Introduction.

Package pre-commit, as stated on the github-page, is a framework for managing and maintaining multi-language pre-commit hooks.

Mostly, it’s used for managing git hook right before commit to run some static checks or formatters.

In my opinion, this package is often misused and I want to show you the alternative way.

Problems of pre-commit.

1. It has separate configuration file with tools versions differ from pyproject.toml and this is developer’s problem to sync them.
2. It uses separate virtual environments which is not very intuitive and can lead to unexpected behaviour.
3. For mypy if we want to check types of external packages, pre-commit by-default is not able to use dependencies from pyproject.toml and we need to list them in config file in additional_dependencies section.

Alternative solution.

1. Using command runner

First of all, I recommend using some command runner, like make, task or just. I prefer just because it has the most convenient parametrised commands, like:

test *args:
    uv run pytest {{ args }}

And this command can be used with any arguments: just test -x -s -vvv

You can describe project commands in configuration file for running tests or static checks or any other routine, like creating database migration.

For example, here is a command from my projects with static checks:

lint:
    uv run ruff format
    uv run ruff check --fix
    uv run mypy .

2. Running just by pre-commit

If you need to run static checks after committing changes to repo, pre-commit can run “just”-commands. Here is the example of .pre-commit-config.yaml

repos:
- repo: local
  hooks:
  - id: lint
    name: lint
    entry: just
    args: [lint]
    language: system
    types: [python] 
    pass_filenames: false

Now run pre-commit install to install pre-commit hook and before commiting just lint command will be called

Summary

By this way we are solving all the problems: we run exactly the same commands in exactly the same environment with all dependencies.