Kernel performance decrease even if variableInspector's UI not opened

[jhgoebbert]

jupyterlab-variableInspector is a fantastic extension ...
.. but it comes with the drawback that it runs (of course) some extra code in the kernel which can lead to a significant slow down.
https://github.com/jupyterlab-contrib/jupyterlab-variableInspector/blob/main/src/kernelconnector.ts#L94

At the moment, jupyterlab-variableInspector runs by default for every Python and R kernel, regardless of whether the user has opened its UI or not. For this reason, jupyterlab-variableInspector cannot be a default extension for a JupyterLab installation that is used by many different users. It must always be an optional extension as it has high impact on the execution performance.

It would be great if jupyterlab-variableInspector could execute extra commands in the kernel only if the users has opened its UI.

[martinRenou]

Thanks for opening an issue and raising this!

It would be great if jupyterlab-variableInspector could execute extra commands in the kernel only if the users has opened its UI.

This sounds fair, I will look into what it takes to do this.

Although I believe it may not be enough, because if the UI for the variable inspector is opened the issue would appear. The slowdown seems to appear after a huge dataframe has been created in the kernel.
I will also look into the pain points in the case where such a dataframe is created and see if we can improve things.

[martinRenou]

@jhgoebbert would you be able to provide a sample notebook that shows this slowdown in your particular case? That would help me see if I can improve other things

[martinRenou]

I guess we can close this now. The situation should be improved now with #309, #311, #313, #319 and #320

[jhgoebbert]

Hi @martinRenou ,
I am sorry that my answer comes late - I was too busy in the last week until now.
Great, that the situation improved with the fixes you listed 👍