sdv 1.34.1.dev0__tar.gz → 1.34.2.dev0__tar.gz

{sdv-1.34.1.dev0/sdv.egg-info → sdv-1.34.2.dev0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sdv
-Version: 1.34.1.dev0
+Version: 1.34.2.dev0
 Summary: Generate synthetic data for single table, multi table and sequential data
 Author-email: "DataCebo, Inc." <info@sdv.dev>
 License-Expression: BUSL-1.1

{sdv-1.34.1.dev0 → sdv-1.34.2.dev0}/pyproject.toml

@@ -150,7 +150,7 @@ namespaces = false
 version = {attr = 'sdv.__version__'}
 
 [tool.bumpversion]
-current_version = "1.34.1.dev0"
+current_version = "1.34.2.dev0"
 parse = '(?P<major>\d+)\.(?P<minor>\d+)\.(?P<patch>\d+)(\.(?P<release>[a-z]+)(?P<candidate>\d+))?'
 serialize = [
     '{major}.{minor}.{patch}.{release}{candidate}',

{sdv-1.34.1.dev0 → sdv-1.34.2.dev0}/sdv/__init__.py

@@ -6,7 +6,7 @@
 
 __author__ = 'DataCebo, Inc.'
 __email__ = 'info@sdv.dev'
-__version__ = '1.34.1.dev0'
+__version__ = '1.34.2.dev0'
 
 
 import sys

{sdv-1.34.1.dev0 → sdv-1.34.2.dev0}/sdv/metadata/multi_table.py

@@ -835,11 +835,14 @@ class MultiTableMetadata:
         """Validate every table of the data has a valid table/metadata pair."""
         errors = []
         warning_dataframes = []
+        tables_with_mismatching_columns_order = []
         for table_name, table_data in data.items():
             table_sdtype_warnings = defaultdict(list)
             try:
                 with warnings.catch_warnings(record=True):
                     self.tables[table_name].validate_data(table_data, table_sdtype_warnings)
+                    if not self.tables[table_name]._check_data_columns_order(table_data.columns):
+                        tables_with_mismatching_columns_order.append(table_name)
 
             except InvalidDataError as error:
                 if INT_REGEX_ZERO_ERROR_MESSAGE in str(error) and len(self.tables) > 1:
@@ -877,6 +880,14 @@ class MultiTableMetadata:
             )
             warnings.warn(warning_msg)
 
+        if len(tables_with_mismatching_columns_order):
+            affected_tables = ', '.join(map(repr, tables_with_mismatching_columns_order))
+            warnings.warn(
+                'The metadata lists columns in a different order than the data. '
+                'This may result in the synthetic data having a different order.\n'
+                f'Affected tables: {affected_tables}.'
+            )
+
         return errors
 
     def _validate_foreign_keys(self, data):
@@ -1143,7 +1154,11 @@ class MultiTableMetadata:
             foreign_key = relationship.get('child_foreign_key')
             primary_key = self.tables.get(parent).primary_key
             edge_label = f'  {foreign_key} → {primary_key}' if show_relationship_labels else ''
-            edges.append((parent, child, edge_label))
+            child_primary_key = self.tables.get(child).primary_key
+            if foreign_key == child_primary_key:
+                edges.append((parent, child, edge_label, 'one-to-one'))
+            else:
+                edges.append((parent, child, edge_label))
 
             if show_table_details is not None:
                 child_node = nodes.get(child)
@@ -1285,12 +1300,12 @@ class MultiTableMetadata:
             filepath (str):
                 String that represents the ``path`` to the ``json`` file.
 
+        Returns:
+            A ``MultiTableMetadata`` instance.
+
         Raises:
             - An ``Error`` if the path does not exist.
             - An ``Error`` if the ``json`` file does not contain the ``METADATA_SPEC_VERSION``.
-
-        Returns:
-            A ``MultiTableMetadata`` instance.
         """
         metadata = read_json(filepath)
         return cls.load_from_dict(metadata)
@@ -1343,11 +1358,11 @@ class MultiTableMetadata:
             filepath (str):
                 String that represents the ``path`` to the old metadata ``json`` file.
 
