Reputation: 7164
According to the pytest documentation, I can generate combinations of multiple parametrized arguments as follows:
@pytest.mark.parametrize("x", [0, 1])
@pytest.mark.parametrize("y", [2, 3])
def test_foo(x, y):
pass
I can also apply marks to individual parameters as such:
@pytest.mark.parametrize("test_input,expected", [
("3+5", 8),
("2+4", 6),
pytest.param("6*9", 42,
marks=pytest.mark.xfail),
])
def test_eval(test_input, expected):
assert eval(test_input) == expected
Is there a reasonable way to combine the two methodologies and apply a mark to a particular combination of parameters? For instance, can I apply a pytest.mark.xfail
ONLY to the test_foo
instance that gets generated with x==0
and y==2
?
Upvotes: 8
Views: 3465
Reputation: 626
pytest.param
does the trick.
Look for It is also possible to mark individual test instances within parametrize
in the link above.
Upvotes: 0
Reputation: 101
A simpler solution that covers many use cases is to mark each component with a "partial tag" and then filter by tests that contain all partial tags.
For example, mark one case tag_2a
and tag_2b
then filter for -m tag_2a and tag_2b
or -m (not tag_2a and not tag_2b)
.
For xfail
you could possibly use a hook via conf.py
to automatically mark tests containing both tag_2a
and tag_2b
as xfail
.
Upvotes: 0
Reputation: 3729
The approach I prefer is to generate my arguments through a simple helper function. Have a look at the following example:
import pytest
def __get_param_xy__(x = [], y = [], xfail = []): # ugly demonstrator ...
out_list = []
for xx in x: # this could be a clever list comprehension ...
for yy in y: # this one, too ...
out_tup = (xx, yy)
if out_tup in xfail: # the ones you expect to fail
out_tup = pytest.param(*out_tup, marks = pytest.mark.xfail)
out_list.append(out_tup)
return out_list
@pytest.mark.parametrize('x,y', __get_param_xy__(
x = [0, 1],
y = [2, 3],
xfail = [(0, 2)]
))
def test_foo(x, y):
assert not (x == 0 and y == 2)
It still uses a single parametrize
decorator, but it comes fairly close to what you want and is easy to read and understand.
EDIT (1): You can actually implement the helper function as a generator. The following works just fine:
def __get_param_xy__(x = [], y = [], xfail = []): # ugly generator ...
for xx in x: # this could be a clever list comprehension ...
for yy in y: # this one, too ...
out_tup = (xx, yy)
if out_tup in xfail: # the ones you expect to fail
out_tup = pytest.param(*out_tup, marks = pytest.mark.xfail)
yield out_tup
EDIT (2): Since it has been asked in the comments, this can actually be generalized for an arbitrary number of parameters and does not conflict with fixtures. Check the following example:
import pytest
class __mock_fixture_class__:
def __init__(self):
self.vector = []
def do_something(self, parameter):
assert parameter != (0, 2)
self.vector.append(parameter)
def fin(self):
self.vector.clear()
@pytest.fixture(scope = 'function')
def mock_fixture(request):
mock_fixture_object = __mock_fixture_class__()
def __finalizer__():
mock_fixture_object.fin()
request.addfinalizer(__finalizer__)
return mock_fixture_object
def __get_param_general_generator__(*_, **kwargs):
xfail = kwargs.pop('xfail') if 'xfail' in kwargs.keys() else []
arg_names = sorted(kwargs.keys())
def _build_args_(in_tup = (), arg_index = 0):
for val in kwargs[arg_names[arg_index]]:
out_tup = (*in_tup, val)
if arg_index < len(arg_names) - 1:
yield from _build_args_(out_tup, arg_index + 1)
else:
if out_tup in xfail:
out_tup = pytest.param(*out_tup, marks = pytest.mark.xfail)
yield out_tup
return ','.join(arg_names), _build_args_()
@pytest.mark.parametrize(*__get_param_general_generator__(
x = [0, 1],
y = [2, 3],
xfail = [(0, 2)]
))
def test_foo_xy(mock_fixture, x, y):
mock_fixture.do_something((x, y))
@pytest.mark.parametrize(*__get_param_general_generator__(
x = [0, 1],
y = [2, 3],
z = [0, 1, 2],
xfail = [(0, 2, 1)]
))
def test_bar_xyz(x, y, z):
assert not (x == 0 and y == 2 and z == 1)
(Thanks to yield from
, this is Python 3.3 and above only.)
Upvotes: 5