Add print message

sdv/single_table/ctgan.py

@@ -138,6 +138,31 @@ def _estimate_num_columns(self, data):
 
         return num_generated_columns
 
+    def _preprocess(self, data):
+        dict_generated_columns = self._estimate_num_columns(data)
+        if sum(dict_generated_columns.values()) > 1000:
+            header = {'Original Column Name  ': 'Est # of Columns (CTGAN)'}
+            dict_generated_columns = {**header, **dict_generated_columns}
+            longest_column_name = len(max(dict_generated_columns, key=len))
+            cap = '<' + str(longest_column_name)
+            lines_to_print = []
+            for column, num_generated_columns in dict_generated_columns.items():
+                lines_to_print.append(f'{column:{cap}} {num_generated_columns}')
+
+            generated_columns_str = '\n'.join(lines_to_print)
+            print(  # noqa: T001
+                'PerformanceAlert: Using the CTGANSynthesizer on this data is not recommended. '
+                'To model this data, CTGAN will generate a large number of columns.'
+                '\n\n'
+                f'{generated_columns_str}'
+                '\n\n'
+                'We recommend preprocessing discrete columns that can have many values, '
+                "using 'update_transformers'. Or you may drop columns that are not necessary "
+                'to model. (Exit this script using ctrl-C)'
+            )
+
+        return super()._preprocess(data)
+
     def _fit(self, processed_data):
         """Fit the model to the table.
 

tests/unit/metadata/test_multi_table.py

@@ -2014,7 +2014,7 @@ def test_detect_table_from_csv(self, load_csv_mock, single_table_mock, log_mock)
         should be created and call the ``detect_from_csv`` method.
 
         Setup:
-            - Mock the ``SingleTableMetadata`` class and the print function.
+            - Mock the ``SingleTableMetadata`` class and the logger.
 
         Assert:
             - Table should be added to ``self.tables``.

tests/unit/single_table/test_ctgan.py

@@ -1,5 +1,9 @@
+import re
 from unittest.mock import Mock, patch
 
+import numpy as np
+import pandas as pd
+
 from sdv.metadata.single_table import SingleTableMetadata
 from sdv.single_table.ctgan import CTGANSynthesizer, TVAESynthesizer
 
@@ -128,6 +132,37 @@ def test_get_parameters(self):
             'cuda': True,
         }
 
+    def test_preprocessing_many_categories(self, capfd):
+        """Test a message is printed during preprocess when a column has many categories."""
+        # Setup
+        metadata = SingleTableMetadata()
+        metadata.add_column('name_longer_than_Original_Column_Name', sdtype='numerical')
+        metadata.add_column('categorical', sdtype='categorical')
+        data = pd.DataFrame({
+            'name_longer_than_Original_Column_Name': np.random.rand(1_001),
+            'categorical': [f'cat_{i}' for i in range(1_001)],
+        })
+        instance = CTGANSynthesizer(metadata)
+
+        # Run
+        instance.auto_assign_transformers(data)
+        instance.preprocess(data)
+
+        # Assert
+        out, err = capfd.readouterr()
+        assert out == re.escape(
+            'PerformanceAlert: Using the CTGANSynthesizer on this data is not recommended. '
+            'To model this data, CTGAN will generate a large number of columns.'
+            ''
+            'Original Column Name                  Est # of Columns (CTGAN)'
+            'name_longer_than_Original_Column_Name 11'
+            'categorical                           1001'
+            ''
+            'We recommend preprocessing discrete columns that can have many values, '
+            "using 'update_transformers'. Or you may drop columns that are not necessary "
+            'to model. (Exit this script using ctrl-C)'
+        )
+
     @patch('sdv.single_table.ctgan.CTGAN')
     @patch('sdv.single_table.ctgan.detect_discrete_columns')
     def test__fit(self, mock_detect_discrete_columns, mock_ctgan):