uploading application from bulkloader, javax.persistence.Query class, 217 286 javax.servlet.http.HttpServlet class, 41 Java Data Objects (see JDO) javax.servlet.http.HttpServletRequest class, 41 Java Persistence API (see JPA) javax.servlet.http.HttpServletResponse class, Java Runtime Environment (JRE) classes, 89 41 Java SE Development Kit (JDK), 20 JCache interface, 52, 234 Java ServerPages (see JSPs) JDO (Java Data Objects), 45, 108 Java ServerPages Standard Tag Library (JSTL), datastore queries, 132 75 essential features provided by, 207 java.io.BufferedReader class, 242 resources for further information, 225 java.io.InputStream class, 266 JID (Jabber ID), 267, 268, 269 java.io.InputStreamReader class, 242 using to get status of Google Talk user, 271 java.io.IOException class, 243 JPA (Java Persistence API), 3, 45, 108–110, java.io.Serializable interface, 53 207–225 java.net.URL class, 242 configuration file, persistence.xml, 45 java.net.URLConnection class, 242 datastore queries, 132 java.security.Principal class, 44 embedded classes, support by App Engine, java.util.Collection class, 114 214 Index | 357
- 379. enhancement process byAccess Platinum, retrieving entities using keys, 116 209 setting up data class in JPA representing entities and keys, 209–212 entities with parents, 216 entity properties, 212–214 specifying name for, 107 queries and JPQL, 217–220 storing as property values, 192 relationships in, 220–225 uniqueness, 107 resources for further information, 225 keys-only queries saving, fetching, and deleting datastore in Java, 133 objects, 214 in Python, 131 setting up, 208 kindless queries, 142 transactions in, 216 ancestor queries, 168 JPQL (Java Persistence Query Language), 217– kinds (datastore entities), 104 220 instantiating GqlQuery from kind class, deleting entities, 220 128 queries for specific fields, 219 queries and, 122 queries on relationship fields, 225 queries without kinds, 142 string prefix trick, 219 query for all entities of a kind, 137 syntax, 218 specifying alternate kind name using JRE (Java Runtime Environment) classes, 89 @Table annotation, 210 JSPs (Java ServerPages), 43, 74 URL mapping for, 74 JSTL (Java ServerPages Standard Tag Library), L 75 Launcher JVM (Java virtual machine), 3, 87 overview, 18 uploading an application, 334 uploading app from, 60 K LIMIT clause (GQL), 126 Key class (Java), 114, 211 link model method, 197 toString( ) method, 212 Linux Key class (Python), 114 downloading and installing Python SDK, complete keys, 193 17 from_path( ) method, 115, 167, 284 installing Java SDK with Google Plugin for key list method, 196 Eclipse, 21 Key objects, 114–116 Python on, 16 key prefixes versus namespaces, 231 testing App Engine Java SDK installation, KeyFactory class (Java), 172 23 createKey( ) method, 115 List class (java.util), 114, 236 keyToString( ) and stringToKey( ) methods, list properties, 190 212 using key list method, 196 keyToString( ) method, 116 ListProperty class (Python), 190, 202 KeyFactory.Builder class (Java), 172 key list method, 196 keys (datastore entity), 104 lists complete key, 193 datastore entity properties with multiple creation by bulkloader, 284 values, 113 data classes defined in JPA, 210 multivalued properties in Python, 154 entities and entity groups, 166 load balancer, application frontend, 64 in indexes, 134 Loader class, 282 and Key objects, 114–116 generate_key( ) method, 284 queries on, 141 handle_entity( ) method, 285 query results returning, 122 loader classes, 282–286 358 | Index
- 380. local transactions, 7 caching UserPrefs objects in Java log files, browsing and downloading, 339–341 application, 51–54 Logger class (java.util.logging), 94 Java API, 234–238 logging, 93–96 Python API in Java applications, 94 cache expiration, 231 in Python applications, 94 cache statistics, 233 Java development server logging email, 254 deleting keys, 232 log levels supported by App Engine, 93 memcache counters, 233 login memcache namespaces, 231 configuring login element in app.yaml, 84 setting and getting multiple values, 230 required login for Python application, 36 setting and getting values in Python, required login restricting access to app 229 URLs, 59 Python application caching user preferences login URL, 32 objects, 37 logout URL, 32 Python development console viewer for, 54 Python memcache API, 228–234 service limits and, 98 M updates to, 228 Mac OS X user preferences stored in, 25 downloading and installing Python SDK, memcache