Why use pre-commit

There are a number of things that CI needs to run:

• pytest

• black

• mypy

• flake8

• isort

The initial approach this module took was to integrate everything under pytest that had a plugin, and isort under flake8:

• pytest

  • pytest-black

  • pytest-mypy

  • pytest-flake8

    • flake8-isort

This had the advantage that a pipenv run tests run in CI would catch and report all test failures, but made each run take longer than it needed to. Also, flake8 states that it does not have a public, stable, Python API so did not recommend the approach taken by pytest-flake8.

To address this, the tree was rearranged:

• pytest

• black

• mypy

• flake8

  • flake8-isort

If using VSCode, this will still run black, flake8 and mypy on file save, but for those using other editors and for CI another solution was needed. Enter pre-commit. This allows hooks to be run at git commit time on just the files that have changed, as well as on all tracked files by CI. The graph now looks like:

• pytest

• pre-commit

  • black

  • mypy

  • flake8

  • flake8-isort

Now the workflow looks like this:

• Save file, VSCode runs black, flake8 and mypy on it

• Run pytest until tests pass

• Commit files and pre-commit runs black, flake8 and mypy on them

• Push to remote and CI runs black, flake8, mypy once on all files, then pytest multiple times in a test matrix