Packaging improvements, Use pytest fixtures, Support Python3.11

[Zeitsperre]

Hi,

I was looking at what your package needed in order to support Python3.11, and I realized there were a few things that could be improved. There's a lot going on here, so here's a summary:

Changes

• Added package classifiers to better describe pysheds on PyPI, including support on Python3.11.
• Added installer recipes (pysheds[dev] and pysheds[recipes]), to install the necessary packages for running tests and notebooks, respectively.
• Removed several bare exceptions. Now using ModuleNotFound and ImportError where appropriate.
• Update GitHub Actions to their newest version and package-building workflow to support Python3.11 builds

I was having issues with pytest segfaulting as well, so l I also made some significant changes to how pytest tests are configured and run.

Previously, the expectation was that tests are run in sequence, with some tests mutating a base class used further down. This design made it so that all tests are co-dependent, going against that design principles of pytest.

The approach I've done here is to create a few base objects that are instantiated on every test run when they are called (not ideal and can very likely probably be optimized). What this does is make it so that test failures in some tests should not impact subsequent tests. For tests that create temporary files, I've applied the tmp_path fixture so that they do not create artifacts that need to be removed by the user.

My impression from all these changes is that pytest is much quicker to load and no longer segfaults on my local machine.

For more information about pytest fixtures: https://docs.pytest.org/en/7.1.x/explanation/fixtures.html#about-fixtures

I didn't want to go overboard, but if you would appreciate further changes, such as more uniform numpy-docstring conventions, I don't mind continuing with this work in this PR or in another.

If you'd like any clarifications, please let me know.

Best,

[mdbartos]

Thank you for submitting this PR. I should be able to review this weekend.

[mdbartos]

Thanks for taking the time to write this PR @Zeitsperre. I am not quite sure why the tests are not running with GitHub actions in this PR. Any idea why?

[Zeitsperre]

Hi there!

Please excuse the tardiness (I was on vacation). I think there must be a setting in the repo that prevents non-maintainers from being able to run GitHub Actions. Maybe check the permissions?

See also: https://docs.github.com/en/repositories/managing-your-repositorys-settings-and-features/enabling-features-for-your-repository/managing-github-actions-settings-for-a-repository

[mdbartos]

Ok, that last commit seemed to do it.

[Zeitsperre]

Great! I'll fix my merge conflicts tomorrow (I'm still technically on vacation). Best!

[mdbartos]

Hi @Zeitsperre, can you update your summary to reflect the commits added after the PR was initiated? Thanks.

[Zeitsperre]

Hi @mdbartos,

I've updated the description. My last commit undoes the previous commits to try and get the development version of llvmlite (a numba dependency) working. I'm not sure if you squash commits, but if you'd like me to try dropping those commits manually, I can do that.

IMO this PR is ready for proper review. Thanks!

[mdbartos]

Quick request: can you change the version number in pysheds/init.py to 0.3.4?

Thanks, MDB

[mdbartos]

Also, I think the status badges on the readme (and github pages) are outdated. If you could fix with this PR, it would be much appreciated.

[Zeitsperre]

Badge is fixed and the version number seems to be already at v0.3.4!

For the install recipes, I went with dev and recipes (to reflect the requirements needed to run the notebooks under "recipes"). Are these names to your liking?

[mdbartos]

Looks good to me! Ready to merge when you give the go-ahead.

[Zeitsperre]

Great! Yes, LGTM!