Update to hatch-based build system and fix CI

[ianthomas23]

This PR switches the build system from jupyter_packaging to the recommended hatch_jupyter_builder, plus some other smaller changes, in an attempt to get CI passing. When it succeeds and this PR is merged, I will deal with other improvements such as bringing dependencies up to date.

Summary of changes made:

• Switch to hatch-based builder.
• Update micromamba github actions.
• Use python -m build to build the package rather than the deprecated python setup.py ...
• Use jlpm rather than yarn for TypeScript package management and building.

With these changes I can build and run the tests locally, and also install from the generated sdist and binary wheel.

[github-actions]

👈 Launch a binder notebook on branch ianthomas23/ipympl/hatch_and_ci

[martinRenou]

LGTM! Thanks for working on this!

Would be nice to setup the jupyter-releaser after that

[martinRenou]

Should we add Python 3.12 here?

[ianthomas23]

Yes, I don't know why I stopped at 3.11.

We should probably decide on what Python versions we are supporting, heading toward SPEC0 eventually. Maybe 3.9 to 3.12 is appropriate now. Then I could ensure that CI is running using both the earliest and latest Python versions that we claim to support.

[martinRenou]

Sounds good 👍🏽

[ianthomas23]

Would be nice to setup the jupyter-releaser after that

I'll take a look at jupyter-releaser.

[martinRenou]

Bot update galata references

[martinRenou]

The bot is failing for some other reason. Anyway I believe the CI failures are unrelated. It seems the visual regression diff is super small and probably due to some update on the matplotlib side. I can look into getting this green separately.

[ianthomas23]

I am going to look at excluding text from the Matplotlib plots. This is done in Matplotlib CI as there can be tiny differences depending on platform, OS and compiler.

[ianthomas23]

Excluding text from the Matplotlib test images looks successful so far, just need to wait for the last 2 CI runs.

Things left to do in future PRs:

• Look into jupyter-releaser
• Fix update_galata_references.yaml

In other words look at all github actions other than main.yml.

[martinRenou]

Thank you! I guess this PR is good to go as-is!

Fix update_galata_references.yaml

The fix could probably be just https://github.com/jupytercad/jupytercad/blob/main/.github/workflows/update_galata_references.yaml#L31

[martinRenou]

Are we missing an os entry here? I wonder if this is the reason why the workflow does not start

[ianthomas23]

Yes, I've corrected this now.

Could you cancel the old CI runs as I don't have permissions to do that?

[ianthomas23]

Thanks!

[ianthomas23]

I changed the test that installed from sdist from python 3.12 to 3.11 so now it passes. With python 3.12 it was trying to compile zmq from source. I don't think that should be necessary, but it is not a problem for us to solve today.

[martinRenou]

I changed the test that installed from sdist from python 3.12 to 3.11 so now it passes. With python 3.12 it was trying to compile zmq from source. I don't think that should be necessary, but it is not a problem for us to solve today.

We may be able to workaround this issue by installing from source with --no-deps and making sure all dependencies are installed using conda/mamba from conda-forge.

Thank you for getting this green! Merging and we can iterate

[martinRenou]

Thank you!