module (Python), 38 17 MemcacheService class (Java), 234 installing Java App Engine SDK, 20 setErrorHandler( ) method, 238 installing Java SDK with Google Plugin for SetPolicy enumeration, 235 Eclipse, 21 setrNamespace( ) method, 236 Python on, 16 MemcacheServiceException class (Java), 238 testing App Engine Java SDK installation, MemcacheServiceFactory class (Java), 234 23 memory cache (see memcache) mail (see email and instant messages) Message class (Java), 270, 275 mail module, 258 methods accessing Message fields, 276 InboundEmailMessage object, 264 Message class (Python), 274 send_mail( ) method, 258 MessageBuilder class (Java), 270 Mail service API, 258 methods assembling a Message object, 271 mailing lists, 346 MessageType enumeration (Java), 271 main( ) function methods of class instance as callable objects, app caching and, 91 248 calling in Python script handler using app MIME types caching, 91 email messages, 263 main.py, 26, 318 for email message attachments, 257 manage.py, 318 static files in Java applications, 80 deploying and managing App Engine static files in Python applications, 77 application, 320 MimeBodyPart class (javax.mail.internet), 262 features added by Django Helper, 319 MimeMessage class (Java), 262 startapp action, 320 MimeMessage class (javax.mail.internet), 261, many-to-many relationships, 196–198 266 JPA, not supported by App Engine, 221 MimeMultipart class (javax.mail.internet), using key list method, 196 262 using link model method, 197 Model class (Python), 34, 184 Map class (java.util), 236 BaseModel class, 322 mappedBy annotation argument, 223 classes based on, inheritance from, 198 memcache, 8, 227–238 Index | 359
- 381. creating datastore entityclasses, 106 specifying in JPA, 223 is_saved( ) method, 117 support in JPA, 221 put( ) method, 38 one-to-one relationships ModelForm class (djangoforms, Python), 329 in Python, 195 models.py, 33 support in JPA, 221 editing to use BaseModel, 322 OneToMany class (javax.persistence), 223 multiple inheritance, 198 OneToOne class (javax.persistence), 221 multivalued properties, 113, 284 optimistic concurrency control, 7, 165 collection types in fields stored as, 213 ORDER BY clause (GQL), 124 property declaration class, db.ListProperty, org.datanucleus.jpa.annotations package 190 (Java), 211 queries and, 153–159 OverQuotaError class (Python), 98 exploding indexes, 159 owned relationships, 221 MVPs and equality filters, 155 MVPs and inequality filters, 156 MVPs and sort orders, 157 P MVPs in Python, 154 parameter substitution, GQL queries in simple example in Python, 153 Python, 128 using to store entity paths, 168 parameters in JPQL queries, 218 multiversion concurrency control, 169 pass keyword (Python), 106 MVPs (see multivalued properties) paths deriving key of entity with a parent, using ancestor path, 167 N entities and entity groups, 166 name argument to @Entity annotation on data from_path( ) method, Key class (Python), class, 217 115, 284 name argument, property validator, 187 key path in Java, 172 namespaces in memcache, 231 using multivalued properties to store, 168 using in Java API, 236 pattern matching NCSA Combined logfile format, 340 mapping URL path pattern to a servlet, 74 NeedIndexError class, 149 for URL paths, 64, 72 NeedIndexError exception (Python), 148 using regular expressions, 72 nested functions, 248 PermanentTaskFailure class (Python), 304 nonindexed properties, 150 PERSIST action cascades, 222 declaring in entity properties in JPA, 213 Persistence class (javax.persistence), 208 establishing using property declaration, persistence.xml file, 45, 208 188 PersistenceManager class, 51 nonindexed types, 152 pickle module, 229 null values plain old Java objects (POJOs), 46 datastore property values, 112 PolyModel class (Python), 183, 199 property values set to None, 187 storing inheritance hierarchy information, 200 polymorphism in queries, 199 O PreparedQuery class (Java), 132 object databases, 5 keys-only queries, 134 OFFSET clause (GQL), 126 Presence class (Java), 272 offset parameter for db.Query fetch( ) method, primary key for data class, 210 129 Principal class (java.security), 44 one-to-many relationships processing rate for task queues, 295 in Python, 195 properties, 5, 104 360 | Index