-        Raises:
-            Raises a ``ValueError`` if the filepath does not exist.
-
         Returns:
             A ``MultiTableMetadata`` instance.
+
+        Raises:
+            Raises a ``ValueError`` if the filepath does not exist.
         """
         old_metadata = read_json(filepath)
         tables_metadata = {}

{sdv-1.34.1.dev0 → sdv-1.34.2.dev0}/sdv/metadata/single_table.py

@@ -1272,6 +1272,17 @@ class SingleTableMetadata:
 
         return []
 
+    def _check_data_columns_order(self, data_columns):
+        data_columns = [column for column in data_columns if column in self.columns]
+        return data_columns == list(self.columns)
+
+    def _warn_data_column_order_mismatch(self, data_columns):
+        if not self._check_data_columns_order(data_columns):
+            warnings.warn(
+                'The metadata lists columns in a different order than the data. '
+                'This may result in the synthetic data having a different order.'
+            )
+
     def validate_data(self, data, sdtype_warnings=None):
         """Validate the data matches the metadata.
 
@@ -1304,6 +1315,9 @@ class SingleTableMetadata:
         # Both metadata and data must have the same set of columns
         self._validate_metadata_matches_data(data.columns)
 
+        # Warn if metadata columns order mismatches data column order
+        self._warn_data_column_order_mismatch(data.columns)
+
         errors = []
         # Primary, sequence and alternate keys can't have missing values
         errors += self._validate_keys_dont_have_missing_values(data)
@@ -1490,12 +1504,12 @@ class SingleTableMetadata:
             filepath (str):
                 String that represents the ``path`` to the ``json`` file.
 
+        Returns:
+            A ``SingleTableMetadata`` instance.
+
         Raises:
             - An ``Error`` if the path does not exist.
             - An ``Error`` if the ``json`` file does not contain the ``METADATA_SPEC_VERSION``.
-
-        Returns:
-            A ``SingleTableMetadata`` instance.
         """
         metadata = read_json(filepath)
         if 'METADATA_SPEC_VERSION' not in metadata:
@@ -1519,11 +1533,11 @@ class SingleTableMetadata:
             filepath (str):
                 String that represents the ``path`` to the old metadata ``json`` file.
 
-        Raises:
-            Raises a ``ValueError`` if the filepath does not exist.
-
         Returns:
             A ``SingleTableMetadata`` instance.
+
+        Raises:
+            Raises a ``ValueError`` if the filepath does not exist.
         """
         old_metadata = read_json(filepath)
         if 'tables' in old_metadata:

{sdv-1.34.1.dev0 → sdv-1.34.2.dev0}/sdv/metadata/visualization.py

@@ -102,8 +102,20 @@ def visualize_graph(nodes, edges, filepath=None):
     for name, label in nodes.items():
         digraph.node(name, label=_replace_special_characters(label))
 
-    for parent, child, label in edges:
-        digraph.edge(parent, child, label=_replace_special_characters(label), arrowhead='oinv')
+    for edge in edges:
+        parent, child, label = edge[0], edge[1], edge[2]
+        relation_type = edge[3] if len(edge) > 3 else 'one-to-many'
+        if relation_type == 'one-to-one':
+            digraph.edge(
+                parent,
+                child,
+                label=_replace_special_characters(label),
+                arrowhead='noneteeodot',
+                arrowtail='nonetee',
+                dir='both',
+            )
+        else:
+            digraph.edge(parent, child, label=_replace_special_characters(label), arrowhead='oinv')
 
     if filename:
         digraph.render(filename=filename, cleanup=True, format=graphviz_extension)

{sdv-1.34.1.dev0 → sdv-1.34.2.dev0}/sdv/multi_table/hma.py

@@ -471,6 +471,7 @@ class HMASynthesizer(BaseHierarchicalSampler, BaseMultiTableSynthesizer):
                         enforce_min_max_values=True
                     )
                     self.extended_columns[child_name][column].fit(extension, column)
+
                 table = table.merge(extension, how='left', right_index=True, left_index=True)
                 num_rows_key = f'__{child_name}__{foreign_key}__num_rows'
                 table[num_rows_key] = table[num_rows_key].fillna(0)
@@ -485,11 +486,10 @@ class HMASynthesizer(BaseHierarchicalSampler, BaseMultiTableSynthesizer):
 
                 tables[table_name] = table
                 self._learned_relationships += 1
-        self._augmented_tables.append(table_name)
 
