Development

Setup

Docker

You can use the provided Dockerfile for running tests and static typing & linting hints via containerized LSP service.

Shortcut to run tests:

make docker-tests

Running PyLSP server:

docker build -t relaton-py-lsp . && \
docker container run \
  --interactive --rm --network=none \
  --workdir="$(pwd)" --volume="$(pwd):$(pwd):z" \
  relaton-py-lsp

Virtual environment

Create a virtual Python 3.10 environment and pip install -r requirements_dev.txt -r requirements.txt within it.

Expanding Relaton model coverage

  • Consult Relaton specs (LutaML models, RNC grammar), and available Relaton YAML bibliographic data sources for the kinds of data we deal with.

    Any new information you want to add to bibliographic item model must already exist in the LutaML spec.

  • For consistency, any fields and types you define must use field names that correspond to those defined in LutaML.

  • If specifications conflict (e.g., LutaML and RNC define different types of a property, which sometimes happens), file a consolidation issue and confirm with maintainers which specification should take precedence.

Process

  • Update documentation, describing what’s being added and any potential backwards compatibility issues. Ideally, do this even before you change the codebase.

  • Always run mypy (preferably, configure your IDE to run it automatically) to check types.

  • You’re welcome to run flake8, though in cases where default flake8 configuration obviously differs from project conventions it’s recommended to stick to the latter.

Marking new release

Pick the next version. Follow semantic versioning guidelines: post-v1, increment major version in case of backwards-incompatible changes.

  1. Update the __version__ string in relaton/__init__.py to your chosen version.

  2. Ensure you updated the documentation, including describing the new version in Changelog.

  3. Tag new release in Git as the same version and push tags:

    git tag -s "v0.0.0" -m "Short message"
git push --follow-tags

  4. The repository is set up to build and publish to PyPI automatically on matching tag push.