Python Programming: Data Structure

PythonIntermediateProgramming
임찬식(chanshik@gmail.com)
1

PythonIntermediateProgramming
타입과 객체
함수와함수형프로그래밍
클래스와객체지향프로그래밍
데이터구조
튜닝과 최적화
2

데이터구조
프로그램을작성할때도움이되는효율적인데이터구조모듈소개
array
bisect
collections
itertools
3

abc
abc모듈은새로운추상기반클래스를정의하는데사용하는
메타클래스와장식자한쌍을가지고 있음
ABCMeta
추상기반클래스를나타내는메타클래스
Python3
>>> import abc
>>> class Stackable(metaclass=abc.ABCMeta):
... pass
Python2
class Stackable:
__metaclass__ = abc.ABCMeta
4

abc
추상기반클래스의특징
abstracemethod, abstractproperty 장식자를사용하여
메서드나프로퍼티를정의하면실제구현을제공할경우에만
인스턴스생성가능
어떤타입을논리적하위클래스로등록하는데사용하는클래스메서드
register(subclass) 를가지고 있음
__subclasshook__(cls, subclass) 를추가로정의가능
타입subclass 가 하위클래스로간주되면True
subclass 가 하위클래스가 아니면False
아무런정보가 없을경우NotImplemented예외를발생
5

abc
abstractmethod(method)
method를추상메서드로선언하는장식자
직접상속을통해정의한파생클래스에서실제구현을제공해야
인스턴스생성가능
register() 메서드로등록한하위클래스에는영향없음
abstractproperty(fget [, fset [, fdel[, doc]]])
추상프로퍼티를생성
매개변수들은일반property() 함수와동일
파생클래스에서는해당프로퍼티구현을제공해야인스턴스생성가능
6

abc
추상기반클래스정의
>>> from abc import ABCMeta, abstractmethod, abstractproperty
>>> class Stackable(metaclass=ABCMeta):
... @abstractmethod
... def push(self, item):
... pass
... @abstractmethod
... def pop(self):
... pass
... @abstractproperty
... def size(self):
... pass
7

abc
Stackable을상속하여클래스생성
>>> class Stack(Stackable):
... def __init__(self):
... self.items = []
... def push(self, item):
... self.items.append(item)
... def pop(self):
... return self.items.pop()
...
>>> a = Stack()
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
TypeError: Can't instantiate abstract class Stack with
abstract methods size
size() 프로퍼티가 없어인스턴스를생성할수없음
8

abc
size() 프로퍼티를Stack에서상속받은클래스에추가
>>> class CompleteStack(Stack):
... @property
... def size(self):
... return len(self.items)
...
>>> s = CompleteStack()
>>> s.push("foo")
>>> s.push("bar")
>>> s.size
2
>>> s.pop()
'bar'
9

array
한종류의타입만담는객체. 공간 효율적인자료구조
array(typecode, [, initializer])
타입코드목록
코드 설명 C 타입 크기
'b' 8비트정수 signedchar 1
'B' 8비트부호없는정수 unsignedchar 1
'u' 유니코드문자 PY_UNICODE 2 / 4
'h' 16비트정수 short 2
'H' 16비트부호없는정수 unsignedshort 2
정수형은머신아키텍처에의해결정
10

array
코드 설명 C 타입 크기
'i' 정수 int 4 / 8
'I' 부호없는정수 unsignedint 4 / 8
'l' 긴 정수 long 4 / 8
'L' 부호없는긴 정수 unsignedlong 4 / 8
'q' 더욱긴 정수 longlong 8 (Python3.3)
'Q' 부호없는더욱긴 정수 unsignedlonglong 8 (Python3.3)
'f' 단일정밀도실수 float 4
'd' 배정밀도실수 double 8
11

array
array 객체와생성기 표현식을이용한데이터생성
>>> import array
>>> a = array.array("i", [1, 2, 3, 4, 5])
>>> b = array.array(a.typecode, (x * x for x in a))
>>> a
array('i', [1, 2, 3, 4, 5])
>>> b
array('i', [1, 4, 9, 16, 25])
12

array
enumerate() 함수를이용한제자리(in‑place) 연산을통해
공간을절약하는방식의프로그래밍활용
>>> for i, x in enumerate(a):
... a[i] = x * x
...
>>> a
array('i', [1, 4, 9, 16, 25])
13

bisect
bisect 모듈은리스트를정렬된순서로유지하는데필요한기능제공
bisect(list, item [, low [, high]])
insort(list, item [, low [, high]])
bisect(): list 를정렬된순서로유지하기 위해서item을
list 어디에삽입해야하는지위치반환
insort(): item을list 에정렬된순서에맞게 삽입
item이이미존재하면기존항목오른쪽에삽입
14

