Ensure that numpy API __array_function__ for external data types is not activated in sklearn flows(scikit-learn#15865)

alexshacked · alexshacked · commit 0f703879c414 · 2020-08-29T17:08:36.000+03:00
diff --git a/sklearn/inspection/tests/test_permutation_importance.py b/sklearn/inspection/tests/test_permutation_importance.py
@@ -23,9 +23,21 @@
 from sklearn.preprocessing import scale
 from sklearn.utils import parallel_backend
 from sklearn.utils._testing import _convert_container
+from sklearn.utils.estimator_checks import _NotAnArray
 
 
-
+def test_array_function_not_called():
+    X = np.array([[1, 1], [1, 2], [1, 3], [1, 4],
+                  [2, 1], [2, 2], [2, 3], [2, 4],
+                  [3, 1], [3, 2], [3, 3], [3, 4]])
+    X = _NotAnArray(X)
+    y = _NotAnArray([1, 1, 1, 1, 2, 2, 2, 2, 1, 1, 2, 2])
+    estimator = LogisticRegression()
+    estimator.fit(X, y)
+    rng = np.random.RandomState(42)
+    permutation_importance(estimator, X, y, n_repeats=5,
+                           random_state=rng, n_jobs=1)
+    
 @pytest.mark.parametrize("n_jobs", [1, 2])
 def test_permutation_importance_correlated_feature_regression(n_jobs):
     # Make sure that feature highly correlated to the target have a higher
diff --git a/sklearn/model_selection/tests/test_validation.py b/sklearn/model_selection/tests/test_validation.py
@@ -14,6 +14,7 @@
 
 from sklearn.model_selection.tests.test_search import FailingClassifier
 
+from sklearn.utils.estimator_checks import _NotAnArray
 from sklearn.utils._testing import assert_almost_equal
 from sklearn.utils._testing import assert_raises
 from sklearn.utils._testing import assert_raise_message
@@ -226,6 +227,17 @@ def get_params(self, deep=False):
         return {'a': self.a, 'allow_nd': self.allow_nd}
 
 
+def test_array_function_not_called():
+    X = np.array([[1, 1], [1, 2], [1, 3], [1, 4],
+                  [2, 1], [2, 2], [2, 3], [2, 4],
+                  [3, 1], [3, 2], [3, 3], [3, 4]])
+    X = _NotAnArray(X)
+    y = _NotAnArray([1, 1, 1, 1, 2, 2, 2, 2, 1, 1, 2, 2])
+    estimator = LogisticRegression()
+    grid = GridSearchCV(estimator, param_grid={'C': [1, 10]})
+    cross_validate(grid, X, y, n_jobs=2)
+
+
 # XXX: use 2D array, since 1D X is being detected as a single sample in
 # check_consistent_length
 X = np.ones((10, 2))
