ipycanvas does not work with JupyterLab 4

[martinRenou]

cc. @davidbrochart who spotted this

ipycanvas works by sending binary buffers through custom comm messages.

It looks like there is a difference in behavior between JupyterLab 3 and JupyterLab 4:

• JupyterLab 3: buffers are received as DataView objects
• JupyterLab 4: buffers are received as ArrayBuffer objects

This happens in this.on('msg:custom', (command: any, buffers: any) => {} where buffers is either a DataView[] in Lab 3 or a ArrayBuffer[] in Lab 4.

This might be an issue/change in JupyterLab 4 actually? cc. @jtpio @fcollonval

If it's a wanted change, we should update ipycanvas to support JLab 4

Note: This was all tested using @jupyter-widgets/jupyterlab-manager v5.0.5 and ipycanvas v0.13.1 labextensions

[martinRenou]

It looks like starting JupyterLab's master branch with jupyter lab --dev-mode --extensions-in-dev-mode --ZMQChannelsWebsocketConnection.kernel_ws_protocol='' fixes ipycanvas.

[jtpio]

It looks like starting JupyterLab's master branch with jupyter lab --dev-mode --extensions-in-dev-mode --ZMQChannelsWebsocketConnection.kernel_ws_protocol='' fixes ipycanvas.

This is probably also reproducible with 4.0.0a32 then (could be more convenient for testing than building lab from source).

And likely a change from https://github.com/jupyterlab/jupyterlab/pull/11841 at first glance.

[davidbrochart]

Fixed in https://github.com/jupyterlab/jupyterlab/pull/13730.

[martinRenou]

Thanks a lot!