Moving the collision test to test_feature_hasher.py

rth · rth · commit 4c1f7dfa54ce · 2016-09-28T20:00:18.000+02:00
diff --git a/sklearn/feature_extraction/tests/test_feature_hasher.py b/sklearn/feature_extraction/tests/test_feature_hasher.py
@@ -107,3 +107,15 @@ def test_hasher_zeros():
     # Assert that no zeros are materialized in the output.
     X = FeatureHasher().transform([{'foo': 0}])
     assert_equal(X.data.shape, (0,))
+
+
+def test_hash_collision():
+    # Ensure that hash collision does not produce zero elements
+    # in the output sparse array (issue #3637)
+    raw_X = ["ab", "ac", "ad", "ae"]
+    fh = FeatureHasher(non_negative=False, input_type="string", n_features=1)
+    X = fh.transform(raw_X)
+    assert_true(len(X.data) <= 2*len(raw_X))  # we have feature collisions
+    assert_equal((X.data == 0.).sum(), 0)
+
+
diff --git a/sklearn/feature_extraction/tests/test_text.py b/sklearn/feature_extraction/tests/test_text.py
@@ -944,15 +944,6 @@ def func():
     assert_raise_message(exception, message, func)
 
 
-def test_hashingvectorizer_hash_collision():
-    # Ensure that hash collision does not produce zero elements
-    # in the output sparse array (issue #3637)
-    text = 'investigation need records'
-    hv = HashingVectorizer(ngram_range=(1, 2), non_negative=True)
-    X = hv.transform([text])
-    assert_equal((X.data == 0.).sum(), 0)
-
-
 def test_tfidfvectorizer_binary():
     # Non-regression test: TfidfVectorizer used to ignore its "binary" param.
     v = TfidfVectorizer(binary=True, use_idf=False, norm=None)