+        self._augmented_tables.append(table_name)
         foreign_keys = self.metadata._get_all_foreign_keys(table_name)
         self._clear_nans(table, ignore_cols=foreign_keys)
-
         return table
 
     def _augment_tables(self, processed_data):
@@ -533,6 +533,7 @@ class HMASynthesizer(BaseHierarchicalSampler, BaseMultiTableSynthesizer):
         keys = {}
         for fk in foreign_keys:
             keys[fk] = table_data.pop(fk).to_numpy()
+
         return keys
 
     def _model_tables(self, augmented_data):
@@ -730,6 +731,7 @@ class HMASynthesizer(BaseHierarchicalSampler, BaseMultiTableSynthesizer):
             extended_columns = getattr(self, '_parent_extended_columns', {}).get(table_name, [])
             if extended_columns:
                 self._set_extended_columns_distributions(synthesizer, table_name, extended_columns)
+
             synthesizer._set_parameters(parameters)
             try:
                 likelihoods[parent_id] = synthesizer._get_likelihood(table_rows)

{sdv-1.34.1.dev0 → sdv-1.34.2.dev0}/sdv/sampling/hierarchical_sampler.py

@@ -317,6 +317,7 @@ class BaseHierarchicalSampler:
             if num_rows <= 0:
                 send_min_sample_warning = True
                 num_rows = 1
+
             synthesizer = self._table_synthesizers[table]
             LOGGER.info(f'Sampling {num_rows} rows from table {table}')
             sampled_data[table] = self._sample_rows(synthesizer, num_rows)
@@ -341,5 +342,12 @@ class BaseHierarchicalSampler:
                 )
                 added_relationships.add((parent_name, child_name))
 
-        sampled_data = self._reverse_transform_constraints(sampled_data)
+        # During this step,  `drop_unknown_references` triggers `metadata.validate(data)`
+        # which now warns if the column order in the data does not match the metadata.
+        with warnings.catch_warnings():
+            warnings.filterwarnings(
+                'ignore', message='The metadata lists columns in a different order than the data.'
+            )
+            sampled_data = self._reverse_transform_constraints(sampled_data)
+
         return self._finalize(sampled_data)

{sdv-1.34.1.dev0 → sdv-1.34.2.dev0/sdv.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sdv
-Version: 1.34.1.dev0
+Version: 1.34.2.dev0
 Summary: Generate synthetic data for single table, multi table and sequential data
 Author-email: "DataCebo, Inc." <info@sdv.dev>
 License-Expression: BUSL-1.1

{sdv-1.34.1.dev0 → sdv-1.34.2.dev0}/sdv/metadata/metadata.py

@@ -24,12 +24,12 @@ class Metadata(MultiTableMetadata):
             filepath (str):
                 String that represents the ``path`` to the ``json`` file.
 
+        Returns:
+            A ``Metadata`` instance.
+
         Raises:
             - An ``Error`` if the path does not exist.
             - An ``Error`` if the ``json`` file does not contain the ``METADATA_SPEC_VERSION``.
-
-        Returns:
-            A ``Metadata`` instance.
         """
         metadata = read_json(filepath)
         if metadata.get('METADATA_SPEC_VERSION') == 'SINGLE_TABLE_V1':

{sdv-1.34.1.dev0 → sdv-1.34.2.dev0}/sdv/single_table/ctgan.py

@@ -51,12 +51,12 @@ class LossValuesMixin:
     def get_loss_values(self):
         """Get the loss values from the model.
 
-        Raises:
-            - ``NotFittedError`` if synthesizer has not been fitted.
-
         Returns:
             pd.DataFrame:
                 Dataframe containing the loss values per epoch.
+
+        Raises:
+            - ``NotFittedError`` if synthesizer has not been fitted.
         """
         if not self._fitted:
             err_msg = 'Loss values are not available yet. Please fit your synthesizer first.'
@@ -71,12 +71,12 @@ class LossValuesMixin:
             title (pd.DataFrame):
                 Title string for the graph created.
 
-        Raises:
-            - ``NotFittedError`` if synthesizer has not been fitted.
-
         Returns:
             plotly.graph_objects._figure.Figure:
                 1D marginal distribution plot (i.e. a histogram) of the columns.
+
+        Raises:
+            - ``NotFittedError`` if synthesizer has not been fitted.
         """
         if not self._fitted:
             err_msg = 'Loss values are not available yet. Please fit your synthesizer first.'