Hacker News Reader

I’m really curious how places with heavy Python usage solve the dependency tree, packaging, distribution, release problems.

I’m always lost about what the best practice is when I occasionally need to write Python. I just wish there was one way to do it all.

View on HN

We have a lot of python code, along with C++ and Java, and it is an absolute nightmare dealing with python. If we include conda-forge as a channel, dependency resolves can take hours with the right/wrong combination of packages. We can't easily replicate environment across machines, and build times for our conda packages can take tens of minutes, just to copy a few py files around. We hate it, but our ecosystem is built on it (data science/ML/etc), so it is what we have to deal with. I always look at these threads with hope, but then envision the nightmare that would be involved in moving to poetry or some other system.

Reply on news.ycombinator.com