Challenges of Transpiling Smalltalk to JavaScript

Challenges of
Transpiling Smalltalk to JavaScript
Noury Bouraqadi & Dave Mason
International Workshop on Smalltalk Technologies Gdansk, Poland; July 1st to 4th, 2025

Why Transpile Smalltalk to JavaScript?
2

ST
Simple & clean semantics
Rich IDE & core libraries
3
JS
Ubiquitous deployment
Performance
Large ecosystem

ST
Simple & clean semantics
Rich IDE & core libraries
4
Wednesday
July 2nd 2025
JS
Ubiquitous deployment
Performance
Large ecosystem
● ST + JS = Best of both worlds
● Smalltalk community interest in the Web/JavaScript
○ ESUG Main Track Talk Smalltalk for the Web

5
Transpilation = TRANSlation & comPILATION
Transpiler
100%
Smalltalk
Program
100%
JavaScript
Program

6
Transpiler
100%
Smalltalk
Program
100%
JavaScript
Program
Valid Valid

7
Transpiler
100%
Smalltalk
Program
100%
JavaScript
Program
Semantically
Equivalent
Valid Valid

Output
Output
Input Input
8
Semantic Equivalence
Transpiler
100%
Smalltalk
Program
100%
JavaScript
Program

9
How to Transpile Smalltalk to JavaScript?

○ Map Smalltalk’s syntax to JavaScript?
○ Map Smalltalk’s reflective kernel to JavaScript?
○ Handle Smalltalk dependencies with the runtime?
10

11
Empirical Study
based on
10 Year
experience with

12
Challenges Catalog
1. Primitive Types &
Literals
2. Messages
3. Block Closures
4. Classes
5. Reflection
Empirical Study
based on
10 Year
experience with

● JS 𝑢𝑛𝑑𝑒𝑓𝑖𝑛𝑒𝑑 and 𝑛𝑢𝑙𝑙 are Not Objects.
● JS has an Impoverished Numeric Stack.
● ST Automatically Converts
○ Between Small Integers & Large Ones.
● ST Supports Fixed-Point Arithmetic.
● ST has Literal Symbols.
14
Challenges: Primitive Types and Literals

● JS has a 𝑆𝑦𝑚𝑏𝑜𝑙 class, but:
○ No literal symbols
■ Simple to fix
● Map ST symbols to instances of JS Symbol
○ JS Symbol class is not related to String!
■ Complex to fix
● Make Symbol subclass of String
○ Side-effects?
● Override some String methods
15
Example: ST Literal Symbols

● Non-Alphanumeric Characters in ST Message Selectors
● JS Math-Like Message Priorities
● ST Message Cascading
16
Challenges: Messages

● Simple to fix:
○ Replace non-alphanumeric characters with their Ascii code
○ ST keyword selectors: 1 string concatenating keywords
● Complex to fix: None
17
Example: Non-Alphanumeric Characters in Selectors

Smalltalk Blocks:
● Always Bind the Outer Context
● Always Answer Some Value
● Support Non-Local Returns
18
Challenges: Block Closures

● Simple to fix: ST blocks always answer some value
○ ST result := [123]
○ JS result = () =>{return 123}
● Complex to fix: ST blocks support non-local returns
○ ST condition ifTrue: [^123]
○ JS rely on exceptions
19
Example: Block Returns

● Class Variables
● Pool Variables
● Class Extensions
● Stateful Traits
20
Challenges: ST Classes - 1

● Class Initialization & Startup/Shutdown Lists
● Methods Always Have a Return Value
● Methods can Have Pragmas
○ Primitive Pragmas Refer to the Virtual Machine
21
Challenges: ST Classes - 2

● Simple to fix: Class Variables
○ JS Classes are objects with attributes
○ JS encapsulation is optional
class A{
anInstanceMethod(){
x = A.someClassVariable + 42; }
}
A.someClassVariable = 37;
22
Examples - 1

● Complex to fix: Pragmas such as #primitive:
○ Implement the primitive behavior
23
Examples - 2

● ST Reifies Messages Upon Handling Type Errors
● ST Reifies Execution Contexts
● ST and JS have different solutions for Intercepting
Method Evaluation
● Pharo ST Reifies Slots
● Pharo ST classes define Object Format/Layout
24
Challenges: Reflection

● Simple to fix: ST DNU Reifies Messages
○ Extend JS Object with
■ doesNotUnderstand() method
■ default methods for every sent message in ST Code
Object.prototype.zork = function(arg1, arg2){
return this.doesNotUnderstand(“zork”, arguments)}
25
Examples - 1

● Complex to fix: ST Reifies Execution Contexts
○ ST thisContext
26
Examples

Smalltalk & JavaScript
100% semantic equivalence
via transpilation is
difficult if not impossible!
27
Conclusion

● Some challenges are “easy” to address
○ Primitive Types & Literals,
○ Messages,
○ Block Closures
● Several ST capabilities are complex to implement in JS
○ Primitives,
○ thisContext,
○ Reified Slots…
28
Conclusion: Not all challenges are equal

● Beyond Transpilation
○ ST and JS Run-time interoperability (Production)
○ Live-coding with JS objects from the ST (Development)
○ Reuse JS libraries (code + globals) in code transcribed from ST
○ Transpile ST to produce JS libraries for 3rd parties.
● Methodology?
○ Generalizable to other language pairs? Smalltalk & Python?
29
Future Work

Develop in Pharo, Run on JavaScript
PharoJS.org
Kindly Supported by
Thanks to all the contributors!
M
IT
License

Challenges of Transpiling Smalltalk to JavaScript

More Related Content

Similar to Challenges of Transpiling Smalltalk to JavaScript

More from ESUG

Recently uploaded

Challenges of Transpiling Smalltalk to JavaScript