Clarification about actions/setup-python usage and potential gains #724
Description
I saw several open source projects following the recommendation from #207 and combining setup-python and setup-uv. I was wondering if there is any noticable gains with this approach and tested it: ulgens/ruffen-docs#11 Assuming there wasn't any rounding in the result values, it only saves a second in some cases - which I don't think is enough to rationalize the use of a second layer. I do understand some could care about even a second, but I think that a GitHub Actions error margin is above a second so the difference is not meaningful.
Yes the documentations don't match up. I will fix that. In the meantime: It is perfectly fine to use actions/setup-python and might even save ~1s. But for most usecases you don't have to use it. I will make that more clear in the documentation.
It seems @eifinger had a similar take in his comment, but
and might even save ~1s. But for most usecases you don't have to use it.
didn't make it into the docs.
Also, I'm confused about
because GitHub caches the Python versions alongside the runner.
Does GitHub Actions have multiple cache mechanism and uses a "alongside the runner" cache for setup-python but a different one for setup-uv? Do we have any resources explaining what is expressed here?
My question is, can we please further clarify the docs by
- Explaining the potential gain is razor-thin
- For most usecases, it's not necessary
- How "alongside the runner" caching is different than the general behaviour defined here?