pylint incorrectly issues error E1123 on numpy.unique kwargs (numpy v1.15.4)

[NiaX84]

Steps to reproduce

1. create code calling numpy.unique with keyword arguments return_counts, return_index
2. run pylint

Current behavior

E1123 is raised.

E1123: Unexpected keyword argument 'return_counts' in function call (unexpected-keyword-arg)

Expected behavior

no error is raised, since the keyword arguments are present since numpy v1.9.0 and numpy v1.15.4 is used

pylint --version output

pylint 2.6.0
astroid 2.4.2
Python 3.7.2 (default, Mar 27 2019, 08:41:08)
[GCC 6.3.0 20170516]

[hippo91]

@NiaX84 thanks for the report.
Can you give an example where the problem arises ?

[AWhetter]

#3736 is a duplicate issue that contains an example for this problem.

[hippo91]

@NiaX84 @igozali i'm unable to reproduce the bug.
I'm linting following code:

import numpy as np
xs = np.array([1, 1, 2])
np.unique(xs, return_index=False)

and i obtain:

************* Module bug_pylint_3825
bug_pylint_3825.py:1:0: C0114: Missing module docstring (missing-module-docstring)

------------------------------------------------------------------
Your code has been rated at 6.67/10 (previous run: 6.67/10, +0.00)

I'm using:

• pylint 2.6
• astroid 2.4
• numpy 1.19.1

[DevinShanahan]

@hippo91 It appears the numpy version is important. I am seeing this with numpy<=1.15.4 but I get the same result as you for numpy>=1.16.

numpy 1.15.4

************* Module bug_pylint_3825
bug_pylint_3825.py:1:0: C0114: Missing module docstring (missing-module-docstring)
bug_pylint_3825.py:3:0: E1123: Unexpected keyword argument 'return_index' in function call (unexpected-keyword-arg)

---------------------------------------------------------------------
Your code has been rated at -10.00/10 (previous run: 6.67/10, -16.67)

[hippo91]

@DevinShanahan thanks for your answer.
Maintaining the astroid brain that handle numpy stuff is a lot of work. Honestly i don't think we are able to maintain it for all versions of numpy. I think the solution here is to upgrade you numpy version.

[guy4261]

Same for E1102: torch.tensor is not callable (not-callable) with

• python==3.6
• torch==1.7.0
• pylint==2.6.0

My pylintrc file doesn't block this, even though it contains:

$ cat pylintrc
[TYPECHECK]
# List of members which are set dynamically and missed by Pylint inference
# system, and so shouldn't trigger E1101 when accessed.
generated-members=torch
ignored-classes=torch.tensor
ignored-modules=torch

I can't make this error disappear (for code such as torch.tensor([True, False]) which is valid in the repl.

[hippo91]

@guy4261 thanks for your report.
I think that astroid, the pylint inference engine, needs a brain to plainly support torch.
If you are interested in, it will be a pleasure to help you through.

[guy4261]

#wat
Thanks @hippo91 but why do these ignores don't work? I don't might the false positives (hey, everyone has bugs) - they reflect something that's hard to do (identifying stuff coming from C++ I guess)? But making the ignores work is something I'd expected of pylint (=properly filter out errors, not because they are incorrect but because I had the ignored-modules=torch directive in my pylintrc file).

I ended code such as torch.tensor([1,2,3], dtype=torch.int64) to torch.Tensor([1,2,3]).to(torch.int64).
So thanks! No help required (but I'd be glad to hear your suggestions if any!).

[PCManticore]

@guy4261 ignored-modules is not taken in consideration for not-callable or other errors you were getting here; we only use it for no-member and maybe some other checks as well. We should update the documentation or use the flag in other locations as well, but that's the reason why the ignored-modules=torch option did not do anything here.

[guy4261]

Thanks for the clarification @PCManticore !