[FIXED] Why is my phyton mock only working with a specific way of import?

Issue

I have a utils class that besides other stuff has a function to get an id:

# utils.py
def get_id():
    return ...

Also, there is another class, lets say its about cars, that uses this function:

# car.py
from utils import get_id

class Car:
    ...
    def setup(self):
        self.id = get_id()

I want to run a unittest that should provide a specific id which is why I am mocking this function:

# test_car.py
import unittest
from unittest.Mock import Mock
import utils

class TestCar(unittest.TestCase):
    ...

    def test_car_setup(self):
        utils.get_id = Mock(callable, return_value="my-testing-id")
        car = Car()
        car.setup()
        ...

But this mock is not working. I still get the behavior of the original function. However, I was able to get it working by adjusting the car class like this:

# car.py
from . import utils

class Car:
    ...
    def setup(self):
        self.id = utils.get_id()

The question is now, why is that working, and the previous version not? I initially thought that this could be a call by reference/call by value issue, but the debugger showed that now matter how I adjust the imports, the utils function is always the same object.

I understand that this might not be a typical question because my projest is working, but I am failing to understand the workings of python that lead to it and I would appreciate if someone could enlighten me.

Thanks in advance!

Solution

Change test_car.py

It is enough to make some changes to your file test_car.py, while car.py and utils.py remain without changes; the file car.py include the original import from utils import get_id as showed below:

# car.py
from utils import get_id

class Car:
    ...
    def setup(self):
        self.id = get_id()

Below I show you the file test_car.py with the needed modifications:

#test_car.py
import unittest
from unittest.mock import Mock
#import utils                      # <----- this import is now useless
from car import Car                # <----- add this import
import car as module_car           # <----- add this import (added the alias
                                   #        module_car to avoid name collision)

class TestCar(unittest.TestCase):

    def test_car_setup(self):
        # I have commented your instruction
        #utils.get_id = Mock(callable, return_value="my-testing-id")
        
        # USE THIS INSTRUCTION 
        module_car.get_id = Mock(callable, return_value="my-testing-id")
        car = Car()
        car.setup()

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

Explanation

The most important change is the following:

utils.get_id = Mock(callable, return_value="my-testing-id") ---> module_car.get_id = Mock(callable, return_value="my-testing-id")

• The instruction module_car.get_id = Mock(...) substitutes the object pointed by the name get_id used inside the module car.py; this means that when is executed the instruction self.id = get_id() in the method setup() is executed the Mock object which returns my-testing-id

• Your instruction (utils.get_id = Mock(...)) substitutes the object pointed by name get_id, but inside the file utils.py; the error is that the object used by the method setup() is pointed by the name get_id defined inside car.py.

Why your second solution works

Your second solution is using the import from . import utils in the file car.py as showed below:

# car.py
from . import utils

class Car:
    ...
    def setup(self):
        self.id = utils.get_id()

In this way, in your test code, the object pointed by utils.get_id() has been substituted by the Mock object (this is thanks to the instruction utils.get_id = Mock(callable, return_value="my-testing-id") present in your code).
So it happens:

utils.get_id() returns a reference which (points to) ----> Mock object

In the file car.py you use exactly the instruction: utils.get_id() and it returns a reference to the Mock.

Why your first solution doesn't work

The file test_car.py is the same both in the first and second solutions, so it is always true that:

utils.get_id() returns a reference which (points to) ----> Mock object

But your original file car.py is:

# car.py
from utils import get_id

class Car:
    ...
    def setup(self):
        self.id = get_id()

So in it, with the instruction from utils import get_id you define the name get_id and get_id() returns a reference to the object get_id defined inside the module utils.py.

In the method setup() you use directly get_id() (and not utils.get_id()); get_id() returns the reference to the real object and not to the mock object; the Mock object is returned only by utils.get_id().

Answered By - User051209