bisect
>>> from bisect import bisect
>>> grades = "FEDCBA"
>>> breakpoints = [30, 44, 66, 75, 85]
>>> grades[bisect(breakpoints, 50)]
'D'
>>> grades[bisect(breakpoints, 90)]
'A'
15

bisect
>>> from bisect import insort
>>> l = [10, 30, 50, 70]
>>> insort(l, 60)
>>> l
[10, 30, 50, 60, 70]
>>> insort(l, 40)
>>> l
[10, 30, 40, 50, 60, 70]
16

collections
유용한컨테이너타입에대한고성능구현함수와
다양한컨테이너를위한추상기반클래스등을제공
deque
defaultdict
namedtuple
17

deque
양끝을가진객체를나타내는컨테이너
>>> from collections import deque
>>> d = deque('hello')
>>> d.pop()
'o'
>>> d.pop()
'l'
>>> d
deque(['h', 'e', 'l'])
>>> d.append('a')
>>> d.append('b')
>>> d.appendleft('c')
>>> d.appendleft('d')
>>> d
deque(['d', 'c', 'h', 'e', 'l', 'a', 'b'])
>>> d.popleft()
'd'
>>> d.popleft()
'c'
>>> d
deque(['h', 'e', 'l', 'a', 'b'])
18

deque
deque를생성할때maxlen인수를지정하면정해진크기를가진
원형버퍼로동작
새로운항목을추가하는데공간이부족하면반대편에있는항목삭제
deque([1, 2, 3], maxlen=5)
>>> d.append(4)
>>> d.append(5)
>>> d
deque([1, 2, 3, 4, 5], maxlen=5)
>>> d.append(6)
>>> d.append(7)
>>> d
deque([3, 4, 5, 6, 7], maxlen=5)
19

deque
deque는매우효율적인자료구조
뒤쪽끝에항목을추가하는작업은list 보다약간 느림
하지만, 앞쪽에항목을추가하는작업은list 보다매우빠름
deque에새로운항목을추가하는작업은스레드에서도안전
pickle모듈로직렬화가능
20

defaultdict
없는키를처리하는부분을제외하고는dict 와동일
키를찾지못하는경우에default_factory 로지정한함수호출
함수는기본값을생성하고 이값을해당키의값으로사용
defaultdict 객체는데이터를추적하기 위해서
사전을컨테이너로쓸때매우유용
21

defaultdict
>>> from collections import defaultdict
>>> s = "yeah but no but yeah but no but yeah"
>>> words = s.split()
>>> word_locations = defaultdict(list)
>>> for n, w in enumerate(words):
... word_locations[w].append(n)
...
>>> word_locations
defaultdict(<class 'list'>, {'yeah': [0, 4, 8],
'but': [1, 3, 5, 7], 'no': [2, 6]})
22

이름있는튜플
튜플을사용할때불편한점은개별항목을숫자로만접근할수있다는것
이러한불편을줄이기 위해이름있는튜플(namedtuple)을제공
namedtuple(typename, fieldnames [, verbose])
이름이typename인tuple하위클래스생성
fieldnames 는속성이름목록지정
유효한파이썬식별자여야함
튜플에나타날항목과 동일한순서로지정
23

이름있는튜플
>>> from collections import namedtuple
>>> NetworkAddr = namedtuple("NetworkAddr", "hostname port")
>>> a = NetworkAddr("www.python.org", 80)
>>> a.hostname
'www.python.org'
>>> a.port
80
>>> type(a)
<class '__main__.NetworkAddr'>
>>> len(a)
2
>>> isinstance(a, tuple)
True
24

이름있는튜플
이름있는튜플은데이터구조로만사용될객체를정의할때유용
name, shares, price변수를가진데이터구조가 필요할경우
>>> Stock = namedtuple("Stock", "name shares price")
>>> a = Stock("APL", 30, 45.50)
>>> a
Stock(name='APL', shares=30, price=45.5)
25

heapq
heapq 모듈은힙(heap)으로우선순위큐를구현
힙조건에따라정렬된항목들의리스트
heap[0] 은항상가장작은항목을담고 있음
0 에서시작하는모든n에대해서
heap[n] <= heap[2 * n+ 1]
heap[n] <= heap[2 * n+ 2] 모두만족
26

heapq
heapify(x)
리스트x 를제자리에서힙으로변환
heappop(heap)
힙조건을유지하면서가장작은항목을반환하고 제거
힙이비어있으면IndexError 발생
heappush(heap, item)
힙조건을유지하면서item을힙에추가
heappushpop(heap, item)
item을힙에추가하고 힙에서가장작은항목을제거하는
동작을한번에수행
27

heapq
heapreplace(heap, item)
힙에서가장작은아이템을반환하고 제거하면서
새로운항목item을추가
힙이비어있으면IndexError 발생
merge(s1, s2, ...)
정렬된반복가능한객체s1, s2 등을하나의정렬된순서열로생성
nlargest(n, iterable[, key])
iterable에있는가장큰n개의항목으로구성된리스트생성
nsmallest(n, iterable[, key])
iterable에있는가장작은n개의항목으로구성된리스트생성
28

