WIP trying to convert everything to string before processing

sdv/data_processing/data_processor.py

@@ -862,6 +862,7 @@ def reverse_transform(self, data, reset_keys=False):
             for column in self.metadata.columns.keys() - set(sampled_columns + self._keys)
             if self._hyper_transformer.field_transformers.get(column)
         ]
+        print(missing_columns)
         if missing_columns and num_rows:
             anonymized_data = self._hyper_transformer.create_anonymized_columns(
                 num_rows=num_rows,

sdv/metadata/multi_table.py

@@ -1025,10 +1025,11 @@ def _set_metadata_dict(self, metadata):
                 Python dictionary representing a ``MultiTableMetadata`` object.
         """
         for table_name, table_dict in metadata.get('tables', {}).items():
-            self.tables[table_name] = SingleTableMetadata.load_from_dict(table_dict)
+            self.tables[str(table_name)] = SingleTableMetadata.load_from_dict(table_dict)
 
         for relationship in metadata.get('relationships', []):
-            self.relationships.append(relationship)
+            type_safe_relationships = {key: str(value) if not isinstance(value, str) else value for key, value in relationship.items()}
+            self.relationships.append(type_safe_relationships)
 
     @classmethod
     def load_from_dict(cls, metadata_dict):

sdv/metadata/single_table.py

@@ -1238,7 +1238,8 @@ def load_from_dict(cls, metadata_dict):
         for key in instance._KEYS:
             value = deepcopy(metadata_dict.get(key))
             if value:
-                setattr(instance, f'{key}', value)
+                type_safe_value = {str(key) if not isinstance(key, str) else key: value for key, value in value.items()}
+                setattr(instance, f'{key}', type_safe_value)
 
         return instance
 

sdv/multi_table/base.py

@@ -417,9 +417,14 @@ def fit(self, data):
                 Dictionary mapping each table name to a ``pandas.DataFrame`` in the raw format
                 (before any transformations).
         """
+        type_safe_data = {str(key) if not isinstance(key, str) else key: value for key, value in data.items()}
         total_rows = 0
         total_columns = 0
-        for table in data.values():
+        for table, dataframe in type_safe_data.items():
+            dataframe.columns = dataframe.columns.astype(str)
+            type_safe_data[table] = dataframe
+
+        for table in type_safe_data.values():
             total_rows += len(table)
             total_columns += len(table.columns)
 
@@ -440,10 +445,10 @@ def fit(self, data):
             self._synthesizer_id,
         )
         check_synthesizer_version(self, is_fit_method=True, compare_operator=operator.lt)
-        _validate_foreign_keys_not_null(self.metadata, data)
+        _validate_foreign_keys_not_null(self.metadata, type_safe_data)
         self._check_metadata_updated()
         self._fitted = False
-        processed_data = self.preprocess(data)
+        processed_data = self.preprocess(type_safe_data)
         self._print(text='\n', end='')
         self.fit_processed_data(processed_data)
 

sdv/single_table/base.py

@@ -448,7 +448,7 @@ def fit(self, data):
             len(data.columns),
             self._synthesizer_id,
         )
-
+        data.columns = data.columns.astype(str)
         check_synthesizer_version(self, is_fit_method=True, compare_operator=operator.lt)
         self._check_metadata_updated()
         self._fitted = False