Does Positron not play well with Jupyter notebooks?

[marcdotson]

This may be more Python ignorance on my part (see #4588), but I'm wondering if there is something under the hood that makes Positron not play well with Jupyter notebooks?

Jupyter notebooks are not my favorite, but I have students who use them with Positron and we have started running into some issues. To start with, I'm trying to import libraries in a code cell at the beginning of a Jupyter notebook:

import pandas as pd
import numpy as np
from matplotlib import pyplot as plt
import statsmodels.api as sm

And I'm consistently running into a ModuleNotFoundError. The modules are installed. I'm using the global environment -- no project-specific virtual environment. I can run this same code in the console, from a Python script, from a Quarto code cell, even the same Jupyter notebook code cell using VS Code and it works -- just not when running the Jupyter notebook code cell in Positron.

Any insight or suggestions are welcome!

[juliasilge]

We definitely intend that Jupyter notebooks work great in Positron, and appreciate any reports of them not doing so!

Can you make sure that the kernel the notebook thinks it is using is the same one that you have running in the console? You'll want to click "Select Kernel" here:

Then choose "Python Environments..." and then choose the specific Python you want to be using.

If you can share any screenshots showing specifics (for example, showing that you are using a specific Python environment that really does have these packages installed), that would be really helpful.

[marcdotson]

That's it! Thank you, Julia! My suspicions of Python ignorance was correct -- I didn't even know there was a kernel selection option in a Jupyter notebook.

FWIW, the selected kernel in the notebook was the system version of Python. I've installed Python using pyenv, and the global pyenv setting wasn't transferred to the notebook kernel -- and maybe not to the interpreter selected in Positron either. I suppose I should just set up a virtual environment and avoid the problem all together. :)

[Btibert3]

Chiming in on this. This was a point of friction with my students.

Perhaps I am overlooking a setting, but as a user of Positron, I expect the notebook kernel to match the selected interpreter. After choosing an interpreter, everything I do in Python should use that interpreter. This should only be set once.

[Btibert3]

A screenshot:

[marcdotson]

Very much agree. Positron playing nice would make that Jupyter selection at the same time/override it.

[petetronic]

Thank you @Btibert3 and @marcdotson, improvements in this area are definitely on our roadmap. The following issue tracks some of the thinking here #5584

[seeM]

I'd consider this a bug in automatic kernel selection. @Btibert3 is Positron opened in a workspace? If not, it could be #6280.

[Btibert3]

@seeM

Steps to reproduce:

1. Start Positron (no folders currently open)
2. Open a folder to give the app context
3. Open up a .ipynb notebook
4. Interpreter and Notebook are not the same

NOTE: After selecting the same environment for both the interpreter and the kernel, seemingly I can close the folder and later come back, and the application remembers both for this folder/project. Putting that aside, if I were writing this feature, a user should only have to select the interpreter within the folder/project, after which all areas of compute (terminal, notebook) should use that selection.