pkpy is a lightweight(~15K LOC) Python 3.x interpreter for game scripting, written in C11.

It aims to be an alternative to lua for game scripting, with elegant syntax, powerful features and competitive performance. pkpy is extremely easy to embed via a single header file pocketpy.h, without external dependencies.

Please see https://pocketpy.dev for details and try the following resources.

• Live Python Demo: Run Python code in your browser

pkpy should work on any platform with a C11 compiler. These platforms are officially tested.

• Windows 64-bit
• Linux 64-bit / 32-bit
• macOS 64-bit
• Android 64-bit / 32-bit
• iOS 64-bit
• Emscripten 32-bit
• Raspberry Pi OS 64-bit

You have two options to integrate pkpy into your project.

Download the pocketpy.h and pocketpy.c on our GitHub Release page. And #include it in your project.

Clone the whole repository as a submodule into your project, In your CMakelists.txt, add the following lines:

add_subdirectory(pocketpy)
target_link_libraries(<your_target> pocketpy)

See CMakeLists.txt for details.

It is safe to use main branch in production if CI badge is green.

To compile it with your project, these flags must be set:

• --std=c11 flag must be set
• For MSVC, /utf-8 flag must be set
• NDEBUG macro should be defined for release build, or you will get poor performance

For amalgamated build, run python amalgamate.py to generate pocketpy.c and pocketpy.h in amalgamated/ directory.

#include "pocketpy.h"
#include <stdio.h>

static bool int_add(int argc, py_Ref argv) {
    PY_CHECK_ARGC(2);
    PY_CHECK_ARG_TYPE(0, tp_int);
    PY_CHECK_ARG_TYPE(1, tp_int);
    py_i64 a = py_toint(py_arg(0));
    py_i64 b = py_toint(py_arg(1));
    py_newint(py_retval(), a + b);
    return true;
}

int main() {
    // Initialize pocketpy
    py_initialize();

    // Hello world!
    bool ok = py_exec("print('Hello world!')", "<string>", EXEC_MODE, NULL);
    if(!ok) goto __ERROR;

    // Create a list: [1, 2, 3]
    py_Ref r0 = py_getreg(0);
    py_newlistn(r0, 3);
    py_newint(py_list_getitem(r0, 0), 1);
    py_newint(py_list_getitem(r0, 1), 2);
    py_newint(py_list_getitem(r0, 2), 3);

    // Eval the sum of the list
    py_Ref f_sum = py_getbuiltin(py_name("sum"));
    py_push(f_sum);
    py_pushnil();
    py_push(r0);
    ok = py_vectorcall(1, 0);
    if(!ok) goto __ERROR;

    printf("Sum of the list: %d\n", (int)py_toint(py_retval()));  // 6

    // Bind native `int_add` as a global variable
    py_newnativefunc(r0, int_add);
    py_setglobal(py_name("add"), r0);

    // Call `add` in python
    ok = py_exec("add(3, 7)", "<string>", EVAL_MODE, NULL);
    if(!ok) goto __ERROR;

    py_i64 res = py_toint(py_retval());
    printf("Sum of 2 variables: %d\n", (int)res);  // 10

    py_finalize();
    return 0;

__ERROR:
    py_printexc();
    py_finalize();
    return 1;
}

Currently, pkpy is as fast as cpython 3.9. Performance results for cpython 3.9 are applicable to for pkpy.

See https://pocketpy.dev/performance/ for details.

And these are the results of the primes benchmark on Intel i5-12400F, WSL (Ubuntu 20.04 LTS), which roughly reflects the performance among c++, lua, pkpy and cpython.

Submit a pull request to add your project here.

All kinds of contributions are welcome.

• Submit a Pull Request
  • fix a bug
  • add a new feature
• Open an Issue
  • any suggestions
  • any questions

If you find pkpy useful, consider star this repository (●'◡'●)

You can sponsor this project via these ways.

• Github Sponsors
• Buy me a coffee

Your sponsorship will help us develop pkpy continuously.

• cpython

  The official implementation of Python programming language.

• byterun

  An excellent learning material. It illustrates how Python's virtual machine works.

MIT License