Request: pulling latest dictionary in pre-commit hook

[jamesbraza]

I use codespell as a pre-commit hook, and I love it. It's really quite useful.

Normally, a new release of codespell comes with the latest dictionary.txt. I would like to be able to pull the latest dictionary.txt without relying on a codespell release. #116 made the same request, and ultimately made these docs: https://github.com/codespell-project/codespell#updating-the-dictionaries

However, these docs are not applicable to pre-commit users, as the dictionary.txt is ultimately squirreled away somewhere in ~/.cache/pre-commit/:

> find ~/.cache/pre-commit -name "dictionary.txt"
/Users/user/.cache/pre-commit/repo4_vfx6q6/codespell_lib/data/dictionary.txt
/Users/user/.cache/pre-commit/repo4_vfx6q6/build/lib/codespell_lib/data/dictionary.txt
/Users/user/.cache/pre-commit/repo4_vfx6q6/py_env-python3/lib/python3.12/site-packages/codespell_lib/data/dictionary.txt
/Users/user/.cache/pre-commit/repoa0vhi7kx/codespell_lib/data/dictionary.txt

The request is, can codespell expose some CLI flag that triggers codespell to use the latest dictionary.txt on master (or some arbitrary codespell repo's commit ID).

[jamesobutler]

The request is, can codespell expose some CLI flag that triggers codespell to use the latest dictionary.txt on master

@jamesbraza Using main as the target wouldn't be something supported as per https://pre-commit.com/#using-the-latest-version-for-a-repository as the configuration is ultimately supposed to be a repeatable experience.

(or some arbitrary codespell repo's commit ID).

Yes you can already specify to use latest (6ed5df1) as completed by pre_commit autoupdate --bleeding-edge. See below:

 - repo: https://github.com/codespell-project/codespell
-  rev: v2.3.0
+  rev: 6ed5df16e9ddd436d6c093669e6d5e73de7c3dd1
   hooks:
   - id: codespell

[jamesbraza]

Thanks for the suggestion, though it's not totally what I was looking for. I think with bleeding edge:

1. The dictionary.txt always gets better: this is good
2. The core codespell package may get issues on its HEAD branch that would my repo's CI's stability

What I want is a way to get the benefits of item 1 without the downsides of item 2. In other words, I mainly want the latest dictionary.txt, but actually don't want the latest codespell package version (it can stay at rev of v2.3.0).

So I am hoping there can be a better way, perhaps using pre-commit's additional_dependencies, where the dictionary.txt is a standalone PyPI package that gets autoreleased weekly with CalVer