Bug Report

Steps to Reproduce:

1. Create a dataframe with a categorical axis.
2. Attempt to load into a PerspectiveWidget or Table.

Expected Result:

A valid widget or Table.

Actual Result:

---------------------------------------------------------------------------
PerspectiveError                          Traceback (most recent call last)
<ipython-i

ipywidgets/packages/jupyterlab-manager/README.md implies JupyterLab 1.0 is the current default:

Prerequisites

• JupyterLab (see package.json for appropriate version, currently JupyterLab 1.0)

ipywidgets/packages/jupyterlab-manager/package.json appears to show otherwise:

  "dependencies": {
    ...
    "@jupyterlab/application": "^2.0.2",

Since PyTorch tensors are very similar to Numpy arrays, is it possible to get the same inspection features for them? Their shape in particular would be great to visualize.

I can work on this if someone can point me in the right direction. I think we just need to add some lines to [inspectorscripts.ts](https://github.com/lckr/jupyterlab-variableInspector/blob/ebad24fad6d421e39674c4fc9fe6b07ae0

At the moment the user has to press Ctrl (or a different modifier key if changed in settings) to bring up the hover tooltip. We could allow specifying a delay as an alternative/additional activation method. This is how it is implemented in VScode I think.

Sketch for setting chema:

"delay": {
  "title": "Hover delay",
  "type": ["number", "null"],
  "default": 300,
  "

When a TF serving model returns a failure instead of a prediction, pass the failure string to the front-end for display instead of the generic http error we see now.

Great extension - thank you.

Could I suggest you place the text from advanced.md in the ReadMe.md file? As it took me a while to get the markdown package working (which requires shell-escape).

I finally found the solution in this closed question: jupyterlab/jupyterlab-latex#112