unittest-example

An example of how to use the unittest module in python.

Unit Testing

Unit testing is a technique for isolating "units" of code and sujecting them to tests which help identify the unintended consiquences of changes or keep track of errors in an actively developing codebase.

One of the benefits of isolating code units is that they can be tested independently.

Unit Testing in Python

Python's standard libraries come with the unittest package which can be used to build whole unit test suites. It provides a simple api for writing tests along with the ability to run the whole suite and recursively discover tests.

Semantics, Syntax and Style

In our current example we have the following files.

└── src
    └── foo
        ├── __init__.py
        ├── tests.py
        └── utils.py

With the function in our utils method.

# src/foo/utils.py
def standard_deviation(values):
    n = float(len(values))
    mean = sum(values) / n
    sum_squared_variances = sum([(x - mean) ** 2 for x in values])
    return (sum_squared_variances / n) ** .5

Suppose we wanted to ensure that this function worked as we imagined it. To do this we write a unittest in the tests.py module.

import unittest
import scipy
from .utils import standard_deviation


class StandardDeviationTest(unittest.TestCase):

    def test_simple(self):
        self.assertEqual(
            standard_deviation([1,2,3]),
            0.816496580927726
        )

    def test_against_scipy(self):
        for t in range(10):
            n = scipy.random.randint(10, 100)
            values = scipy.randn(n)
            self.assertTrue(
                abs(standard_deviation(values) - scipy.std(values)) < 10 ** -14
            )

This test file illustrates two different ways to write a test.

Please note that both tests make use of the self.assertXxxx pattern. This pattern is available for a variety of different assertions and you should make use of them. They will give you useful error messages.

In the second test example, the self.assertTrue function gets called because scipy and my naive implementation of starndard_deviation don't round in the same ways.

Running Tests

Finally, to run all the tests in your src folder can execute this from the root of the repository.

ipython -m unittest discover src

Note that if you are already in the src directory you can omit the last argument.