Unit Tests — Concept & Examples

What They Test

One function, one method, one class. No database, no API calls, no browser. Code calls code, result is checked. That's it.

Function → [input] → Unit test → assert result

---

Why They Are the Foundation

Property	Value
Speed	0.001–0.01 s per test
Count	Thousands on a real project
Feedback	Instant — developer runs before committing
Tool	pytest

Fast feedback is the entire point. A developer changes the password validation logic, runs the tests, sees a failure in 2 seconds — not after deploy, not after manual QA.

---

What to Test at Unit Level

• Business logic (discounts, calculations, rules)
• Validation (input checks, error messages)
• Data transformations (formatters, parsers, serializers)
• Edge cases (empty input, None, zero, max value)
• Error paths (expected exceptions with correct messages)

---

Example: Password Validator

import pytest


def validate_password(password: str) -> list[str]:
    errors = []
    if len(password) < 8:
        errors.append("at least 8 characters")
    if not any(char.isupper() for char in password):
        errors.append("one uppercase letter")
    if not any(char.isdigit() for char in password):
        errors.append("one digit")
    if not any(char in "!@#$%^&*" for char in password):
        errors.append("one special character")
    return errors


@pytest.mark.unit
class TestValidatePassword:
    def test_valid_password(self) -> None:
        assert validate_password("Str0ng!Pass") == []

    def test_too_short(self) -> None:
        assert "at least 8 characters" in validate_password("Ab1!")

    def test_no_uppercase(self) -> None:
        assert "one uppercase letter" in validate_password("weak1!pass")

    def test_no_digit(self) -> None:
        assert "one digit" in validate_password("NoDigits!Here")

    def test_no_special_char(self) -> None:
        assert "one special character" in validate_password("NoSpecial1")

    def test_empty_string_returns_all_errors(self) -> None:
        assert len(validate_password("")) == 4

Six tests, under a second. Covers valid case, each rule separately, and worst-case input.

---

Toolchain Requirements

Without these enforced in CI, unit test suites quietly rot:

Tool	Purpose
coverage	Track which lines are covered; reject PRs below threshold
ruff	Lint — catch style issues and common bugs
mypy	Type checking — catch type mismatches before runtime

# Run with coverage
uv run pytest --cov=src --cov-report=term-missing tests/unit/

# Enforce threshold
uv run pytest --cov=src --cov-fail-under=95 tests/unit/

---

Pytest Marks for Layer Separation

# conftest.py
import pytest

def pytest_configure(config: pytest.Config) -> None:
    config.addinivalue_line("markers", "unit: unit tests — no I/O")
    config.addinivalue_line("markers", "integration: integration tests — real services")
    config.addinivalue_line("markers", "e2e: end-to-end browser tests")

Run only unit tests in the fast feedback loop:

uv run pytest -m unit          # only unit
uv run pytest -m "not e2e"    # everything except E2E

---

Isolation Contract

A unit test must not: - connect to a database - make HTTP requests - read from the filesystem (unless testing I/O code specifically) - depend on environment variables being set

If it does any of these, it is an integration test — slow, fragile, and failing on CI because the dev's local Postgres isn't running in the container. Use mocks (unittest.mock, pytest-mock) to isolate external dependencies at this level.