Testing the Backoff-Utils¶
There are many schools of thought when it comes to test design. When building the Backoff-Utils, we decided to focus on practicality. That means:
- DRY is good, KISS is better. To avoid repetition, our test suite makes extensive use of fixtures, parametrization, and decorator-driven behavior. This minimizes the number of test functions that are nearly-identical. However, there are certain elements of code that are repeated in almost all test functions, as doing so will make future readability and maintenance easier.
- Coverage matters…kind of. We have documented the primary intended behavior of every method in the Backoff-Utils library, and the most-likely failure modes that can be expected. At the time of writing, we have about 80% code coverage. Yes, yes: We know that is less than 100%. But there are edge cases which are almost impossible to bring about, based on confluences of factors in the wide world. Our goal is to test the key functionality, and as bugs are uncovered to add to the test functions as necessary.
Each individual test module (e.g.
test_decorator.py) corresponds to a
conceptual grouping of functionality. For example:
test_decorator.pytests the decorator functions found in
Certain test modules may be tightly coupled, as the behavior in one test module may
have implications on the execution of tests in another. These test modules use
a numbering convention to ensure that they are executed in their required order,
test_1_NAME.py is always executed before
The Backoff-Utils do not use any custom command-line options in their test suite.
For a full list of the CLI options, including the defaults available, try:
backoff-utils $ cd tests/ backoff-utils/tests/ $ pytest --help
Because the Backoff-Utils has a very simple test suite, we have not
pytest.ini configuration file.
Because of the simplicity of the Backoff-Utils, the test suite does not currently support any test skipping.
The Backoff-Utils test suite does support incremental testing using, however at the moment none of the tests designed rely on this functionality.
A variety of test functions are designed to test related functionality. As a
result, they are designed to execute incrementally. In order to execute tests
incrementally, they need to be defined as methods within a class that you decorate
@pytest.mark.incremental decorator as shown below:
@pytest.mark.incremental class TestIncremental(object): def test_function1(self): pass def test_modification(self): assert 0 def test_modification2(self): pass
This class will execute the
TestIncremental.test_function1() test, execute and
fail on the
TestIncremental.test_modification() test, and automatically fail
TestIncremental.test_modification2() because of the
To pass state between incremental tests, add a
state argument to their method
definitions. For example:
@pytest.mark.incremental class TestIncremental(object): def test_function(self, state): state.is_logged_in = True assert state.is_logged_in = True def test_modification1(self, state): assert state.is_logged_in is True state.is_logged_in = False assert state.is_logged_in is False def test_modification2(self, state): assert state.is_logged_in is True
Given the example above, the third test (
test_modification2) will fail because
test_modification updated the value of
state is instantiated at the level of the entire test session (one run of
the test suite). As a result, it can be affected by tests in other test modules.