Disposable Python environments for ad-hoc analyses

ericmjl.github.io

12 points·sebg··4 comments

The Python community appears hostile to thinking ahead; designing stable, durable APIs; and sticking with them. It's gotten to the point where you're lucky if you can solve the dependencies for a single version of a single program.

The response to the resulting chaos has been to normalize setting up a complete custom "environment" for every. Single. Script. And they keep doubling down with more elaborate ways to do that.

Not expecting, or even respecting, discipline is probably one of the reasons the language has gotten so popular.

I don't know what you're supposed to do if you have a security issue. Not using Python for anything important comes to mind.

I don't understand how no language has really figured out dependencies in a way that is truly sharable. It seems like docker is the only way to create applications that can be build and run everywhere without headache.

Agreed. However containers make maintenance much more difficult. A year after a container was deployed, it needs an update. And then you discover there are a set of deprecated libraries that need updating. It can get messy. Not sure any solution is perfect.

Great to have these improvements! However, it seems the overhead to create the shareable files exceeds the utility for most cases. Would love this work continues until we get a one line „pip share“ and „pip load“ type tooling.

Reply on news.ycombinator.com