matt

uv: Python packaging in Rust
https://astral.sh/blog/uv
They're aiming to build a cargo-like experience for #python. That will be welcome and I'm hopeful.

It does bring up a question for me though. I primarily use #conda (via #micromamba) for managing projects. How's that going to work with uv, and not?

uv: Python packaging in Rust

uv is an extremely fast Python package installer and resolver, designed as a drop-in alternative to pip and pip-tools.

Feb 21, 2024 at 6:32amWeb
Show threadgramFeb 21, 2024

@maphew

Currently, uv re-implements pip, pip-tools, and venv. Regular python packages that work with pip will work with uv. Conda packages are out of scope, at least at this point.

Show threadmattFeb 21, 2024
@orsinium thanks!
I guess my underlying questioning is if I should be using venv in some of the places I now use conda. An idiosyncratic no-clear view place that no one else can reasonably advise on. LoL
Show threadgramFeb 21, 2024

@maphew

Conda is quite messy and people outside of the scientific world rarely use it. In my long career, I haven't worked on any project that would use conda. We usually solve the question of system dependencies with Docker.

I'd say, if your project has lot of non-Python system dependencies and Docker isn't an option, conda might be a good choice. If not, well, there are slightly better alternatives.

Show threadmattFeb 22, 2024
@orsinium I work amid science although not actually in science (geospatial, arcgis). Primarily in Windows which didn't used to be docker friendly, better recently, but now there's a licensing hurdle. That said, the centrality of conda in my personal projects is more an accident of history than considered decision.
Show threadAlexandre B A Villares 🐍Feb 26, 2024
@maphew @orsinium the thing is, the scientific Python community is huge, it is thanks to them we have numpy, scipy etc... so there really are a lot of people using conda. https://www.youtube.com/watch?v=ZyjCqQEUa8o
Jake Vanderplas - Keynote - PyCon 2017

"Speaker: Jake VanderplasKeynoteSlides can be found at: https://speakerdeck.com/pycon2017 and https://github.com/PyCon/2017-slides"

YouTube
Show threadMarkFeb 21, 2024
@maphew maybe Pixi might be interesting? https://prefix.dev/
7 Reasons to Switch from Conda to Pixi

Pixi is conda-compatible and comes with more speed, lockfiles and tasks. It is the next-gen package manager for Python and R and more.

prefix.dev
Show threadmattFeb 22, 2024
@cofiem ooh, looks interesting 🤔
Show threadRuben ArtsFeb 26, 2024
@maphew @cofiem
We're implementing uv as we speak: https://prefix.dev/blog/uv_in_pixi
7 Reasons to Switch from Conda to Pixi

Pixi is conda-compatible and comes with more speed, lockfiles and tasks. It is the next-gen package manager for Python and R and more.

prefix.dev
Show threadmattFeb 27, 2024

@rarts @cofiem more interesting the more I read: "On a global level, you can install tools that you would otherwise download from apt-get, Homebrew or WinGet - just much faster and with the same tool on all platforms."

So if a tool exists in both Winget and Conda-forge repos preferentially use Pixi? (If it's to be part of a reproducible project or ecosystem)

Show threadRuben ArtsFeb 27, 2024
@maphew @cofiem yep, if conda forge has the tool you can globally install it with pixi or even better lock it into your project. So your coworkers will be guaranteed to also have the same tools available. If that would be too much tools in the environment for instance in CI, you can create multiple environments, to separate concerns and install/download times.