PTS 7: pytest fixtures

09.12.2019 - Download Code Snippet - Back to top

As code goes by my hands, my spider sense gets refined and tingles sooner than before on situations where there must be a better way to do it!.

Today insight has been on pytest fixtures. I am already used to pytest parametrization but today I was looking to my test functions with more than one assert statement - spider sense tingles…, there must be a better way!.

Multiple assert statements get wrapped around the same test umbrella and are not highlighted as test counts and, more importantly, multiple assert statements inside the same test_function sum to coupling instead of cohesion.

So, looking for a proper method to separate those multiple asserts inside my parametrized test_functions, I found the fixture implementation in pytest.

Descriptions to this approach are found when one searches for ways to reuse variables from test function to test function. However, as far as my current readings, this is a bad practice because tests should be completely independent from each other. Therefore, the best to way to reuse a variable (or treatment) in several tests is to recreate it for every test; and the best way (I found so far) to achieve this with pytest is using its fixture implementation.

Bellow is a simple transformation of a parametrized test split into two fixtures:

# pseudo code

@pytest.mark.parametrize(
    'fname',
    [
        (Path('file1')),
        (Path('file2')),
        ],
    )
def test_my_file_parser(fname):
    parser = MyFileParser(fname)
    data1 = parser.data_foo()
    assert isinstance(data1, list)
    assert len(data1) == 50


# bellow the fixture approach
@pytest.fixture(
    params=[
        Path('file1'),
        Path('file2'),
        ],
    )
def parser_data1(request):
    parser = MyFileParser(request.param)
    data1 = parser.data_foo()
    return data1


def test_parser_data1_type(parser_data1):
    assert isinstance(parser_data1, list)


def test_parser_data1_len(parser_data1):
    assert len(parse_data1) == 50

This setup will created a product of four tests: two for each of the two params in the fixture decorator.

Pytest fixtures provide much more advanced options that I don’t fully understand at this point but which I will be willing to explore!

Is there something that could be improved on this entry?i Share your thoughts, comment on this post here.

Further readings

• pytest docs fixture
• fixtures and parameters at Medium
• StackOverFlow 47402435

tags

• python
• pytest
• test-driven-development
• pytest-fixtures
• pytest-parametrize