CLN Fixes PendingDeprecationWarning in CountVectorizer (#19299)

thomasjpfan · ogrisel · glemaitre · commit e9b9f2390e36 · 2021-04-28T09:40:11.000+02:00
Co-authored-by: Olivier Grisel &lt;olivier.grisel@ensta.org&gt;
diff --git a/sklearn/feature_extraction/tests/test_text.py b/sklearn/feature_extraction/tests/test_text.py
@@ -773,18 +773,30 @@ def test_vectorizer_inverse_transform(Vectorizer):
     vectorizer = Vectorizer()
     transformed_data = vectorizer.fit_transform(data)
     inversed_data = vectorizer.inverse_transform(transformed_data)
+    assert isinstance(inversed_data, list)
+
     analyze = vectorizer.build_analyzer()
     for doc, inversed_terms in zip(data, inversed_data):
         terms = np.sort(np.unique(analyze(doc)))
         inversed_terms = np.sort(np.unique(inversed_terms))
         assert_array_equal(terms, inversed_terms)
 
-    # Test that inverse_transform also works with numpy arrays
-    transformed_data = transformed_data.toarray()
-    inversed_data2 = vectorizer.inverse_transform(transformed_data)
+    assert sparse.issparse(transformed_data)
+    assert transformed_data.format == "csr"
+
+    # Test that inverse_transform also works with numpy arrays and
+    # scipy
+    transformed_data2 = transformed_data.toarray()
+    inversed_data2 = vectorizer.inverse_transform(transformed_data2)
     for terms, terms2 in zip(inversed_data, inversed_data2):
         assert_array_equal(np.sort(terms), np.sort(terms2))
 
+    # Check that inverse_transform also works on non CSR sparse data:
+    transformed_data3 = transformed_data.tocsc()
+    inversed_data3 = vectorizer.inverse_transform(transformed_data3)
+    for terms, terms3 in zip(inversed_data, inversed_data3):
+        assert_array_equal(np.sort(terms), np.sort(terms3))
+
 
 def test_count_vectorizer_pipeline_grid_selection():
     # raw documents
diff --git a/sklearn/feature_extraction/text.py b/sklearn/feature_extraction/text.py
@@ -1270,22 +1270,20 @@ def inverse_transform(self, X):
             List of arrays of terms.
         """
         self._check_vocabulary()
-
-        if sp.issparse(X):
-            # We need CSR format for fast row manipulations.
-            X = X.tocsr()
-        else:
-            # We need to convert X to a matrix, so that the indexing
-            # returns 2D objects
-            X = np.asmatrix(X)
+        # We need CSR format for fast row manipulations.
+        X = check_array(X, accept_sparse='csr')
         n_samples = X.shape[0]
 
         terms = np.array(list(self.vocabulary_.keys()))
         indices = np.array(list(self.vocabulary_.values()))
         inverse_vocabulary = terms[np.argsort(indices)]
 
-        return [inverse_vocabulary[X[i, :].nonzero()[1]].ravel()
-                for i in range(n_samples)]
+        if sp.issparse(X):
+            return [inverse_vocabulary[X[i, :].nonzero()[1]].ravel()
+                    for i in range(n_samples)]
+        else:
+            return [inverse_vocabulary[np.flatnonzero(X[i, :])].ravel()
+                    for i in range(n_samples)]
 
     def get_feature_names(self):
         """Array mapping from feature integer indices to feature name.
