How to Contribute

Everyone is welcome to contribute to PyTango project. If you don’t feel comfortable with writing core PyTango we are looking for contributors to documentation or/and tests.


A Git feature branch workflow is used. More details can be seen in this tutorial. Good practices:

  • For commit messages the first line should be short (50 chars or less) and contain a summary of all changes. Provide more detail in additional paragraphs unless the change is trivial.

  • Merge requests (MRs) should be ALWAYS made to the develop branch.

reStructuredText and Sphinx

Documentation is written in reStructuredText and built with Sphinx - it’s easy to contribute. It also uses autodoc importing docstrings from tango package. Theme is not important, a theme prepared for Tango Community can be also used.

To test the docs locally (work in a virtualenv):
  • $ cd /path/to/pytango

  • $ python -m pip install -r doc/requirements.txt

  • $ python -m sphinx doc build/sphinx

To test the docs locally in a Sphinx Docker container:
  • (host) $ cd /path/to/pytango

  • (host) $ docker run --rm -ti -v $PWD:/docs sphinxdoc/sphinx bash

  • (container) $ python -m pip install gevent numpy packaging pillow psutil sphinx_rtd_theme

  • (container) $ python -m sphinx doc build/sphinx

After building, open the build/doc/index.html page in your browser.

Source code standard

All code should be PEP8 compatible. We have set up checking code quality with pre-commit which runs ruff, a Python linter written in Rust. pre-commit is run as first job in every gitlab-ci pipeline and will fail if errors are detected.

It is recommended to install pre-commit locally to check code quality on every commit, before to push to GitLab. This is a one time operation:

That’s it. pre-commit will now run automatically on every commit. If errors are reported, the commit will be aborted. You should fix them and try to commit again.

Note that you can also configure your editor to run ruff. See ruff README.

Using Conda for development

For local development, it is recommended to work in a Conda environment.

To run the tests locally (after activating your Conda environment):
  • $ pytest

To run only some tests, use a filter argument, -k:
  • $ pytest -k test_ping

Using Docker for development

Docker containers are useful for developing, testing and debugging PyTango. See the folder .devcontainer in the root of the source repo. It includes instructions for building the Docker images and using them for development.

For direct usage, rather than PyTango developement, Docker images with PyTango already installed are available from the Square Kilometre Array Organisation’s repository.

For example:
  • docker run --rm -ti

Releasing a new version

Starting from 9.4.2 pytango tries to follow cpptango releases with the delay up to ~1 month. The basic steps to make a new release are as follows:

Pick a version number
  • A 3-part version numbering scheme is used: <major>.<minor>.<patch>

  • Note that PyTango does not follow Semantic Versioning. API changes can occur at minor releases (but avoid them if at all possible).

  • The major and minor version fields (e.g., 9.4) track the TANGO C++ core version.

  • Small changes are done as patch releases. For these the version number should correspond the current development number since each release process finishes with a version bump.

  • Patch release example:
    • 9.4.4.devN or 9.4.4rcN (current development branch)

    • changes to 9.4.4 (the actual release)

    • changes to 9.4.5.dev0 (bump the patch version at the end of the release process)

  • Minor release example:
    • 9.4.4.devN or 9.4.4rcN (current development branch)

    • changes to 9.5.0 (the actual release)

    • changes to 9.5.1.dev0 (bump the patch version at the end of the release process)

Check which versions of Python should this release support
  • Follow the version policy and modify correspondingly requires-python, classifiers, and minimum runtime dependencies for NumPy in pyproject.toml.

Create an issue in GitLab
  • This is to inform the community that a release is planned.

  • Use a checklist similar to the one below:

    Task list:
    - [ ] Read steps in the how-to-contribute docs for making a release
    - [ ] Release candidate testing and fixes complete
    - [ ] Merge request to update changelog and bump version
    - [ ] Merge MR (this is the last MR for the release)
    - [ ] Make sure CI is OK on develop branch
    - [ ] Make sure the documentation is updated for develop (readthedocs)
    - [ ] Create an annotated tag from develop branch
    - [ ] Push stable to head of develop
    - [ ] Make sure the documentation is updated for release (readthedocs)
    - [ ] Check the new version was automatically uploaded to PyPI
    - [ ] Bump the version with “-dev” in the develop branch
    - [ ] Create and fill in the release description on GitLab
    - [ ] Build conda packages
    - [ ] Advertise the release on the mailing list
    - [ ] Close this issue
  • A check list in this form on GitLab can be ticked off as the work progresses.

Make a branch from develop to prepare the release
  • Example branch name: prepare-v9.4.4.

  • Edit the changelog (in docs/revision.rst). Include all merge requests since the version was bumped after the previous release. Reverted merge requests can be omitted. A command like this could be used to see all the MR numbers, just change the initial version: git log --ancestry-path v9.4.3..develop | grep "merge request" | sort

  • Find the versions of the dependencies included in our binary PyPI packages, and update this in docs/news.rst.
  • Bump the versions (tango/, pyproject.toml and CMakeLists.txt). E.g. version_info = (9, 4, 4), version = "9.4.4", and VERSION 9.4.4 for a final release. Or, for a release candidate: version_info = (9, 4, 4, "rc", 1), version = "9.4.4.rc1", and VERSION 9.4.4.

  • Create a merge request to get these changes reviewed and merged before proceeding.

Make sure CI is OK on develop branch
  • On Gitlab CI all tests, on all versions of Python, must be passing. If not, bad luck - you’ll have to fix it first, and go back a few steps…

Make sure the documentation is updated
Create an annotated tag for the release
  • GitLab’s can be used to create the tag, but a message must be included. We don’t want lightweight tags.

  • Alternatively, create tag from the command line (e.g., for version 9.4.4):
    • $ git checkout develop

    • $ git pull

    • $ git tag -a -m "tag v9.4.4" v9.4.4

    • $ git push -v origin refs/tags/v9.4.4

Push stable to head of develop
  • Skip this step for release candidates!

  • Merge stable into the latest develop. It is recommended to do a fast-forward merge in order to avoid a confusing merge commit. This can be done by simply pushing develop to stable using this command:

    $ git push origin develop:stable

    This way the release tag corresponds to the actual release commit both on the stable and develop branches.

  • In general, the stable branch should point to the latest release.

Upload the new version to PyPI
  • The source tarball and binary wheels are automatically uploaded to PyPI by Gitlab CI on tag.

Bump the version with “-dev” in the develop branch
  • Make a branch like bump-dev-version from head of develop.

  • In tango/, change version_info, e.g. from (9, 4, 4) to (9, 4, 5, "dev", 0).

  • In pyproject.toml, change version, e.g. from "9.4.4" to "9.4.5.dev0".

  • In CMakeLists.txt, change VERSION, e.g. from 9.4.4 to

  • Create MR, merge to develop.

Create and fill in the release description on GitLab
  • Go to the Tags page:

  • Find the tag created above and click “Edit release notes”.

  • Content must be the same as the details in the changelog. List all the merge requests since the previous version.

Build conda packages
Advertise the release on the mailing list
Close off release issue
  • All the items on the check list should be ticked off by now.

  • Close the issue.