tbenthompson / okada_wrapper

MATLAB and Python wrappers of the Okada Green's functions.
MIT License
67 stars 33 forks source link

Request for help: Upgrade `okada_wrapper` to work with Python 3.12 and up. #8

Closed tbenthompson closed 5 months ago

tbenthompson commented 6 months ago

numpy.distutils has been deprecated and this will kill okada_wrapper as currently written

See the discussion and explanations here:

If someone wants to migrate okada_wrapper to build with scikit-build/CMake or Meson as explained in the link above, I would be happy to give you assistance. But I don't have time to lead that effort.

dxiao commented 6 months ago

I made a partial attempt at migrating to meson in #9, but it doesn't pass tests yet. I kinda did this on a lark and over a spare evening and that evening is now over, so if someone sees this and wants to take the last mile, I'm sure there are other folks out there that would appreciate it :)

tobiscode commented 6 months ago

Happy to help @tbenthompson. I use okada_wrapper for simple point sources in my DISSTANS package so there's no need for that to upgrade to TDEs. I think I have to migrate that package to meson anyway as well, and I should be able to find some spare time starting next week.

@dxiao in the meantime, could you post the error messages you got / which tests you were running?

Cheers

dxiao commented 6 months ago

Hey tobiscode!

You should see it in the description of the PR I linked above (#9). It's just when running the test_okada.py script provided in the repo.

tobiscode commented 5 months ago

Cross-linking my PR: https://github.com/tbenthompson/okada_wrapper/pull/11

tobiscode commented 5 months ago

Hi @tbenthompson, it turns out I overlooked something. While test_okada.py works, it does so because it ends up using a local import for okada_wrapper. Globally the package is installed as DC3D, so I have to figure out how to force the original package name. Could you please reopen the issue, and I will submit a new PR as soon as I can?

Also, in the meantime, what would you like the new version number to be? 24.03.22.1, e.g.?

Cheers

tbenthompson commented 5 months ago

Ah, good catch! Thanks! Glad I hadn't done the PyPI release yet.

I think 24.03.22 should be good. I just added you as a collaborator on the project so you should have full permissions to merge PRs, etc now.

tobiscode commented 5 months ago

Thanks, just saw it! Alright I'll go ahead and merge the updates. It should all work now (tested both reinstalling locally and from my GitHub in an old environment with Python 3.11 as well as installing locally in a new environment with 3.12) but it would probably be good to double-check on your end before you push it to PyPI.