Incorrectly determines mock objects are of type dict when mocking a module

[pylint-bot]

Originally reported by: Anonymous

---

Give these two files:

sut.py:

#!python

import os

def silly_function(arg):
    return os.path.exists(arg)

and test.py:

#!python

import unittest
import mock

import sut

class SillyTest(unittest.TestCase):

    def setUp(self):
        self.os_patcher = mock.patch("sut.os")
        self.os_mock = self.os_patcher.start()

    def tearDown(self):
        self.os_patcher.stop()

    def test_mock_is_mock(self):
        self.assertIsInstance(self.os_mock, mock.MagicMock)
        self.assertNotIsInstance(self.os_mock, dict)

    def test_silly_function(self):
        self.os_mock.path.exists.return_value = "hello"
        self.assertEqual("hello", sut.silly_function("world"))

The code works fine:

#!shell

$ python -m unittest -v test
test_mock_is_mock (test.SillyTest) ... ok
test_silly_function (test.SillyTest) ... ok

----------------------------------------------------------------------
Ran 2 tests in 0.001s

OK

But pylint incorrectly determines that self.os_mock is of type dict:

#!shell

$ pylint -E test
************* Module test
E: 20, 8: Instance of 'dict' has no 'path' member (no-member)

The test_mock_is_mock test above shows that this is explicitly not the case.

This is using python 2.7.6

---

• Bitbucket: https://bitbucket.org/logilab/pylint/issue/697

[pylint-bot]

Original comment by Claudiu Popa (BitBucket: PCManticore, GitHub: @PCManticore):

---

I don't think there's anyway for pylint to figure out what mock is doing behind the scene in a sane way. These kind of libraries which overuses dynamic features of the language aren't well suited for a static analysis understanding.

But this might be solvable with an astroid brain plugin.

[pylint-bot]

Original comment by Oliver Jeeves (BitBucket: ojeeves):

---

This problem only occurs when you're mocking a module. Pylint correctly finds the type of the mock if you're mocking a function, class or method.