Fixture fails with "module" scope and numpy array as value

[kohr-h]

Minimal failing example:

import numpy as np
import pytest

@pytest.fixture(scope="module", params=[0, np.ones(2)])
def my_fixture(request):
    return request.param

def test_stuff(my_fixture):
    pass

pytest output:

=============================================== test session starts ================================================
platform linux -- Python 3.8.1, pytest-5.3.3, py-1.8.1, pluggy-0.13.1
rootdir: /home/hkohr/git
collected 2 items                                                                                                  

tmp.py .E                                                                                                    [100%]

====================================================== ERRORS ======================================================
____________________________________ ERROR at setup of test_stuff[my_fixture1] _____________________________________

cls = <class '_pytest.runner.CallInfo'>, func = <function call_runtest_hook.<locals>.<lambda> at 0x7f8b16fb03a0>
when = 'setup', reraise = (<class '_pytest.outcomes.Exit'>, <class 'KeyboardInterrupt'>)

    @classmethod
    def from_call(cls, func, when, reraise=None) -> "CallInfo":
        #: context of invocation: one of "setup", "call",
        #: "teardown", "memocollect"
        start = time()
        excinfo = None
        try:
>           result = func()

../miniconda/envs/tmp/lib/python3.8/site-packages/_pytest/runner.py:235: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
../miniconda/envs/tmp/lib/python3.8/site-packages/_pytest/runner.py:208: in <lambda>
    lambda: ihook(item=item, **kwds), when=when, reraise=reraise
../miniconda/envs/tmp/lib/python3.8/site-packages/pluggy/hooks.py:286: in __call__
    return self._hookexec(self, self.get_hookimpls(), kwargs)
../miniconda/envs/tmp/lib/python3.8/site-packages/pluggy/manager.py:93: in _hookexec
    return self._inner_hookexec(hook, methods, kwargs)
../miniconda/envs/tmp/lib/python3.8/site-packages/pluggy/manager.py:84: in <lambda>
    self._inner_hookexec = lambda hook, methods, kwargs: hook.multicall(
../miniconda/envs/tmp/lib/python3.8/site-packages/_pytest/runner.py:120: in pytest_runtest_setup
    item.session._setupstate.prepare(item)
../miniconda/envs/tmp/lib/python3.8/site-packages/_pytest/runner.py:365: in prepare
    col.setup()
../miniconda/envs/tmp/lib/python3.8/site-packages/_pytest/python.py:1449: in setup
    fixtures.fillfixtures(self)
../miniconda/envs/tmp/lib/python3.8/site-packages/_pytest/fixtures.py:294: in fillfixtures
    request._fillfixtures()
../miniconda/envs/tmp/lib/python3.8/site-packages/_pytest/fixtures.py:473: in _fillfixtures
    item.funcargs[argname] = self.getfixturevalue(argname)
../miniconda/envs/tmp/lib/python3.8/site-packages/_pytest/fixtures.py:483: in getfixturevalue
    return self._get_active_fixturedef(argname).cached_result[0]
../miniconda/envs/tmp/lib/python3.8/site-packages/_pytest/fixtures.py:499: in _get_active_fixturedef
    self._compute_fixture_value(fixturedef)
../miniconda/envs/tmp/lib/python3.8/site-packages/_pytest/fixtures.py:580: in _compute_fixture_value
    fixturedef.execute(request=subrequest)
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

self = <FixtureDef argname='my_fixture' scope='module' baseid='tmp.py'>
request = <SubRequest 'my_fixture' for <Function test_stuff[my_fixture1]>>

    def execute(self, request):
        for argname in self._dependee_fixture_argnames(request):
            fixturedef = request._get_active_fixturedef(argname)
            if argname != "request":
                fixturedef.addfinalizer(functools.partial(self.finish, request=request))
    
        my_cache_key = self.cache_key(request)
        cached_result = getattr(self, "cached_result", None)
        if cached_result is not None:
            result, cache_key, err = cached_result
>           if my_cache_key == cache_key:
E           ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()

../miniconda/envs/tmp/lib/python3.8/site-packages/_pytest/fixtures.py:895: ValueError
============================================ 1 passed, 1 error in 0.35s ============================================

I'm not sure if there's a good way to fix this. If not, perhaps the docs should mention that arrays should not be used as parameter values, or only with function scope.

[asottile]

simpler example not involving numpy

import pytest

class Unequatable:
    def __eq__(self, _):
        raise TypeError('no!')

@pytest.fixture(scope="module", params=[0, Unequatable()])
def my_fixture(request):
    return request.param

def test_stuff(my_fixture):
    pass

definitely think this is a bug

[kohr-h]

@asottile You're right, it's not specific to Numpy, rather due to == not doing the job it's intended for. Perhaps, though, the Numpy case is the most common one for people to stumble over the issue.

A simple fix would be to first test with is, if that makes sense with pytest's caching mechanism. I don't know if there's any copying going on.
And then, == should be guarded with try..except, but what exception types to catch? TypeError probably, and ValueError for NumPy, others? Particularly poorly implemented classes may raise an AttributeError.