def _invalid_check(data):
with self.assertRaisesRegex(
TypeError,
"Data must be imported using the "
"data_readers, pd.DataFrames, "
"np.ndarrays, or lists.",
):
BaseDataLabeler._check_and_return_valid_data_format(data)
def _valid_check(data):
try:
print("\tChecking data format: {}".format(str(type(data))))
data = BaseDataLabeler._check_and_return_valid_data_format(
data, fit_or_predict='predict')
except Exception as e:
self.fail("Exception raised on input of accepted types.")
return data
def test_load_with_components(self, *mocks):
mock_preprocessor = mock.Mock(spec=data_processing.BaseDataPreprocessor)
mock_preprocessor._parameters = {"test": 1}
mock_postprocessor = mock.Mock(
spec=data_processing.BaseDataPostprocessor)
mock_postprocessor._parameters = {"test": 2}
mock_model = mock.Mock(spec=BaseTrainableModel)
mock_model._parameters = {"test": 3}
data_labeler = BaseDataLabeler.load_with_components(
preprocessor=mock_preprocessor,
model=mock_model,
postprocessor=mock_postprocessor)
self.assertIsInstance(data_labeler, BaseDataLabeler)
self.assertEqual('CustomDataLabeler', data_labeler.__class__.__name__)
self.assertEqual(mock_preprocessor, data_labeler.preprocessor)
self.assertEqual({"test": 1}, data_labeler.preprocessor._parameters)
self.assertEqual(mock_model, data_labeler.model)
self.assertEqual({"test": 3}, data_labeler.model._parameters)
self.assertEqual(mock_postprocessor, data_labeler.postprocessor)
self.assertEqual({"test": 2}, data_labeler.postprocessor._parameters)
def test_save_labeler(self, mock_load_data_labeler, mock_open,
mock_load_model):
# setup mocks
mock_file = setup_save_mock_open(mock_open)
base_data_labeler = BaseDataLabeler('fake_path')
base_data_labeler._model = mock.Mock()
base_data_labeler._preprocessor = mock.Mock()
base_data_labeler._postprocessor = mock.Mock()
base_data_labeler.save_to_disk('test')
self.assertEqual(
'{"model": {"class": "Mock"}, "preprocessor": {"class": "Mock"}, '
'"postprocessor": {"class": "Mock"}}', mock_file.getvalue())
mock_open.assert_called_with('test/data_labeler_parameters.json', 'w')
base_data_labeler._model.save_to_disk.assert_called_with('test')
base_data_labeler._preprocessor.save_to_disk.assert_called_with('test')
base_data_labeler._postprocessor.save_to_disk.assert_called_with(
'test')
# close mock
StringIO.close(mock_file)
def _invalid_check(data):
with self.assertRaisesRegex(
TypeError, "Data must either be imported using "
"the data_readers or pd.DataFrame."):
BaseDataLabeler._check_and_return_valid_data_format(data)
def test_check_and_return_valid_data_format(self):
# test incorrect fit_or_predict value
with self.assertRaisesRegex(ValueError, '`fit_or_predict` must equal '
'`fit` or `predict`'):
BaseDataLabeler._check_and_return_valid_data_format([], 'oops')
# test incorrect data type
with self.assertRaisesRegex(TypeError, "Data must be imported using the"
" data_readers, pd.DataFrames, "
"np.ndarrays, or lists."):
BaseDataLabeler._check_and_return_valid_data_format('oops')
# test proper conversion of 2 dimensional structured data
two_dim = [["this", "is"], ["two", "dimensions"]]
two_dim_pred = np.array(["this", "is", "two", "dimensions"])
# for fit
self.assertTrue(
np.array_equal(np.array(two_dim),
BaseDataLabeler._check_and_return_valid_data_format(
two_dim, fit_or_predict='fit')))
self.assertTrue(
np.array_equal(np.array(two_dim),
BaseDataLabeler._check_and_return_valid_data_format(
pd.DataFrame(two_dim), fit_or_predict='fit')))
self.assertTrue(
np.array_equal(np.array(two_dim),
BaseDataLabeler._check_and_return_valid_data_format(
np.array(two_dim), fit_or_predict='fit')))
# for predict
self.assertTrue(
np.array_equal(two_dim_pred,
BaseDataLabeler._check_and_return_valid_data_format(
two_dim, fit_or_predict='predict')))
self.assertTrue(
np.array_equal(two_dim_pred,
BaseDataLabeler._check_and_return_valid_data_format(
pd.DataFrame(two_dim), fit_or_predict='predict')))
self.assertTrue(
np.array_equal(two_dim_pred,
BaseDataLabeler._check_and_return_valid_data_format(
np.array(two_dim), fit_or_predict='predict')))
# test proper conversion of 1 dimensional data
one_dim = ["this", "is", "one", "dimension"]
one_dim_pred = np.array(one_dim)
# for fit
self.assertTrue(
np.array_equal(np.array(one_dim),
BaseDataLabeler._check_and_return_valid_data_format(
one_dim, fit_or_predict='fit')))
self.assertTrue(
np.array_equal(np.array(one_dim),
BaseDataLabeler._check_and_return_valid_data_format(
pd.Series(one_dim), fit_or_predict='fit')))
self.assertTrue(
np.array_equal(np.array(one_dim),
BaseDataLabeler._check_and_return_valid_data_format(
np.array(one_dim), fit_or_predict='fit')))
# for predict
self.assertTrue(
np.array_equal(one_dim_pred,
BaseDataLabeler._check_and_return_valid_data_format(
one_dim, fit_or_predict='predict')))
self.assertTrue(
np.array_equal(one_dim_pred,
BaseDataLabeler._check_and_return_valid_data_format(
pd.DataFrame(one_dim), fit_or_predict='predict')))
self.assertTrue(
np.array_equal(one_dim_pred,
BaseDataLabeler._check_and_return_valid_data_format(
np.array(one_dim), fit_or_predict='predict')))
# test proper conversion of unstructured labels
labels = [[(0, 4, "UNKNOWN"), (4, 10, "ADDRESS")],
[(0, 5, "SSN"), (5, 8, "UNKNOWN")]]
validated_labels = \
BaseDataLabeler._check_and_return_valid_data_format(labels)
self.assertIsInstance(validated_labels, np.ndarray)
self.assertEqual(len(validated_labels), 2)
self.assertEqual(len(validated_labels[0]), 2)
self.assertEqual(len(validated_labels[0][0]), 3)
self.assertEqual(validated_labels[0][0][0], 0)
self.assertEqual(validated_labels[0][1][1], 10)
self.assertEqual(validated_labels[1][0][2], "SSN")
# test proper conversion of data reader objects
for dt in ["csv", "json", "parquet"]:
data_obj = dp.Data(data=pd.DataFrame(two_dim), data_type=dt)
val = BaseDataLabeler._check_and_return_valid_data_format(data_obj)
self.assertTrue(np.array_equal(np.array(two_dim), val))
def setUpClass(cls) -> None:
cls.data = np.array([
'123 Fake St.', '1/2/2020', 'nice.', '4/3/22', 'abc', '333-44-2341'
]).reshape((-1, ))
cls.data_labeler = BaseDataLabeler.load_from_disk(
os.path.join(default_labeler_dir, 'regex_model'))
def setUpClass(cls) -> None:
cls.data = np.array([
"123 Fake St.", "1/2/2020", "nice.", "4/3/22", "abc", "333-44-2341"
]).reshape((-1, ))
cls.data_labeler = BaseDataLabeler.load_from_disk(
os.path.join(default_labeler_dir, "regex_model"))
