pytest.mark.parametrize on multiple fixtures

[lodagro]

A little bit of demo code showing three different version of the same test.
Difference between the versions is how the parametrization is done. v0 and v1 run as expected, but on v2 the pytest.param objects are passed into the test instead of the data as for the other two versions. For v2 the ids are not used in the test name.

No idea if this a bug or an unsupported feature. It would be handy to be able to use also v2 version of the test.

from itertools import product

import pytest

COUNTRIES = [("Italy", "IT"), ("Germany", "DE")]
DISHES = [("Caprese Salad", "CS"), ("Sauerkraut", "SK")]

COUNTRY_PARAMS = [pytest.param(country, id=id) for country, id in COUNTRIES]
DISH_PARAMS = [pytest.param(dish, id=id) for dish, id in DISHES]
COUNTRY_DISH_PARAMS = [
    pytest.param(country, dish, id=f"{country_id}-{dish_id}")
    for (country, country_id), (dish, dish_id) in product(COUNTRIES, DISHES)
]


@pytest.mark.parametrize("country", COUNTRY_PARAMS)
@pytest.mark.parametrize("dish", DISH_PARAMS)
def test_dish_v0(country, dish):
    print(f"Eat {dish} in {country}")


@pytest.mark.parametrize("country,dish", COUNTRY_DISH_PARAMS)
def test_dish_v2(country, dish):
    print(f"Eat {dish} in {country}")


@pytest.mark.parametrize("country,dish", product(COUNTRY_PARAMS, DISH_PARAMS))
def test_dish_v3(country, dish):
    print(f"Eat {dish} in {country}")

❯ pytest test_food.py -s -v
======================================================== test session starts ========================================================
platform linux -- Python 3.10.0, pytest-7.4.0, pluggy-1.2.0 -- <snip>
cachedir: .pytest_cache
metadata: {'Python': '3.10.0', 'Platform': 'Linux-3.10.0-1160.42.2.el7.x86_64-x86_64-with-glibc2.17', 'Packages': {'pytest': '7.4.0', 'pluggy': '1.2.0'}, 'Plugins': {'typeguard': '3.0.2', 'repeat': '0.9.1', 'metadata': '3.0.0', 'xdist': '3.3.1', 'html': '3.2.0'}}
rootdir: /lwork/OLT/wo904355/sandbox
plugins: typeguard-3.0.2, repeat-0.9.1, metadata-3.0.0, xdist-3.3.1, html-3.2.0
collected 12 items                                                                                                                  

test_food.py::test_dish_v0[CS-IT] Eat Caprese Salad in Italy
PASSED
test_food.py::test_dish_v0[CS-DE] Eat Caprese Salad in Germany
PASSED
test_food.py::test_dish_v0[SK-IT] Eat Sauerkraut in Italy
PASSED
test_food.py::test_dish_v0[SK-DE] Eat Sauerkraut in Germany
PASSED
test_food.py::test_dish_v2[IT-CS] Eat Caprese Salad in Italy
PASSED
test_food.py::test_dish_v2[IT-SK] Eat Sauerkraut in Italy
PASSED
test_food.py::test_dish_v2[DE-CS] Eat Caprese Salad in Germany
PASSED
test_food.py::test_dish_v2[DE-SK] Eat Sauerkraut in Germany
PASSED
test_food.py::test_dish_v3[country0-dish0] Eat ParameterSet(values=('Caprese Salad',), marks=(), id='CS') in ParameterSet(values=('Italy',), marks=(), id='IT')
PASSED
test_food.py::test_dish_v3[country1-dish1] Eat ParameterSet(values=('Sauerkraut',), marks=(), id='SK') in ParameterSet(values=('Italy',), marks=(), id='IT')
PASSED
test_food.py::test_dish_v3[country2-dish2] Eat ParameterSet(values=('Caprese Salad',), marks=(), id='CS') in ParameterSet(values=('Germany',), marks=(), id='DE')
PASSED
test_food.py::test_dish_v3[country3-dish3] Eat ParameterSet(values=('Sauerkraut',), marks=(), id='SK') in ParameterSet(values=('Germany',), marks=(), id='DE')
PASSED

[RonnyPfannschmidt]

currently this is intentionally not supported, im -1 on adding support as in practice it means aritrary nesting of a intentionally not nested marker structure

we might however consider adding support for the + operator of parameterset

[lodagro]

A + operator of ParameterSet would help.

[RonnyPfannschmidt]

there is the caveat that setting the id is only partially valid/safe for the pus operator, so only no id or all id can be supported safely

[lodagro]

No id or all id is fine.
Would the same apply for the marks? We have for some ParameterSet marks, others not.

[RonnyPfannschmidt]

For marks its concatenate

[lodagro]

perfect