- 382. creating your ownproperty classes, 200 application development (see developing accepting arguments, 205 applications) customizing default values, 204 application IDs and versions, 71, 335 marshaling value types, 202 bulk loader tool, 280 validating property values, 201 bulkloader tool, backup and restore, 281 data model classes inheriting from configuring frontend, 67 db.Model versus db.Expando, CPU usage by request handler, analyzing, 184 98 datastore entities defined in JPA, 212–214 data modeling library, 183 datastore entity datastore API, 105–108 inspecting and modifying, 117 datastore entities values assigned by Model class, 106 deleting, 119 datastore property values, 110–114 getting using keys, 116 datastore types and equivalent language inspecting, 117 types, 110 saving, 118 strings, text, and blobs, 112 datastore entity keys and Key objects, 114 declarations, 185–192 datastore entity property values, 110–114 duplicate property error, 198 code setting, 111 inheritance of, 198 listing of equivalent types, 110 list properties, 190 multivalued properties, 113 models and schema migration, 191 strings, text, and blobs, 112 nonindexed properties, 188 unset and null values, 113 property value types, 186 datastore queries reference properties, 192–194 multivalued properties, 154 setting automatic values, 189 using GQL, 123 validator arguments, 187 deploying an application, 334 declaring in class definition, 184 deployment limits on files, 99 entity created in Python, 106 developing applications (see developing fields of JPA embedded classes stored as, applications) 214 development console, 54 multiple sort orders in a query, 143 Django (see Django web application name property value in indexes, 134 framework) names beginning with _ (underscore), 284 downloading logs, 339 names of, 187 enabling inbound services, 253 queries and multivalued properties, 153– fetching URLs, 240–242 159 asynchronous requests, 244–249 query with filters on multiple properties, frontend configuration for authentication, 144 84 reserved names in Python data model API, Google Accounts API, 32 185 if __name__ ... condition, 92 unset and nonindexed properties, queries index configuration, 160 and, 150 index configuration for query with filters on value types, 283 multiple properties, 145 Property class (Python), 201 index configuration for query with multiple __init__( ) method, 205 sort orders, 144 property descriptors, 186 installing App Engine SDK, 16 PyDev, 10 logging, 94 Python memcache API, 228–234 App Engine runtime environment, 3 cache expiration, 231 Index | 361
- 383. cache statistics, 233 run_in_transaction( ) function, 171 deleting keys, 232 unindexed properties, 151 memcache counters, 233 version 2.5, 17 namespaces for memcache keys, 231 web application frameworks, 28 setting and getting multiple values, 230 python -V command, 16 setting and getting values, 229 query API, 126–131 GqlQuery class, 128 Q queries, 121 iterable query objects, 131 (see also datastore queries) keys-only queries, 131 datastore entity kinds and, 122 kindless queries, 143 Query class (Java), 132 Query class, 127 setKeysOnly( ) method, 133 retrieving results, 129 Query class (javax.persistence), 217 query on keys, using inequality filter, 141 Query class (Python), 127, 323 receiving email messages, 264–266 count( ) method, 130 parsing POST content into fetch( ) method, 129 InboundEmailMessage object, filter method, 143 264 get( ) method, 130 receiving XMPP messages, 273–275 iteraction with, 131 remote API key-only queries, 131 setting up, 278 Queue class (Java), 304 using from a script, 291–292 Queue class (Python), 301 remote shell tool, 290–291 queue.xml file, 306 request handlers, 72 QueueFactory class (Java), 307 request limits, 96 queues (see task queues) runtime environment, 86 quotas and limits, 96–102 app caching, 90 billable quotas, 100 scheduled tasks, 309 CPU limits, 97 Cron Jobs in development console, 310 deployment limits, 98 SDK (software development kit), 10 quotas and billing for an application, 344 secure connections, 82 request limits, 96 sending email, 258–261 service limits, 98 example using several features, 259 using EmailMessage object, 258 using mail module, send_mail( ), 258 R sending XMPP messages, 269 range of results, fetching in JPQL query, 218 chat message, 270 ranges in query filters, 139 checking Google Talk user’s status, 271 recurrence of scheduled tasks, 310 service limits, 98 redirects sqlite3 module, 281 Django application, 328 static files in, 76–79 following when using Java URLConnection, task queues in, 299–304 243 deferring work with tasks, 302–304 follow_redirects argument of urlfetch transactions in, 169–171 module, fetch( ) function, 242 batch updates, 179 reference properties, 192 commits and rollbacks, 171 back-references created by entities and entity groups, 166 db.ReferenceProperty, 194 function objects handling transactions, declaring with db.ReferenceProperty, 192 170 362 | Index
