invalid-unary-operand-type error false positive

[pbaughman]

Steps to reproduce

1. create a file called test.py containing the following:

def frob(value=None):
    if value is not None:
        return -value
    return value

2. pylint test.py

Current behavior

pylint output contains E: 3,15: bad operand type for unary -: NoneType (invalid-unary-operand-type)

Expected behavior

There should be no invalid-unary-operator-type error for this line because it is not possible for value to be None here

pylint --version output

pete@dev:~$ pylint --version
No config file found, using default configuration
pylint 1.7.1, 
astroid 1.5.2
Python 2.7.6 (default, Oct 26 2016, 20:30:19) 
[GCC 4.8.4]

[nschloe]

Just bumped into this with numpy.arrays:

import numpy
def foobar(value=None):
    a = numpy.array([1, 2, 3])
    return +a, -a

[JVillella]

Same issue, e.x.

import numpy as np

def position_camera(dq=(0, 0, 0, 0), dT=(0, 0, 0)):
    q = -np.array((0.0, 1.0, 0.15, 0.35)) + dq

    # ...

[ScriptingSquirrel]

When using @pbaughman's snipped from the issue description for testing, it seems the false positive appeared with Pylint release 1.7.0. In Pylint 1.6.5 (=the last release before 1.7.0) this false positive does not get triggered.

Using the older release 1.6.5 until this is fixed.

[olysonek]

I have a similar issue with the following code:

try:
    open('file')
except OSError as ex:
    print(-ex.errno)

Pylint output:
E: 4, 7: bad operand type for unary -: errno (invalid-unary-operand-type)

[PCManticore]

Folks, not all these examples are the same problem. :) I would suggest for some of them to be in their separate issues, since we don't need flow control for correctly checking numpy arrays or OSError instances. Creating different issues for these would mean that we could focus on smaller ones first rather than being concentrated on a major blob of work which is control flow inference

[dovahcrow]

After two years this looks like still an issue. Any ideas on how to solve this?

[hippo91]

@nschloe @JVillella the bug you were facing with numpy does not occur anymore with latest pylint and master as soon as you import numpy with an alias (import numpy as np for example). For direct numpy import the situation will be correct once pylint-dev/astroid#918 will be merged.

I let this issue open because initial report was dealing with control-flow.

Please do not add any reports here unless it is directly related to control-flow.

[flaviovs]

Not sure if this qualifies as control-flow (apologies if it does not), but I can easily replicate the issue with:

# func.py
def func(moo = None):
    return -moo if moo else 0

$ pylint func
************* Module func
func.py:2: [E1130(invalid-unary-operand-type), func] bad operand type for unary -: NoneType

----------------------------------------------------------------------
Your code has been rated at -15.00/10 (previous run: -15.00/10, +0.00)

$ pylint --version
pylint 2.7.2
astroid 2.5.1
Python 3.7.3 (default, Jul 25 2020, 13:03:44) 
[GCC 8.3.0]