| *** jph5 is now known as jph | 13:45 | |
| opendevreview | Ghanshyam proposed openstack/project-config master: End gate and update acl for retiring openstack-chef projects https://review.opendev.org/c/openstack/project-config/+/909134 | 20:59 |
|---|---|---|
| JayF | I'm not exactly suggesting we do it, but I'm extremely curious if https://astral.sh/blog/uv could have a meaningful impact on runtime for tox jobs | 21:29 |
| JayF | actually, I can probably test that locally... maybe a fun Friday project | 21:30 |
| fungi | i'm following the discuss.python.org topic on it, waiting to see when the obvious bugs are ironed out | 21:31 |
| fungi | JayF: so far someone said extras didn't seem to be working | 21:32 |
| fungi | also concerns were raised that the `uv pip install ...` invocation could lead to user confusion, so i wouldn't assume the cli syntax is set in stone just yet | 21:32 |
| JayF | I mean, even if it's a good idea it's much too early, obviously | 21:34 |
| fungi | but overall it will be interesting to see if there are really any substantial performance gains. i won't be surprised if speedups are just in areas that aren't particularly noticeable, for example when there's no wheel available it's not going to make an sdist build happen any faster | 21:34 |
| JayF | I'm just curious as to if package install/resolution performance is something that matters enough fo rus to even care | 21:35 |
| JayF | exactly | 21:35 |
| clarkb | also constraints makes dep solving much quicker than it would be without constraints | 21:38 |
| fungi | good thing about zuul is speculative testing, you can try to adjust the roles that install and invoke pip and then run other jobs with a depends-on to that series and check for comparative task/playbook/job duration vs what's using pip | 21:38 |
| clarkb | similar to how openstack was also the only thing seeing the pypa index errors we're isolated from a lot of dep solver madness | 21:39 |
| JayF | I'll note similarly, their other tool, ruff, is awesome ... but our use of hacking, and it not supporting all the config styles we use makes it tough | 22:26 |
| JayF | these tools that cover 95% of the use case; we almost always have something in the 5% | 22:26 |
| fungi | such is openstack | 22:38 |
| JayF | I mean, yeah. There's a reason we were out in front on these features: you need 'em to solve the kinda problems you have with literally hundreds of integrated python repos | 22:54 |
| fungi | we were the reason for some of the current features in popular python packaging tools | 22:54 |
| fungi | we don't talk about that enough | 22:54 |
| fungi | for example, pip's constraints feature was implemented by openstack contributors | 22:55 |
| fungi | former and current pypi maintainers were also openstack contributors, and our use cases informed design choices there | 22:56 |
| JayF | I talk about it frequently in channels where there aren't other folks to say it :) | 23:18 |
| clarkb | I'm a fan of trading overly specific things like special hacking rules for someone else writing a tool that is 95% of the way there | 23:58 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!