Add notebook.workingDirectory setting

[austin3dickey]

Addresses #7988. Adds a notebook.workingDirectory setting and hides the jupyter.notebookFileRoot setting, which doesn't work in Positron.

Release Notes

New Features

• Added notebook.workingDirectory setting, which can be used to set the default working directory for notebooks (jupyter.notebookFileRoot setting has no effect #7988)

Bug Fixes

• N/A

QA Notes

@:notebooks @:vscode-settings @:console @:extensions @:interpreter @:win

I added some unit tests but no e2e tests.

Some things to try:

• The jupyter.notebookFileRoot setting should no longer exist
• Open a workspace, and a new notebook in a subdirectory of that workspace, with a cell containing %pwd, and see its value for different values of the new setting (note you need to close/open the notebook for the change to take effect):
  • (unset)
  • ${workspaceFolder}
  • ${fileDirname}
  • ${userHome}
  • /some/path/that/exists
  • /some/path/that/doesnt/exist
  • whatever else you want to try (variables are defined here)
• The Console and Terminal working directories shouldn't change
• Filename completions should correspond to the correct working directory
• This should work for R (getwd()) or Python (os.getcwd()) notebooks, and for VSC or Positron notebooks

[github-actions]

E2E Tests 🚀
This PR will run tests tagged with: @:critical @:notebooks @:vscode-settings @:console @:extensions @:interpreter @:win

^{readme  valid tags}

[nstrayer]

This is some good stuff. The tests are great and you even remembered to update the dates in the headers! (2025/2025 points).

Some comments on the config setting itself.

[nstrayer]

My only thought is it may get confusion when we're all-in on positron notebooks that this setting is in the original vscode notebooks contrib. I'm not sure we will respect every one of the original vscode notebook config options. I don't think it should go in the positron notebooks one though... So this is probably good for now and then maybe in the future we will split out to a new Switzerland positron(the editor)-specific-notebook-configs section?

[austin3dickey]

yeah, makes sense! We probably will still at least want to use the notebooks. prefix right? (I put the setting here because I thought it should start with that prefix since it only applies to notebooks)

[seeM]

Nicely done! Works as expected when changing the setting to ${workspaceFolder} as well as ${workspaceFolder}/mypkg where mypkg exists, in both R and Python.

When I set it to a nonexistent folder like ${workspaceFolder}/mypkg2 or a bad template like ${workspaceFolder2} it seemed to default to ${workspaceFolder}, although I don't see where that's handled. Should we handle and/or test those cases?

[austin3dickey]

When I set it to a nonexistent folder like ${workspaceFolder}/mypkg2 or a bad template like ${workspaceFolder2} it seemed to default to ${workspaceFolder}, although I don't see where that's handled. Should we handle and/or test those cases?

Oh nice catch, I thought I tried this but that's different behavior from what I'd expect (I think I'd expect it to fall back to the parent dir). Turns out this is happening in kallichore. If the directory doesn't exist, it defaults to the pwd, which is the workspace root.

Maybe I'll verify that the directory exists at the end of resolveWorkingDirectory() so that if it doesn't we can fall back to the old behavior, the parent dir.

[austin3dickey]

Okay, I just moved the default behavior of "if the setting doesn't resolve to an existing directory, use the notebook's parent" up the stack to when we're creating the RuntimeSession metadata (in resolveNotebookWorkingDirectory()). That way we can delete the code that handles that default in KallichoreSession and the Jedi implementation. This fixes the bug @seeM was seeing.

[austin3dickey]

I'll have to fix the Python tests and add an e2e test.