Merge branch 'cau/legacy-add-payload-field' of github.com:DS4SD/docling-core into feat-add-legacy-convert

Author: cau-git

docling_core/types/doc/document.py

@@ -1576,7 +1576,7 @@ def iterate_items(
         self,
         root: Optional[NodeItem] = None,
         with_groups: bool = False,
-        traverse_pictures: bool = True,
+        traverse_pictures: bool = False,
         page_no: Optional[int] = None,
         _level: int = 0,  # fixed parameter, carries through the node nesting level
     ) -> typing.Iterable[Tuple[NodeItem, int]]:  # tuple of node and level
@@ -1593,30 +1593,31 @@ def iterate_items(
         if not root:
             root = self.body
 
+        # Yield non-group items or group items when with_groups=True
         if not isinstance(root, GroupItem) or with_groups:
             if isinstance(root, DocItem):
-                if page_no is not None:
-                    for prov in root.prov:
-                        if prov.page_no == page_no:
-                            yield root, _level
-                else:
+                if page_no is None or any(
+                    prov.page_no == page_no for prov in root.prov
+                ):
                     yield root, _level
             else:
                 yield root, _level
 
+        # Handle picture traversal - only traverse children if requested
+        if isinstance(root, PictureItem) and not traverse_pictures:
+            return
+
         # Traverse children
         for child_ref in root.children:
             child = child_ref.resolve(self)
-
             if isinstance(child, NodeItem):
-                # If the child is a NodeItem, recursively traverse it
-                if not isinstance(child, PictureItem) or traverse_pictures:
-                    yield from self.iterate_items(
-                        child,
-                        _level=_level + 1,
-                        with_groups=with_groups,
-                        page_no=page_no,
-                    )
+                yield from self.iterate_items(
+                    child,
+                    with_groups=with_groups,
+                    traverse_pictures=traverse_pictures,
+                    page_no=page_no,
+                    _level=_level + 1,
+                )
 
     def _clear_picture_pil_cache(self):
         """Clear cache storage of all images."""

test/data/doc/dummy_doc.yaml.dt

@@ -4,7 +4,6 @@
 <location><page_1><loc_59><loc_0><loc_91><loc_75></location>
 <caption>Figure 1: Four examples of complex page layouts across different document categories</caption>
 </figure>
-<section_header><location><page_1><loc_42><loc_57><loc_49><loc_61></location>OPERATION (cont.)</section_header>
 <table>
 <location><page_1><loc_42><loc_57><loc_49><loc_61></location>
 </table>

test/data/doc/dummy_doc.yaml.et

@@ -1,6 +1,5 @@
 0: unspecified with name=_root_
  1: title
  2: picture
-  3: section_header
- 4: caption
- 5: table
+ 3: caption
+ 4: table

test/data/doc/dummy_doc.yaml.html

@@ -57,6 +57,5 @@
     </head>
 <h1>DocLayNet: A Large Human-Annotated Dataset for Document-Layout Analysis</h1>
 <figure><figcaption>Figure 1: Four examples of complex page layouts across different document categories</figcaption></figure>
-<h2>OPERATION (cont.)</h2>
 
 </html>

test/data/doc/dummy_doc.yaml.md

@@ -2,6 +2,4 @@
 
 Figure 1: Four examples of complex page layouts across different document categories
 
-<!-- image -->
-
-## OPERATION (cont.)
+<!-- image -->