Which Python Virtual Environment Should You Use?

I've found virtualenv to be superior to alternatives. There's a couple of edge cases that venv doesn't cover. Poetry has issues integrating with other tools and has an opinionated setup. Pipenv is a coin flip on whether or not it's gonna break the next time you update it and it's had a ton of bugs.

Arjan thank you so much for this!!

I asked you this very question a couple of days ago!!

I will check out poetry.

I do use venv, but I wasn’t sure what other people use as that was something I stumbled on.

Docker is also on my to-do list and I’m sure I’ll get to it at some point.

I came from venv then had to move to anaconda when I switched to mac and now docker. You're right. It reduces the redundancy since I can already set the nginx in docker too.

Could you please do a full episode on poetry? In particular about using it in conjunction with conda and getting your IDE to understand the full setup.

Peotry doesn't really solve running multiple python versions, so docker is the way to go for me. Also cleaning up your dev environment is easier with docker.

I watched the docker video by you which is great, but it would be nice to see more examples of docker with python in the future videos. Thank you!

Hatch is an amazing python packaging tool which has a great environment management system. I can only recommend taking a look! Getting started is super easy and everything is very intuitiv.

i really think the target audience vs the subject matter is in stark contrast here. people usually watch shorts for entertainment and even if you have people trying to learn from shorts, this kind of stuff is probably way too high level for them and needs more explanation

Conda was what I was trained on, but I moved to venv. Would love to see a docker example/tutorial.

this should be a whole video, not a short! More on topic, though, at my company (one of the biggest in the NL), we use poetry mainly for its ability to solve the version interdependency hell that is managing requirements.txt manually. No more conflicting version numbers - just have poetry figure it out for you. Highly recommended.

No love for conda?

From my humble experience, if you are in scientific computing miniconda saves tons of time due to its ability pull from its registry compiled linear programming solvers, ML frameworks writen in C out of the box. Packages from Pypi for the same modules are usually just wrappers expect you to have the binaries pre-installed and in your path, which is quite hard for a junior data scientist or analyst to wrap their head around.

How do you usually develop inside a container? (That’s what you mean by using Docker, right?) :) in the past I’ve tried that with VS code, but the connection often crashed :/ could you maybe do a short on that? :)

Maybe time for video about virtual env? :) It would be interested.