do quarto (qmd) files run through VSCode/Cursor work with nbstata?

[ocolluphid]

I am having trouble using the quarto <> VSCode/Cursor connection (can't find the kernel), should I try using ipynb directly if I'm doing interactive development via VSCode/Cursor?

[hugetim]

I think I've gotten it to work via a .qmd file. Troubleshooting the kernel recognition is probably more of a Quarto or VSCode/Cursor issue. But I'm happy to try to help.

Do you have it working with .ipynb files in VSCode?

[hugetim]

Ok, refreshing my memory with the GitHub issue history, Stata has been added as a special case, so interactive .qmd code execution should work, with an important caveat:

note that I found it was necessary to manually switch to the nbstata Kernel to make this work as the Jupyter extension may assume Python by default

quarto-dev/quarto#336 (comment)

Here's how the manual switch works. When you add a stata code cell and click Run Cell, it initially assumes Python for some reason. (I'm not sure, but that may be possible to change via the Jupyter extension config.) It looks like this:

Click the kernel selector in the interactive window (circled in yellow above) and select a Stata kernel (first "Select Another Kernel..." then "Jupyter Kernel..." then hopefully you have an nbstata kernel listed to select). Then if you click "Run Cell" again (and, when it asked whether I wanted a new interactive window, I said "no"), it works:

[hugetim]

(In case you are only reading these responses in your email, note that my initial reply above was wrong, so be sure to view the edited version online.)

[ocolluphid]

thanks will give this a shot in cursor later today!

[ocolluphid]

Looks like it's working, thanks!
One more thing -- if I want to run a preview I get the following error in the VSC/Cursor terminal:

quarto preview stata_quarto_demo.qmd --no-browser --no-watch-inputs
ERROR: Jupyter kernel 'nbstata' not found. Known kernels: python3. Run 'quarto check jupyter' with your python environment activated to check python version used.

Stack trace:
    at jupyterKernelspecFromMarkdown (file:///opt/quarto/bin/quarto.js:32003:35)
    at eventLoopTick (ext:core/01_core.js:183:11)
    at async quartoMdToJupyter (file:///opt/quarto/bin/quarto.js:31835:36)
    at async createNotebookforTarget (file:///opt/quarto/bin/quarto.js:38601:16)
    at async Object.target (file:///opt/quarto/bin/quarto.js:38286:18)
    at async fileExecutionEngineAndTarget (file:///opt/quarto/bin/quarto.js:41258:20)
    at async renderContexts (file:///opt/quarto/bin/quarto.js:72183:32)
    at async renderFormats (file:///opt/quarto/bin/quarto.js:72234:22)
    at async file:///opt/quarto/bin/quarto.js:98062:24
    at async Command.fn (file:///opt/quarto/bin/quarto.js:98059:25)

[hugetim]

Did you install nbstata in a specific environment? If so, it may be an issue of activating it, which is straightforward in a terminal but can be a hassle to get working via a button.

[hugetim]

By the way, though, if what you want is to use Cursor with Stata, you may be better off using https://github.com/sepinetam/stata-mcp I haven't tried it yet myself, but their direct goal is Cursor use. (There's also a separate Stata MCP project with similar goals. See #59 )

[ocolluphid]

thanks, that's super helpful!