https://github.com/matthewdeanmartin/recython

use ai to assist in turning ordinary python into cython python
https://github.com/matthewdeanmartin/recython

Last synced: 1 day ago
JSON representation

use ai to assist in turning ordinary python into cython python

• Host: GitHub
• URL: https://github.com/matthewdeanmartin/recython
• Owner: matthewdeanmartin
• License: mit
• Created: 2024-01-27T18:45:15.000Z (11 months ago)
• Default Branch: main
• Last Pushed: 2024-02-11T04:17:39.000Z (11 months ago)
• Last Synced: 2024-04-16T04:51:43.930Z (8 months ago)
• Language: Python
• Size: 341 KB
• Stars: 0
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# recython

Use AI to assist in turning ordinary python into cython python.

Don't expect it to work unassisted. It is less ambitious than other coding assistants that try to do it all.

## Why cython?

- Compilation for speed.

- Compilation for obfuscation

- Compilation for calling from c/c++

Recython might help with the above.

- Wrap c/c++ code for calling from python

Recython has no features, yet, for creating wrappers.

## Approaches

### Compile as is with no changes 

This used to require renaming the file to .pyx and then compiling. I think the file renaming is optional.

See build.py, setup.py and pyproject.toml and build.yml for pointers.

This is the least likely to create interesting performance improvements.

One idea is to use cython [compilation as an obfuscator](https://www.youtube.com/watch?v=A1CqUVLda4g).

### Convert to .pxd and .pyx files

The bot gets a file by file prompt to create pyx and pxd headers using cython syntax. You can skip over problematic files.

### Convert to pure python cython

This differs from compiling as-is, in that here we import cython and use cython decorators and other patterns to 

recreate what the cython special syntax was doing in pyx files.

[Pure python cython](https://cython.readthedocs.io/en/stable/src/tutorial/pure.html) is a somewhat new approach and the bots are less familiar with it.

You import cython and use certain decorators and cython type annotations to get the performance improvements.

## Example

```python

from pathlib import Path

from dotenv import load_dotenv

import recython

load_dotenv()

project = Path("../examples/src_hangman/hangman")

recython.cythonize_classic_project(

    folder_path=project, target_folder=Path("./classic/"), never_translate=["__init__", "tests"]

)

```

## TODO

- Add config section to pyproject.toml

- Add a whole condensed manual to the prompts.

- switch between completion and chat models

- switch for particular model

- autoload .env file

- Attempt a compile and give the bot the error messages

- Code review bot.

- Log commentary as it is stripped off

- Maybe provide option to run a continuous conversation for additional context.

- Add a system prompt