Skip to content

Latest commit

 

History

History
54 lines (36 loc) · 1.84 KB

README-pyproject.md

File metadata and controls

54 lines (36 loc) · 1.84 KB

Migration from setup.py to pyproject.toml

This project has been migrated from a classic setup.py configuration to a modern pyproject.toml configuration. Here are the key changes and how to work with the new setup:

Package Structure

  • pyproject.toml: Main configuration file for building, dependencies, and metadata
  • setup.py: Minimal file that provides compatibility with older tools
  • Automated versioning using setuptools_scm based on git tags

Building and Development

For development:

pip install -e .

For building:

pip install build
python -m build

Conda Build

The conda recipe has been updated to work with the new pyproject.toml structure. You can build a conda package using:

conda build conda.recipe

Version Management with setuptools_scm

The project now uses setuptools_scm for automatic version determination from git tags:

  1. Tag your release: git tag 2025.05.23
  2. The version will automatically be determined from the git metadata
  3. Format: {tag}.post{commits}+{hash}.d{date} (e.g., 2025.05.23.post2+g7250da5.d20250526)

Dependencies

All dependencies are now specified in pyproject.toml. Make sure to keep both the pyproject.toml and conda.recipe/meta.yaml files in sync when adding new dependencies.

CI/CD with GitHub Actions

The GitHub workflow has been updated to work with setuptools_scm:

  1. It checks out the complete git history with fetch-depth: 0 to ensure setuptools_scm has access to all tags
  2. It sets the SETUPTOOLS_SCM_PRETEND_VERSION environment variable to ensure consistent versioning
  3. The version is extracted from the git tag and used for both conda and PyPI (when enabled) builds

When a new version is tagged and pushed to GitHub, the workflow will automatically:

  • Build a conda package with the correct version
  • Upload it to the specified Anaconda channel