Nathaniel Starkman

GH isn't letting me attach this doc so here's a snippet Details 2025-07-01T17:40:32.2249438Z tests/test_package.py . [100%] 2025-07-01T17:40:32.2250023Z 2025-07-01T17:40:32.2250231Z =================================== FAILURES =================================== 2025-07-01T17:40:32.2251037Z ____________________ frame_transforms.py line=186 column=1 _____________________ 2025-07-01T17:40:32.2251528Z 2025-07-01T17:40:32.2252925Z Example...

Thanks @wesselb, that would be great. I'd love to be using plum with Generic types. Especially in return annotations.

> For dev-only extras, I'm assuming the reason was that no equally convenient way existed to define them without exposing them Yes. IMO fine to move.

See https://results.pre-commit.ci/run/github/2081289/1764606584.fu00TP8jS9SQ90KQ7SxgUA in #19012 -- having a `dev` dependency-group is now officially suggested by the scientific-python repo-review tool!

I personally prefer `copier` and posted the cruft link as an example, though I suppose adding this update feature would necessitate supporting both.

@patrick-kidger, I managed to isolate the problem somewhat and Equinox is involved. It doesn't happen when I use a `dataclasses.dataclass` instead. Here'a MWE that raises a warning ```python from collections.abc...

Sorry @Patrick-kidger, that may hav been a spurious finding. When I updated equinox this wasn't sourced by equinox any more.

I'm still getting the issue everywhere in https://github.com/GalacticDynamics/coordinax/pull/128. It's just not (maybe?) caused by Equinox. I'm suspicious of `jaxtyping.install_import_hook` ATM, but have no proof.

Great. I'll probably use that in `coordinax` so I can unpin `plum`. Then I can continue trying to find a MWE.

Nothing should change. this is all refactoring of a private variable or adding type annotations.