Provide way to determine if running in VS Code Interactive Notebook

[MarcSkovMadsen]

I am a user of and contributor to HoloViz Panel. Panel is a framework for interactive exploratory data analysis and for building powerful analytics apps (see https://awesome-panel.org).

Panel builds on the Bokeh ecosystem and Panel provides full interactivity in the VS Code Interactive Notebook environment if I set pn.extension(comms='vscode') at the top of my code. See the Panel - VS Code docs.

My pain is that I and the users I develop notebooks for work in many environments Notebook, Labs, VS Code, PyCharm, Colab and deploy to many servers Bokeh/ Panel, Voila, Flask, Django etc. And the line of code pn.extension(comms='vscode') should not be added in any environment but the VS Code Interactive Notebook environment.

So I need a way to programmatically determine if the code is running in the VS Code Interactive Notebook environment.

Solution

Please tell me how to programmatically from python determine if I am running in VS Code Interactive Notebook or develop something that can tell me.

Thanks

Additional Context

I tried printing sys.modules to see if something vs code python extension specific is loaded. But without luck.

I have also tried Google and Github to find an answer without luck.

Not working with 3.0

Panel is currently not working in VS Code with jupyter >= 3.0 c.f. bokeh/jupyter_bokeh#131. Some js package needs to be deployed to npm.

Panel and VS Code is a super powerful combination

The below is something you can develop with Panel. It can also run interactively inside Notebooks and VS Code. And it even lets you start additional servers via .show() method on Panel objects so that can open and update windows outside of VS Code from the Interactive Notebook. I.e. circumventing the problem of VS Code not really supporting working efficiently with multiple windows".

deckgl_binder.mp4

It also has a powerful --autoreload function for users for hot reloading of .py and .ipynb files. Which makes it super powerful for working with VS Code and Python.

[rchiodo]

Right now the only way to detect this is to look at the environment of the kernel. It should inherit env variables from VS code. For example, a kernel on my own machine will have these values:

We could inject something into the kernel though to make this easier to detect. For example the env variables won't be there if the kernel is remote.

[MarcSkovMadsen]

Thanks so much.

It would be great if there could be something that was 100% certain. So yes. Please inject something so that it can be determined.

If that something that is aligned with Jupyter in general like some kind of way of determining the "front end" (Notebook/ Labs, VS Code, Colab, ...) it would be really cool.

Thanks again.

[brettcannon]

Moved to microsoft/vscode-jupyter#7380