From 0c653a03b31392cffd953037717735d4dfb5b6c8 Mon Sep 17 00:00:00 2001
From: Temrjan <x.temrjan@gmail.com>
Date: Wed, 25 Mar 2026 09:52:37 +0500
Subject: [PATCH] fix: guard against empty regex matches in WordAug to prevent
 IndexError

__replace() and __text2emoji() crash with IndexError when a token
contains only characters not matched by the regex (e.g. parentheses,
em-dashes). Return the original token unchanged when re.findall()
produces an empty list.

Added regression tests reproducing the exact scenario from issue #20.

Closes #20

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---
 augmentex/word.py      | 20 +++++++-----
 tests/__init__.py      |  0
 tests/test_word_aug.py | 71 ++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 83 insertions(+), 8 deletions(-)
 create mode 100644 tests/__init__.py
 create mode 100644 tests/test_word_aug.py

diff --git a/augmentex/word.py b/augmentex/word.py
index 4f44b02..10a99fa 100644
--- a/augmentex/word.py
+++ b/augmentex/word.py
@@ -102,11 +102,13 @@ def __text2emoji(self, word: str) -> str:
         Returns:
             str: Emoji that matches this word.
         """
-        word = re.findall("[а-яА-ЯёЁa-zA-Z0-9']+|[.,!?;-]+", word)
-        words = self.text2emoji_map.get(word[0].lower(), [word[0]])
-        word[0] = np.random.choice(words)
+        tokens = re.findall("[а-яА-ЯёЁa-zA-Z0-9']+|[.,!?;-]+", word)
+        if not tokens:
+            return word
+        words = self.text2emoji_map.get(tokens[0].lower(), [tokens[0]])
+        tokens[0] = np.random.choice(words)
 
-        return "".join(word)
+        return "".join(tokens)
 
     def __split(self, word: str) -> str:
         """Divides a word character-by-character.
@@ -130,11 +132,13 @@ def __replace(self, word: str) -> str:
         Returns:
             str: A misspelled word.
         """
-        word = re.findall("[а-яА-ЯёЁa-zA-Z0-9']+|[.,!?;]+", word)
-        word_probas = self.orfo_dict.get(word[0].lower(), [[word[0]], [1.0]])
-        word[0] = np.random.choice(word_probas[0], p=word_probas[1])
+        tokens = re.findall("[а-яА-ЯёЁa-zA-Z0-9']+|[.,!?;]+", word)
+        if not tokens:
+            return word
+        word_probas = self.orfo_dict.get(tokens[0].lower(), [[tokens[0]], [1.0]])
+        tokens[0] = np.random.choice(word_probas[0], p=word_probas[1])
 
-        return "".join(word)
+        return "".join(tokens)
 
     def __delete(self) -> str:
         """Deletes a random word.
diff --git a/tests/__init__.py b/tests/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/tests/test_word_aug.py b/tests/test_word_aug.py
new file mode 100644
index 0000000..ed73532
--- /dev/null
+++ b/tests/test_word_aug.py
@@ -0,0 +1,71 @@
+"""Regression tests for WordAug — covers IndexError on unmatchable tokens."""
+
+import unittest
+
+from augmentex import WordAug
+
+
+class TestWordAugReplace(unittest.TestCase):
+    """Tests for the 'replace' action."""
+
+    def setUp(self):
+        self.aug = WordAug(
+            unit_prob=1.0,
+            min_aug=1,
+            max_aug=5,
+            lang="rus",
+            platform="pc",
+            random_seed=42,
+        )
+
+    def test_replace_does_not_crash_on_parentheses(self):
+        """Regression for #20: tokens like '(' don't match the regex."""
+        text = "один из партнёров ( партнёршу)"
+        # Should not raise IndexError
+        result = self.aug.augment(text=text, action="replace")
+        self.assertIsInstance(result, str)
+
+    def test_replace_stress_with_original_issue_text(self):
+        """Exact reproduction from issue #20 — 100 iterations."""
+        text = (
+            "это когда в отношениях, один из партнёров насилует и истязает "
+            "своего партнёра ( партнёршу) бывает абъюзив и по отношению "
+            "родителей к своим детям"
+        )
+        for _ in range(100):
+            result = self.aug.augment(text=text, action="replace")
+            self.assertIsInstance(result, str)
+
+    def test_replace_pure_special_chars(self):
+        """Tokens consisting entirely of unmatchable characters."""
+        text = "hello — world"
+        result = self.aug.augment(text=text, action="replace")
+        self.assertIsInstance(result, str)
+
+
+class TestWordAugText2Emoji(unittest.TestCase):
+    """Tests for the 'text2emoji' action — same bug pattern as replace."""
+
+    def setUp(self):
+        self.aug = WordAug(
+            unit_prob=1.0,
+            min_aug=1,
+            max_aug=5,
+            lang="rus",
+            platform="pc",
+            random_seed=42,
+        )
+
+    def test_text2emoji_does_not_crash_on_special_chars(self):
+        text = "привет — мир"
+        result = self.aug.augment(text=text, action="text2emoji")
+        self.assertIsInstance(result, str)
+
+    def test_text2emoji_with_parentheses(self):
+        text = "слово ( другое)"
+        result = self.aug.augment(text=text, action="text2emoji")
+        self.assertIsInstance(result, str)
+
+
+if __name__ == "__main__":
+    unittest.main()