xstrumodeler

Python bindings (Boost.Python) for the xstrucore cross-section modeler.

The repository is self-contained:

Boost 1.88 is downloaded automatically at configure time (CMake FetchContent) and Boost.Python is compiled statically into the module, so no system Boost is required. The wheel depends only on the C++ runtime.

Requirements

• Python >= 3.10 (with headers)
• a C++17 compiler:
  • Linux: GCC or Clang (g++)
  • Windows: Visual Studio 2022 Build Tools with the Desktop development with C++ workload
• network access on first build (≈100 MB Boost download)

CMake is provided by scikit-build-core if not installed.

Build

pip wheel . -w dist          # build a wheel
pip install .                # or install directly

from xstrumodeler import xstrumodeler
modeler = xstrumodeler.XStruModeler()

Faster rebuilds (development)

By default pip builds in a temporary directory, so every build downloads Boost again and recompiles everything (about a minute). To reuse a build directory, install the build tools once and disable build isolation:

pip install "scikit-build-core>=0.11" ninja
pip install --no-build-isolation . -Cbuild-dir=build/{wheel_tag}

Following builds only recompile the changed files (seconds). Build isolation must be off: with it, pip uses a different temporary ninja on each run and the build directory is discarded.

Offline build

Download and extract boost-1.88.0-cmake.tar.xz then:

pip wheel . -w dist -Ccmake.define.FETCHCONTENT_SOURCE_DIR_BOOST=/path/to/boost-1.88.0

Plain CMake (development)

cmake -S . -B build
cmake --build build -j
PYTHONPATH=build python -c "import xstrumodeler; print(xstrumodeler.XStruModeler())"

Here the compiled module is imported directly (import xstrumodeler), while the installed package is from xstrumodeler import xstrumodeler.

Wheels for all Python versions (like CI)

pipx run cibuildwheel --platform linux     # requires Docker
pipx run cibuildwheel --platform windows   # on Windows

Wheels are written to wheelhouse/. Locally cibuildwheel uses its default manylinux image, so Linux wheels may be tagged manylinux_2_28 instead of the manylinux_2_34 built by CI.

Releasing

GitLab CI builds the sdist, manylinux_2_34 wheels (cp310–cp314) and Windows wheels (cibuildwheel):

• on pushes to master and merge requests that change build files (sources, CMake, pyproject.toml, CI), or from Run pipeline in the GitLab UI: build only, to catch problems before a release;
• on version tags: it also checks that the tag matches version in pyproject.toml and, if every build succeeds, uploads them all:

Uploads use Trusted Publishing: no token is stored in GitLab.

Release steps:

1. Set version = "X.Y.ZrcN" in pyproject.toml, commit, push, then git tag vX.Y.ZrcN && git push origin vX.Y.ZrcN.
2. Check the release candidate: pip install -i https://test.pypi.org/simple/ xstrumodeler==X.Y.ZrcN.
3. Set version = "X.Y.Z" in pyproject.toml and CITATION.cff, commit, push, then git tag vX.Y.Z && git push origin vX.Y.Z.

A version can be uploaded only once: if a release fails after uploading, use the next rc number.

One-time setup:

• PyPI → xstrumodeler → Settings → Publishing → add a GitLab publisher: namespace luigi_paone, project xstrumodeler, pipeline file .gitlab-ci.yml, environment pypi.
• TestPyPI → Account → Publishing → add a pending publisher with the same values, project name xstrumodeler, environment testpypi.
• GitLab → Settings → Repository → Protected tags: v*, Maintainers only.

Provenance

Extracted from the exagon repository (commit 5455d2d) with history preserved: src/app/xstrumodeler → repository root, src/lib/coreobj and src/lib/xstrucore → lib/. The copies in lib/ are now independent from exagon; changes are not synchronized automatically.

License

Licensed under the Apache License, Version 2.0. Redistributions must retain the NOTICE file. If you use xstrumodeler in published work, please cite it as described in CITATION.cff.

Boost is distributed under the Boost Software License, Version 1.0.