问题
I am trying to test a math function that I wrote. I would like to supply data to it from a number of different fixtures. The issue is that all of the fixtures accept different fixture parameters of their own.
The test that I run is always the same (test_myfunc in the example), and the fixtures that I want to plug into it all have the same compatible return values (clean_data and noisy_data in the code). I would therefore like to "chain" these two fixtures together so that one or the other will provide inputs to the test.
Here is what the setup looks like:
import numpy as np
import pytest
from scipy import stats
def myfunc(x, y):
return True
_noises = {
'normal': lambda scale, n: np.random.normal(scale=scale, size=n),
'uniform': lambda scale, n: np.random.uniform(-scale, scale, size=n),
'triangle': lambda scale, n: np.random.triangular(-scale, 0, scale, size=n),
}
@pytest.fixture(params=[10**x for x in range(1, 4)])
def x_data(request):
""" Run the test on a few different densities """
return np.linspace(-10, 10, request.param)
@pytest.fixture(params=[0, 1, 0xABCD, 0x1234])
def random_seed(request):
""" Run the test for a bunch of datasets, but reporoducibly """
np.random.seed(request.param)
@pytest.fixture(params=np.arange(0.5, 5.5, 0.5))
def shape(request):
""" Run the test with a bunch of different curve shapes """
return request.param
@pytest.fixture()
def clean_data(x_data, shape):
""" Get a datset with no noise """
return shape, stats.gamma.pdf(x_data, shape)
@pytest.fixture(params=["triangle", "uniform", "normal"])
def noisy_data(request, clean_data, random_seed):
shape, base = clean_data
noise = _noises[request.param](10, base.shape)
return shape, base + noise
def test_myfunc(x_data, data):
shape, y_data = data
assert myfunc(x_data, y_data)
The reason that I am using so many fixtures is that I want to run the complete matrix of tests, with the ability to enable, disable, xfail, etc. any of them at will.
Since the clean_data and noisy_data fixtures return the same type of result, I would like to be able to use both of them for my test, one after the other. How do I run a single test with multiple fixtures that accept arguments?
If possible, I would like to avoid test generation. I am familiar with the idea of indirectly parametrizing the test, for example as in Running the same test on two different fixtures. I have tried to create a meta-fixture that can execute the y-data providers by name:
@pytest.fixture()
def data(request):
""" Get the appropriate datset based on the request """
return request.getfuncargvalue(request.param)
@pytest.mark.parametrize('data', ['clean_data', 'noisy_data'], indirect=True)
def test_myfunc(x_data, data):
shape, y_data = data
assert myfunc(x_data, y_data)
When I run the tests with
pytest -v pytest-parametrized.py
I get a slew of errors, which all seem to point at the fact that the indirected fixture require parameters, which aren't supplied:
_________________ ERROR at setup of test_myfunc[10-clean_data] _________________
self = <_pytest.python.CallSpec2 object at 0x7f8a4ff06518>, name = 'shape'
def getparam(self, name):
try:
> return self.params[name]
E KeyError: 'shape'
/usr/lib/python3.6/site-packages/_pytest/python.py:684: KeyError
During handling of the above exception, another exception occurred:
self = <SubRequest 'clean_data' for <Function 'test_myfunc[10-clean_data]'>>
fixturedef = <FixtureDef name='shape' scope='function' baseid='pytest-parametrized.py' >
def _compute_fixture_value(self, fixturedef):
"""
Creates a SubRequest based on "self" and calls the execute method of the given fixturedef object. This will
force the FixtureDef object to throw away any previous results and compute a new fixture value, which
will be stored into the FixtureDef object itself.
