Recommended way to reliably detect if jupyter is running inside vscode?

[songololo]

When running jupyter inside vscode the tqdm module's auto notebook functionality will break. See this related issue: tqdm/tqdm#747

Hence this question: Is there a recommended (and future-proof) way to reliably detect when jupyter is running inside vscode so that imported modules can be set-up accordingly?

From running os.environ inside a jupyter notebook running inside vscode the following environment variables look like potential candidates for flagging these situations:

'VSCODE_NODE_CACHED_DATA_DIR': '...'
'VSCODE_PREVENT_FOREIGN_INSPECT': 'true',
'VSCODE_LOGS': '...',
'VSCODE_HANDLES_UNCAUGHT_ERRORS': 'true',
'VSCODE_IPC_HOOK_EXTHOST': '...',
'VSCODE_NLS_CONFIG': '{"locale":"en-gb","availableLanguages":},"_languagePackSupport":true}',
'VSCODE_IPC_HOOK': '.../1.34.0-main.sock',
'VSCODE_PID': '94309'

Would one of these, e.g. VSCODE_PID be sufficient for this purpose?

[rchiodo]

Yes I believe VSCODE_PID would only be valid if the jupyter notebook was started from within VS Code.

[songololo]

@rchiodo do you possibly know in what ways the vscode functionality for jupyter works differently from the vanilla version running inside a browser?

I ask from the point of view of trying to understand why the tqdm auto notebook progress bar functionality doesn't work the same way.

[rchiodo]

That's a hard question to answer. It depends.

Some of the differences are CSS styles that we apply to output.
Some of the differences are our VS Code extension isn't handling all of the Jupyter communication correctly.

Do you have some example code for the tqdm auto progress bar? I may have just fixed the tqdm progress bar:
https://github.com/microsoft/vscode-python/issues/5801