PyPOTS icon indicating copy to clipboard operation
PyPOTS copied to clipboard

refactor: explicit channels in conda env ymls

Open MaciejSkrabski opened this issue 3 years ago • 9 comments
trafficstars

Hi! You may have noticed that, when creating a new conda environment from *.yml file, it takes ages to solve package dependencies. I attempt to speed the process up by explicitly defining channel in which to search for a package. I also defined minimal pandas version to be 1.4.1 - the things were weird before that. I also allow for python versions newer than 3.7.13 and I believe you'll find it acceptable.

Please let me know if this is in any way helpful.

MaciejSkrabski avatar Jul 22 '22 11:07 MaciejSkrabski

Hi Maciej,

Yes, I noticed that building an environment for PyPOTS does take some time. So your PR looks helpful indeed. Thank you for your idea!

But GitHub auto-testing met errors with your PR. Please check it out here.

WenjieDu avatar Jul 22 '22 12:07 WenjieDu

Yeah, I noticed that so i closed the PR and am working on fixing it :)

MaciejSkrabski avatar Jul 22 '22 12:07 MaciejSkrabski

Cool. Please reopen your PR when you think it's ready. Then we can run a test to see if your commits can speed up the process of environment building.

WenjieDu avatar Jul 22 '22 12:07 WenjieDu

If the tests pass, we could probably speed it up a lot by determining possible version range per package, but I guess that would make us conda dependency resolvers :wink:

MaciejSkrabski avatar Jul 22 '22 13:07 MaciejSkrabski

This commit works fine. You can try to specify the version of each package to see how much it can speed up. 👀

WenjieDu avatar Jul 22 '22 14:07 WenjieDu

I froze the test env, each package with one version explicitly set. If the time gain is not huge I don't think it is worth testing each dependency and searching for minimal requirements.

MaciejSkrabski avatar Jul 22 '22 14:07 MaciejSkrabski

The tests failed, i capitulate :sob: :white_flag:

MaciejSkrabski avatar Jul 22 '22 14:07 MaciejSkrabski

last try with versions that conda found for me with python3.7.13

MaciejSkrabski avatar Jul 22 '22 15:07 MaciejSkrabski

No worries. I'll give your idea a try by myself when I have time 😃. I add this idea in GitHub Discussion, you can check it out here #8 Many thanks for your suggestion and your efforts ;-)

WenjieDu avatar Jul 22 '22 15:07 WenjieDu

I'm closing this PR due to the problem has been solved in the latest release v0.0.10. Please feel free to reopen it if anyone meets the same issue again.

WenjieDu avatar Apr 04 '23 05:04 WenjieDu