Support jupyterlite-core 0.2.0a0

[bollwyvl]

References

• fixes Support jupyterlab-core >=0.2.0a0 #57

Changes

• update pins

  • prettier is... undecided on trailing commas for this PR. none of the settings provide a stable output 😿

• collapse more root ./.dotwhatever noise into ./build/.cache
• update all the versions

  • i know the releaser, version bumpers, etc. love me and want me to be happy, but it's really hard to reason about this stuff vs new, nested, semi-undefined yarn/nx/lerna/esbuild/hatch/gha behavior

  • either way, it shouldn't make a difference if we skip releasing a0

  • among these, the widgetsnbextension stuff needs to be backported to 0.1.x, as it is causing active breakage downstream vs ipywidgets 8.1.1
    • [0.1.x] pyodide 0.24.0, widget shim patches #62
    • Backport PR #62 on branch main ([0.1.x] pyodide 0.24.0, widget shim patches) #63
• update docs
  • add compatibility matrix and prerelease install instructions
• ???

Downstreams

The hot sdist and wheel are up on the RTD preview:

https://jupyterlite-pyodide-kernel--58.org.readthedocs.build/en/58/_static/jupyterlite_pyodide_kernel-0.2.0a0.tar.gz
https://jupyterlite-pyodide-kernel--58.org.readthedocs.build/en/58/_static/jupyterlite_pyodide_kernel-0.2.0a0-py3-none-any.whl

[github-actions]

👈 Try it on ReadTheDocs

[jtpio]

Bumping the version is done when making the release with the releaser so we can revert that changes and other bumps in the package.json: https://github.com/jupyterlite/pyodide-kernel/blob/main/RELEASE.md#specifying-a-version-spec

[bollwyvl]

yep, i know: see above. I wanted to actually see something working with the interplay between the new versions.

and the check-release is showing it can handle going up to 0.2.0a1.

[jtpio]

Not sure to understand what bumping the version in this PR helps with?

[bollwyvl]

It is for my understanding, and presumably lerna/yarn/esbuild/webpack's, as i was having issues with cache collisions, pulling the old packages, etc. even after blowing everything away, new envs all that.

also this leaves accurately-numbered wheels/sdist in the RTD build, as i'm trying to do downstream testing as well, due to the limited surface in this test suite and docs site.

and again, the bumper does fine on check-release, correctly going to 0.2.0a1, so won't hurt anybody.

[jtpio]

also this leaves accurately-numbered wheels/sdist in the RTD build

Are they not already "scoped" per sub-domain, for example https://jupyterlite-pyodide-kernel--58.org.readthedocs.build

[bollwyvl]

downstream, i want roughly accurate epochal version numbers for e.g. pip check.

[jtpio]

Should the version bump be reverted if downstream testing is done now?

[bollwyvl]

sorry, too many things broke yesterday to get much more work in. Taking a look now at merge conflicts. I still don't really trust non-trivial packages that span the lab3/lab4 divide, and kind of feel like we're going to inherit problems if this package does.