- 384. declaring with db.SelfReferenceProperty, configuring frontends, 66–85 193 app IDs and versions, 70 representing one-to-many relationships, authorization with Google Accounts, 195 83–85 representing one-to-one relationships, 195 domain names, 69 ReferenceProperty class (Python), 192 Java app, 68 back-references from referenced object to Python app, 67 objects referring to it, 194 request handlers, 72–75 registering applications, 56 secure connections, 81–83 application ID and title, 57 static files and resource files, 75–80 Google Apps and authentication, 59 limit for datastore queries, 129 setting up domain name, 58 quotas and limits on system resources used, regular expressions, 72 96–102 URL pattern in static_files element, 76 request handler for deferred work in Python, using with app caching with re module, 92 303 relational databases, 5 request handler for tasks, 294 relationships, 183 request handler script invoking Django, in JPA, 220–225 316 field on owned class pointing back to running the application, 85–96 owner, 222 logging, 93–96 lazy fetching of related objects, 224 Python runtime environment, 86 owned one-to-one relationship, 221 running the application app caching, 89– propagating changes to related objects, 93 222 separation from data storage, 103 queries on relationship fields, 225 task handlers and retries, 297 specifying one-to-many relationship, request limits, 96 223 RequestHandler class (Python), 275 modeling in Python, 192 required argument, property validator, 187 many-to-many relationships, 196–198 resource files, 75 one-to-many relationships, 195 Java applications, 79 one-to-one relationships, 195 deployment limits on, 99 remote access in WEB-INF directory, 80 architecture of remote API, 277 security requirements for accessing, 83 setting up remote API for Java, 279 limits on, 98 setting up remote API for Python, 278 Python applications, 77 using remote API from a script, 291–292 stored in ZIP archive, 99 using remote shell tool, 290–291 resource usage headers, 101 RemoteApiServlet class (Java), 279 resources, billable, 100 reply-to address for email, 256 ResponseTooLargeError exception (Python), request handlers, 24 242 mapping to URL/prefs in Python ResponseTooLargeException (Java), 243 application, 36 results for datastore queries, 122 simple Python request handler script, 26 retrieving with Python query API, 129 simple Python request handler using retries for task handlers, 298 webapp framework, 29 watching the count, 299 static file, specifying in Python applications, Rollback exception (Python), 171 76 root entity, 166, 170 request handling, 63 creating key for, in Java, 172 App Engine architecture, 64–66 deletion of, entity group and, 167 Index | 363
- 385. RPC objects, 246 SELECT statements in JPQL, 218 nested function as callback accessing RPC SelfReferenceProperty class (Python), 193 object, 248 sender addresses object method as callback accessing RPC email messages, 254 object, 248 XMPP messages, 268 wait( ) method, 249 Sendmail, configuring Python development runtime environment, 2, 65 server for, 254 app caching, 89 SendResponse class (Java), 270 in Python, 90 Serializable interface (Java), 234 components of, 85 serialization Java applications, 87 objects in memcache, 53 app caching, 92 pickle module (Python), 229 logging, 93–96 serializable objects as key or value in Python applications, 86 memcache, 234 quotas and limits on system resources used serializable objects in JPA data object fields, by application, 96 212 sandbox, 88 Server Name Identification (SNI), 81 specifying name and version in Python service configuration, 337 app.yaml file, 67 service limits, 98 run_in_transaction( ) function (Python), 171 services, 8 servlet class, 73 servlets, 74 S (see also JSPs) sandbox, 2, 88 ClockServlet (example), 39–54 Java sandbox, 89 controlling loading order, 93 Python sandbox, 89 declaring, 73 scheduled tasks, 10, 293, 308–311 mapping incoming email URL path to, 266 Python configuration file, cron.yaml, 309 mapping remote API servlet with URL, 279 recurrence of the task, 310 mapping to URL path called by XMPP time zone information for, 310 service, 275 updating schedule, 337 mapping to