Looks like uv released platform agnostic lock files which I believe was a blocker for pex. https://astral.sh/blog/uv-unified-python-packaging . Is there a way to adopt this for package resolution?

There are still parts of pex that uv doesn’t support (subsetting is a big one that comes to mind) - but I still think we could try using uv for things like generating lockfiles + exporting resolves. I have a branch trying to hack some of this together, so hopefully soon 🤞

I'd be curious to see a side-by-side for the main functional requirements. As in, forgetting backwards compat for a moment, could the functionality

pex

be replaced by

uv

at dev time, and maybe rely on pex at package time (ala Vite + Rollup) ... Completely not thought through concept....

I am looking at this for the new Python backend

Is there a way to adopt this for package resolution?

For resolution performance? Or some other reason?

@curved-manchester-66006 Yes, generate-lockfiles is too slow at the moment

Note: There are some upcoming pip improvements which should speed up dep resolution, lockfiles, etc - but no ETA exactly

I would love to do more stuff with

uv

, but probably as an alternative to

pex

👍 I would love some sort of opt-in solution, but that can be very complicated (but kinda future proofs against needing to change back, or forth, or if a new challenger springs up 😆 )