heapq
>>> l = [19, 9, 4, 10, 11, 8, 2]
>>> heapq.heapify(l)
>>> l
[2, 9, 4, 10, 11, 8, 19]
>>> heapq.heappop(l)
2
>>> l
[4, 9, 8, 10, 11, 19]
>>> heapq.heappop(l)
4
>>> l
[8, 9, 19, 10, 11]
>>> heapq.heappush(l, 5)
>>> l
[5, 9, 8, 10, 11, 19]
>>> heapq.heappush(l, 25)
>>> l
[5, 9, 8, 10, 11, 19, 25]
>>>
29

heapq
>>> heapq.nlargest(3, l)
[25, 19, 11]
>>> heapq.nsmallest(3, l)
[5, 8, 9]
>>> a = [3, 5, 1, 9, 7]
>>> b = [6, 4, 8, 2, 10]
>>> heapq.heapify(a)
>>> heapq.heapify(b)
>>> h = heapq.merge(a, b)
>>> h.__next__()
1
>>> h.__next__()
2
>>> h.__next__()
4
>>> h.__next__()
5
30

itertools
itertools 모듈은데이터에대해서다양한본복을수행하는데
사용할수있는효율적인반복자를생성하는함수들을제공
모듈에있는모든함수는for 문이나기타반복자와관련있는
함수와함께사용할수있는반복자를반환
31

chain
chain(iter1, inter2, ..., interN)
반복자그룹이주어질때모든반복자를연결하는새로운반복자생성
>>> for x in chain([1, 2, 3], [4, 5, 6]):
... print(x)
...
1
2
3
4
5
6
32

combinations
combinations(iterable, r)
iterable에서길이r 인모든하위순서열을반환하는반복자생성
>>> list(combinations([1, 2, 3, 4], 2))
[(1, 2), (1, 3), (1, 4), (2, 3), (2, 4), (3, 4)]
>>> list(combinations([1, 2, 3, 4], 3))
[(1, 2, 3), (1, 2, 4), (1, 3, 4), (2, 3, 4)]
33

count
count([n])
n에서시작하는연속적인정수를만드는반복자생성
n을생략하면0 에서시작
sys.maxint 를넘으면‑sys.maxint ‑ 1 에서다시시작
>>> g = count()
>>> g.__next__()
0
>>> g.__next__()
1
>>> g.__next__()
2
>>> g.__next__()
3
34

cycle
cycle(iterable)
iterable에있는원소들을계속해서순환하는반복자생성
>>> g = cycle([1, 2, 3, 4])
>>> g.__next__()
1
>>> g.__next__()
2
>>> g.__next__()
3
>>> g.__next__()
4
>>> g.__next__()
1
>>> g.__next__()
2
35

dropwhile
dropwhile(predicate, iterable)
함수predicate(item) 이True로평가되는동안
iterable에서항목을버리는반복자생성
일단predicate가 False를반환하면해당항목과
iterable에있는나머지항목이생성
>>> s = [10, 20, -3, 5, 3, 6, -5, 7, 15]
>>> list(dropwhile(lambda x: x >= 0, s))
[-3, 5, 3, 6, -5, 7, 15]
36

groupby
groupby(iterable [, key])
iterable에서생성되는연속적인항목을그룹짓는반복자생성
그룹을짓기 위해중복된항목검색
iterable에서동일한항목이여러번나오면그룹이생성
>>> g = groupby("aaaabbccddeeefffgg")
>>> g.__next__()
('a', <itertools._grouper object at 0x10673cd30>)
>>> g.__next__()
('b', <itertools._grouper object at 0x10673ccf8>)
>>> g.__next__()
('c', <itertools._grouper object at 0x10673ce80>)
37

ifilter (Python3: filter)
ifilter(predicate, iterable)
iterable에서predicate(item) 이True인항목만추출하는반복자생성
predicate가 None이면iterable에있는모든항목이True로평가
>>> s = [10, 15, -3, 5, -6, 7, 15]
>>> g = filter(lambda x: x < 0, s)
>>> list(g)
[-3, -6]
38

ifilterfalse(Python3: filterfalse)
ifilterfalse(predicate, iterable)
iterable에서predicate(item) 이False인항목만추출하는반복자생성
predicate가 None이면iterable에있는모든항목이False로평가
>>> from itertools import filterfalse
>>> s = [10, 15, -3, 5, -6, 7, 15]
>>> g = filterfalse(lambda x: x < 0, s)
>>> list(g)
[10, 15, 5, 7, 15]
39

