Skip to content

Latest commit

 

History

History
166 lines (122 loc) · 6.6 KB

README.md

File metadata and controls

166 lines (122 loc) · 6.6 KB
Raw

matrix

GitHub Action that returns a dynamic test matrix. Currently, it supports projects using:

  • python and tox

Supported optional arguments:

  • min_python - Minimal version of python to be tested against, default is "3.8". The maximum value is currently "3.14"
  • other_names- A list of other tox environments to include in the matrix. We plan to read them from envlist field in tox.ini.
  • platforms - Default to linux only but can be linux, windows, macos or a combination of them (comma separated).
  • linux: matrix expansion strategy for Linux, full or minmax.
  • windows: matrix expansion strategy for Windows, full or minmax.
  • macos: matrix expansion strategy for MacOS, full or minmax.
  • skip_explode: pass 1 if you want to avoid generating implicit pyXY jobs.

Upgrading action from v2 to v3

The returned tox environment name from returned passed_name was replaced by command which would now contain the full command to be executed like tox -e py39.

# old v2 syntax:
- run: tox -e ${{ matrix.passed_env }}
# new v3 syntax:
- run: ${{ matrix.command }}

Returned values

This action returns a list of actions to be executed, each of them containing the following fields:

  • name of the job to run

  • command, and optional command2, command3, ... which are the commands to be executed using run: step.

  • python_version is a string compatible with the expected format used by actions/setup-python github action, like 3.12 or 3.11\n3.12 when multiple python versions are to be installed.

  • os the name of an github runner, should be passed to runs_on:

Examples

Simple workflow using coactions/dynamic-matrix

# .github/workflows/tox.yml (your workflow file)
---
jobs:
  pre: # <-- this runs before your real matrix job
    name: pre
    runs-on: ubuntu-24.04
    outputs:
      matrix: ${{ steps.generate_matrix.outputs.matrix }}
    steps:
      - name: Determine matrix
        id: generate_matrix
        uses: coactions/dynamic-matrix@v3
        with:
          other_names: |
            lint
            pkg
            py39-all:tox -f py39
            py39-arm64:tox -e py39:runner=ubuntu-24.04-arm64
          # ^ job-visible name : optional command : optional arguments
          # command can use ; as separator to run multiple commands
          # the only recognized argument is now 'runner'

  build:
    name: ${{ matrix.name }}
    runs-on: ${{ matrix.os || 'ubuntu-24.04' }}
    needs: pre
    strategy: # this the magic part, entire matrix comes from pre job!
      matrix: ${{ fromJson(needs.pre.outputs.matrix) }}

    steps: # common steps used to test with tox
      - uses: actions/checkout@main
        with:
          fetch-depth: 0

      - name: Set up python ${{ matrix.python_version }}
        uses: actions/setup-python@v4
        with:
          python-version: ${{ matrix.python_version }}

      - name: Install tox
        run: |
          python -m pip install -U pip
          pip install tox

      - run: ${{ matrix.command }}

Q&A

Which projects using tox would benefit from this GitHub Action?

If your tox envlist is simple, like lint,packaging,py{36,37,38,39} you are among the best candidates to make use of it as that is the primary usage case it covers. If you use environments combining multiple factors, you will need to specify them in other_names argument.

Why this action does not just load envlist values?

We plan to add support for this in the future but it might not be as simple as one would assume. For historical reasons, envlist does very often already include Python versions instead of generic py entry or they are outdated. The repository code is not available at the time this action runs.

Why only Linux testing is enabled by default?

Linux runners are the fastest ones and many Python projects do not need to support platforms like Windows or macOS. That is why the default platform contains only lines. Still, you can enable all of them by specifying platforms: linux,windows,macos in the action arguments.

Why Windows and MacOS matrix expansion strategy is different than the Linux one?

The defaults for macOS and Windows are minmax while for Linux is full. This limit resource usage low while still providing a good level of testing. If your pythons are py38,py39,py310,py311 unless you specify windows: full you will see only two Windows based jobs in the generated matrix: py38 and py311.

Why is other_names a multiline string instead of being a comma-separated one?

We wanted to allow users to chain (group) multiple tox environments in a single command like tox run -e lint,packaging, and this means that we needed to allow users to use commas as part of a valid name, without splitting on it.

How to use custom test commands for some jobs.

In v3 we allow users to add entries like py39-all:tox -f py39 inside other-names. This would be translated into returning the job name py39-all and the command tox -f py39.

This is especially useful as it allows users to make use of labels (-m) and factor filtering (-f) to select groups of tox environments instead of just using the environments (-e) selector.

This allows running other test frameworks instead of tox.

Generating multiple test commands for the same job

In some cases, you might want to have separate test steps inside the same job, as this makes it easier to debug them. As GHA does not have any support for step looping, you are forced to manually add several steps if you want to use this feature.

Use ; as a separator inside the other_names to achieve this:

# Return two commands instead of one for `all-tests` job
uses: coactions/dynamic-matrix@v3
with:
  other_names: |
    all-tests:tox -e unit;tox -e integration
  # ^ job-name ':' 1st command ';' 2nd command ...
---
# Inside matrix job:
steps:
  - run: "${{ matrix.command }}"
  # Execute second command *if* it does exist
  - run: "${{ matrix.command2 }}"
    if: "${{ matrix.command2 || false }}"

How to use custom test commands for some jobs.

In v3 we allow users to add entries like py39-all:tox -f py39 inside other-names. This would translated into returning the job name py39-all and the command tox -f py39.

This is especially useful as it allows users to make use of labels (-m) and factor filtering (-f) to select groups of tox environments instead of just using the environments (-e) selector.

This allows running other test frameworks instead of tox.