Jupyter: Variables explorer extremely slow for certain data types

[ejohnsson6]

Applies To

• Notebooks (.ipynb files)
• Interactive Window and/or Cell Scripts (.py files with #%% markers)

What happened?

The following cell runs in < 1s and show up instantly in the variables explorer.

import numpy as np

bla = np.random.rand(10000000, 10)

This also runs in < 1s. However, it takes several minutes for bla2 to show up in the variables explorer.

import numpy as np

bla = np.random.rand(10000000, 10)
bla2 = [i for i in bla]

While the variable explorer is loading, all further cell executions are blocked

Logs don't show much. If the varaibles window is closed all is fast again.

I'm geussing the reason it is so slow is because it's fetching more metadata if it's a list and not a numpy array? Would be awesome if it was possbile to disable this for lists above a certain size. Even having the option to disable showing lists and dicts in the variable explorer would be preferable as it currently takes 10 min to rerun cells.

The problem is also present for dicts.

import numpy as np

bla = np.random.rand(10000000, 10)
bla2 = {i[0]: i[1] for i in bla}

VS Code Version

1.91.0-insider

Jupyter Extension Version

v2024.6.2024070901 (pre-release)

Jupyter logs

N/A

Coding Language and Runtime Version

python 3.12.4

Language Extension Version (if applicable)

v2024.11.2024070901 (pre-release)

Anaconda Version (if applicable)

No response

Running Jupyter locally or remotely?

Local

[ejohnsson6]

Adding "jupyter.variableExplorerExclude": "module;function;builtin_function_or_method;list;dict" to the settings is a temp fix to make it usable while working with large lists/dicts. But obviously not ideal.

[kale-ha]

Is this issue solved? I still having trouble for same issue..

[amunger]

It should be resolved in the new variables view that shows up in the debug sidebar with "notebook.experimental.variablesView": true.
I'll re-open this until we start trying to move over to using that view by default.