Fixtures: dependency injection for tests

A fixture is a factory function pytest injects into any test that names it as an argument.

Why they matter

Without fixtures, every test starts with 5 lines of setup. With fixtures:

• Setup lives in one place.
• Each test gets a fresh instance (default scope="function").
• Fixtures can depend on other fixtures — composition for free.

Scopes

• function (default) — new per test.
• class — shared across tests in one class.
• module — shared across a file.
• session — created once. Use for expensive stuff (real DB, browser).

Mocking with monkeypatch

The best mocking tool comes built into pytest:

def test_something(monkeypatch):
    monkeypatch.setattr(mymodule, "send_email", fake_send)
    # ... test uses fake_send until test ends, then auto-restored

Cleaner than unittest.mock.patch for most cases. Auto-undoes at test teardown.

unittest.mock when you need it

For call-tracking, MagicMock is unbeatable:

from unittest.mock import MagicMock
mock_client = MagicMock()
mock_client.fetch.return_value = {"ok": True}
service = MyService(mock_client)
service.do_thing()
mock_client.fetch.assert_called_once_with("some-arg")

Try it

1. Add a bob fixture that depends on repo.
2. Write a test that monkeypatches send_welcome_email to raise, and asserts signup still returns a valid uid (or handles the error — your choice, document it).