Dataviewer fails when builtins are overridden with user variables

[WillCoulterKorrus]

Applies To

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

What happened?

Pre-amble:
Data viewer fails if you ever re-assign a default class as a variable. There is no warning to avoid doing so, either in "Black", "Pylint", or in the jupyter extension failure error.
This only happens when vs code is launched via github desktop by right clicking the repo name --> "open in VS Code".
The data viewer simply becomes empty if you do the following but open VS Code via anaconda navigator.

Any of these lines placed anywhere in a jupyter .ipynb file cause the data viewer creation error.

list = 1

dict = 3

tuple =3

Expected Result:
I expected to have a new "Data Viewer - var_name" appear as a new window.

Result:
I received the error pop-up, ""Failed to create the Data Viewer. Check the Jupyter tab of the Output window for more info.".
The entire output log of jupyter is pasted below when opening VS Code via github desktop and running on a script within that repo.

Minimally-Reproducible Example:

1. Launch vscode via Github Desktop / Anaconda Navigator (I don't think this matters)
2. Create a two-line .ipynb script

some_list = [1,2,3,4]
list = 2

3. Restart Jupyter kernel
4. Run script
5. Go to "JUPYTER: VARIABLES" window tab
6. Click the "show variable in data viewer" button
   6a) If VS Code was opened via github desktop:
   

6b) If VS Code was opened via anaconda navigator:

VS Code Version

Version: 1.69.2 (user setup) Commit: 3b889b090b5ad5793f524b5d1d39fda662b96a2a Date: 2022-07-18T16:12:52.460Z Electron: 18.3.5 Chromium: 100.0.4896.160 Node.js: 16.13.2 V8: 10.0.139.17-electron.0 OS: Windows_NT x64 10.0.19043

Jupyter Extension Version

v2022.6.1201981810

Jupyter logs

info 14:28:37.558: Execute Cell 0 c:\Users\William.coulter\Documents\example_jupyter_dataviewer_fail.ipynb
info 14:28:37.623: Cell 0 executed with state Success
info 14:29:9.833: Executing silently Code (idle) = def _VSCODE_getVariable(what_to_get, is_debugging, *args):\n# Query Jupyter server for the info
info 14:29:9.875: Executing silently Code (completed) = def _VSCODE_getVariable(what_to_get, is_debugging, *args):\n# Query Jupyter server for the info with 1 output(s)
info 14:29:11.294: Loading webview. View is notset
info 14:29:11.295: Loading web view...
info 14:29:11.296: Webview panel created.
info 14:29:11.381: Process Execution: > ~\Anaconda3\python.exe -c "import pandas;print(pandas.__version__)"
> ~\Anaconda3\python.exe -c "import pandas;print(pandas.__version__)"
info 14:29:11.410: Executing silently Code (idle) = def _VSCODE_getVariable(what_to_get, is_debugging, *args):\n# Query Jupyter server for the info
info 14:29:11.410: Executing silently Code (idle) = def _VSCODE_getVariable(what_to_get, is_debugging, *args):\n# Query Jupyter server for the info
info 14:29:11.437: Executing silently Code (completed) = def _VSCODE_getVariable(what_to_get, is_debugging, *args):\n# Query Jupyter server for the info with 1 output(s)
info 14:29:11.465: Executing silently Code (completed) = def _VSCODE_getVariable(what_to_get, is_debugging, *args):\n# Query Jupyter server for the info with 1 output(s)
info 14:29:11.645: Web view react rendered
info 14:29:12.139: Executing silently Code (idle) = def _VSCODE_getDataFrame(what_to_get, is_debugging, *args):\nimport pandas as _VSCODE_pd\nimp
info 14:29:12.831: Executing silently Code (completed) = def _VSCODE_getDataFrame(what_to_get, is_debugging, *args):\nimport pandas as _VSCODE_pd\nimp with 0 output(s)
error 14:29:12.831: [Error: Failed to fetch variable info from the Jupyter server.
	at g.extractJupyterResultText (c:\Users\William.coulter\.vscode\extensions\ms-toolsai.jupyter-2022.6.1201981810\out\extension.node.js:24:299215)
	at g.deserializeJupyterResult (c:\Users\William.coulter\.vscode\extensions\ms-toolsai.jupyter-2022.6.1201981810\out\extension.node.js:24:299313)
	at getDataFrameInfo (c:\Users\William.coulter\.vscode\extensions\ms-toolsai.jupyter-2022.6.1201981810\out\extension.node.js:24:296311)
	at async r.ensureInitialized (c:\Users\William.coulter\.vscode\extensions\ms-toolsai.jupyter-2022.6.1201981810\out\extension.node.js:39:283280)
	at async r.getDataFrameInfo (c:\Users\William.coulter\.vscode\extensions\ms-toolsai.jupyter-2022.6.1201981810\out\extension.node.js:39:282040)
	at async E.prepDataFrameInfo (c:\Users\William.coulter\.vscode\extensions\ms-toolsai.jupyter-2022.6.1201981810\out\extension.node.js:39:267659)
	at async E.showData (c:\Users\William.coulter\.vscode\extensions\ms-toolsai.jupyter-2022.6.1201981810\out\extension.node.js:39:265433)
	at async f.create (c:\Users\William.coulter\.vscode\extensions\ms-toolsai.jupyter-2022.6.1201981810\out\extension.node.js:39:279718)]

Coding Language and Runtime Version

No response

Language Extension Version (if applicable)

v2022.10.1

Anaconda Version (if applicable)

2.2.0

Running Jupyter locally or remotely?

Local

[rchiodo]

Thanks for the issue.

We might be able to detect this and put up a warning. I think this would more an error for a linter to detect though.

[WillCoulterKorrus]

Thanks for the issue.

We might be able to detect this and put up a warning. I think this would more an error for a linter to detect though.

No problem! I'm migrating from Matlab to Python, so it's a struggle to get all of these pieces (linter, formatting, github repo, visual code studio, anaconda, python interpreter in vs code) all working, let alone understand them fully.

I just saw the error code appear and thought I should mention the headache I went through. I'm using Pylint (because that's most popular I guess), so should I post this issue there?

[rchiodo]

I just saw the error code appear and thought I should mention the headache I went through. I'm using Pylint (because that's most popular I guess), so should I post this issue there?

Yeah that would probably be a better place to capture this. This would mess up other things too (not just dataframe viewing) that would be more generic to all python code.

[DonJayamanne]

I think we should use the builtin alias _VSCODE_builtins instead of just using list, dict, etc. We ran into similar issues in the past and used an alias for referencing these builtins, i think we haven't completed that work, list and a few others are still pointing to the globally imported items, instead of the scoped alias _VSCODE_builtins.
Same applies to str, range & others.

[WillCoulterKorrus]

Yeah that would probably be a better place to capture this. This would mess up other things too (not just dataframe viewing) that would be more generic to all python code.

Ok, I'll post it there. I was surprised to find there was no warning for overriding standard classes!
I don't know how it is causing a problem with vs code jupyter, but we could probably agree that "Failed to fetch variable info from the Jupyter server" is fairly misleading.". Maybe there could be some guidance on how to track down the issue. Maybe there are many paths leading to the error.

Anyway, I'll post on Pylint. I'm impressed with the fast responses!