Update to JupyterLab 2.0

[jtpio]

Fixes #336.
Fixes #208.

Summary of the changes

Here is a list of all the changes, so this PR can also serve as reference when updating other third-party extensions to JupyterLab 2.0

• Update README.md to install JupyterLab 2.0 beta
• Switch to Lumino: Switch to Lumino #208:
  • This is mostly a search / replace across the code base.
  • It is also useful to remove the existing yarn.lock file and node_modules folder.
• Rename iconClassName to iconClass
• Access toolbar from the console MainAreaWidget to avoid special case for the ConsolePanel.
• Fix tests config to handle svg
  • See jupyterlab/jupyterlab@acd909b#diff-e0a6c733b0412c3c723f4e521c837291 for more info
• Handle breaking changes from the kernel refactoring
  • Go through the changes in https://github.com/jupyterlab/jupyterlab/pull/7252/files
  • One notable change: await client.kernel.ready becomes await sessionContext.session.kernel.info to check when a kernel is ready
• Update other dependency version numbers to match JupyterLab

[jtpio]

Getting there :)

[jtpio]

This also requires adding the jest-raw-loader as the transform for svg files.

See jupyterlab/jupyterlab@acd909b#diff-e0a6c733b0412c3c723f4e521c837291 for more info

[afshin]

This is so much nicer ❤️

Thanks @jasongrout!

[jtpio]

There is now a default svg added to the dom when creating a new ToolbarButton:

We might want to investigate using the new JLIcon at some point instead of manipulating the element CSS directly.

[afshin]

Agreed.

[jtpio]

Tests pass locally but are failing on CI. Investigating.

[jtpio]

Updated the summary of the changes in the first comment, so this can serve as reference for other extensions and also help checking the JupyterLab changelog for 2.0.

[jtpio]

There is now a warning logged to the dev tools console:

This is probably a side-effect of the bump to React 16.9 in jupyterlab/jupyterlab#7661, and seems to originate from react-inspector (that defines the TreeView component).

We might want to give a second try at upgrading react-inspector to the latest version (4.0.0) in a separate change.

[jtpio]

Marking this one as ready for review.

Two things to investigate after as follow-ups:

• Switching to and from a "No Kernel" state seems to be broken, and requires changing the current active shell widget to trigger the update. We should handle the kernelChanged signal at the session context level for the notebook and console panels (instead of session connection level).
• Check whether there is a better way to handler kernel restarts (Better way to handle kernel restarts #322)

We can take a look at these two points separately to keep the current PR focused on migrating to JupyterLab 2.0 only.

cc @afshin

[jtpio]

@afshin let's do a 0.1.2 release before merging this one?

[jtpio]

Tests seem to be flakier than before. But this might as well be an issue with the tests and the kernel readiness not being awaited correctly.

[jtpio]

Adding a missing await sessionContext.session.kernel.info like in https://github.com/jupyterlab/jupyterlab/blob/master/tests/test-apputils/src/toolbar.spec.ts#L321 should do the trick.

[afshin]

A clean build is failing for me, have you seen this?

git clean -fdx
pip uninstall jupyterlab
pip install --upgrade --pre jupyterlab
jlpm install
jlpm build
jupyter labextension uninstall @jupyterlab/debugger
jupyter labextension install .
juoyter lab

[jtpio]

To do the upgrade, I created a new conda environment locally from scratch but haven't seen this.

[jtpio]

And also:

rm -rf node_modules/
rm yarn.lock

But that should be covered by the git clean command.

[afshin]

The git clean -fdx removes those files. I will try in a new conda environment to check again.

[afshin]

It works in a brand new conda environment. I wonder what my issue is.

[jtpio]

Good question.

We could try to reproduce a "manual upgrade" if that is worth the effort.

[afshin]

The issue came from ipywidgets. The old version of @jupyter-widgets/jupyterlab-manager was still installed and the JupyterLab 2 beta accepted it but failed at run-time. I think this issue was fixed in a recent PR Steve did, but I'm not 100% sure of that. (cc: @blink1073 @jasongrout)

This fixed it:

[afshin]

Thank you! This looks good and works locally for me.

[afshin]

This PR is included in the JupyterLab 1.x to 2.x migration guide: jupyterlab/jupyterlab#7808