Rely on a public sklearn attribute instead of a private one

Signed-off-by: Pierre Bartet <[email protected]>

Author: Pierre Bartet

skl2onnx/shape_calculators/feature_selection.py

@@ -9,7 +9,7 @@ def calculate_sklearn_select(operator):
     check_input_and_output_numbers(operator, output_count_range=1)
     i = operator.inputs[0]
     N = i.get_first_dimension()
-    C = operator.raw_operator._get_support_mask().sum()
+    C = operator.raw_operator.get_support().sum()
     operator.outputs[0].type = i.type.__class__([N, C])
 
 

tests/test_sklearn_feature_selection_converters.py

@@ -51,6 +51,26 @@ def test_generic_univariate_select_int(self):
             X, model, model_onnx, basename="SklearnGenericUnivariateSelect"
         )
 
+    def test_generic_univariate_select_kbest_int(self):
+        model = GenericUnivariateSelect(mode="k_best", param=2)
+
+        X = np.array(
+            [[1, 2, 3, 1], [0, 3, 1, 4], [3, 5, 6, 1], [1, 2, 1, 5]], dtype=np.int64
+        )
+        y = np.array([0, 1, 0, 1])
+        model.fit(X, y)
+
+        model_onnx = convert_sklearn(
+            model,
+            "generic univariate select",
+            [("input", Int64TensorType([None, X.shape[1]]))],
+            target_opset=TARGET_OPSET,
+        )
+        self.assertTrue(model_onnx is not None)
+        dump_data_and_model(
+            X, model, model_onnx, basename="SklearnGenericUnivariateSelect"
+        )
+
     def test_rfe_int(self):
         model = RFE(estimator=SVR(kernel="linear"))
         X = np.array(