Sum and multiply

[hippo91]

Steps

• [ * ] For new features or bug fixes, add a ChangeLog entry describing what your PR does.
• [ * ] Write a good description on what the PR does.

Description

This PR tries to solve the issue #669. In fact the call to some functions of the numpy brains are inferred not only as numpy.ndarray but also as Uninferable. This lead to false negative for such function calls.

The origin of this bug has been hard to find but seems to be due to the fact that the context.push method return True as soon as the node in argument has been visited once for the current context.
However for some particular cases, such as those encountered in the numpy library, a node could be visited more than once due to peculiar imports.
The modification made here ensure that all calls to numpy ufunc functions are inferred only as numpy.ndarray and modifies two inference unit tests. Those ones seem more pertinent but a close look from the reviewer would be appreciated.

Type of Changes

	Type
✓	🐛 Bug fix

Related Issue

Closes #669

[PCManticore]

Hey @hippo91 The code looks good. I tested this against pylint and it causes a regression test to fail. Can you determine why that would be the case with this patch? It's regression_property_no_member_2641. I believe it's something to do with the new inference limit but haven't been able to investigate more closely.

[hippo91]

@PCManticore thanks for looking at this. I confirm that there is a problem with this unit test. I'm trying to fix it.

[hippo91]

@PCManticore ,

i had a look at the regression_property_no_member_2641 unittest and i think that the problem is not directly linked to this PR.
In fact, the following code triggers also the no-member message :

# pylint: disable=missing-docstring,unused-argument,too-few-public-methods
class Person:
    def __init__(self, name):
        self.name = name

    @property
    def name(self):
        return self.__name

    @name.setter
    def name(self, value):
        self.__name = value

A = Person('toto')
Person.name.fset(A, 'toto')

In case fset is replaced by fget then no problem arises. I think the problem is a lack of attr_fset method in the PropertyModel class (objectmodel module).
@PCManticore do you thinks there is a reason for the absence of attr_fset and attr_del method in this class?
Is it wise to add them?
If yes i'll make another PR.
Thanks in advance.

[PCManticore]

@hippo91 Yeah, that makes sense, let's add that in separate PRs.

[hippo91]

Replaced by sum_and_multiply_clean. Same sources with a cleaner log history