pytest-dependency

Warnings

• breaking Version 0.6.0 (released Dec 2023) dropped support for Python 2. Ensure your test environment uses Python 3.
  Fix: Upgrade your Python environment to Python 3.x if you are still on Python 2.x.
• breaking Version 0.5.0 (released Feb 2020) requires `pytest` version 3.7.0 or newer. Older `pytest` versions are incompatible.
  Fix: Upgrade `pytest` to version 3.7.0 or newer using `pip install --upgrade pytest`.
• breaking Version 0.3 (released Dec 2017) changed the default naming convention for test methods within test classes. The class name is now prepended to the method name to form the default test name. If you were relying on default names for dependencies of class methods in older versions, you may need to update your `depends` references.
  Fix: If referencing test methods by their default names in `depends`, prepend the class name (e.g., `Classname::test_method`). Alternatively, use the `name` argument in `pytest.mark.dependency(name='custom_name')` for explicit naming.
• gotcha `pytest-dependency` is incompatible with `pytest-xdist` when parallelization is enabled, as dependencies assume a sequential execution order. Tests may fail or be skipped incorrectly in parallel runs.
  Fix: Do not use `pytest-xdist`'s parallelization features (e.g., `-n`) when using `pytest-dependency`. If parallelization is critical, consider refactoring tests to be independent or exploring alternative dependency management strategies.
• gotcha Test names in `depends` lists must precisely match the registered test names. For parameterized tests or tests within classes, this often means using the full `pytest` node ID (e.g., `module.py::TestClass::test_method[param]`) or explicitly assigning a `name` via `@pytest.mark.dependency(name='my_test_name')`.
  Fix: Use explicit `name` arguments for `pytest.mark.dependency` to define clear, stable dependency names. If relying on default names, inspect `pytest -v --collect-only` output to find the exact node IDs.
• gotcha Applying the `@pytest.mark.dependency()` marker multiple times to the same test will result in only the *last* invocation being effective. This can silently override previously defined dependencies or custom names.
  Fix: Ensure each test function or class has only one `@pytest.mark.dependency()` marker. Combine all `depends` arguments into a single list if necessary.
• gotcha `pytest-dependency` manages skipping based on dependency outcomes but does not reorder tests to ensure dependencies are run first. If pytest's default collection order does not satisfy the dependencies, tests might be skipped unexpectedly because their dependencies haven't been executed yet.
  Fix: Consider using `pytest-order` in conjunction with `pytest-dependency` if your test suite requires explicit reordering to satisfy dependencies. Alternatively, structure your test files and classes such that dependencies naturally run before dependent tests.

Install

• pip install pytest-dependency Install latest version

Imports

• pytest.mark.dependency

  import pytest
  @pytest.mark.dependency()

  The `dependency` marker is exposed directly via `pytest.mark`, not a separate import from `pytest_dependency` itself.

Quickstart

This example demonstrates basic test dependency management. `test_a` is marked to deliberately fail. `test_c` depends on `test_a` and will be skipped. `test_e` depends on both `test_b` (which passes) and `test_c` (which is skipped), leading to `test_e` also being skipped.

import pytest

@pytest.mark.dependency()
@pytest.mark.xfail(reason="deliberate fail", raises=AssertionError)
def test_a():
    assert False

@pytest.mark.dependency()
def test_b():
    assert True

@pytest.mark.dependency(depends=["test_a"])
def test_c():
    assert True

@pytest.mark.dependency(depends=["test_b"])
def test_d():
    assert True

@pytest.mark.dependency(depends=["test_b", "test_c"])
def test_e():
    assert True

# To run this, save as `test_deps.py` and run `pytest -rsx test_deps.py`