:param FixtureDef fixturedef:
"""
# prepare a subrequest object before calling fixture function
# (latter managed by fixturedef)
argname = fixturedef.argname
funcitem = self._pyfuncitem
scope = fixturedef.scope
try:
> param = funcitem.callspec.getparam(argname)
/usr/lib/python3.6/site-packages/_pytest/fixtures.py:484:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
self = <_pytest.python.CallSpec2 object at 0x7f8a4ff06518>, name = 'shape'
def getparam(self, name):
try:
return self.params[name]
except KeyError:
if self._globalparam is NOTSET:
> raise ValueError(name)
E ValueError: shape
/usr/lib/python3.6/site-packages/_pytest/python.py:687: ValueError
During handling of the above exception, another exception occurred:
request = <SubRequest 'data' for <Function 'test_myfunc[10-clean_data]'>>
@pytest.fixture()
def data(request):
""" Get the appropriate datset based on the request """
> return request.getfuncargvalue(request.param)
pytest-parametrized.py:55:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
/usr/lib/python3.6/site-packages/_pytest/fixtures.py:439: in getfuncargvalue
return self.getfixturevalue(argname)
/usr/lib/python3.6/site-packages/_pytest/fixtures.py:430: in getfixturevalue
return self._get_active_fixturedef(argname).cached_result[0]
/usr/lib/python3.6/site-packages/_pytest/fixtures.py:455: in _get_active_fixturedef
self._compute_fixture_value(fixturedef)
/usr/lib/python3.6/site-packages/_pytest/fixtures.py:526: in _compute_fixture_value
fixturedef.execute(request=subrequest)
/usr/lib/python3.6/site-packages/_pytest/fixtures.py:778: in execute
fixturedef = request._get_active_fixturedef(argname)
/usr/lib/python3.6/site-packages/_pytest/fixtures.py:455: in _get_active_fixturedef
self._compute_fixture_value(fixturedef)
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
self = <SubRequest 'clean_data' for <Function 'test_myfunc[10-clean_data]'>>
fixturedef = <FixtureDef name='shape' scope='function' baseid='pytest-parametrized.py' >
def _compute_fixture_value(self, fixturedef):
"""
Creates a SubRequest based on "self" and calls the execute method of the given fixturedef object. This will
force the FixtureDef object to throw away any previous results and compute a new fixture value, which
will be stored into the FixtureDef object itself.
:param FixtureDef fixturedef:
"""
# prepare a subrequest object before calling fixture function
# (latter managed by fixturedef)
argname = fixturedef.argname
funcitem = self._pyfuncitem
scope = fixturedef.scope
try:
param = funcitem.callspec.getparam(argname)
except (AttributeError, ValueError):
param = NOTSET
param_index = 0
if fixturedef.params is not None:
frame = inspect.stack()[3]
frameinfo = inspect.getframeinfo(frame[0])
source_path = frameinfo.filename
source_lineno = frameinfo.lineno
source_path = py.path.local(source_path)
if source_path.relto(funcitem.config.rootdir):
source_path = source_path.relto(funcitem.config.rootdir)
msg = (
"The requested fixture has no parameter defined for the "
"current test.\n\nRequested fixture '{0}' defined in:\n{1}"
"\n\nRequested here:\n{2}:{3}".format(
fixturedef.argname,
getlocation(fixturedef.func, funcitem.config.rootdir),
source_path,
source_lineno,
)
)
> fail(msg)
E Failed: The requested fixture has no parameter defined for the current test.
E
E Requested fixture 'shape' defined in:
E pytest-parametrized.py:27
E
E Requested here:
E /usr/lib/python3.6/site-packages/_pytest/fixtures.py:526
/usr/lib/python3.6/site-packages/_pytest/fixtures.py:506: Failed
If supplying the missing parameters somehow is the answer, that's great, but I don't want to ask the question that way because I think I may be running into a massive XY problem here.
回答1:
Passing fixture as parameters in the parametrization marker is not supported by pytest. See issue #349 for more details: Using fixtures in pytest.mark.parametrize. When in need of parametrizing with fixtures, I usually resort to creating an auxiliary fixture that accepts all the parameter fixtures and then parametrizing that indirectly in the test. Your example would thus become:
@pytest.fixture
def data(request, clean_data, noisy_data):
type = request.param
if type == 'clean':
return clean_data
elif type == 'noisy':
return noisy_data
else:
raise ValueError('unknown type')
@pytest.mark.parametrize('data', ['clean', 'noisy'], indirect=True)
def test_myfunc(x_data, data):
shape, y_data = data
assert myfunc(x_data, y_data)
来源：https://stackoverflow.com/questions/52307317/parametrizing-test-with-multiple-fixtures-that-accept-arguments