mypy plugin for @mark.parametrize

The idea is that a plugin would allow the arguments to the @pytest.mark.parametrize decorator to be type-checked.

For example:

import pytest

@pytest.mark.parametrize(
    "arg",
    [
        123,        # correct type int
        456.789,    # wrong type float, mypy plugin to complain
    ]
)
def test_foo(
    arg: int   # plugin reads `int` annotation here
) -> None:
    ...

That seems like a good idea - but I don't think it should be part of pytest core.

Nice idea. I found that there is typeguard project which can be used to perform run-time type checks.

# conftest.py
import typeguard


def pytest_collection_modifyitems(session, config, items):
    for item in items:
        item.obj = typeguard.typechecked(item.obj)

pytest -vvv          
==================================== test session starts ====================================
platform linux -- Python 3.10.1, pytest-6.2.4, py-1.10.0, pluggy-0.13.1 -- /home/karthikeyan/stuff/python/myenv/bin/python
cachedir: .pytest_cache
rootdir: /tmp/testdir1
plugins: forked-1.4.0, typeguard-2.13.3, anyio-3.5.0, mypy-plugins-1.9.3, xdist-2.5.0
collected 2 items                                                                           

tests/test_foo.py::test_foo[123] PASSED                                               [ 50%]
tests/test_foo.py::test_foo[456.789] FAILED                                           [100%]

========================================= FAILURES ==========================================
_____________________________________ test_foo[456.789] _____________________________________

args = (), kwargs = {'arg': 456.789}, memo = <typeguard._CallMemo object at 0x7fd9fc7d2620>

    def wrapper(*args, **kwargs):
        memo = _CallMemo(python_func, _localns, args=args, kwargs=kwargs)
>       check_argument_types(memo)

/home/karthikeyan/stuff/python/myenv/lib/python3.10/site-packages/typeguard/__init__.py:1032: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

memo = <typeguard._CallMemo object at 0x7fd9fc7d2620>

    def check_argument_types(memo: Optional[_CallMemo] = None) -> bool:
        """
        Check that the argument values match the annotated types.
    
        Unless both ``args`` and ``kwargs`` are provided, the information will be retrieved from
        the previous stack frame (ie. from the function that called this).
    
        :return: ``True``
        :raises TypeError: if there is an argument type mismatch
    
        """
        if memo is None:
            # faster than inspect.currentframe(), but not officially
            # supported in all python implementations
            frame = sys._getframe(1)
    
            try:
                func = find_function(frame)
            except LookupError:
                return True  # This can happen with the Pydev/PyCharm debugger extension installed
    
            memo = _CallMemo(func, frame.f_locals)
    
        for argname, expected_type in memo.type_hints.items():
            if argname != 'return' and argname in memo.arguments:
                value = memo.arguments[argname]
                description = 'argument "{}"'.format(argname)
                try:
                    check_type(description, value, expected_type, memo)
                except TypeError as exc:  # suppress unnecessarily long tracebacks
>                   raise TypeError(*exc.args) from None
E                   TypeError: type of argument "arg" must be int; got float instead

/home/karthikeyan/stuff/python/myenv/lib/python3.10/site-packages/typeguard/__init__.py:875: TypeError
================================== short test summary info ==================================
FAILED tests/test_foo.py::test_foo[456.789] - TypeError: type of argument "arg" must be in...
================================ 1 failed, 1 passed in 0.14s ================================

I think that the mypy plugin couldn't work for indirect parametrisation, e.g. from the documentation:

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

This is where typeguard can help.

This one seems to not be so activate anymore, but would something like this be good practice when working with type checkers for this?
For the following above i do get incompatible types error from mypy:

import pytest

cases: list[tuple[str, str]] = [
    ("text", False),
    ("other_text", 123),
]


@pytest.mark.parametrize("input_value, expected_text", cases)
def test_example_types(
    input_value: str, expected_text: str
) -> None:
    print(input_value)
    print(expected_text)

But no errors when i run something like the following below:

import pytest

@pytest.mark.parametrize(
    "input, expected_text",
    [
        ("text", False),
        ("other_text", 123),
    ],
 )
def test_example_types(
    input: str, expected_text: str,
) -> None:
    print(input)
   print(expected_text)

Either way i have not tested the plugin but that probably works too -- also maybe this should be explained in the typing docs perhaps?
Since this can technically have "silent" errors like i showed above for certain codebases if not careful enough...

Any thoughts?

pytest-beartype is a good runtime option for this problem.

I have published mypy-pytest-parametrize with support for this (not indirect parametrisation).

I'm happy to transfer ownership if pytest-dev folks want to own it.

hard reject - at first glance this looks like a quickly agent strung name squat

we are not going to take ownership of something that looks prompt-dropped - whats wrong with people - we barely have time to react to all of the agent based bombard

Navigation Menu

Repository navigation

mypy plugin for `@mark.parametrize` #9334

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

mypy plugin for @mark.parametrize #9334

Description

Activity

The-Compiler commented on Nov 24, 2021

tirkarthi commented on Jan 15, 2022

adamtheturtle commented on Aug 21, 2024

phalberg commented on Oct 4, 2026

adamtheturtle commented on Oct 5, 2026

adamtheturtle commented on Oct 5, 2026

RonnyPfannschmidt commented on Oct 5, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

mypy plugin for `@mark.parametrize` #9334