Mocking python function based on input arguments

Question

We have been using Mock for python for a while.

Now, we have a situation in which we want to mock a function

def foo(self, my_param):
    #do something here, assign something to my_result
    return my_result

Normally, the way to mock this would be (assuming foo being part of an object)

self.foo = MagicMock(return_value="mocked!")

Even, if i call foo() a couple of times i can use

self.foo = MagicMock(side_effect=["mocked once", "mocked twice!"])

Now, I am facing a situation in which I want to return a fixed value when the input parameter has a particular value. So if let's say "my_param" is equal to "something" then I want to return "my_cool_mock"

This seems to be available on mockito for python

when(dummy).foo("something").thenReturn("my_cool_mock")

I have been searching on how to achieve the same with Mock with no success?

Any ideas?

Answer 1

If side_effect_func is a function then whatever that function returns is what calls to the mock return. The side_effect_func function is called with the same arguments as the mock. This allows you to vary the return value of the call dynamically, based on the input:

>>> def side_effect_func(value):
...     return value + 1
...
>>> m = MagicMock(side_effect=side_effect_func)
>>> m(1)
2
>>> m(2)
3
>>> m.mock_calls
[call(1), call(2)]

http://www.voidspace.org.uk/python/mock/mock.html#calling

Answer 2

As indicated at Python Mock object with method called multiple times

A solution is to write my own side_effect

def my_side_effect(*args, **kwargs):
    if args[0] == 42:
        return "Called with 42"
    elif args[0] == 43:
        return "Called with 43"
    elif kwargs['foo'] == 7:
        return "Foo is seven"

mockobj.mockmethod.side_effect = my_side_effect

That does the trick

Answer 3

Side effect takes a function (which can also be a lambda function), so for simple cases you may use:

m = MagicMock(side_effect=(lambda x: x+1))

Answer 4

If you "want to return a fixed value when the input parameter has a particular value", maybe you don't even need a mock and could use a dict along with its get method:

foo = {'input1': 'value1', 'input2': 'value2'}.get

foo('input1')  # value1
foo('input2')  # value2

This works well when your fake's output is a mapping of input. When it's a function of input I'd suggest using side_effect as per Amber's answer.

You can also use a combination of both if you want to preserve Mock's capabilities (assert_called_once, call_count etc):

self.mock.side_effect = {'input1': 'value1', 'input2': 'value2'}.get

Answer 5

I've ended up here looking for "how to mock a function based on input arguments" and I finally solved this creating a simple aux function:

def mock_responses(responses, default_response=None):
  return lambda input: responses[input] if input in responses else default_response

Now:

my_mock.foo.side_effect = mock_responses(
  {
    'x': 42, 
    'y': [1,2,3]
  })
my_mock.goo.side_effect = mock_responses(
  {
    'hello': 'world'
  }, 
  default_response='hi')
...

my_mock.foo('x') # => 42
my_mock.foo('y') # => [1,2,3]
my_mock.foo('unknown') # => None

my_mock.goo('hello') # => 'world'
my_mock.goo('ey') # => 'hi'

Hope this will help someone!

Answer 6

Although side_effect can achieve the goal, it is not so convenient to setup side_effect function for each test case.

I write a lightweight Mock (which is called NextMock) to enhance the built-in mock to address this problem, here is a simple example:

from nextmock import Mock

m = Mock()

m.with_args(1, 2, 3).returns(123)

assert m(1, 2, 3) == 123
assert m(3, 2, 1) != 123

It also supports argument matcher:

from nextmock import Arg, Mock

m = Mock()

m.with_args(1, 2, Arg.Any).returns(123)

assert m(1, 2, 1) == 123
assert m(1, 2, "123") == 123

Hope this package could make testing more pleasant. Feel free to give any feedback.

Answer 7

You can also use partial from functools if you want to use a function that takes parameters but the function you are mocking does not. E.g. like this:

def mock_year(year):
    return datetime.datetime(year, 11, 28, tzinfo=timezone.utc)

@patch('django.utils.timezone.now', side_effect=partial(mock_year, year=2020))

This will return a callable that doesn't accept parameters (like Django's timezone.now()), but my mock_year function does.

Answer 8

Just to show another way of doing it:

def mock_isdir(path):
    return path in ['/var/log', '/var/log/apache2', '/var/log/tomcat']

with mock.patch('os.path.isdir') as os_path_isdir:
    os_path_isdir.side_effect = mock_isdir

Answer 9

You can also use @mock.patch.object:

Let's say a module my_module.py uses pandas to read from a database and we would like to test this module by mocking pd.read_sql_table method (which takes table_name as argument).

What you can do is to create (inside your test) a db_mock method that returns different objects depending on the argument provided:

def db_mock(**kwargs):
    if kwargs['table_name'] == 'table_1':
        # return some DataFrame
    elif kwargs['table_name'] == 'table_2':
        # return some other DataFrame

In your test function you then do:

import my_module as my_module_imported

@mock.patch.object(my_module_imported.pd, "read_sql_table", new_callable=lambda: db_mock)
def test_my_module(mock_read_sql_table):
    # You can now test any methods from `my_module`, e.g. `foo` and any call this 
    # method does to `read_sql_table` will be mocked by `db_mock`, e.g.
    ret = my_module_imported.foo(table_name='table_1')
    # `ret` is some DataFrame returned by `db_mock`