Fix the extended Arabic ranges in HARAKAT_RE

.github/workflows/claude-code-review.yml

@@ -12,6 +12,7 @@ on:
 
 jobs:
   claude-review:
+    if: github.event.pull_request.head.repo.full_name == github.repository
     # Optional: Filter by PR author
     # if: |
     #   github.event.pull_request.user.login == 'external-contributor' ||

arabic_reshaper/arabic_reshaper.py

@@ -27,8 +27,8 @@
     '\u06d6-\u06dc'
     '\u06df-\u06e8'
     '\u06ea-\u06ed'
-    '\u08d4-\u08e1'
-    '\u08d4-\u08ed'
+    '\u0897-\u089f'
+    '\u08ca-\u08e1'
     '\u08e3-\u08ff'
     ']',
 

arabic_reshaper/tests/test_006_harakat_ranges.py

@@ -0,0 +1,30 @@
+import unittest
+
+import arabic_reshaper
+
+
+class TestHarakatRanges(unittest.TestCase):
+    def setUp(self):
+        self.keep = arabic_reshaper.ArabicReshaper(
+            configuration={'delete_harakat': False}
+        )
+
+    def test_extended_marks_do_not_break_joining(self):
+        # Combining marks added in later Unicode versions must not split
+        # the letters around them. U+FE91 and U+FE90 are initial and
+        # final beh.
+        for mark in ('\u0897', '\u089f', '\u08ca', '\u08d0', '\u08d3',
+                     '\u08e1', '\u08e3', '\u08ff'):
+            with self.subTest(mark='U+%04X' % ord(mark)):
+                self.assertEqual(
+                    self.keep.reshape('\u0628' + mark + '\u0628'),
+                    '\ufe91' + mark + '\ufe90'
+                )
+
+    def test_disputed_end_of_ayah_is_not_a_haraka(self):
+        # U+08E2 is a format character, so deleting harakat keeps it.
+        self.assertIn('\u08e2', arabic_reshaper.reshape('\u0628\u08e2'))
+
+
+if __name__ == '__main__':
+    unittest.main()