Quick checks with assert

20 min · Intermediate

After you trace a bug once, write a test so it cannot return. An assert is the smallest test in Python. This lesson covers assert, tables of cases, property checks and the float trap.

assert says what must be true

assert is a statement that checks a condition and raises AssertionError, with an optional message, when the condition is false. It is used to test your own code.

After you write a function, write down what it must return. An assert statement checks that claim and stops the program with an error if it is false. When it is true, nothing happens.

Python

1def total(prices):2    return sum(prices)3
4assert total([10, 20]) == 305assert total([]) == 06assert total([-5, 5]) == 0, "negatives must cancel"7print("all good")

Example 1 · Three asserts and a message

Python

1def largest(values):2    return max(values)3
4assert largest([3, 9, 4]) == 95assert largest([-2, -7]) == -26assert largest([6]) == 67print("3 checks passed")

Output

3 checks passed

What this code does

No output appears from a passing assert; the final print shows the run reached the end.
The all-negative and one-item cases are in the list on purpose.

Each assert holds a condition. The text after the comma is shown only when the condition is false.

A table of cases

Writing one assert per case gets long. Put the inputs and the expected answers in a list of pairs, and check them in one loop. A new case is then one new line.

Python

1def clamp(x, low, high):2    return max(low, min(x, high))3
4cases = [5    ((5, 0, 10), 5),6    ((-3, 0, 10), 0),7    ((99, 0, 10), 10),8    ((0, 0, 10), 0),9]10for args, expected in cases:11    got = clamp(*args)12    assert got == expected, f"{args}: got {got}, expected {expected}"13print("passed", len(cases))

Example 2 · A runner that reports every failure

Python

1def double(x):2    return x * x3
4cases = [(2, 4), (3, 6), (0, 0), (5, 10)]5failed = 06for x, expected in cases:7    got = double(x)8    if got != expected:9        failed += 110        print(f"FAIL double({x}): got {got}, expected {expected}")11print(failed, "failed of", len(cases))

Output

FAIL double(3): got 9, expected 6
FAIL double(5): got 25, expected 10
2 failed of 4

What this code does

A plain assert stops at the first failure. This runner counts them all.
The bug here is x * x where x * 2 was meant; it passes for 2 and 0.

The message names the failing input and both values, so a failure tells you what to trace. *args unpacks the tuple into the three parameters.

Check properties, not only values

For some functions you cannot write every expected answer. You can still check facts that must hold. A sorted list must be in order and must keep the same items.

Python

1import random2
3random.seed(3)4for _ in range(200):5    data = [random.randint(-9, 9) for _ in range(random.randint(0, 8))]6    out = sorted(data)7    assert len(out) == len(data)8    assert all(out[i] <= out[i + 1] for i in range(len(out) - 1))9    assert sorted(out) == sorted(data)10print("properties hold")

Example 3 · Properties of a deduplicate function

Python

1def dedupe(items):2    seen = set()3    out = []4    for x in items:5        if x not in seen:6            seen.add(x)7            out.append(x)8    return out9
10for data in ([], [1], [2, 2, 2], [3, 1, 3, 2, 1]):11    out = dedupe(data)12    assert len(out) == len(set(data))13    assert set(out) == set(data)14    assert all(data.index(a) < data.index(b) for a, b in zip(out, out[1:]))15    print(data, "->", out)

Output

[] -> []
[1] -> [1]
[2, 2, 2] -> [2]
[3, 1, 3, 2, 1] -> [3, 1, 2]

What this code does

Three facts must hold: no duplicates, no lost items and the first-seen order.
The loop covers empty, one item, all equal and a mixed list.

Three properties together catch a lost item, a wrong order and a changed value. The seed makes any failure repeatable.

assert is for tests, not for user input

Python can run with -O, which removes every assert. Use assert to test your own code and to state things that must be true. For bad input from a user, raise an exception instead.

Python

1def safe_div(a, b):2    if b == 0:3        raise ValueError("b must not be zero")4    return a / b5
6try:7    safe_div(4, 0)8except ValueError as err:9    print("caught:", err)

A raised ValueError always runs. An assert might be removed, so it must never guard real input.

Compare floats with math.isclose

Decimal fractions are stored in binary, so some sums are not exact. Never assert that two floats are equal with ==. Ask whether they are close.

Python

import math
assert 0.1 + 0.2 != 0.3assert math.isclose(0.1 + 0.2, 0.3)print(0.1 + 0.2)

Example 4 · Floats need isclose

Python

import math
print(0.1 + 0.2 == 0.3)print(math.isclose(0.1 + 0.2, 0.3))print(round(0.1 + 0.2, 2) == 0.3)

Output

False
True
True

What this code does

Direct == is False because 0.1 + 0.2 is stored as 0.30000000000000004.
Both isclose and rounding give True; isclose is the clearer choice.

math.isclose allows a tiny relative error, which is what you want for averages and prices.

Next step

Check your understanding

11 quick questions

assert says what must be true

assert is a statement that checks a condition and raises AssertionError, with an optional message, when the condition is false. It is used to test your own code.

After you write a function, write down what it must return. An assert statement checks that claim and stops the program with an error if it is false. When it is true, nothing happens.

1def clamp(x, low, high):2 return max(low, min(x, high))3 4cases = [5 ((5, 0, 10), 5),6 ((-3, 0, 10), 0),7 ((99, 0, 10), 10),8 ((0, 0, 10), 0),9]10for args, expected in cases:11 got = clamp(*args)12 assert got == expected, f"{args}: got {got}, expected {expected}"13print("passed", len(cases))

1def double(x):2 return x * x3 4cases = [(2, 4), (3, 6), (0, 0), (5, 10)]5failed = 06for x, expected in cases:7 got = double(x)8 if got != expected:9 failed += 110 print(f"FAIL double({x}): got {got}, expected {expected}")11print(failed, "failed of", len(cases))

1import random2 3random.seed(3)4for _ in range(200):5 data = [random.randint(-9, 9) for _ in range(random.randint(0, 8))]6 out = sorted(data)7 assert len(out) == len(data)8 assert all(out[i] <= out[i + 1] for i in range(len(out) - 1))9 assert sorted(out) == sorted(data)10print("properties hold")

1def dedupe(items):2 seen = set()3 out = []4 for x in items:5 if x not in seen:6 seen.add(x)7 out.append(x)8 return out9 10for data in ([], [1], [2, 2, 2], [3, 1, 3, 2, 1]):11 out = dedupe(data)12 assert len(out) == len(set(data))13 assert set(out) == set(data)14 assert all(data.index(a) < data.index(b) for a, b in zip(out, out[1:]))15 print(data, "->", out)