pythran mit Homebrew, apk, dnf, apt installieren

Projektlinks

• Homepage
• Repository
• Paketmanager-Seite

Befehle und Aliase

• pythran
• pythran-config

Projektgeschichte

Pythran emerged from scientific-Python work on compiling numerical kernels without requiring developers to rewrite entire modules in C, C++, or Fortran. Its 2013 SciPy paper presented a static compiler that translates annotated Python modules into native extension modules.

The project developed into a Python-to-C++17 translator with interprocedural optimization, partial NumPy support, SIMD and multicore facilities, package-build integration, configurable compiler and BLAS settings, and an interface for producing either Python modules or raw C++.

Adoptionsgeschichte

Pythran remained a specialized compiler rather than a general application runtime, but it found a durable role in scientific Python. SciPy began accepting Pythran-based performance improvements and made it a default build dependency in the SciPy 1.7 development cycle, with an opt-out during the transition.

Its distribution through PyPI, conda-forge, Homebrew, and Linux distributions reflects use both as a developer-facing compiler and as build infrastructure behind native scientific-Python extensions.

Wie es verwendet wird

Developers add export declarations or supported type annotations to selected Python functions and invoke pythran to create an importable native module. It is best suited to numerical kernels expressible within Pythran's supported Python and NumPy subset.

Compiler, linker, optimization, and BLAS choices can be stored in .pythranrc. The documented XDG location is $XDG_CONFIG_HOME/.pythranrc, ~/.pythranrc is also documented, and PYTHRANRC can override the selected file; command-line settings take precedence.

Warum Paket-Nerds sich dafür interessieren

Pythran is interesting to package maintainers because it occupies the boundary between Python source, generated C++, compiler toolchains, NumPy APIs, BLAS selection, SIMD libraries, and platform-specific extension modules. Reproducible compilation and compatibility with new CPython and NumPy releases are therefore central concerns.

Unlike a general Python interpreter, Pythran is normally a build-time accelerator: users preserve a Python-facing import interface while packaging native artifacts produced from a statically analyzable subset of the language.

Zeitleiste

• 2013: Pythran is presented to the SciPy community as a static compiler for scientific Python.
• 2021: SciPy 1.7 adopts Pythran as a default build dependency with an opt-out path.
• 2024: Pythran documents Meson integration and improved BLAS detection.
• 2026: Pythran adds compatibility work for CPython 3.15 and continues reproducible-build improvements.

Related projects

• NumPy: the array library and expression vocabulary used by many Pythran kernels.
• SciPy: a scientific-computing library that incorporated Pythran into its build and optimization workflow.
• Cython: another compiler-based route to native Python extensions, using a Python superset and optional static types.
• CPython: the runtime into which ordinary Pythran-generated extension modules are imported.

Quellen

• https://conference.scipy.org/proceedings/scipy2013/serge_guelton.html
• https://docs.scipy.org/doc/scipy/release/1.7.0-notes.html
• https://pythran.readthedocs.io/en/latest/
• https://pythran.readthedocs.io/en/latest/Changelog.html
• https://pythran.readthedocs.io/en/latest/MANUAL.html