Nuitka — Compile Python to Standalone Executables

Introduction

Nuitka is a Python compiler that translates Python source into C code and links it against libpython to produce native executables. It supports the full Python language including dynamic features, while delivering measurable speedups and true standalone distribution.

What Nuitka Does

• Compiles Python modules and packages into C code via a whole-program optimizer
• Produces standalone executables or single-file binaries with all dependencies bundled
• Supports CPython 3.4 through 3.12 and the complete standard library
• Applies constant folding, dead code elimination, and type inference optimizations
• Handles C extensions, data files, and Qt/Tk/PyGame resources automatically

Architecture Overview

Nuitka parses Python source into its own AST, performs optimization passes (constant propagation, escape analysis, type specialization), then emits C source files. These are compiled with gcc or MSVC and linked against the CPython runtime. For standalone mode, Nuitka traces all imports, bundles shared libraries, and optionally compresses everything into a single executable using AppImage or NSIS.

Self-Hosting & Configuration

• Install via pip: pip install nuitka
• Requires a C compiler (gcc on Linux, MSVC or MinGW on Windows, clang on macOS)
• Use --standalone to bundle all dependencies into a dist folder
• Use --onefile to produce a single executable archive
• Configure via nuitka-project: comment directives inside your source files

Key Features

• Full Python compatibility — handles metaclasses, decorators, generators, and async
• Whole-program optimization with measurable 2-4x speedups on CPU-bound code
• Standalone and onefile output modes for easy distribution
• Plugin system for NumPy, Qt, Tk, PyGame, and other frameworks
• Commercial-friendly Apache 2.0 license

Comparison with Similar Tools

• PyInstaller — bundles the interpreter and bytecode but does not compile to native code; no speedup
• cx_Freeze — similar to PyInstaller with cross-platform freeze but no compilation
• Cython — compiles annotated Python to C but requires type annotations for best results
• mypyc — compiles type-annotated Python via mypy but limited to a subset of the language
• PyOxidizer — embeds Python in a Rust binary; focuses on packaging, not optimization

FAQ

Q: Does Nuitka support the full Python language? A: Yes. It compiles all CPython-compatible code including dynamic imports, eval, and metaclasses.

Q: How much faster is compiled code? A: CPU-bound code often sees 2-4x improvement. IO-bound code benefits less since the bottleneck is external.

Q: Can I compile packages with C extensions? A: Yes. Nuitka bundles C extensions as-is alongside the compiled Python code.

Q: Is the output binary portable? A: Standalone binaries include all dependencies and run on the same OS and architecture without Python installed.

Sources

• https://github.com/Nuitka/Nuitka
• https://nuitka.net/doc/user-manual.html