def test_smlbobject_unused_initalizer_arguments():
"""Tests the unused-initializer-parameters functionality provided by SmlbObject."""
# diamond inheritance example
with pytest.raises(smlb.InvalidParameterError):
class A(smlb.SmlbObject):
pass
class B(smlb.SmlbObject):
pass
class C(A, B):
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
C(None)
# example from distributions
with pytest.raises(smlb.InvalidParameterError):
smlb.DeltaPredictiveDistribution([1, 2, 3], invalid=None)
# example from evaluation metrics
with pytest.raises(smlb.InvalidParameterError):
smlb.MeanContinuousRankedProbabilityScore(hellohello=99)
def test_interface_erroneous_arguments():
"""Tests whether errors are raised for wrong keyword arguments."""
# PredictiveDistribution can not be instantiated because it is abstract
# this test also ensures that PredictiveDistribution.__init__, if defined, calls super().__init__
with pytest.raises(Exception):
smlb.DeltaPredictiveDistribution([1, 2, 3],
badarg=None) # spelling error
def test_interface_decomposition():
d = smlb.NormalPredictiveDistribution([1, 2, 3], [0.5, 0.5, 1])
assert not (d.has_noise_part or d.has_signal_part)
dd = smlb.DeltaPredictiveDistribution([1, 2, 3], noise_part=d)
assert dd.has_noise_part and not dd.has_signal_part
dd = smlb.DeltaPredictiveDistribution([1, 2, 3], signal_part=d)
assert dd.has_signal_part and not dd.has_noise_part
dd = smlb.DeltaPredictiveDistribution([1, 2, 3],
noise_part=d,
signal_part=d)
assert dd.has_signal_part and dd.has_noise_part
with pytest.raises(smlb.BenchmarkError):
d.noise_part
with pytest.raises(smlb.BenchmarkError):
d.signal_part