Testing with pytest: Writing Good Tests

Why this matters

Up to now the lessons tested your code for you. Real programmers write their own tests, because tests let you change code without fear: if you break something, a test tells you at once. pytest is the most popular tool for it, and you have been using it all along.

The idea

A test is an ordinary function whose name starts with test_, containing one or more assert statements. If an assert fails, the test fails:

def add(a, b):
    return a + b

def test_add_positive_numbers():
    assert add(2, 3) == 5

def test_add_negative_numbers():
    assert add(-2, -3) == -5

Run pytest in a terminal and it finds every test_* function in files named test_*.py, runs them, and prints a summary. A failure shows the line and both sides of the comparison.

Good tests:

• are small: one idea per test, with a name that says which idea;
• cover the normal case and the edge cases: empty input, one item, negative numbers, the biggest or smallest value;
• check errors with pytest.raises:

import pytest

def test_divide_by_zero():
    with pytest.raises(ZeroDivisionError):
        1 / 0

How do you know your tests are good? Ask: “would they notice if the function were broken?” Programmers sometimes check this by deliberately breaking the code (a mutant) and confirming that some test fails. A suite that passes on broken code is not protecting you.

Try it

Run the example, then change add to return a - b. Which tests fail? Now try return abs(a) + abs(b): do both still pass or not? What does that tell you about the first test?

Your turn

The starter contains a working function slugify(text): it lowercases the text, trims it, joins the words with single hyphens, and raises TypeError if text is not a string.

Write at least four tests for it, as functions starting with test_ in the same file. The checks do two things:

1. Your tests must all pass on the correct slugify.
2. Your tests must fail on four deliberately broken versions: one that forgets to lowercase, one that mishandles spaces at the ends, one that mishandles several spaces in a row, and one that accepts non-text input.

Recap

• A test is a test_* function with assert; pytest finds and runs them.
• Test the normal case, the edge cases and the errors (pytest.raises).
• Name each test after the behaviour it checks.
• Good tests fail when the code is broken.

Common mistakes

• Testing only the happy path.
• A test with no assert, which can never fail.
• One giant test that checks ten things, so a failure does not tell you which.