"dag" is not a known attribute of module "networkx"

[AbdealiLoKo]

I am using python 3.9 and am seeing pyright throwing an error with networkx in CLI - but NOT in vscode.

I am unable to figure out what the difference is ! Any suggestions ?

Things I have tried:

• Tried explicitly setting useLibraryCodeForTypes: true
• Tried with python 3.8

---

Here is a MRE file I am testing with:

$ cat run.py
from __future__ import annotations

import networkx as nx


graph = nx.DiGraph({})
nx.dag.topological_sort(graph)

And the CLI command:

$ venv/bin/pyright run.py
/home/ali/run.py
  /home/ali/run.py:7:4 - error: "dag" is not a known attribute of module "networkx" (reportAttributeAccessIssue)
1 error, 0 warnings, 0 informations

But in vscode, there is no error (and intellisense and so on work fine)

And intellisense works too:

This is the pyright_config:

$ cat pyrightconfig.json
{
  "pythonVersion": "3.7",
  "exclude": ["**/dist", "**/build", "**/venv*"],
  "venvPath": ".",
  "venv": "venv"
}

Versions:

$ venv/bin/python -V
Python 3.9.19
$ venv/bin/pyright --version
pyright 1.1.382
$ venv/bin/pip list | grep networkx
networkx                  3.2.1

[rchiodo]

My guess is this is because Pylance ships a bunch of bundled stubs and Pyright doesn't. You can read about the differences here:
https://github.com/microsoft/pylance-release/blob/main/USING_WITH_PYRIGHT.md

[erictraut]

It appears that Pylance ships its own stubs for the networkx library, and those override the typeshed stubs for the same. The Pylance team may want to consider removing these custom stubs and instead incorporating any improvements into the typeshed version. This will help the community as a whole, whereas shipping competing stubs for the same library harms usability.

[rchiodo]

Transferred. Going to make an issue out of this.

[rchiodo]

Created issue #6508

[AbdealiLoKo]

Interesting, I was so sure this was a usage issue on my part :D

I also see a similar thing with pandas then.
When I have the code:

import pandas as pd
val = 1
if pd.api.types.is_scalar(val) and pd.isna(val):
    print("scalar")

I get:

$ venv/bin/pyright run.py
/home/ali/run.py
  /home/ali/run.py:7:11 - error: Cannot access attribute "types" for class "type[Panel]"
    Attribute "types" is unknown (reportAttributeAccessIssue)
  /home/ali/run.py:7:11 - error: Cannot access attribute "types" for class "type[datetime]"
    Attribute "types" is unknown (reportAttributeAccessIssue)
  /home/ali/run.py:7:11 - error: Cannot access attribute "types" for class "type[SparseArray]"
    Attribute "types" is unknown (reportAttributeAccessIssue)
  /home/ali/run.py:7:17 - error: "is_scalar" is not a known attribute of module "numpy.testing" (reportAttributeAccessIssue)
  /home/ali/run.py:7:17 - error: Cannot access attribute "is_scalar" for class "type[NoseTester]"
    Attribute "is_scalar" is unknown (reportAttributeAccessIssue)

[rchiodo]

Are you sure you saved the file? I don't get any errors myself when doing the same thing.