Speed up rye test

[dylwil3]

Pytest is wonderful but it can be slow at times. Is it on Astral's roadmap to improve the experience around testing in Python?

That's probably a big project, but I wonder if there is some low-ish hanging fruit around just the collection phase?

Certainly just identifying the functions to be tested would be faster leveraging what Ruff already has available after scanning the package - maybe that alone would give a little boost.

A bigger boost could be obtained if there was a smarter way to deal with module imports - those, I believe, are the next big culprit for time spent in the collection phase.

Finally, as long as one is still tied to pytest for post-collection, there is the overhead of packaging up what you've collected into an object suitable for pytest's consumption, which hopefully doesn't cancel out the above two performance gains.

Anyway- have folks thought about this?

Apologies if this is not the appropriate place for this discussion.