CorePython

A Python to WebAssembly compiler written in Rust.

Features:

• Very minimal subset of the Python language. Only the core of it, nothing fancy.
• CorePython compiler itself is embeddable in browser (small WebAssembly download).

Phase

This project is in prototype phase.

Usage

First clone this repository, and check the Python to WebAssembly compiler options:

$ git clone <this repo>
$ cargo run -- -h  # All options after -- are passed to the CorePython compiler
...

Say, you start with an annotated function like this:

def myAdd(a: int, b: int) -> int:
    return a + b

Put this function in demo.py, and compile demo.py into a WebAssembly file:

$ cargo run -- demo.py
...  # No time for coffee at this point :(
$ ls demo.wasm

Now you can run this WebAssembly file as usual, using one of the many runtimes, such as nodejs or ppci:

$ node run_demo.js
$ python run_demo.py

Or view the wasm in the browser:

$ python -m http.server
$ open test_page.html in the browser

Since the CorePython compiler is written in rust, you can use the WebAssembly build of CorePython, and use it client side:

$ cd corepython-wasm
$ wasm-pack build
$ cd www
$ npm install
$ npm run start

Performance

Well, okay, that's nice, but how fast is it?

Good question! I did some completely non-scientific early tests with the mandelbrot example (mandel.py). Compiled it to WebAssembly and ran it with both node and ppci.

• Python version: 17 ms.
• Python -> WebAssembly -> native (using ppci): 2.7 ms.
• Python -> WebAssembly -> node: 20 ms.

You can try this using the run_mandel.js and run_mandel.py scripts:

$ cargo run -- mandel.py  # Compile mandel.py
$ python run_mandel.py
$ node run_mandel.js

Design topics

Below is a list of python concepts and how they map to WebAssembly. This will also contain idea's behind implementation of various language constructs.

How is Python's int implemented?

For now, it is mapped to WebAssembly i32. Other options are i64 or support infinite size integers (how?).

How is Python's str implemented?

Open topic. TODO, check javascript bindgen for rust?

How is Python's float implemented?

It is mapped to f64.

How is Python's list implemented?

This is an open topic. Initial idea is to go for Python's new (3.9+) syntax:

def myFunc(x: list[int]):
    y = [1, 2, 3]  # Type of y will be list[int]

Representation of the list in memory is a single i32 with the length of the list, followed by the elements of the list. A list object is passed around as a single i32 value which points to the memory where the list is residing.

This might change in the future, when support for list extending is added. Open issues:

• How to append items? Reallocate the memory?

How are Python's magic functions like eval, exec, sys.setprofile implemented?

They are not implemented. In order to stick to the essence of what the Python language is, those function are not available. If it cannot be compiled to WebAssembly, it cannot be supported.

How is Python's class supported?

This is an open topic. Classes will be analyzed at compile time, and there fields and methods will be determined. Then they will be layed out in memory in some way.

Planning

• Python int support.
• Python float support.
• function def support.
• Python str support.
• Python list support.
• Python class support.
• The rest (TM).

Motivation

This is a prototype to answer this question: https://snarky.ca/what-is-the-core-of-the-python-programming-language/