Testing

Programming is not just about writing code. It is also about verifying that the code does what it is supposed to do and, if needed, correcting it. This process of verification is called testing.

You have probably already tested your programs by executing them. When you test your program, you usually enter some input data and look if the result is correct. This is okay for a small program, but it gets harder as the program gets bigger. Bigger programs have more options what they can do based on the possible user input and configuration. Their manual testing becomes time-consuming, especially when it needs to be repeated after every change, and it becomes more likely errors slip unnoticed into our code.

Humans are not very good at performing boring repetitive tasks, that is the domain of computers. And, not surprisingly, that is the reason why developers write the code that verifies their programs.

Automated tests are functions checking, with no manual intervention, that all features of the tested program work correctly. If made right, the tests should be lean and give quick response whether the program has issues or not. The testing does not give us 100% proof that the code is without errors but it is still better than no testing at all.

The automated tests make modification of the code easier as you can faster find possible bugs in the existing functionality (aka regressions).

Installing the pytest library

Up to now, we have used only the modules that come installed with Python, for example, modules such as math or turtle. There are many more libraries that are not included in Python but you can install them to your Python environment and use them.

There is a built-in library for testing called unittest. It is not particularly easy to use and we decided to go with the pytest library instead which is not only easier to use but also popular among Python developers.

To install the library run the following command in the command line:

> python3 -m pip install pytest

$ python3 -m pip install pytest

Writing tests

We will show testing with a very simple example. There is a function add that adds two numbers. There is another function test_add that tests if the add function returns correct results for specific numbers.

Make a copy of the code into a file named test_addition.py in a new empty directory.

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

def test_add():
    assert add(1, 2) == 3

What does the test function do?

The assert statement evaluates the logical expression that follows it. If the result is not True then it raises the AssertionError exception which is interpreted by pytest as a failing test.

E.g, the assert a == b command is equivalent to:

if not (a == b):
    raise AssertionError

Running tests

You execute tests with the command python -m pytest -v <path> followed by the path to the file containing the tests.

This command scans the given file and calls all functions that start with the test_ prefix. It executes them and checks if they raise any exception, e.g., raised by the assert statement.

$ python3 -m pytest -v test_addition.py
============================= test session starts ==============================
platform linux -- Python 3.8.3, pytest-7.1.2, pluggy-1.0.0
rootdir: /tmp/test_example
collected 1 item

test_addition.py .                                                       [100%]

============================== 1 passed in 0.00s ===============================

If an exception occurs, pytest shows a red message with additional details that can help you find the bug and fix it:

============================= test session starts ==============================
platform linux -- Python 3.8.3, pytest-7.1.2, pluggy-1.0.0
rootdir: /tmp/test_example
collected 1 item

test_addition.py F                                                       [100%]

=================================== FAILURES ===================================
___________________________________ test_add ___________________________________

    def test_add():
>       assert add(1, 2) == 3
E       assert 4 == 3
E        +  where 4 = add(1, 2)

test_addition.py:5: AssertionError
=========================== short test summary info ============================
FAILED test_addition.py::test_add - assert 4 == 3
============================== 1 failed in 0.01s ===============================

Try to run the test yourself. Modify the add function or (its test) so that the test fails.

Test modules

It is common to write tests separately (in another file) from the tested regular code. This way, the code is easier to read and it can be distributed without the tests, if desired.

Let's split the test_addition.py into two separate files, moving the add() function into a new module addition.py and keeping the tests in the old test_addition.py file. Import the tested function from the new module.

Run the test and see if it is passing.

