def test_fit_not_contant(self, select_mock):
"""if not constant call select_univariate and fit the returned instance.
Check that candidates are passed down to select_univariate
and that the returned instance is fitted on the input data.
"""
# Setup
candidate = MagicMock()
candidates = [candidate]
distribution = Univariate(candidates)
# Run
data = np.array([1, 2, 3, 4, 5])
distribution.fit(data)
# Assert
assert distribution.fitted
assert distribution.constant_value is None
# candidates are passed down
assert select_mock.call_count == 1
expected_call = call(data, candidates)[1:]
actual_call = select_mock.call_args
compare_nested_iterables(expected_call, actual_call)
# the returned instance is fitted
instance = select_mock.return_value
assert instance.fit.call_count == 1
expected_call = call(data)[1:]
actual_call = instance.fit.call_args
compare_nested_iterables(expected_call, actual_call)
def test_fit_not_constant(self):
"""if constant values, replace methods."""
# Setup
distribution = Univariate()
# Run
distribution.fit(np.array([1, 2, 3, 4, 1]))
# Assert
assert distribution.fitted
assert not distribution._instance._is_constant()
def fit(self, data, cols):
"""Fits univariate distributions (automatically selected).
Args:
data (DataFrame):
Data, where each column in `cols` is a continuous column.
cols (list[str]):
Column names.
"""
for col in cols:
col_data = np.array(data[col])
dist_model = Univariate()
dist_model.fit(col_data)
self.cdfs.append(dist_model)
def test_fit_contant(self):
"""if constant values, replace methods."""
# Setup
distribution = Univariate()
replace_mock = MagicMock()
distribution._replace_constant_methods = replace_mock
# Run
distribution.fit(np.array([1, 1, 1, 1, 1]))
# Assert
assert distribution.fitted
assert distribution.constant_value == 1
replace_mock.assert_called_once_with()
def test_fit_selection_sample_size_small(self, select_mock):
"""if selection_sample_size is smaller than data, subsample the data before selecting."""
# Setup
distribution = Univariate(selection_sample_size=3)
# Run
distribution.fit(np.array([1, 1, 1, 1, 1]))
# Assert
assert distribution.fitted
assert distribution._instance == select_mock.return_value
call_args = select_mock.call_args_list
selection_sample = call_args[0][0][0]
np.testing.assert_array_equal(selection_sample, np.array([1, 1, 1]))
fit_call_args = select_mock.return_value.fit.call_args_list
np.testing.assert_array_equal(fit_call_args[0][0][0],
np.array([1, 1, 1, 1, 1]))
def fit(self, data, cols):
for col in cols:
col_data = np.array(data[col])
dist_model = Univariate()
dist_model.fit(col_data)
self.cdfs.append(dist_model)