如何在同一个测试中将 pytest 固定装置和参数化参数实现为参数？

Question

I want to create a test for a class called Graph . I've created a fixture called my_graph that just creates a Graph instance. Now I want to create a test that accepts the fixture and parametrized arguments that contain a list of inputs and expected outputs.

Here is what I've tried below but I keep getting that the vertex fixture is not found.

@pytest.fixture
def my_graph() -> adjacency_list.Graph:
    g = adjacency_list.Graph()    
    return g

@pytest.mark.parametrize("my_graph, vertex, expected",
[('A', ['A']), (1, [1])])
def test_add_vertex(my_graph: adjacency_list.Graph, vertex: Union[str, int], expected: List[Union[str, int]]):
    my_graph.add_vertex(vertex)
    assert all([a == b for a, b in zip(my_graph.get_vertices(), expected))

Answer 1

The mistake is including the my_graph fixture in the .parametrize(arg_names,...) .

Since my_graph is already a fixture, and you seem to be creating a Graph instance in the same way for all the tests, you don't need to parametrize it. Just declare it as a parameter of the test function and pytest will find your fixture:

# Remove "my_graph" from here ---▼
@pytest.mark.parametrize("vertex, expected", [("A", ["A"]), (1, [1])])
def test_add_vertex(my_graph: Graph, vertex: Union[str, int], expected: List[Union[str, int]]):
    ...

I think you got confused from the @pytest.mark.parametrize example, wherein it shows that all inputs to the test are parametrized. It doesn't have to be like that. Only those inputs that will have multiple possible values should be included in the .parametrize declaration . For other fixtures ( @pytest.fixture ), just declare them as-is in the test function parameters.

Note also that the signature for parametrize expects that the number of argnames and the number of argvalues should match:

@pytest.mark.parametrize("input_a, input_b, input_c", [
    ("param_a1", "param_b1", "param_c1"),
    ("param_a2", "param_b2", "param_c2"),
    ("param_a3", "param_b3", "param_c3"),
    ...    
])

Here, there are 3 parametrized inputs, so the list should be 3-item tuples. This is a common gotcha when parametrizing tests, and this will produce errors such as " fixture not found " or " must be equal to the number of values ".

So, if for example, you need to instantiate the Graph object differently based on the parametrized vertex and expected , then you need to also include the instantiated Graph object in the .parametrize list of tuples:

class Graph:
    def __init__(self, name: str):
        self._name = name
        self._vertices = []

...

@pytest.mark.parametrize(
    "input_graph, vertex, expected",
    [
        (Graph("Graph A"), "A", ["A"]),
        (Graph("Graph 1"), 1, [1]),
    ],
)
def test_add_vertex(input_graph, vertex: Union[str, int], expected: List[Union[str, int]]):
    input_graph.add_vertex(vertex)
    assert all([a == b for a, b in zip(input_graph.get_vertices(), expected)])

Here, instead of using the my_graph fixture, parametrize declares a differently named input_graph , and each set of test inputs instantiates a Graph object. The renaming is mostly to avoid confusion with your my_graph fixture.