confluence-mdx: implement Phase 4 container reconstruction

Author: jk-kim0

confluence-mdx/bin/reverse_sync/mapping_recorder.py

@@ -18,6 +18,7 @@ class BlockMapping:
 HEADING_TAGS = {'h1', 'h2', 'h3', 'h4', 'h5', 'h6'}
 
 CALLOUT_MACRO_NAMES = frozenset({'tip', 'info', 'note', 'warning', 'panel'})
+CONTAINER_MACRO_NAMES = CALLOUT_MACRO_NAMES | frozenset({'expand'})
 
 # backward-compat aliases
 _CALLOUT_MACRO_NAMES = CALLOUT_MACRO_NAMES
@@ -102,15 +103,15 @@ def record_mapping(xhtml: str) -> List[BlockMapping]:
                              block_type='code')
             else:
                 # Callout 매크로: body 텍스트만 추출 (파라미터 메타데이터 제외)
-                if macro_name in CALLOUT_MACRO_NAMES:
+                if macro_name in CONTAINER_MACRO_NAMES:
                     rich_body = child.find('ac:rich-text-body')
                     plain = get_text_with_emoticons(rich_body) if rich_body else child.get_text()
                 else:
                     plain = child.get_text()
                 _add_mapping(mappings, counters, f'macro-{macro_name}', str(child), plain,
                              block_type='html_block')
-                # Callout 매크로: 자식 요소 개별 매핑 추가
-                if macro_name in CALLOUT_MACRO_NAMES:
+                # callout/details 매크로: 자식 요소 개별 매핑 추가
+                if macro_name in CONTAINER_MACRO_NAMES:
                     parent_mapping = mappings[-1]
                     _add_rich_text_body_children(
                         child, parent_mapping, mappings, counters)

confluence-mdx/bin/reverse_sync/patch_builder.py

@@ -192,6 +192,24 @@ def _mapping_block_family(mapping: BlockMapping) -> str:
     return _xpath_block_family(mapping.xhtml_xpath)
 
 
+def _strategy_for_resolved_mapping(
+    change: BlockChange,
+    mapping: BlockMapping,
+) -> str:
+    """이미 찾은 mapping에 대해 처리 전략을 다시 계산한다."""
+    if change.old_block.type == 'callout':
+        return 'containing'
+    if mapping.children:
+        if change.old_block.type == 'list':
+            return 'list'
+        return 'containing'
+    if change.old_block.type == 'list':
+        return 'list'
+    if is_markdown_table(change.old_block.content):
+        return 'table'
+    return 'direct'
+
+
 def _flush_containing_changes(
     containing_changes: dict,
     used_ids: 'set | None' = None,
@@ -330,7 +348,9 @@ def _mark_used(block_id: str, m: BlockMapping):
             idx, mdx_to_sidecar, xpath_to_mapping)
         if mapping is None:
             continue
-        sidecar_block = xpath_to_sidecar_block.get(mapping.xhtml_xpath)
+        sidecar_block = _find_roundtrip_sidecar_block(
+            del_change, mapping, roundtrip_sidecar, xpath_to_sidecar_block,
+        )
         if _is_clean_block(
             add_change.new_block.type,
             mapping,
@@ -346,6 +366,18 @@ def _mark_used(block_id: str, m: BlockMapping):
             _paired_indices.add(idx)
             _mark_used(mapping.block_id, mapping)
             continue
+        if sidecar_block_requires_reconstruction(sidecar_block):
+            patches.append(
+                _build_replace_fragment_patch(
+                    mapping,
+                    add_change.new_block,
+                    sidecar_block=sidecar_block,
+                    mapping_lost_info=mapping_lost_info,
+                )
+            )
+            _paired_indices.add(idx)
+            _mark_used(mapping.block_id, mapping)
+            continue
         old_plain = normalize_mdx_to_plain(
             del_change.old_block.content, del_change.old_block.type)
         new_plain = normalize_mdx_to_plain(
@@ -391,13 +423,37 @@ def _mark_used(block_id: str, m: BlockMapping):
             change, old_plain, mappings, used_ids,
             mdx_to_sidecar, xpath_to_mapping)
 
+        if strategy == 'skip' and roundtrip_sidecar is not None and change.old_block.content:
+            identity_sidecar = find_sidecar_block_by_identity(
+                roundtrip_sidecar.blocks,
+                sha256_text(change.old_block.content),
+                (change.old_block.line_start, change.old_block.line_end),
+            )
+            if identity_sidecar is not None:
+                mapping = xpath_to_mapping.get(identity_sidecar.xhtml_xpath)
+                if mapping is not None:
+                    strategy = _strategy_for_resolved_mapping(change, mapping)
+
+        if strategy == 'list' and mapping is None and roundtrip_sidecar is not None and change.old_block.content:
+            identity_sidecar = find_sidecar_block_by_identity(
+                roundtrip_sidecar.blocks,
+                sha256_text(change.old_block.content),
+                (change.old_block.line_start, change.old_block.line_end),
+            )
+            if identity_sidecar is not None:
+                mapping = xpath_to_mapping.get(identity_sidecar.xhtml_xpath)
+
         if strategy == 'skip':
             continue
 
-        if strategy == 'list':
-            list_sidecar = _find_roundtrip_sidecar_block(
+        sidecar_block = None
+        if mapping is not None:
+            sidecar_block = _find_roundtrip_sidecar_block(
                 change, mapping, roundtrip_sidecar, xpath_to_sidecar_block,
             )
+
+        if strategy == 'list':
+            list_sidecar = sidecar_block
             # roundtrip sidecar가 있지만 이 list에 매칭되는 block이 없을 때
             # (cross-type 거부 또는 mapping drift) clean list는 whole-fragment 재생성으로 처리
             should_replace_clean_list = (
@@ -409,7 +465,8 @@ def _mark_used(block_id: str, m: BlockMapping):
             if (mapping is not None
                     and not _contains_preserved_anchor_markup(mapping.xhtml_text)
                     and (
-                        sidecar_block_requires_reconstruction(list_sidecar)
+                        list_sidecar is not None
+                        or sidecar_block_requires_reconstruction(list_sidecar)
                         or should_replace_clean_list
                     )):
                 _mark_used(mapping.block_id, mapping)
@@ -446,6 +503,18 @@ def _mark_used(block_id: str, m: BlockMapping):
                         mdx_to_sidecar, xpath_to_mapping))
             continue
 
+        if strategy == 'containing' and sidecar_block_requires_reconstruction(sidecar_block):
+            _mark_used(mapping.block_id, mapping)
+            patches.append(
+                _build_replace_fragment_patch(
+                    mapping,
+                    change.new_block,
+                    sidecar_block=sidecar_block,
+                    mapping_lost_info=mapping_lost_info,
+                )
+            )
+            continue
+
         new_plain = normalize_mdx_to_plain(
             change.new_block.content, change.new_block.type)
 
@@ -465,9 +534,6 @@ def _mark_used(block_id: str, m: BlockMapping):
                 and collapse_ws(new_plain) == collapse_ws(mapping.xhtml_plain_text)):
             continue
 
-        sidecar_block = _find_roundtrip_sidecar_block(
-            change, mapping, roundtrip_sidecar, xpath_to_sidecar_block,
-        )
         if _can_replace_table_fragment(change, mapping, roundtrip_sidecar):
             patches.append(
                 _build_replace_fragment_patch(