Do not enforce min/max on sequence index column (#2043)

sdv/sequential/par.py

@@ -194,6 +194,28 @@ def _transform_sequence_index(self, data):
 
         return data
 
+    def auto_assign_transformers(self, data):
+        """Automatically assign the required transformers for the given data and constraints.
+
+        This method will automatically set a configuration to the ``rdt.HyperTransformer``
+        with the required transformers for the current data.
+
+        Args:
+            data (dict):
+                Mapping of table name to pandas.DataFrame.
+
+        Raises:
+            InvalidDataError:
+                If a table of the data is not present in the metadata.
+        """
+        super().auto_assign_transformers(data)
+
+        # Ensure that sequence index does not get auto assigned with enforce_min_max_values
+        if self._sequence_index:
+            sequence_index_transformer = self.get_transformers()[self._sequence_index]
+            if sequence_index_transformer.enforce_min_max_values:
+                sequence_index_transformer.enforce_min_max_values = False
+
     def _preprocess(self, data):
         """Transform the raw data to numerical space.
 

tests/integration/sequential/test_par.py

@@ -282,3 +282,45 @@ def test_par_missing_sequence_index():
     # Assert
     assert sampled.shape == data.shape
     assert (sampled.dtypes == data.dtypes).all()
+
+
+def test_par_unique_sequence_index_with_enforce_min_max():
+    """Test to see if there are duplicate sequence index values
+    when sequence_length is higher than real data
+    """
+    # Setup
+    test_id = list(range(10))
+    s_key = [0, 0, 0, 0, 0, 1, 1, 1, 1, 1]
+    visits = [
+        '2021-01-01', '2021-01-03', '2021-01-05', '2021-01-07', '2021-01-09',
+        '2021-09-11', '2021-09-17', '2021-10-01', '2021-10-08', '2021-11-01'
+    ]
+    pre_date = [
+        '2020-01-01', '2020-01-02', '2020-01-03', '2020-01-04', '2020-01-05',
+        '2021-04-01', '2021-04-02', '2021-04-03', '2021-04-04', '2021-04-05'
+    ]
+    test_df = pd.DataFrame({
+        'id': test_id,
+        's_key': s_key,
+        'visits': visits,
+        'pre_date': pre_date
+    })
+    test_df[['visits', 'pre_date']] = test_df[['visits', 'pre_date']].apply(
+        pd.to_datetime, format='%Y-%m-%d', errors='coerce')
+    metadata = SingleTableMetadata()
+    metadata.detect_from_dataframe(test_df)
+    metadata.update_column(column_name='s_key', sdtype='id')
+    metadata.set_sequence_key('s_key')
+    metadata.set_sequence_index('visits')
+    synthesizer = PARSynthesizer(metadata, enforce_min_max_values=True,
+                                 enforce_rounding=False, epochs=100, verbose=True)
+
+    # Run
+    synthesizer.fit(test_df)
+    synth_df = synthesizer.sample(num_sequences=50, sequence_length=50)
+
+    # Assert
+    for i in synth_df['s_key'].unique():
+        seq_df = synth_df[synth_df['s_key'] == i]
+        has_duplicates = seq_df['visits'].duplicated().any()
+        assert not has_duplicates

tests/unit/sequential/test_par.py

@@ -340,6 +340,9 @@ def test_preprocess(self, base_preprocess_mock):
         par._transform_sequence_index = Mock()
         par.auto_assign_transformers = Mock()
         par.update_transformers = Mock()
+        get_transform_mock = Mock()
+        get_transform_mock.return_value = {'time': Mock()}
+        par.get_transformers = get_transform_mock
         par._data_processor._prepared_for_fitting = True
         data = self.get_data()