Native Interactive Window - Pylance errors with variables imported in jupyter startup script

[jaymegordo]

Environment data

• Version: 1.60.0-insider
• Commit: 7c25c174726ab969db3f99058cdf24bcfae1cfbe
• Date: 2021-08-11T05:13:17.497Z
• Jupyter Extension version: v2021.8.2025984474
• Python Extension version: v2021.9.1117361286-dev
• OS: Darwin x64 20.6.0
• Python and/or Anaconda version: 3.9.2
• Type of virtual environment used: poetry
• Jupyter server running: Local

One main part of my workflow is to have certain variables/modules imported at jupter startup, defined in a startup.py file at:

• ~/.ipython/profile_default/startup/startup.py

Expalained fully here

Expected behaviour

• Jupyter/native interactive window/pylance recognizes variables imported through a startup script.

Actual behaviour

For example I'd have

#! startup.py
import pandas as pd

The module is imported, and pd is valid in the globals(), but pylance doesn't seem to recognize it. I'll also import lots of custom objects from the current project I'm working on, so you can imagine it's kinda hard to work when almost every variable is "undefined" (and also doesn't have intellisense).

In screenshots:

Not sure what's actually going on here, but just expanding the code from the previous image (before executing anything) made pylance/problems go crazy:

pd is recognized after explicitely importing in the current session:

[greazer]

@jaymegordo Thanks for the feedback!

[rchiodo]

Sorry but this isn't currently supported. This is basically the same bug as this one:
#6987

[jaymegordo]

@rchiodo thanks for letting me know. Do you know if there will be any plans to support this in the future? Do you know of any other workarounds to have certain modules importd to the IW automatically which would support intellisense?

[rchiodo]

@jaymegordo you can track #6987 to see if we end up supporting dynamic modules in pylance or not. My guess is it won't be supported for a long time as pylance doesn't run any python code. It's all static analysis based. We might be able to fudge it by concating the code for a run startup commands onto the top of the notebook, but we haven't tried yet.