nick-sullivan / pytest-topo Goto Github PK

pip install pytest-topo

The pytest-topo library offers a topological ordering feature for pytest tests, ensuring that dependent tests are run in the correct sequence. This is particularly useful for integration testing libraries that use pytest for orchestration.

Please note that this library is not recommended for unit tests, as unit tests should ideally be independent of each other.

To sequence your tests, apply the dependency and depends_on markers.

import pytest

@pytest.mark.dependency("One")
def test_one():
    # This test is executed first

@pytest.mark.depends_on("Two")
def test_three():
    # This test is executed third

@pytest.mark.depends_on("One")
@pytest.mark.dependency("Two")
def test_two():
    # This test is executed second

A test will be skipped if any of their dependencies do not pass. This will also skip any fixture creation if there are no remaining tests that need to use them.

import pytest

@pytest.mark.dependency("Fail")
def test_one():
    assert False  # force a failure

@pytest.fixture(scope="function")
def my_fixture():
    # This fixture is never created

@pytest.mark.depends_on("Fail")
def test_two(my_fixture):
    # This test will be skipped

Consider software where customers register with an account API, and submit purchases to a purchases API. These are two services with API tests that could look like this:

tests
  - account
    - test_create_and_delete
    - ...
  - purchases
    - test_make_purchase
    - ...

The purchases tests may run a pre-test setup to create an account, execute tests, then run a post-test teardown. If there is an issue in the account creation or deletion, all these tests will error. By adding dependency links, the purchases tests will be skipped, reducing test failure noise.

There is some support for operating with pytest-xdist. When using the --dist loadgroup option, groups of connected tests will register as independent load groups. For example, when running the following tests with pytest-xdist installed and with the command pytest -n 2 --dist loadgroup, it will run two workers in parallel:

import pytest

@pytest.mark.dependency("A-1")
def test_a_one():
    # Runs on worker A

@pytest.mark.depends_on("A-1")
def test_a_two():
    # Runs on worker A

@pytest.mark.dependency("B-1")
def test_b_one():
    # Runs on worker B

@pytest.mark.depends_on("B-1")
def test_b_two():
    # Runs on worker B