Wrong error report (false positive) in numpy complex array value assignment

[wldhg]

When I pylinted a script contains numpy.complex array, E1137 appeared on correct code, and not in incorrect code.

Steps to reproduce

"""
This demonstrates numpy complex E1137 error
"""
# pylint: disable=C0103
# pylint: disable=W0702

import numpy as np

# Create a numpy complex array
comp_zeros = np.zeros(2, dtype=np.complex)
print(comp_zeros)

# Correct assignment
try:
    comp_zeros.imag[0] = 1
    comp_zeros.real[1] = 2
except:
    print("Exception in correct assignment!")
finally:
    print(comp_zeros)

# Incorrect assignment
try:
    comp_zeros[0].imag = 3
    comp_zeros[1].real = 4
except:
    print("Exception in incorrect assignment!")
finally:
    print(comp_zeros)

Current behavior

I saved above script in test.py.

Script Output

$ python test.py
[0.+0.j 0.+0.j]
[0.+1.j 2.+0.j]
Exception in incorrect assignment!
[0.+1.j 2.+0.j]

Lint Output

$ python -m pylint test.py
************* Module test
test.py:15:4: E1137: 'comp_zeros.imag' does not support item assignment (unsupported-assignment-operation)
test.py:16:4: E1137: 'comp_zeros.real' does not support item assignment (unsupported-assignment-operation)

------------------------------------------------------------------
Your code has been rated at 4.12/10

Expected behavior

I expected E1137 to be appeared in line 24 and 25 (incorrect code).

Lint Output

test.py:24:4: E1137: 'comp_zeros[0].imag' does not support item assignment (unsupported-assignment-operation)
test.py:25:4: E1137: 'comp_zeros[1].real' does not support item assignment (unsupported-assignment-operation)

Versions

pylint 2.4.4
astroid 2.3.3
Python 3.7.5 (tags/v3.7.5:5c02a39a0b, Oct 15 2019, 01:31:54) [MSC v.1916 64 bit (AMD64)]

and

$ pip show numpy
Name: numpy
Version: 1.18.0
... (omitted)

[hippo91]

@wldh-g thanks for this report!

[hippo91]

@wldh-g there are two problems here: the wrong positive one and the wrong negative one.
For the wrong positive the fix will be quite easy (i have got already a draft) however fixing the wrong negative would necessitate a huge work that is, to my mind, not worthy.

[wldhg]

@hippo91 Thanks for your effort! I understand your mind. Does that wrong negative problem is only trouble for this #3322 issue?

[hippo91]

@wldh-g i don't have in mind if other issues are also due to this wrong negative problem. I think so but have not more precise information.