MyPy does not find packages installed by Conda

[MaxG87]

I use Anaconda to manage everything related to Python. Thus, I have a Conda environment where I installed mypy, keras and numpy.

My PYTHONPATH reads:

/home/user/workspace/project/:/home/user/workspace/project/python/:

My MYPYPATH is just a copy of this, created by:

export MYPYPATH="$PYTHONPATH"

My minimum working example reads:

#!/usr/bin/env python

import keras
import numpy

print("Hello World!")

When I execute mypy module-test.py I get:

module-test.py:3: error: Cannot find module named 'keras'
module-test.py:3: note: See https://mypy.readthedocs.io/en/latest/running_mypy.html#missing-imports
module-test.py:4: error: No library stub file for module 'numpy'
module-test.py:4: note: (Stub files are from https://github.com/python/typeshed)

Note that mypy does not complain about numpy. However, numpy should not be installed system-wide:

$ apt search numpy | egrep -i insta

WARNING: apt does not have a stable CLI interface. Use with caution in scripts.

Now I wonder why mypy finds numpy but not keras and how to resolve this. Since I am using several Conda environments, manually tweaking MYPYPATH is not an option.

[JukkaL]

Mypy also complains about numpy, but the error message is different because mypy recognizes it as a third-party package:

module-test.py:4: error: No library stub file for module 'numpy'
module-test.py:4: note: (Stub files are from https://github.com/python/typeshed)

Mypy doesn't look into installed Python packages unless they advertise that they support type checking (PEP 561). You normally need library stub files to type check code using a third party library, unless you just ignore the relevant errors from mypy (this is explained in mypy docs).

For numpy you can find stubs at https://github.com/numpy/numpy-stubs (I haven't tried them myself). I'm not aware any stub files for keras. You can perhaps experiment with the stubgen utility shipped with mypy and see if it can generate stub files for you.

[MaxG87]

Thank you very much for the helpful answer.