Testing

Environment Variables §

Use the unittest.mock library to overwrite environment variables for tests or test suites.

import os
from unittest import mock
 
 
@mock.patch.dict(os.environ, {'SOME_VARIABLE': 'value for testing'})
class TestFoo:
    ...

Mocks §

Official documentation

Mock vs patch §

If you are passing an instance of a class to the unit under test (UUT), e.g. via a constructor argument, then in your tests you can create a new Mock(spec=SomeClass) instance and pass that instead.

If your UUT creates its own instances of dependencies rather than accepting them as arguments, then you have to patch(autospec=True) the imported class to globally replace the library class with a mock.

Mocking an imported class §

package1/file1.py

from somelib import SomeClass
 
def do_something(myid):
    x = SomeClass(myid=myid)
    x.doit()

test/package1/test_file1.py

from unittest import mock
 
import pytest
 
from package1 import file1
from package1.file1 import do_something
 
 
@pytest.fixture
def mock_some_class():
    with mock.patch.object(file1, 'SomeClass', autospec=True) as mock_some_class:
        yield mock_some_class
 
 
class TestFile1:
    def test_do_something(self, mock_some_class):
        do_something('someid')
        mock_some_class.assert_called_once_with(myid='someid')
        mock_some_class.return_value.doit.assert_called_once()

How to call mock.patch and its methods depends on how the library is imported in file1.py. Generally you have to patch whatever’s present in file1.py, not the somelib.SomeClass class itself. Further details in the documentation.

Setting up failures §

The simplest way to raise an exception when a mocked method is called is like this:

mock.patch.object(
    file1,
    'SomeClass',
    side_effect=RuntimeError('fail'),
    autospec=True
)

Here’s a more complicated example in which we conditionally fail calls to the constructor of the mocked class:

test/package1/test_file1.py

from unittest import mock
from unittest.mock import call, DEFAULT
 
import pytest
 
from package1 import file1
from package1.file1 import do_something
 
 
def constructor_side_effect(myid):
    if myid == 'bad':
        raise RuntimeError('bad id')
    return DEFAULT  # ensures that mock_some_class() returns a mock as expected
 
 
@pytest.fixture
def mock_some_class():
    with mock.patch.object(
        file1,
        'SomeClass',
        side_effect=constructor_side_effect,
        autospec=True,
    ) as mock_some_class:
        yield mock_some_class
 
 
class TestFile1:
    def test_do_something(self, mock_some_class):
        do_something('someid')
 
        with pytest.raises(RuntimeError):
            do_something('bad')
 
        expected_calls = [
            call(myid='someid'),
            call(myid='bad'),
        ]
        # use any_order=True becase calls to SomeClass.doit() seem to count too
        mock_some_class.assert_has_calls(expected_calls, any_order=True)
        mock_some_class.return_value.doit.assert_called_once()

#python