Hacker News Reader

Comment by the_mitsuhiko | Hacker News Reader

I’m a bit confused why uv is not an option for you. You don’t need to compile Python, it manages virtualenvs for you, you can use them with Jupyter and vscode. What are you missing?

View on HN

So the only difference is that Conda also isolates "system" libraries (like libcublasLt.so), or does uv also do this?

It's not that uv is not an option for me, I made this move to miniforge before uv was on my radar because it wasn't popular, but I'm still at a point where I'm not sure if uv can do what I need.

Try Pixi (https://prefix.dev/). It uses uv for Python while also managing your other libraries from Conda. It has a migration path from Conda.

According to these docs

https://docs.astral.sh/uv/pip/environments/

I think uv supports conda envs

uv does not ship system libraries because pypi does not have them. There is a philosophical difference between pypi and conda today. I believe over time pypi will likely ship some system libraries but we will see.

So uv is resticted to pypi but does offer isolated Python installations, with precompiled Python binaries?

Yeah it has precompiled Python binaries. They're custom standalone builds of CPython: https://github.com/astral-sh/python-build-standalone

You can also import existing Python versions into uv, for example on my Mac it has imported the Homebrew versions.

It doesn’t “restrict” to pypi but it wants to be rooted in pypi. That means if you install “tensorflow” you get it from there. The state of pypi is the state of pypi and I have some hope that this also will improve. See for instance the efforts that go into “wheel next”.

Reply on news.ycombinator.com