feat: Implementing the form structure based on the new iso-standard

docling_core/types/doc/document.py

@@ -1188,6 +1188,7 @@ class TextItem(DocItem):
         DocItemLabel.REFERENCE,
         DocItemLabel.TEXT,
         DocItemLabel.EMPTY_VALUE,
+        DocItemLabel.FORM_KEY,
     ]
 
     orig: str  # untreated representation
@@ -1920,14 +1921,55 @@ def export_to_document_tokens(
         text = serializer.serialize(item=self).text
         return text
 
+class CheckboxItem(ListItem):
+    """FormTextItem."""
+
+    label: typing.Literal[DocItemLabel.CHECKBOX] = DocItemLabel.CHECKBOX
 
+    checked: bool = False
+
+"""
+class FormHeaderItem(SectionHeaderItem):
+
+    label: typing.Literal[DocItemLabel.FORM_HEADER] = DocItemLabel.FORM_HEADER
+
+class FormTextItem(TextItem):
+
+    label: typing.Literal[DocItemLabel.FORM_TEXT] = DocItemLabel.FORM_TEXT
+"""
+
+class FormListItem(DocItem):
+    """FormListItem."""
+
+    label: typing.Literal[DocItemLabel.FORM_LISTITEM] = DocItemLabel.FORM_LISTITEM
+
+    marker: Optional[TextItem] = None
+
+    key: TextItem
+
+    def add_value(self, item: Union[CheckboxItem, ListItem, TextItem]) -> NodeItem:    
+        item.parent = self.get_ref()
+        self.children.append(item)
+
+        return item
+
+
+
 class FormItem(FloatingItem):
     """FormItem."""
 
     label: typing.Literal[DocItemLabel.FORM] = DocItemLabel.FORM
 
-    graph: GraphData
+    def add(self, item: Union["FormItem", SectionHeaderItem, TextItem, FormListItem]) -> NodeItem:
+        item.parent = self.get_ref()
+        self.children.append(item.get_ref())
+
+        return item
 
+    def add_listitem(self, doc: DoclingDocument, prov: Optional[ProvenanceItem] = None) -> NodeItem:
+        li = FormListItem(self_ref=self.get_ref())
+        return item
+
 
 ContentItem = Annotated[
     Union[
@@ -1940,6 +1982,7 @@ class FormItem(FloatingItem):
         PictureItem,
         TableItem,
         KeyValueItem,
+        FormItem,
     ],
     Field(discriminator="label"),
 ]
@@ -2987,7 +3030,7 @@ def add_key_values(
 
     def add_form(
         self,
-        graph: GraphData,
+        form: Optional[FormItem] = None,
         prov: Optional[ProvenanceItem] = None,
         parent: Optional[NodeItem] = None,
     ):
@@ -3003,11 +3046,12 @@ def add_form(
         form_index = len(self.form_items)
         cref = f"#/form_items/{form_index}"
 
-        form_item = FormItem(
-            graph=graph,
-            self_ref=cref,
-            parent=parent.get_ref(),
-        )
+        if form is None:
+            form = FormItem(
+                self_ref=cref,
+                parent=parent.get_ref(),
+            )
+
         if prov:
             form_item.prov.append(prov)
 

docling_core/types/doc/labels.py

@@ -31,6 +31,11 @@ class DocItemLabel(str, Enum):
     HANDWRITTEN_TEXT = "handwritten_text"
     EMPTY_VALUE = "empty_value"  # used for empty value fields in fillable forms
 
+    # FORM_HEADER = "form_header"
+    FORM_KEY = "form_key"
+    FORM_LISTITEM = "form_listitem"
+    CHECKBOX = "checkbox"
+
     # Additional labels for markup-based formats (e.g. HTML, Word)
     PARAGRAPH = "paragraph"
     REFERENCE = "reference"

test/data/docling_document/unit/FormItem.yaml

@@ -24,7 +24,7 @@ graph:
     source_cell_id: 1
     target_cell_id: 0
 image: null
-label: form
+label: key_value_region
 parent: null
 prov: []
 references: []

test/test_docling_doc.py

@@ -14,6 +14,7 @@
 from docling_core.types.doc.base import BoundingBox, CoordOrigin, ImageRefMode, Size
 from docling_core.types.doc.document import (  # BoundingBox,
     CURRENT_VERSION,
+    CheckboxItem,
     CodeItem,
     ContentLayer,
     DocItem,
@@ -42,6 +43,8 @@
     TableItem,
     TextItem,
     TitleItem,
+    CheckboxItem,
+    FormListItem,    
 )
 from docling_core.types.doc.labels import (
     DocItemLabel,
@@ -491,6 +494,7 @@ def verify(dc, obj):
 
         elif dc is FormItem:
 
+            """
             graph = GraphData(
                 cells=[
                     GraphCell(
@@ -524,7 +528,31 @@ def verify(dc, obj):
                 self_ref="#",
             )
             verify(dc, obj)
+            """
+
+            key_name = TextItem(text="name", orig="name", self_ref="#", label=DocItemLabel.FORM_KEY)
+            val_name = TextItem(text="John Doe", orig="name", self_ref="#", label=DocItemLabel.TEXT)
+
+            form_item_name = FormListItem(key=key_name, self_ref="#")
+            form_item_name.add_value(val_name)
+
+            key_age = TextItem(text="Age", orig="Age", self_ref="#", label=DocItemLabel.FORM_KEY)
+
+            cb_age_0 = CheckboxItem(checked=True, text="0-20", orig="0-20", self_ref="#")
+            cb_age_1 = CheckboxItem(checked=False, text="20-40", orig="20-40", self_ref="#")
+            val_age = TextItem(text="other", orig="other", self_ref="#", label=DocItemLabel.TEXT)
+
+            form_item_age = FormListItem(key=key_age, self_ref="#") #, value=[cb_age_0, cb_age_1, val_age])
+            for _ in [cb_age_0, cb_age_1, val_age]:
+                form_item_age.add_value(_)
+
+            form = FormItem(self_ref="#")
+
+            form.add(form_item_name)
+            form.add(form_item_age)
 
+            verify(dc, obj)
+
         elif dc is TitleItem:
             obj = dc(
                 text="whatever",
@@ -571,8 +599,12 @@ def verify(dc, obj):
                 text="E=mc^2",
             )
             verify(dc, obj)
-        elif dc is GraphData:  # we skip this on purpose
+        elif dc is CheckboxItem:  # we skip this on purpose
+            continue            
+        elif dc is FormListItem:  # we skip this on purpose
             continue
+        elif dc is GraphData:  # we skip this on purpose
+            continue        
         else:
             raise RuntimeError(f"New derived class detected {dc.__name__}")
 
@@ -1002,8 +1034,10 @@ def _construct_doc() -> DoclingDocument:
 
     doc.add_key_values(graph=graph)
 
-    doc.add_form(graph=graph)
+    form_1 = doc.add_form(graph=graph)
 
+    form_1_item_1 = form_1.add_listitem(key="Name")
+
     inline_fmt = doc.add_inline_group()
     doc.add_text(
         label=DocItemLabel.TEXT, text="Some formatting chops:", parent=inline_fmt