Add input validation + tests for checking rounding scheme

Author: frances-h

rdt/transformers/numerical.py

@@ -670,6 +670,11 @@ def __init__(
         max_value=1.0,
         learn_rounding_scheme=False,
     ):
+        if not (isinstance(min_value, int) or isinstance(min_value, float)) or not (
+            isinstance(max_value, int) or isinstance(max_value, float)
+        ):
+            error_msg = 'The min_value and max_value must be of type int or float.'
+            raise TransformerInputError(error_msg)
         if min_value == max_value:
             error_msg = 'The min_value and max_value for the logit function cannot be equal.'
             raise TransformerInputError(error_msg)

tests/integration/transformers/test_numerical.py

@@ -7,6 +7,7 @@
     ClusterBasedNormalizer,
     FloatFormatter,
     GaussianNormalizer,
+    LogitScaler,
 )
 
 
@@ -560,3 +561,59 @@ def test_out_of_bounds_reverse_transform(self):
 
         # Assert
         assert isinstance(reverse, pd.DataFrame)
+
+
+class TestLogitScaler:
+    def test_learn_rounding(self):
+        """Test that transformer learns rounding scheme from data."""
+        # Setup
+        data = pd.DataFrame({'test': [0.2, np.nan, 1.0]})
+        transformer = LogitScaler(
+            missing_value_generation=None,
+            missing_value_replacement='mean',
+            learn_rounding_scheme=True,
+        )
+        expected = pd.DataFrame({'test': [0.2, 0.6, 1.0]})
+
+        # Run
+        transformer.fit(data, 'test')
+        transformed = transformer.transform(data)
+        reversed_values = transformer.reverse_transform(transformed)
+
+        # Assert
+        np.testing.assert_array_equal(reversed_values, expected)
+
+    def test_missing_value_generation_from_column(self):
+        """Test from_column missing value generation with nans."""
+        # Setup
+        data = pd.DataFrame({'test': [0.2, np.nan, 1.0]})
+        transformer = LogitScaler(
+            missing_value_generation='from_column',
+            missing_value_replacement='mean',
+        )
+
+        # Run
+        transformer.fit(data, 'test')
+        transformed = transformer.transform(data)
+        reversed_values = transformer.reverse_transform(transformed)
+
+        # Assert
+        np.testing.assert_array_almost_equal(reversed_values, data)
+
+    def test_missing_value_generation_random(self):
+        """Test random missing_value_generation with nans."""
+        # Setup
+        data = pd.DataFrame({'test': [0.2, np.nan, 1.0, 1.0]})
+        transformer = LogitScaler(
+            missing_value_generation='random',
+            missing_value_replacement='mode',
+        )
+        expected = pd.DataFrame({'test': [0.2, np.nan, 1.0, np.nan]})
+
+        # Run
+        transformer.fit(data, 'test')
+        transformed = transformer.transform(data)
+        reversed_values = transformer.reverse_transform(transformed)
+
+        # Assert
+        np.testing.assert_array_almost_equal(reversed_values, expected)

tests/unit/transformers/test_numerical.py

@@ -1892,13 +1892,19 @@ def test___init__(self):
     def test___init___invalid_inputs(self):
         """Test super() arguments are properly passed and set as attributes."""
         # Setup
-        min_value = 10.0
-        max_value = 10.0
+        same_min_value = 10.0
+        same_max_value = 10.0
+        bad_min_value = '10.0'
+        bad_max_value = (100.0,)
 
         # Run / Assert
         expected_msg = 'The min_value and max_value for the logit function cannot be equal.'
         with pytest.raises(TransformerInputError, match=re.escape(expected_msg)):
-            LogitScaler(max_value=max_value, min_value=min_value)
+            LogitScaler(max_value=same_max_value, min_value=same_min_value)
+
+        expected_msg = 'The min_value and max_value must be of type int or float.'
+        with pytest.raises(TransformerInputError, match=re.escape(expected_msg)):
+            LogitScaler(max_value=bad_max_value, min_value=bad_min_value)
 
     def test__validate_logit_inputs_with_default_settings(self):
         """Test validating data against input arguments."""