Add a .pre-commit-hooks.yaml to enable hooks to refer to this repo

The current documentation suggests to use repo: local to use jupytext as a pre-commit hook with pre-commit. However, this requires the user to have jupytext installed in the environment they are working in when committing.

To avoid version conflicts etc, pre-commit usually creates a virtualenv per repo and runs the hooks from there. To do this, relies on hook providers to have a .pre-commit-hooks.yaml file. Adding this here would allow users to use repo: https://github.com/mwouts/jupytext to make a dedicated jupytext environment within pre-commit, instead of relying on users to have it installed in their current env. As an added bonus, jupytext could provide a “menu” of common hooks, if desired (e.g. “always convert .ipynb to .py”, “always unstage .ipynb”, etc). These are configurable, but not required.

[JohnPaton]

I am against the githook actually modifing the git index on its own

I essentially agree. The reason for it in this case is that pre-commit doesn’t consider changes to files if they aren’t in the index, so in the PR we only put them there if they are new, and then fail the hook to alert the user.

I guess an alternative would just be to fail if the outputs aren’t tracked with some kind of warning telling the user what’s up?

[mwouts]

What a great work done at #698! I’ll close this issue now, do some more testing/doc updates at #722, and soon we’ll be ready for a release…