The purpose of indirect=True in the pytest.mark.parametrize decorator was not very clear to me just by reading pytest documentation:

Using the indirect=True parameter when parametrizing a test allows to parametrize a test with a fixture receiving the values before passing them to a test

import pytest

@pytest.fixture
def fixt(request):
    return request.param * 3

@pytest.mark.parametrize("fixt", ["a", "b"], indirect=True)
def test_indirect(fixt):
    assert len(fixt) == 3

I found this definition rather unclear, and a bit of a mouthful. So I just created a toy example to test it’s purpose… (Could this be meta-testing? 🤔)

Hands-on!

After writing some toy examples, the purpose of indirect becomes apparent when we remove it:

@pytest.mark.parametrize("fixt", ["a", "b"])
def test_indirect(fixt):
    assert len(fixt) == 3

This is what we get after running pytest:

collected 2 items

tests/test_indirect.py::test_indirect[a] FAILED            [ 50%]
tests/test_indirect.py::test_indirect[b] FAILED            [100%]

============================ FAILURES ============================
________________________ test_indirect[a] ________________________

my_fixture = 'a'

    @pytest.mark.parametrize("my_fixture", ["a", "b"])
    def test_indirect(my_fixture):
>       assert len(my_fixture) == 3
E       AssertionError: assert 1 == 3
E        +  where 1 = len('a')

tests/test_indirect.py:12: AssertionError
________________________ test_indirect[b] ________________________

my_fixture = 'b'

    @pytest.mark.parametrize("my_fixture", ["a", "b"])
    def test_indirect(my_fixture):
>       assert len(my_fixture) == 3
E       AssertionError: assert 1 == 3
E        +  where 1 = len('b')

tests/test_indirect.py:12: AssertionError
==================== short test summary info =====================
FAILED tests/test_indirect.py::test_indirect[a] - AssertionError: assert 1 == 3
FAILED tests/test_indirect.py::test_indirect[b] - AssertionError: assert 1 == 3
======================= 2 failed in 0.01s ========================

What we can see is that parametrize is directly passing each parameter value to the fixt test argument. Therefore, it ignores the fixture and parametrizes the test.

Should we want to parametrize the fixture back, we would set indirect=True, and so the tests pass again:

collected 2 items

tests/test_indirect.py::test_indirect[a] PASSED            [ 50%]
tests/test_indirect.py::test_indirect[b] PASSED            [100%]

======================= 2 passed in 0.00s ========================

pytest would still spawn 2 tests cases—one for each parameter—but this time, the test would receive the result of the fixture, and parametrize would send the argument to the fixture instead of the test.

Hence we can conclude that indirect controls whether parametrize sends the arguments to either the test, or a fixture with the same argument’s name. If we tried to use indirect=True, and there was no fixture fixt, we would get an error:

_______________ ERROR at setup of test_indirect[a] _______________
file /Users/alberto/Code/pytest-indirect-parametrization/tests/test_indirect.py, line 10
  @pytest.mark.parametrize("fixt", ["a", "b"], indirect=True)
  def test_indirect(fixt):
E       fixture 'fixt' not found

pytest is great!

pytest’s fixture system is awesome. I can declare test resources in a very modular, flexible fashion. Then, I can use those resources in my tests, and pytest takes care of running those fixtures for me, and cleaning-up those test resources when each test case is finished. With parametrize, I can ensure to cover more cases without having to write more tests. And, if I have fixtures that should be parametrized, I can tweak those parameters from each particular test.

I wrote this piece back in the day as an exercise for me to understand a bit better what I could achieve with pytest. I hope it can be useful to somebody else!

Happy testing! 🧪