From 059636b7bda214e6ea6d77d06a66129fcd2c593d Mon Sep 17 00:00:00 2001
From: satsragclaw <270253513+satsragclaw@users.noreply.github.com>
Date: Tue, 14 Jul 2026 22:35:22 +0800
Subject: [PATCH] Fix normalize_text handling of Nirugu and ZWJ

---
 mongol_norm/shaper.py | 21 ++++++++++++++-------
 tests/test_joiners.py | 15 +++++++++++++++
 2 files changed, 29 insertions(+), 7 deletions(-)

diff --git a/mongol_norm/shaper.py b/mongol_norm/shaper.py
index 0e0e21b..4efb196 100644
--- a/mongol_norm/shaper.py
+++ b/mongol_norm/shaper.py
@@ -84,6 +84,18 @@ def is_mongolian_letter(cp):
     return cp in MONGOLIAN_BLOCK
 
 
+def is_mongolian_word_char(cp):
+    """Return whether *cp* participates in a Mongolian word run.
+
+    Strict word validation and mixed-text segmentation must share this
+    classification. Splitting at Nirugu or ZWJ changes neighbouring letters'
+    isol/init/medi/fina positions and therefore changes normalization.
+    """
+    return (is_mongolian_letter(cp)
+            or cp in FVS_CPS
+            or cp in (MVS_CP, NNBSP_CP, NIRUGU_CP, ZWJ_CP))
+
+
 def _check_word_chars(text):
     """
     Raise ValueError if `text` contains any char that is not a Mongolian
@@ -97,12 +109,7 @@ def _check_word_chars(text):
     """
     for i, ch in enumerate(text):
         cp = ord(ch)
-        if (is_mongolian_letter(cp)
-                or cp in FVS_CPS
-                or cp == MVS_CP
-                or cp == NNBSP_CP
-                or cp == NIRUGU_CP
-                or cp == ZWJ_CP):
+        if is_mongolian_word_char(cp):
             continue
         # Build a helpful error with the offending char's position and id.
         # 报错信息含越位字符的位置和码位。
@@ -1376,7 +1383,7 @@ def normalize_text(self, text):
 
         for i, ch in enumerate(text):
             cp = ord(ch)
-            is_mong = is_mongolian_letter(cp) or cp in FVS_CPS or cp == MVS_CP or cp == NNBSP_CP
+            is_mong = is_mongolian_word_char(cp)
             if current_is_mong is None:
                 current_is_mong = is_mong
             elif is_mong != current_is_mong:
diff --git a/tests/test_joiners.py b/tests/test_joiners.py
index 771cfec..e3a1898 100644
--- a/tests/test_joiners.py
+++ b/tests/test_joiners.py
@@ -84,5 +84,20 @@ def test_single_sided_nirugu_round_trips(self):
             self._round_trips(text)
 
 
+class TestJoinerNormalizeText(_Base):
+    def test_nirugu_word_uses_the_same_joining_context_as_normalize(self):
+        word = NIRUGU + U + NIRUGU
+        self.assertEqual(self.s.normalize_text(word), self.s.normalize(word))
+
+    def test_nirugu_word_inside_mixed_text_uses_the_same_joining_context(self):
+        word = NIRUGU + U + NIRUGU
+        self.assertEqual(self.s.normalize_text('A ' + word + ' B'),
+                         'A ' + self.s.normalize(word) + ' B')
+
+    def test_zwj_word_uses_the_same_joining_context_as_normalize(self):
+        word = ZWJ + D
+        self.assertEqual(self.s.normalize_text(word), self.s.normalize(word))
+
+
 if __name__ == '__main__':
     unittest.main()
\ No newline at end of file