Let's now try to add two different tests for a function for computing perimeter of a rectangle from custom functions

def find_rectangle_perimeter(width, height):
    """ Calculate perimeter of a rectangle from the given sides.
    """
    return  2 * (width  +  height)

# Do not forget to import the find_rectangle_perimeter function.
from addition import find_rectangle_perimeter

def test_find_rectangle_perimeter():
    """ Tests if the function can handle two positive integer values as input.
    """
    assert find_rectangle_perimeter(4, 5) == 18

def test_find_rectangle_perimeter_zero_width():
    """ Tests if the function can handle width set to 0.
    """
    assert find_rectangle_perimeter(0, 3) == 6

Executable modules

Automated tests have to be able to run unattended. They are often executed automatically and the failures are reported via some sort of notification, e.g., by email.

In practical terms, this means that the tests must not depend on live interaction with the user, e.g., the input function will not work in tests.

This can make your work harder sometimes. Let's look at a more complex project, the 1D (one-dimensional) tic-tac-toe.

The structure of the 1D tic-tac-toe code looks roughly like this:

import random  # (and possibly other import statements that are needed)

def move(board, space_number, mark):
    """Return the board with the specified mark placed in the specified position"""
    ...

def player_move(board):
    """Ask the player what move should be done and return the updated board
    with the move played.
    """
    ...
    input('What is your move? ')
    ...

def tic_tac_toe_1d():
    """Start the game

    It creates an empty board and run player_move and computer_move alternately
    until the game is finished.
    """
    while ...:
        ...
        player_move(...)
        ...

# Start the game:
tic_tac_toe_1d()

If you import this module, Python executes all commands in it, from top to bottom:

• The first command, import, initializes the variables and functions of the random module. It is module from the standard Python library it is unlikely that it would have any side effect to worry about.

• The definitions of functions (def statements and everything in them) just define the functions but they do not execute them.

• Calling the tic_tac_toe_1d function starts the game. The tic_tac_toe_1d calls the player_move() function which calls input(). This is an issue.

If you import this module to the tests, the input fails and the module does not get not imported.

The calling of tic_tac_toe_1d is a side-effect and we need to remove it. Okay, but you cannot start the game without it! What can we do about it?

There are two possible solutions. First, we detect if the module is imported from another script or it is itself the main script and start the game only if it runs as the __main__ module:

if __name__ == "__main__":
    # Start the game:
    tic_tac_toe_1d()

Second, we can create a new python file, e.g., game.py and we move the tic_tac_toe_1d() call in it:

import tic_tac_toe

tic_tac_toe.tic_tac_toe_1d()

Obviously, the game.py itself, cannot be tested because it calls input indirectly. But almost empty and you can execute it only if you want to play.

After the fix, we can import the original module into the test module or other scripts.

A test for the original module could look like this:

import tic_tac_toe

def test_move_to_empty_space():
    board = tic_tac_toe.computer_move("-" * 20)
    assert len(board) == 20
    assert board.count("x") == 1
    assert board.count("-") == 19

Positive and negative tests

Tests that verify that a program works correctly under correct conditions are called positive tests. An exception raised during the positive testing lead to failure of the test.

Tests that check behaviour in case of invalid inputs are called negative tests. The purpose of the negative testing is verification of the graceful handling of error states. Raising of an exception is often the expected behaviour of the tested code.

For example, the computer_move function should raise an error (e.g., ValueError) when the board is full.

To test if your code raises an exception, use the with statement and the pytest.raises() context manager.

import pytest
import tic_tac_toe

def test_move_failure():
    with pytest.raises(ValueError):
        tic_tac_toe.computer_move("ox" * 10)

Let's now try to edit the function for getting a perimeter of rectangle so that it raises a ValueError if any of the sides is smaller or equal to zero. Add a negative test checking that the exception is raised as expected.

import pytest

def find_rectangle_perimeter(width, height):
    """ Calculate perimeter of a rectangle from the given sides.
    """
    if width < 0 or height < 0:
        raise ValueError("Rectangle sides must be non-negative.")
    return  2 * (width + height)

def test_find_perimeter_exception_negative():
    """ Tests of negative integer values as input
    """
    with pytest.raises(ValueError):
        find_rectangle_perimeter(-3, 5)