Mocking Boundaries

Your function calls the network, the clock, the filesystem. Tests that hit real services are slow, flaky, and sometimes destructive. Mock the boundary:

from unittest.mock import patch

def fetch_price(url):
    import urllib.request
    with urllib.request.urlopen(url) as r:
        return int(r.read())

class TestPrice(unittest.TestCase):
    @patch("__main__.urllib.request.urlopen")
    def test_fetch_price_parses(self, fake_urlopen):
        fake_urlopen.return_value.__enter__.return_value.read.return_value = b"123"
        self.assertEqual(fetch_price("http://example.com"), 123)

patch swaps the real object for a Mock for the duration of the test and restores it afterwards. The mock records how it was called, so you can assert behavior, not just return values:

fake_send.assert_called_once_with("hello")

What to mock — and what never to mock

Mock edges: network, clock, random, filesystem, external services. Keep domain logic real — mocking the code under test's own calculations makes the test a tautology. Rule of thumb: if you're mocking it, it should be a collaborator, not the subject.

The danger sign

Tests full of mocks that mirror the implementation break on every refactor and prove nothing. Prefer testing through the public API; mock only what you cannot afford to touch for real.