URL path pattern, 74 uploading configuration file, 311 mapping URL to request handler schema migration, data models and, 191 performing tasks, 305 schemas for data, 183 Session class (javax.mail), 261 SDKs (software development kits), 10, 15–24 settings.py, 318 App Engine download site, 13 INSTALLED_APPS variabale, 320 browsing contents, 346 template directories, 321 installing Java SDK, 20 ShortBlob class (Java), 112, 212 without Eclipse, 23 SMPT server to send email, configuring Python with Google Plugin for Eclipse, 21 development server to use, 254 on Mac OS X, 20 SNI (Server Name Identification), 81 installing Python SDK, 16 sort order, 122 secure connections, 12, 81–83 datastore value types and, 152 Java applications, 83 GQL ORDER BY clause, 124 Python applications, 82 multivalued properties and, 157 security constraints queries on keys, 141 authentication constraint in Java web.xml queries with one sort order, 139 file, 85 Query class (Python), order( ) method, 127 specifying for secure connection in Java application, 83 364 | Index
- 386. query combining sortorders with inequality setting dealy for task, 306 filters, 146 using in Python, 299–304 query with multiple sort orders, 143 task, defined, 293 SQL (Structured Query Language), GQL and, TaskOptions class (Java), 305 123 TaskOptions.Builder class (Java), 305 SQLite, 281 tasks, scheduled (see scheduled tasks) SSL (Secure Socket Layer), HTTP over, 81 TCP port 443, secure connections over, 82 enabling or disabling in Java apps, 83 TCP ports for HTTP and HTTPS connections, static file servers, 4, 65 239 static files, 75 templates in Django application, 321 Java applications, 79–80 templating systems, web application cache expiration, 80 frameworks, 31 deployment limits on, 99 TestCase class (django.test, Python), 324 MIME types, 80 fixtures attribute, 326 specifying, 80 TestCase class (unittest, Python), 324 limits on, 98 Text class (Java), 112 Python applications, 76–79 Text class (Python), 112 cache expiration, 77 text editors, 16 MIME types, 77 text strings, storing in datastore, 112 stored in ZIP archive, serving, 99 throughput, 164 treating as resource files also in Python, 77 time limits for URL connections, setting in Java, static properties, 185 243 statistics on caching in memcache, 233 time, setting for cache expiration, 232 Stats class (Java), 237 TimeProperty class (Python), 189 StrictErrorHandler class (Java), 238 timer for requests, 96 string prefix trick in JPQL, 219 title (application), 57 StringProperty class (Python), 185, 202 token buckets, 295 strings, 112 tombstones or tombstoned tasks, 297 converting between Key and string-encoded Transaction class (Java) key, 212 commit( ) method, 172 converting datastore entity keys to, 115 rollback( ) method, 173 property values, validation of, 201 transactional reads, 169 string-encoded key generated by system, TransactionFailedError class (Python), 171 211 transactions (see datastore transactions) strong consistency, 164 Transport class (javax.mail), 261 synchronous URL fetching, 240 system properties, setting, 95 system resources, quotas and limits for U applications, 96–102 Unicode, 112 unit testing and fixtures in Django, 324–327 unittest module, 324 T unowned relationships, 221 task queues, 9, 293–308 unset datastore property values, null versus, elements of a task, 296 113 processing rates and token buckets, 295 uploading applications, 60, 334 task handlers and retries, 297 URL class (java.net), 242 testing and managing tasks, 299 URL Fetch API, 8 transactional task enqueueing, 307 URLConnection class (java.net), 242 updating configuration, 337 urlfetch module, 241 using in Java, 304–307 customizing URL Fetch behaviors, 241 Index | 365
- 387. using from Java,243 value types for properties, 110–114 using from Python for asynchronous calls, marshaling, 202 245 version ID, 70, 335 urllib module (Python), 240 getting using ApiProxy class, 88 urllib2 module (Python), 240 setting in Java applications, 71 URLs setting in Python applications, 71 configuring paths to use secure connections, versions of applications, 335 82 task queues and, 298 development server console, 299 views.py, 321 Django application, 321 editing to query database, 323 fetching, 239 (see also fetching URLs) full URL for use with urlfetch module, W WAR (Web Archive) files, 42, 68 fetch( ) function, 241 files pushed to application servers, 79 mapping incoming email URL path to script WAR directory path, 61 handler in