imap (Python3: map)
imap(function, iter1, iter2, ..., iterN)
function(i1, i2, ..., iN) 들을수행하는반복자생성
>>> a = [1, 2, 4, 10]
>>> b = [5, 10, 15, 20]
>>> m = map(lambda x, y: x + y, a, b)
>>> list(m)
[6, 12, 19, 30]
>>> m = map(lambda x, y: x + y, a, b)
>>> m.__next__()
6
40

islice
islice(iterable, [start,] stop [, step])
iterable[start:stop:step]이반환한것과 비슷한항목을추출하는반복자생성
>>> s = [10, 15, -3, 5, -6, 7, 15]
>>> g = islice(s, 4)
>>> g.__next__()
10
>>> g.__next__()
15
>>> g.__next__()
-3
>>> g.__next__()
5
>>> g.__next__()
StopIteration
41

izip (Python3: zip)
izip(iter1, iter2, ... iterN)
(iter1, iter2, ..., iterN) 등묶어서사용할수있게 해주는반복자생성
주어진반복자중하나라도값을더이상생성하지않으면반복을멈춤
>>> a = [1, 3, 5, 7]
>>> b = [2, 4, 6, 8]
>>> c = [3, 5, 7]
>>> g = zip(a, b, c)
>>> g.__next__()
(1, 2, 3)
>>> g.__next__()
(3, 4, 5)
>>> g.__next__()
(5, 6, 7)
>>> g.__next__()
StopIteration
42

izip_longest (Python3: zip_longest)
izip_longest(iter1, iter2, ..., iterN [,fillvalue=None])
반복자iter1, iter2, iterN 등을모두소진할때까지반복이이어진다는것을
제외하고는izip() 과 동일
>>> g = zip_longest(a, b, c)
>>> g.__next__()
(1, 2, 3)
>>> g.__next__()
(3, 4, 5)
>>> g.__next__()
(5, 6, 7)
>>> g.__next__()
(7, 8, None)
>>> g.__next__()
StopIteration
43

permutaions
permutations(iterable [, r])
iterable에서길이r 인모든순열을반환하는반복자생성
r 을생략하면iterable에있는항목수와동일한길이로가정
>>> a = [1, 3, 5, 7]
>>> permutations(a, 2)
<itertools.permutations object at 0x1066c4f10>
>>> list(permutations(a, 2))
[(1, 3), (1, 5), (1, 7), (3, 1), (3, 5), (3, 7),
(5, 1), (5, 3), (5, 7), (7, 1), (7, 3), (7, 5)]
44

product
product(iter1, iter2, ... iterN [, repeat=1])
iter1, iter2 등에있는항목들의데카르트곱(Cartesianproduct)을
나타내는튜플을만들어내는반복자생성
>>> a = [1, 3, 5, 7]
>>> b = [2, 4]
>>> product(a, b)
<itertools.product object at 0x106742558>
>>> list(product(a, b))
[(1, 2), (1, 4), (3, 2), (3, 4), (5, 2), (5, 4), (7, 2), (
45

repeat
repeat(object [, times])
object 를계속해서만드는반복자를생성
times: 반복횟수(times 가 없으면끝없이반복)
>>> g = repeat([1, 3, 5], 4)
>>> g.__next__()
[1, 3, 5]
>>> g.__next__()
[1, 3, 5]
>>> g.__next__()
[1, 3, 5]
>>> g.__next__()
[1, 3, 5]
>>> g.__next__()
StopIteration
46

starmap
starmap(func [, iterable)
func(*item) 들을만들어내는반복자생성
item은iterable에서가져옴
func() 을제대로호출할수있는iterable에대해서만제대로동작
>>> g = starmap(lambda a, b: a * b, [(1, 2), (2, 3), (3, 4
>>> g.__next__()
2
>>> g.__next__()
6
>>> g.__next__()
12
>>> g.__next__()
StopIteration
47

takewhile
takewhile(predicate [, iterable])
predicate(item) 이True로평가되는동안iterable에서항목을추출하는
반복자생성
predicate가 False로평가되는즉시반복을멈춤
>>> s = [10, 15, -3, 5, -6, 7, 15]
>>> g = takewhile(lambda x: x >= 0, s)
>>> g.__next__()
10
>>> g.__next__()
15
>>> g.__next__()
StopIteration
48

tee
tee(iterable [, n])
iterable에서n개의독립적인반복자생성
생성된반복자들은n개 항목튜플로반환
n: 기본값은2
>>> s = [10, 15, -3, 5, -6, 7, 15]
>>> g = tee(s)
>>> g[0].__next__()
10
>>> g[0].__next__()
15
>>> g[1].__next__()
10
>>> g[1].__next__()
15
49

Python Programming: Data Structure

More Related Content

What's hot

Viewers also liked

Similar to Python Programming: Data Structure

More from Chan Shik Lim

Recently uploaded

Python Programming: Data Structure