def test_cast_to_series_indexed(reference, other):
out_mean, out_sd = cast_to_series(reference, other)
assert reference.equals(out_mean)
assert reference.equals(out_sd)
out_mean, out_sd = cast_to_series(other, reference)
assert reference.equals(out_mean)
assert reference.equals(out_sd)
def test_cast_to_series_mismatched_index():
reference = pd.Series([1, 2, 3], index=['a', 'b', 'c'])
other = pd.Series([1, 2, 3])
with pytest.raises(ValueError, match='identically indexed'):
cast_to_series(reference, other)
with pytest.raises(ValueError, match='identically indexed'):
cast_to_series(other, reference)
def get_parameters(cls,
parameters: Parameters = None,
mean: Parameter = None,
sd: Parameter = None) -> pd.DataFrame:
required_parameters = list(cls.expected_parameters +
('x_min', 'x_max'))
if parameters is not None:
if not (mean is None and sd is None):
raise ValueError(
"You may supply either pre-calculated parameters or"
" mean and standard deviation but not both.")
parameters = format_data(parameters, required_parameters,
'parameters')
else:
if mean is None or sd is None:
raise ValueError(
"You may supply either pre-calculated parameters or"
" mean and standard deviation but not both.")
mean, sd = cast_to_series(mean, sd)
parameters = pd.DataFrame(0,
columns=required_parameters,
index=mean.index)
computable = cls.computable_parameter_index(mean, sd)
parameters.loc[computable,
['x_min', 'x_max']] = cls.compute_min_max(
mean.loc[computable], sd.loc[computable])
# The scipy.stats distributions run optimization routines that handle FloatingPointErrors,
# transforming them into RuntimeWarnings. This gets noisy in our logs.
with warnings.catch_warnings():
warnings.simplefilter('ignore', RuntimeWarning)
parameters.loc[
computable,
list(cls.expected_parameters)] = cls._get_parameters(
mean.loc[computable], sd.loc[computable],
parameters.loc[computable, 'x_min'],
parameters.loc[computable, 'x_max'])
return parameters
def test_cast_to_series_mismatched_length(reference, other):
with pytest.raises(ValueError, match='same number of values'):
cast_to_series(reference, other)
with pytest.raises(ValueError, match='same number of values'):
cast_to_series(other, reference)
def test_cast_to_series_nulls(val, null):
with pytest.raises(ValueError, match='Empty data structure'):
cast_to_series(val, null)
with pytest.raises(ValueError, match='Empty data structure'):
cast_to_series(null, val)
def test_cast_to_series_array_like(mean, sd):
expected_mean, expected_sd = pd.Series([1, 2, 3]), pd.Series([1, 2, 3])
out_mean, out_sd = cast_to_series(mean, sd)
assert expected_mean.equals(out_mean)
assert expected_sd.equals(out_sd)
def test_cast_to_series_single_floats(mean, sd):
expected_mean, expected_sd = pd.Series([1.]), pd.Series([1.])
out_mean, out_sd = cast_to_series(mean, sd)
assert expected_mean.equals(out_mean)
assert expected_sd.equals(out_sd)