chore: bump ruff version from 0.3.3 to 0.9.2 and fix some invalid/dead noqas #3282
Conversation
|
Would it be possible to move forward with this? Is there anything I can do to help speeding things up / simplifying the review? |
There was a problem hiding this comment.
Hey @bluenote10!
Thanks for opening the PR and sorry for delayed response.
Thanks for the detailed description for the PR as well, the second option with using X as X pattern is definitely much more preferred.
Everything looks great, I've just left a comment with regards to the conflicts with main branch. I'll run the PR checks after the branch is up to date with main.
@mihran113 Perfect, I've quickly resolved the tiny conflict, so this should be mergable again. |
|
Thanks a lot for the changes @bluenote10! I'm going to merge the PR, there's a failure in unit-tests, but that's not related to your changes, I'm going to fix it separately. |
I noticed that the ruff version was a bit outdated, and that there where quite a lot of spurious/dead
# noqalingering around.This PR bumps ruff to the latest version, which has lead to a minor re-formatting of some code (mainly around fusing string literals).
I've also enabled a few ruff rules that have helped me to identify dead or malformed (
# noqa <code>instead of the needed# noqa: <code>patter) noqa rules.Regarding
F401: Here we should stick to the library interface conventions from the official specs. Otherwise type checkers will not understand the symbols exported in these modules, which is a frequent paint point on user side, when libraries suppressF401. There are basically two options:__all__list manually.X as Xpattern to mark these symbols as explicitly public (and to differentiate them from actual unused imports).I felt that the latter fits better to how Aim maintains its library interface.
My plan is to fully fix all
F401in a follow-up, because the broken library interface is biting us.