Skip to content

Latest commit

 

History

History
96 lines (76 loc) · 2.14 KB

README.md

File metadata and controls

96 lines (76 loc) · 2.14 KB
Raw

🦾 codex_py2cpp 🤖

OpenAI Codex Python to C++ Code Generator

Your Python Code is too slow? 🐌 You want to speed it up but forgot how to code in C++? ⌨ Convert your Python script to C++ Code using OpenAI Codex.

How it works

Reads a Python file and creates an input prompt which is then fed to OpenAI Codex to generate corresponding C++ code. The generated code is getting compiled using g++ and if compilation is successful the executable is saved.

To generate your own files you need to get access to the Codex API (https://openai.com/blog/openai-codex/).

Installation

git clone https://github.com/alxschwrz/codex_py2cpp.git
cd codex_py2cpp
pip3 install -r requirements.txt

Run example

Reads the file "simpleScript.py", and feeds the corresponding input prompt to OpenAI Codex. Compilable solutions are stored in the form of .cpp and .exe files.

python3 python2cppconverter.py

If the generated C++ code got compiled, test it with

./simpleScript.exe

You hopefully get the same output as when running

python3 simpleScript.py

Check how much faster you are now ...

time ./simpleScript.exe
time python3 simpleScript.py

Example Code Generation:

[In]: Python Snippet [Out]: How the CODEX conversion might look like 
def add_something(x, y):
    print("casually adding some stuff together")
    z = x + y
    return z


if __name__ == "__main__":
    print('Okay, lets go')
    print(add_something(5, 2))
 
// C++ Code generated from Python Code: 
#include <iostream>
using namespace std;

int add_something(int x, int y) {
    cout << "casually adding some stuff together" << endl;
    int z = x + y;
    return z;
}

int main() {
    cout << "Okay, lets go" << endl;
    cout << add_something(5, 2) << endl;
    return 0;
}

Please test your generated code before usage. This does not produce robust code conversions, but is intended to experiment with codex. [WIP]

Credits

This project is based on the OpenAI Codex project. Inspired by https://github.com/tom-doerr