PowerToys makes jupyter notebook so slow with multiple dictionaries

[Fred-Wu]

I have multiple dictionaries with length 10,000 created and each key is associated with a (1, 300) vector. Then suddenly, running cells became so slow on the notebook. It took a few mins to show a single key-value pair, and even took about 4 mins to submit a written function.

So I checked executing notifications, and turned out that it might be caused by PowerToys.

When I submitted a function, and there appeared multiple Executing code running at the same time.

After disabling it, everything went back to normal.

[shism2]

I have the notifications popping up

[shism2]

Restarting the kernel makes it go away. The way for me to reproduce it is to install a kernel and switch tabs from one notebook and back to the same one.

[DonJayamanne]

@Fred-Wu Please could you provide a code sample and steps to repro this issue.

[Fred-Wu]

@Fred-Wu Please could you provide a code sample and steps to repro this issue.

It's been for a while, and it also happens occasionally so I have disabled it.
I will try to find the old codes, and update you later.
Thanks.

[xuxiao2695]

This problem happens to me even without executing any cell. Usually when moving the cursor around or typing out a line from the middle will spawn 20+ "Executing code" notifications. Each can take several seconds to complete.

It looks like contextual help is processing the code under the cursor every time a new key is pressed. The "Code under cursor" shown in the panel actually "replays" me typing the line but very slowly.

[xuxiao2695]

This is especially a problem if the code under the cursor is e.g. a large pandas dataframe. Contextual help will actually evaluate the string form of the dataframe which can take a very long time.