def reverse_transform(self, data):
"""Reverse the transformed data to the original format.
Args:
data (pandas.DataFrame):
Data to be reverse transformed.
Returns:
pandas.DataFrame
"""
if not self.fitted:
raise MetadataNotFittedError()
reversed_data = self._hyper_transformer.reverse_transform(data)
for constraint in self._constraints:
reversed_data = constraint.reverse_transform(reversed_data)
for name, field_metadata in self._fields_metadata.items():
field_type = field_metadata['type']
if field_type == 'id' and name not in reversed_data:
field_data = self._make_ids(name, field_metadata, len(reversed_data))
elif field_metadata.get('pii', False):
faker = self._get_faker(field_metadata['pii_category'])
field_data = pd.Series([faker() for _ in range(len(reversed_data))])
else:
field_data = reversed_data[name]
reversed_data[name] = field_data[field_data.notnull()].astype(self._dtypes[name])
return reversed_data[self._field_names]
def reverse_transform(self, data):
"""Reverse the transformed data to the original format.
Args:
data (pandas.DataFrame):
Data to be reverse transformed.
Returns:
pandas.DataFrame
"""
if not self.fitted:
raise MetadataNotFittedError()
reversed_data = self._hyper_transformer.reverse_transform(data)
for constraint in self._constraints:
reversed_data = constraint.reverse_transform(reversed_data)
fields = self._fields_metadata
for name, dtype in self.get_dtypes(ids=True).items():
field_metadata = fields[name]
field_type = field_metadata['type']
if field_type != 'id':
field_data = reversed_data[name]
elif field_metadata.get('pii', False):
faker = self._get_faker(field_metadata['pii_category'])
field_data = pd.Series(
[faker() for _ in range(len(reversed_data))])
else:
field_data = pd.Series(np.arange(len(reversed_data)))
reversed_data[name] = field_data.dropna().astype(dtype)
return reversed_data[self._field_names]
def transform(self, data):
"""Transform the given data.
Args:
data (pandas.DataFrame):
Table data.
Returns:
pandas.DataFrame:
Transformed data.
"""
if not self.fitted:
raise MetadataNotFittedError()
fields = [
field for field in self.get_dtypes(ids=False)
if field in data.columns
]
LOGGER.debug('Anonymizing table %s', self.name)
data = self._anonymize(data[fields])
LOGGER.debug('Transforming constraints for table %s', self.name)
for constraint in self._constraints:
data = constraint.transform(data)
LOGGER.debug('Transforming table %s', self.name)
return self._hyper_transformer.transform(data)
def reverse_transform(self, data):
"""Reverse the transformed data to the original format.
Args:
data (pandas.DataFrame):
Data to be reverse transformed.
Returns:
pandas.DataFrame
"""
if not self.fitted:
raise MetadataNotFittedError()
try:
reversed_data = self._hyper_transformer.reverse_transform(data)
except rdt.errors.NotFittedError:
reversed_data = data
for constraint in reversed(self._constraints_to_reverse):
reversed_data = constraint.reverse_transform(reversed_data)
for name, field_metadata in self._fields_metadata.items():
field_type = field_metadata['type']
if field_type == 'id' and name not in reversed_data:
field_data = self._make_ids(field_metadata, len(reversed_data))
elif field_metadata.get('pii', False):
field_data = pd.Series(
Table._get_fake_values(field_metadata, len(reversed_data)))
else:
field_data = reversed_data[name]
reversed_data[name] = field_data[field_data.notnull()].astype(
self._dtypes[name])
return reversed_data[self._field_names]
def transform(self, data):
"""Transform the given data.
Args:
data (pandas.DataFrame):
Table data.
Returns:
pandas.DataFrame:
Transformed data.
"""
if not self.fitted:
raise MetadataNotFittedError()
fields = self.get_dtypes(ids=False)
data = self._anonymize(data[fields])
for constraint in self._constraints:
data = constraint.transform(data)
return self._hyper_transformer.transform(data)
def transform(self, data, on_missing_column='error'):
"""Transform the given data.
Args:
data (pandas.DataFrame):
Table data.
on_missing_column (str):
If the value is error, then a `MissingConstraintColumnError` is raised.
If the value is drop, then the columns involved in the constraint that
are present in data will be dropped.
Returns:
pandas.DataFrame:
Transformed data.
Raises:
ConstraintsNotMetError:
If the table data is not valid for the provided constraints.
"""
if not self.fitted:
raise MetadataNotFittedError()
fields = [
field for field in self.get_dtypes(ids=False)
if field in data.columns
]
LOGGER.debug('Anonymizing table %s', self.name)
data = self._anonymize(data[fields])
self._validate_data_on_constraints(data)
LOGGER.debug('Transforming constraints for table %s', self.name)
data = self._transform_constraints(data, on_missing_column)
LOGGER.debug('Transforming table %s', self.name)
return self._hyper_transformer.transform(data)