PIP vs UV: The Future of Python Package Management?

[@JakeCallahan]

Update: I have released an updated review that includes all the progress UV has made since this one. kzbin.info/www/bejne/mKi1mIyuoqikgqs --- I have switched to using uv for virtual environment creation and installation of IPython. With the user-level caching, setup/install went from 5.674s down to 0.451s, and with a complete cache just 0.078s!

[@rafiqrushdan]

impressive

[@0xZarathustra]

issue is tho that it's dependency resolver is not as good as poetry. I don't have a overly complicated setup, but need to downgrade pyyaml during the setup. Poetry does this without issue, `uv pip install pyyaml==5.4.1` just crashes. Well, not only that, the fact that the lock files are not system and platform independent... pretty terrible you gain seconds during the setup of the environment - if it can find the packages that is ... - only to lose all of those seconds and many more hours when you need to ensure compatibility with other systems and platforms 💀

[@JakeCallahan]

The team certainly has areas to improve, and it will be interesting to see how they handle things like resolver optimization. As you saw in the video, it had trouble with one of my team's projects. Thanks for weighing in!

[@felicytatomaszewska]

[EDITED] Yes the new version of UV fixes all the issues shown in this video. So thank you Jake for updating on the topic. I have installed and started test/using it. [OLD] UV is a new shiny tool but apart from installing for simple projects, it cannot compete with the old and tested pip right now..may be in the later versions fix all the issues.

[@JakeCallahan]

Check out my follow-up video linked in the pinned comment. You may be pleasantly surprised at the progress since this video.

[@vortex2210]

Would definitely love to see a comparison between Poetry and uv since all of our projects have been moved to poetry some time back. But this is quite interesting!!

[@JakeCallahan]

That would be a great follow-up once the project is more mature.

[@vortex2210]

@@JakeCallahan yaa, currently I feel the lockfile system works quite well and the envs are easy to manage. I hope uv too gets there soon!

[@dragonkat13]

interesting! look forward to seeing how it unfolds! thank you for covering it! what is your favorite thing about PIP, favorite about UV?

[@JakeCallahan]

pip's compatibility is hard to beat, so it's still the king of production environments. However, I think UV is well positioned to win out in the long term, if Astral can fulfill their vision.

[@B3L13V3R]

I didn't have any of these issues testing the same on an M1 Max in late July 2024. Hope it helps.

[@JakeCallahan]

The team and community has done a great job with uv since then. It is now my daily driver for package management. I'll need to do an update video sometime soon.

[@B3L13V3R]

@@JakeCallahan Awesome! The video is a go-to for the potential for uv.

[@mews75]

I mean the speed is cool but not being able to install all pypi packages is enough for me to not use it

[@JakeCallahan]

It's definitely not mature enough yet, but the team moves pretty fast and has an active community, so hopefully they can get most of the big issues I ran into fixed soon.

[@0730pleomax]

What do you mean by not being able to install all pypi packages?

[@mews75]

@@0730pleomax i haven't tried again since i commented this but when i did try to install the dependencies in some requirements.txt it just said it couldnt install half of them

[@ash1kh]

Thats why i use poetry if i do any python related stuff.