Were there any significant performance improvements for the

pants generate-lockfiles

goal in recent versions? I’m a few versions behind so if I can improve our generation time with a bump that would be great, before I attempt to finish my uv-based lockfile converter.

A reminder for readers: You can upgrade Pex ahead of upgrading Pants itself (as Joe's message implies).

Thanks both. We have bumped Pex separately from Pants previously, just not to 2.32.0

pip 25.1 has significant resovler changes; but it is unclear under which cases that nets out to faster https://pip.pypa.io/en/stable/news/#v25-1

yeah, we didn't see as big of a jump for 25.1

the biggest speed up we gained with resolvers was pinning as many versions as possible

This is true across a lot of dimensions.

interpreter_constraints

and then the pip installed library versions themselves

the biggest speed up we gained with resolvers was pinning as many versions as possible. The biggest culprit in our case was boto. Pex/pip were checking hundreds of versions during resolution because we had a ‘>=’ version constraint in requirements

Is there a nice way of outputting this info to see if there’s any glaring bottlenecks? Previously we were using

uv

to compile a constraints.txt which is almost instant, with the Pex lockfile generation we’re looking at ~ 12-15 minutes for all requirements.

I'd say the pip.log would be a good place to start.

Relevant? https://github.com/pantsbuild/pants/issues/21979