Merge pull request #506 from QData/augment-test

add more tests for augment function

tests/test_misc.py → tests/test_augment_api.py

@@ -61,6 +61,18 @@ def test_easydata_augmenter():
     assert augmented_s in augmented_text_list
 
 
+def test_easydata_augmenter2():
+    from textattack.augmentation import EasyDataAugmenter
+
+    augmenter = EasyDataAugmenter(
+        pct_words_to_swap=0.01, transformations_per_example=64
+    )
+    s = "hello hello hello derek"
+    augmented_text_list = augmenter.augment(s)
+    augmented_s = "derek hello hello hello"
+    assert augmented_s in augmented_text_list
+
+
 def test_wordnet_augmenter():
     from textattack.augmentation import WordNetAugmenter
 
@@ -69,3 +81,13 @@ def test_wordnet_augmenter():
     augmented_text_list = augmenter.augment(s)
     augmented_s = "The firedrake warrior is a panda"
     assert augmented_s in augmented_text_list
+
+
+def test_deletion_augmenter():
+    from textattack.augmentation import DeletionAugmenter
+
+    augmenter = DeletionAugmenter(pct_words_to_swap=0.1, transformations_per_example=10)
+    s = "The United States"
+    augmented_text_list = augmenter.augment(s)
+    augmented_s = "United States"
+    assert augmented_s in augmented_text_list

textattack/constraints/semantics/word_embedding_distance.py

@@ -64,6 +64,13 @@ def _check_constraint(self, transformed_text, reference_text):
                 "Cannot apply part-of-speech constraint without `newly_modified_indices`"
             )
 
+        # FIXME The index i is sometimes larger than the number of tokens - 1
+        if any(
+            i >= len(reference_text.words) or i >= len(transformed_text.words)
+            for i in indices
+        ):
+            return False
+
         for i in indices:
             ref_word = reference_text.words[i]
             transformed_word = transformed_text.words[i]