Brian C. Keegan, Ph.D.Awesome post by @emilydoesastro about alternatives like #pyenv to the enshittification of #conda #anaconda #python https://emilydoesastro.com/posts/240820-pyenv/
Martin Fleischmann@bkeegan @emilydoesastro @mszll if you need access to conda-forge packages, which in certain scientific files you simply do, then none of these is an alternative to conda. But (micro)mamba or pixi are. Those should be on the top of the list when offering replacement for Anaconda or conda. We don’t use conda because it handles environments for us but because it allows us to resolve complex environments based on conda-forge project.
Emily Hunt@martinfleis @bkeegan @mszll that's very fair - how many packages actually need Conda, though? At least for me, I haven't needed Conda in six years of being in science (and I imagine most people in the astronomical data science world don't either). If anything, the incompatibility between pip and Conda has always been a big issue for me, as there are a lot of small and specialist tools not on Conda distros.
@martinfleis @bkeegan @mszll I think I wrote the blog post aimed at the Conda users I know in astro - most people use it and just use numpy, matplotlib, astropy, and pandas; I think it's complete overkill for them.
Also, conda-forge is hosted by Anaconda still, which costs them 'about $100,000 a month.' Given their recent behaviour, I worry about them deciding to stop doing it for free at some point if everyone switches to it 🙁 https://stackoverflow.com/a/74762864
@emilydoesastro @bkeegan @mszll that’s fair but it should be explicit. “Hey folks, I know this works in Astro but you mileage may vary”. Geospatial, oceanography and similar still depend heavily on conda-forge. Also, there’s a conda-forge mirror on prefix.dev if you want purely Anaconda-free experience.
John Howard@martinfleis @emilydoesastro @bkeegan @mszll only a dabbler in this but #uv is looking promising. https://astral.sh/blog/uv-unified-python-packaging
@john_howard @emilydoesastro @bkeegan @mszll uv does not solve anything caused by missing wheels or incorrectly found C dependencies when a library is included in more wheels. Which is the whole purpose of conda-forge.
@john_howard @emilydoesastro @bkeegan @mszll uv does a great job if all you need is PyPI. Unfortunately, that’s not always the case. These days it is way way better than it used to be and in most cases, it will just work as @emilydoesastro writes. But some stuff is still tricky and I don’t think it will improve as packaging is hard and super low-priority for many scientists given they’re fine with wheels for Linux only or rely on conda-forge.
Nick Santos@martinfleis
Agreed on geospatial. GDAL is a major dependency for a ton of other packages that for a long time was awful to get working on Windows without conda. Nowadays I skip conda for it, but you still need to know what you're doing (no specific package manager helps). I also turn to conda for operations research/systems optimization libraries that interface with compiled C packages. I prefer not using it when I can, but in some cases, it's a requirement.
Agreed on geospatial. GDAL is a major dependency for a ton of other packages that for a long time was awful to get working on Windows without conda. Nowadays I skip conda for it, but you still need to know what you're doing (no specific package manager helps). I also turn to conda for operations research/systems optimization libraries that interface with compiled C packages. I prefer not using it when I can, but in some cases, it's a requirement.