Add comment

sdv/data_processing/data_processor.py

@@ -718,13 +718,15 @@ def reverse_transform(self, data, reset_keys=False):
             sampled_columns.extend(self._keys)
 
         for constraint in reversed(self._constraints_to_reverse):
+            sampled_columns = constraint.reverse_transform(sampled_columns)
             reversed_data = constraint.reverse_transform(reversed_data)
 
         # Add new columns generated by the constraint
         new_columns = list(set(reversed_data.columns) - set(sampled_columns))
         sampled_columns.extend(new_columns)
 
-        # Sort the sampled columns in the order of the metadata
+        # Sort the sampled columns in the order of the metadata.
+        # Any extra columns not present in the metadata will be dropped.
         # In multitable there may be missing columns in the sample such as foreign keys
         # And alternate keys. Thats the reason of ensuring that the metadata column is within
         # The sampled columns.