Ctrl+K

Development guidelines for specific languages and frameworks#

Python#

Setting up a project#

  • To set up a new Python project, use the NIU cookiecutter repository. This ensures that the required tooling is in place (e.g. CI) and also that we are all using a familiar project structure. Part of using this cookiecutter is keeping it updated (e.g. Python versions, best practices) and communicating those changes to all users.

  • Whenever possible, we will use pyproject.toml as the main configuration file for the project. Support for setup.cfg and setup.py will be deprecated in future python versions. This will also be the default option for the SWC cookiecutter.

Virtual environments#

  • We will use conda environments for all Python projects.

  • Package installations will be handled by either conda or pip.

Testing and coverage#

  • All Python software should have automated tests written using pytest that run quickly on all operating systems. Tests that take many minutes on low-powered systems (e.g. GitHub actions) are often ignored.

  • We will use codecov as the coverage reporting tool. This is free for open-source projects and integrates with GitHub actions.

Formatting and QC#

  • We use Black, ruff, and mypy to ensure a consistent code style.

  • Currently, the above tools are automated in the SWC cookiecutter using pre-commit. Running pre-commit install once locally will set up the pre-commit hooks to be executed automatically before each commit.

Continuous integration#

  • All pushes and pull requests will be built by GitHub actions. This will usually include linting, testing and deployment.

  • As a rule, actions will run on all operating systems (Linux, macOS, Windows) and on all Python versions that are supported by the project.

  • GitHub actions workflows should be contributed to the NIU actions repository to aid reuse.

Automated versioning#

We use setuptools_scm to automatically version packages. It should be configured in the pyproject.toml file (as per the cookiecutter template). setuptools_scm will automatically infer the version using git. To manually set a new semantic version, create a tag and make sure the tag is pushed to GitHub. Make sure you commit any changes you wish to be included in this version. E.g. to bump the version to 1.0.0:

git add .
git commit -m "Add new changes"
git tag -a v1.0.0 -m "Bump to version 1.0.0"
git push --follow-tags

Dependency support#

Packages have to choose which versions of dependencies they officially support, with minimum supported versions of each dependency used in continuous integration testing. All NIU projects should follow NEP 29 — Recommend Python and NumPy version support as a community policy standard to determine the minimum set of supported package versions:

  • The last 42 months of Python releases

  • The last 24 months of NumPy releases

In addition to this, the last 24 months of other dependencies should also be supported.

Licensing#

  • Unless the software has commercial potential or depends upon other software with a restrictive license, we will default to using The 3-Clause BSD License (BSD-3-Clause).

  • A LICENSE file should be included in the root of the repository (already present in the SWC cookiecutter).

  • To learn more about our licensing policies, see our licensing guide.

Documentation#

  • We will use Sphinx to generate documentation for Python projects. We will build the documentation websites with the PyData Sphinx Theme.

  • Docstrings should be written in numpydoc format.

  • The documentation structure should be informed by the diataxis framework.

Logging#

All Python software should use the inbuilt Python logging module, rather than e.g. print statements. To standardise this, we will use fancylog. This tool should be updated regularly to reflect best practices and ensure it is suitable for use with all projects.

Configuration files#

User-editable configuration files should use the YAML format. As far as possible, machine-readable outputs from software should also use this format.

Release#

  • All Python software should be released on PyPI to enable a simple pip install. At a minimum, each project should have two owners — Adam (adamltyson) and the lead developer. Others can be added as owners/maintainers as appropriate.

  • We will also aim to release projects on conda-forge, especially for packages with non-Python dependencies.

On this page
Show Source