Interesting, but if I have to use Windows then I would consider Conda depending on my dependency situation.
I don’t particularly like Conda, or Windows, but what I like even less is manually finding wheels for my project. For something like GDAL, I wouldn’t even try on Windows without Conda. I think it’s also easy for a beginner to get up and running with this setup.
My preferred setup is pyenv on Linux with poetry :)
I’d say use WSLv2, myself.
I’ve had a particularly difficult time with CUDA/Pytorch in WSL. Also with Windows not reclaiming memory…
But don’t get me wrong, WSL has helped a lot when I’ve needed to use Windows at work.
Oh my. Yeah. I don’t Windows except to test tools so, that’s not surprising.
At one point I moved from pyenv to conda as I found it easier. May I know what draw backsyou see in conda for my curiously?
In my experience:
- No lockfile, and using the third party conda-lock is clunky
- Painfully slow solver, although the libmamba solver came to the rescue
- Conda-forge can lag behind pypi by weeks, depending on the package
