Accept a Mapping of `{id: {arg-name: arg-value, ...}, ...}` for `@pytest.mark.parametrize`

[seandstewart]

What's the problem this feature will solve?

I find myself writing many highly-parametrized tests. One of the pain-points arises when supplying an array of test IDs for a long array of cases in argvalues.

Describe the solution you'd like

One of the ways I work around this is by building a dictionary like so (a trivial example):

test_matrix = dict(
    foo_is_not_bar=("foo", "bar")
)

@pytest.mark.parametrize(
    argnames="input,negation"
    argvalues=test_matrix.values()
    ids=test_matrix.keys()
)
def test_negation(input, negation):
    assert input != negation

A real-world example:

https://github.com/seandstewart/typical/blob/5c609a0091c2500319cf714e3f86e907db8c717f/tests/ext/schema/test_json_schema.py#L41-L413

As you can see, the more cases we try to pass through a parametrized test, the further the test matrix gets from the test definition.

Alternative Solutions

Use Parametrized Fixture(s)

This could also be implemented via a parametrized fixture, but has the same drawback of isolating the test case from from test definition, so it becomes difficult to understand a test from a single look.

Additional context

Essentially, I'm looking for a way to define my test inputs and IDs in a unified manner, as close to the test definition as possible. I think the ability to define these as an object which I pass to parametrize(...) makes sense, but I'm open to alternatives.

[The-Compiler]

Duplicate of #7977