def test_fit_with_analyze(self):
"""Test fit and analyze the transformers"""
# Setup
data = pd.DataFrame({
'integers': [1, 2, 3, 4],
'floats': [1.1, 2.2, 3.3, 4.4],
'booleans': [True, False, False, True]
})
int_mock = Mock()
float_mock = Mock()
bool_mock = Mock()
analyzed_data = {
'integers': int_mock,
'floats': float_mock,
'booleans': bool_mock
}
# Run
transformer = Mock()
transformer.transformers = None
transformer._analyze.return_value = analyzed_data
HyperTransformer.fit(transformer, data)
# Asserts
expect_int_call_count = 1
expect_float_call_count = 1
expect_bool_call_count = 1
self.assertEqual(int_mock.fit.call_count, expect_int_call_count)
self.assertEqual(float_mock.fit.call_count, expect_float_call_count)
self.assertEqual(bool_mock.fit.call_count, expect_bool_call_count)
def test_hypertransformer_with_transformers(faker_mock):
faker_mock.return_value.first_name.side_effect = [
'Jaime', 'Cersei', 'Tywin', 'Tyrion'
]
data = get_input_data()
transformers = get_transformers()
ht = HyperTransformer(transformers)
ht.fit(data)
transformed = ht.transform(data)
expected = get_transformed_data()
np.testing.assert_allclose(
transformed.sort_index(axis=1).values,
expected.sort_index(axis=1).values)
reversed_data = ht.reverse_transform(transformed)
original_names = data.pop('names')
reversed_names = reversed_data.pop('names')
pd.testing.assert_frame_equal(data.sort_index(axis=1),
reversed_data.sort_index(axis=1))
for name in original_names:
assert name not in reversed_names
def test_single_category():
ht = HyperTransformer(transformers={'a': OneHotEncodingTransformer()})
data = pd.DataFrame({'a': ['a', 'a', 'a']})
ht.fit(data)
transformed = ht.transform(data)
reverse = ht.reverse_transform(transformed)
pd.testing.assert_frame_equal(data, reverse)
def test_dtype_category():
df = pd.DataFrame({'a': ['a', 'b', 'c']}, dtype='category')
ht = HyperTransformer()
ht.fit(df)
trans = ht.transform(df)
rever = ht.reverse_transform(trans)
pd.testing.assert_frame_equal(df, rever)
def test_empty_transformers_nan_data():
"""If transformers is an empty dict, do nothing."""
data = get_input_data_with_nan()
ht = HyperTransformer(transformers={})
ht.fit(data)
transformed = ht.transform(data)
reverse = ht.reverse_transform(transformed)
pd.testing.assert_frame_equal(data, transformed)
pd.testing.assert_frame_equal(data, reverse)
def test_subset_of_columns_nan_data():
"""HyperTransform should be able to transform a subset of the training columns.
See https://github.com/sdv-dev/RDT/issues/152
"""
data = get_input_data_with_nan()
ht = HyperTransformer()
ht.fit(data)
subset = data[[data.columns[0]]]
transformed = ht.transform(subset)
reverse = ht.reverse_transform(transformed)
pd.testing.assert_frame_equal(subset, reverse)
def test_hypertransformer_without_transformers_nan_data():
data = get_input_data_with_nan()
ht = HyperTransformer()
ht.fit(data)
transformed = ht.transform(data)
expected = get_transformed_nan_data()
np.testing.assert_allclose(
transformed.sort_index(axis=1).values,
expected.sort_index(axis=1).values)
reversed_data = ht.reverse_transform(transformed)
original_names = data.pop('names')
reversed_names = reversed_data.pop('names')
pd.testing.assert_frame_equal(data.sort_index(axis=1),
reversed_data.sort_index(axis=1))
for name in original_names:
assert name not in reversed_names