How to verify a .__getitem__() call in a Mock mock_calls list during unit testing

Question

When trying to unit test a method that returns a tuple and I am trying to see if the code accesses the correct tuple index, python tries to evaluate the expected call and turns it into a string.

call().methodA().__getitem__(0) ends up getting converted into '().methodA' in my expected_calls list for the assertion.

The example code provided, results in the output and traceback:

expected_calls=[call().methodA(), '().methodA']
  result_calls=[call().methodA(), call().methodA().__getitem__(0)]
======================================================================
ERROR: test_methodB (badMockCalls.Test_UsingToBeMocked_methods)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "C:\dev\workspace\TestCode\src\badMockCalls.py", line 43, in test_methodB
    self.assertListEqual(expected_calls, self.result_calls)
  File "C:\Python33\lib\unittest\case.py", line 844, in assertListEqual
    self.assertSequenceEqual(list1, list2, msg, seq_type=list)
  File "C:\Python33\lib\unittest\case.py", line 764, in assertSequenceEqual
    if seq1 == seq2:
  File "C:\Python33\lib\unittest\mock.py", line 1927, in __eq__
    first, second = other
ValueError: too many values to unpack (expected 2)

----------------------------------------------------------------------
Ran 1 test in 0.006s

FAILED (errors=1)

How do I go about asserting that methodB is calling self.tbm.methodA()[0] properly?

Example code (Python 3.3.2):

import unittest
from unittest.mock import call, patch
import logging

log = logging.getLogger(__name__)
log.setLevel(logging.DEBUG)
_ch = logging.StreamHandler()
_ch.setLevel(logging.DEBUG)
log.addHandler(_ch)

class ToBeMocked():  # external resource that can't be changed
    def methodA(self):
        return (1,)

class UsingToBeMocked():  # project code
    def __init__(self):
        self.tbm = ToBeMocked()

    def methodB(self):
        value = self.tbm.methodA()[0]
        return value

class Test_UsingToBeMocked_methods(unittest.TestCase):
    def setUp(self):
        self.patcher = patch(__name__ + '.ToBeMocked')
        self.mock_ToBeMocked = self.patcher.start()
        self.utbm = UsingToBeMocked()
        # clear out the mock_calls list from the constructor calls
        self.mock_ToBeMocked.mock_calls = []
        # set result to always point to the mock_calls that we are testing
        self.result_calls = self.mock_ToBeMocked.mock_calls

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

    def test_methodB(self):
        self.utbm.methodB()
        # make sure the correct sequence of calls is made with correct parameters
        expected_calls = [call().methodA(),
                          call().methodA().__getitem__(0)]
        log.debug('expected_calls=' + str(expected_calls))
        log.debug('  result_calls=' + str(self.result_calls))
        self.assertListEqual(expected_calls, self.result_calls)

if __name__ == "__main__":
    unittest.main()

Answer 1

To test for mock_object.account['xxx1'].patch(body={'status': 'active'}) I had to use the test:

mock_object.account.__getitem__.assert_has_calls([
    call('xxxx1'),
    call().patch(body={'status': 'active'}),
])

I can't explain why this works, this looks like weird behaviour, possibly a bug in mock, but I consistently get these results and it works.

Answer 2

I've just stumbled upon the same problem. I've used the solution/work-around from here:

http://www.voidspace.org.uk/python/mock/examples.html#mocking-a-dictionary-with-magicmock

namely:

>> mock.__getitem__.call_args_list
[call('a'), call('c'), call('d'), call('b'), call('d')]

You can skip the magic function name misinterpretation and check against its arguments.