Skip to content

Latest commit

 

History

History
167 lines (130 loc) · 6.51 KB

DEVELOPMENT.md

File metadata and controls

167 lines (130 loc) · 6.51 KB

RustPython Development Guide and Tips

RustPython attracts developers with interest and experience in Rust, Python, or WebAssembly. Whether you are familiar with Rust, Python, or WebAssembly, the goal of this Development Guide is to give you the basics to get set up for developing RustPython and contributing to this project.

The contents of the Development Guide include:

Setting up a development environment

RustPython requires the following:

  • Rust latest stable version (e.g 1.38.0 at Oct 1st 2019)
    • To check Rust version: rustc --version
    • If you have rustup on your system, enter to update to the latest stable version: rustup update stable
    • If you do not have Rust installed, use rustup to do so.
  • CPython version 3.7.4 or higher
    • CPython can be installed by your operating system's package manager, from the Python website, or using a third-party distribution, such as Anaconda.
  • [Optional] The Python package, pytest, is used for testing Python code snippets. To install, enter python3 -m pip install pytest.

Code style

The Rust code style used is the default rustfmt codestyle. Please format your code accordingly. We also use clippy to detect rust code issues.

Python code should follow the PEP 8 style. We also use flake8 to check Python code style.

Testing

To test RustPython's functionality, a collection of Python snippets is located in the tests/snippets directory and can be run using pytest:

$ cd tests
$ pytest -v

Rust unit tests can be run with cargo:

$ cargo test --all

Profiling

To profile RustPython, build it in release mode with the flame-it feature. This will generate a file flamescope.json, which can be viewed at https://speedscope.app.

$ cargo run --release --features flame-it script.py
$ cat flamescope.json
{<json>}

You can specify another file name other than the default by using the --output-file option to specify a file name (or stdout if you specify -). The --output-format option determines the format of the output file. The speedscope json format (default), text, or raw html can be passed. There exists a raw html viewer which is currently broken, and we welcome a PR to fix it.

Code organization

Understanding a new codebase takes time. Here's a brief view of the repository's structure:

  • bytecode/src: python bytecode representation in rust structures
  • compiler/src: python compilation to bytecode
  • derive/src: Rust language extensions and macros specific to rustpython
  • parser/src: python lexing, parsing and ast
  • Lib: Carefully selected / copied files from CPython sourcecode. This is the python side of the standard library.
  • vm/src: python virtual machine
    • builtins.rs: Builtin functions
    • compile.rs: the python compiler from ast to bytecode
    • obj: python builtin types
    • stdlib: Standard library parts implemented in rust.
  • src: using the other subcrates to bring rustpython to life.
  • docs: documentation (work in progress)
  • py_code_object: CPython bytecode to rustpython bytecode converter (work in progress)
  • wasm: Binary crate and resources for WebAssembly build
  • tests: integration test snippets

Understanding Internals

The RustPython workspace includes the rustpython top-level crate. The Cargo.toml file in the root of the repo provide configuration of the crate and the implementation is found in the src directory (specifically, src/main.rs).

The top-level rustpython binary depends on several lower-level crates including:

  • rustpython-parser (implementation in parser/src)
  • rustpython-compiler (implementation in compiler/src)
  • rustpython-vm (implementation in vm/src)

Together, these crates provide the functions of a programming language and enable a line of code to go through a series of steps:

  • parse the line of source code into tokens
  • determine if the tokens are valid syntax
  • create an Abstract Syntax Tree (AST)
  • compile the AST into bytecode
  • execute the bytecode in the virtual machine (VM).

rustpython-parser

This crate contains the lexer and parser to convert a line of code to an Abstract Syntax Tree (AST):

  • Lexer: parser/lexer.rs converts Python source code into tokens
  • Parser: parser/parser.rs takes the tokens generated by the lexer and parses the tokens into an AST (Abstract Syntax Tree) where the nodes of the syntax tree are Rust structs and enums.
    • The Parser relies on LALRPOP, a Rust parser generator framework.
    • More information on parsers and a tutorial can be found in the LALRPOP book.
  • AST: parser/ast.rs implements in Rust the Python types and expressions represented by the AST nodes.

rustpython-compiler

The rustpython-compiler crate's purpose is to transform the AST (Abstract Syntax Tree) to bytecode. The implementation of the compiler is found in the compiler/src directory. The compiler implements Python's peephole optimizer implementation, Symbol table, and streams in Rust.

Implementation of bytecode structure in Rust is found in the bytecode/src directory. The bytecode/src/bytecode.rs contains the representation of instructions and operations in Rust. Further information about Python's bytecode instructions can be found in the Python documentation.

rustpython-vm

The rustpython-vm crate has the important job of running the virtual machine that executes Python's instructions. The vm/src directory contains code to implement the read and evaluation loop that fetches and dispatches instructions. This directory also contains the implementation of the Python Standard Library modules in Rust (vm/src/stdlib). In Python everything can be represented as an Object. vm/src/obj directory holds the Rust code used to represent a Python Object and its methods.

Questions

Have you tried these steps and have a question, please chat with us on gitter.