Strip numbering from revised_text when violation_text is stripped

• Track strip mode instead of boolean
• Apply same stripping to revised_text
• Only strip for replace actions
• Add regression test coverage
• Ensure consistent numbering handling

Author: danielaskdd

skills/doc-audit/scripts/docx_edit/item_search_mixin.py

@@ -513,15 +513,15 @@ def _locate_item_match(
                         match_pos = -1
                         matched_text = violation_text
                         matched_override = None
-                        matched_is_stripped = False
+                        matched_strip_mode: Optional[str] = None
 
-                        for search_text, is_stripped in search_attempts:
+                        for search_text, strip_mode in search_attempts:
                             match_pos, matched_override = self._find_in_runs_with_normalization(
                                 all_runs, search_text
                             )
                             if match_pos != -1:
                                 matched_text = matched_override or search_text
-                                matched_is_stripped = is_stripped
+                                matched_strip_mode = strip_mode
                                 break
 
                         if match_pos != -1:
@@ -530,7 +530,12 @@ def _locate_item_match(
                             matched_start = match_pos
                             is_cross_paragraph = len(body_paras_data) > 1
                             violation_text = matched_text
-                            numbering_stripped = matched_is_stripped
+                            numbering_stripped = matched_strip_mode is not None
+                            if matched_strip_mode and item.fix_action == 'replace':
+                                stripped_revised, _ = strip_numbering_by_mode(
+                                    revised_text, matched_strip_mode
+                                )
+                                revised_text = stripped_revised
 
                             if self.verbose:
                                 if is_cross_paragraph:

tests/test_apply_audit_edits_core.py

@@ -803,6 +803,60 @@ def test_no_strip_normal(self):
         assert strip_auto_numbering("Normal text") == ("Normal text", False)
 
 
+class TestLocateItemMatchNumbering:
+    """Regression tests for numbering stripping during item locate."""
+
+    def test_boundary_body_match_strips_revised_text(self):
+        """When violation_text is stripped by mode, revised_text should be stripped too."""
+        applier = create_mock_applier()
+        para = create_paragraph_xml(
+            "三防漆涂覆\n依据文件ADYU0.005.000《三防漆涂覆通用工艺规程》和"
+            "ADYU2.088.1192ZD6《信息处理组件三防漆涂覆作业指导书》中要求执行。",
+            para_id="AAAA1111",
+        )
+        item = create_edit_item(
+            uuid="AAAA1111",
+            uuid_end="AAAA1111",
+            fix_action="replace",
+            violation_text=(
+                "h) 三防漆涂覆\n依据文件ADYU0.005.000《三防漆涂覆通用工艺规程》和"
+                "ADYU2.088.1192ZD6《信息处理组件三防漆涂覆作业指导书》中要求执行。"
+            ),
+            revised_text=(
+                "h) 三防漆涂覆\n依据文件ADYU0.005.000《三防漆涂覆通用工艺规程》和"
+                "ADYU2.088.1192ZD5《信息处理组件三防漆涂覆作业指导书》中要求执行。"
+            ),
+        )
+
+        # Force fallback route:
+        # single-para original miss -> numbering variant miss -> boundary_crossed ->
+        # body segment original miss -> body segment numbering variant hit
+        search_results = iter([(-1, None), (-1, None), (-1, None), (0, None)])
+        applier._find_in_runs_with_normalization = lambda runs, text: next(search_results)
+        applier._iter_paragraphs_in_range = lambda start_para, uuid_end: iter([para])
+        applier._collect_runs_info_across_paragraphs = (
+            lambda start_para, uuid_end: ([], "", False, "boundary_crossed")
+        )
+        applier._find_tables_in_range = lambda start_para, uuid_end: []
+        applier._is_paragraph_in_table = lambda _para: False
+        applier._apply_fallback_comment = lambda *args, **kwargs: None
+
+        context = applier._locate_item_match(
+            item=item,
+            anchor_para=para,
+            violation_text=item.violation_text,
+            revised_text=item.revised_text,
+        )
+
+        assert context["target_para"] is para
+        assert context["matched_start"] == 0
+        assert context["numbering_stripped"] is True
+        assert context["violation_text"].startswith("三防漆涂覆")
+        assert context["revised_text"].startswith("三防漆涂覆")
+        assert not context["revised_text"].startswith("h) ")
+        assert "ZD5" in context["revised_text"]
+
+
 # ============================================================
 # Tests: _find_revision_ancestor
 # ============================================================