Python, 264 web application frameworks, 313, 314 mapping servlet to URL path pattern, 74 (see also Django web application mapping to Django handler script, 317 framework) mapping to JSPs (Java ServerPages), 74 Python, 28, 314 mapping to script calling web application CGI communication with application framework, 314 servers, handling, 86 mapping URL /prefs to PrefsServlet templating systems, 31 (example), 50 web forms remote API URL in Python, 279 Java servlet handling user preferences form, remote API, mapping to servlet in Java, 50 279 request handler for preferences form, 35 scheduled task URL path, 309 user preferences form displayed by Java task URL, default, 305 servlet application, 49 urls.py, 321 using Django forms, 327–331 user accounts, 9 web hooks, 293 user preferences web.xml file application customizing time zone, 24 ClockServlet application (example), 41 Python application storing preferences in example, 68 datastore, 33 mapping URL /prefs to PrefsServlet UserProperty class (Python), 190 (example), using security users constraint, 50 Google Accounts information and links mapping URL to task request handler, 305 displayed by Python app, 30–33 webapp framework, 28 Google Accounts information, displaying BaseHandler class, 275 with Java servlet, 43 CommandHandler class, 275 users module (Python), 32 InboundMailHandler class, 265 UserService class (Java), 44 simple Python request handler using, 29 UserServiceFactory class (Java), 44 webapp.RequestHandler class, 29 webapp.WSGIApplication class, 29 V WHERE clause (GQL), 124 validation of property values, 201 WHERE clause in JPQL, 219 validator arguments in property declarations, wildcards in URL path patterns, 74 187 Windows value types (datastore), sort orders and, 152 366 | Index
- 388. downloading and installingPython SDK, 17 installing Java SDK with Google Plugin for Eclipse, 21 testing App Engine Java SDK installation, 23 X XMPP addresses, 253 XMPP messages, 12, 251–276 addresses, 267 enabling inbound services, 253 flow of incoming messages, 252 receiving, 272 in Java, 275 in Python, 273–275 sender address, 268 sending, 267–272 chat invitation in Python, 269 chat message in Python, 270 checking GoogleTalk user’s status, 271 sending in applications, testing in development console, 55 xmpp module, 269 XMPP protocol, 8, 251 XMPPService class (Java) getPresence( ) method, 272 parseMessage( ) method, 275 sendMessage( ) method, 270 XMPPService class (Python), 269 XMPPServiceFactory class (Python), 269 Y YAML format for configuration files, 26, 67 Z ZIP archive files for Python application files, 99 Index | 367
- 390. About the Author DanSanderson is a technical writer and software engineer at Google Inc. He has worked in the web industry for more than 10 years as a software engineer and technical writer for Google, Amazon.com, and the Walt Disney Internet Group. He lives in Seattle, Washington. For more information about Dan, visit his website at http://www .dansanderson.com. Colophon The animal on the cover of Programming Google App Engine is a waterbuck (Kobus ellipsiprymnus), a type of antelope found in western, eastern, and southern Africa. Waterbucks stand at about five feet at the shoulder and have reddish-brown coats that become darker in color as the animals age. Long, sinuous horns distinguish male waterbucks, while a ring of white hair around the tail distinguishes both genders from other antelopes. Waterbucks live in savannas and other vegetative areas, where they graze on rough grass and leaves from trees and bushes. Contrary to its name, the waterbuck spends most of its time on land, but it will often take refuge in a body of water to avoid predators. African myth claims that the meat of the waterbuck is inedible, but this isn’t so. Al- though the waterbuck’s sweat glands produce a strong odor reminiscent of turpentine in order to better protect itself from predators, the animal’s meat—while not especially flavorful—is safe to consume. Male waterbucks frequently use their horns as a means of defense against their enemies. As the males are polygamous and highly possessive of their mates, they are especially prone to fatally goring other male waterbucks who enter their territories and try to steal members of their harems. The cover image is from Wood’s Animate Creations. The cover font is Adobe ITC Garamond. The text font is Linotype Birka; the heading font is Adobe Myriad Con- densed; and the code font is LucasFont’s TheSansMonoCondensed.