docling-core 2.35.0__tar.gz → 2.37.0__tar.gz

{docling_core-2.35.0 → docling_core-2.37.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: docling-core
-Version: 2.35.0
+Version: 2.37.0
 Summary: A python library to define and validate data types in Docling.
 Author-email: Cesar Berrospi Ramis <ceb@zurich.ibm.com>, Panos Vagenas <pva@zurich.ibm.com>, Michele Dolfi <dol@zurich.ibm.com>, Christoph Auer <cau@zurich.ibm.com>, Peter Staar <taa@zurich.ibm.com>
 Maintainer-email: Panos Vagenas <pva@zurich.ibm.com>, Michele Dolfi <dol@zurich.ibm.com>, Christoph Auer <cau@zurich.ibm.com>, Peter Staar <taa@zurich.ibm.com>, Cesar Berrospi Ramis <ceb@zurich.ibm.com>

{docling_core-2.35.0 → docling_core-2.37.0}/docling_core/transforms/serializer/html.py

@@ -340,7 +340,7 @@ class HTMLTableSerializer(BaseTableSerializer):
 
                     content = html.escape(cell.text.strip())
                     celltag = "td"
-                    if cell.column_header:
+                    if cell.column_header or cell.row_header or cell.row_section:
                         celltag = "th"
 
                     opening_tag = f"{celltag}"

{docling_core-2.35.0 → docling_core-2.37.0}/docling_core/transforms/visualizer/layout_visualizer.py

@@ -40,6 +40,7 @@ class LayoutVisualizer(BaseVisualizer):
         """Layout visualization parameters."""
 
         show_label: bool = True
+        content_layers: set[ContentLayer] = {cl for cl in ContentLayer}
 
     base_visualizer: Optional[BaseVisualizer] = None
     params: Params = Params()

{docling_core-2.35.0 → docling_core-2.37.0}/docling_core/transforms/visualizer/reading_order_visualizer.py

@@ -5,6 +5,7 @@ from typing import Optional
 
 from PIL import ImageDraw
 from PIL.Image import Image
+from pydantic import BaseModel
 from typing_extensions import override
 
 from docling_core.transforms.visualizer.base import BaseVisualizer
@@ -14,7 +15,16 @@ from docling_core.types.doc.document import ContentLayer, DocItem, DoclingDocume
 class ReadingOrderVisualizer(BaseVisualizer):
     """Reading order visualizer."""
 
+    class Params(BaseModel):
+        """Layout visualization parameters."""
+
+        show_label: bool = True
+        content_layers: set[ContentLayer] = {
+            cl for cl in ContentLayer if cl != ContentLayer.BACKGROUND
+        }
+
     base_visualizer: Optional[BaseVisualizer] = None
+    params: Params = Params()
 
     def _draw_arrow(
         self,
@@ -71,7 +81,7 @@ class ReadingOrderVisualizer(BaseVisualizer):
         my_images: dict[Optional[int], Image] = images or {}
         prev_page = None
         for elem, _ in doc.iterate_items(
-            included_content_layers={ContentLayer.BODY, ContentLayer.FURNITURE},
+            included_content_layers=self.params.content_layers,
         ):
             if not isinstance(elem, DocItem):
                 continue

{docling_core-2.35.0 → docling_core-2.37.0}/docling_core/transforms/visualizer/table_visualizer.py

@@ -23,8 +23,23 @@ class TableVisualizer(BaseVisualizer):
 
         # show_Label: bool = False
         show_cells: bool = True
-        # show_rows: bool = False
-        # show_cols: bool = False
+        show_rows: bool = False
+        show_cols: bool = False
+
+        cell_color: tuple[int, int, int, int] = (256, 0, 0, 32)
+        cell_outline: tuple[int, int, int, int] = (256, 0, 0, 128)
+
+        row_color: tuple[int, int, int, int] = (256, 0, 0, 32)
+        row_outline: tuple[int, int, int, int] = (256, 0, 0, 128)
+
+        row_header_color: tuple[int, int, int, int] = (0, 256, 0, 32)
+        row_header_outline: tuple[int, int, int, int] = (0, 256, 0, 128)
+
+        col_color: tuple[int, int, int, int] = (0, 256, 0, 32)
+        col_outline: tuple[int, int, int, int] = (0, 256, 0, 128)
+
+        col_header_color: tuple[int, int, int, int] = (0, 0, 256, 32)
+        col_header_outline: tuple[int, int, int, int] = (0, 0, 256, 128)
 
     base_visualizer: Optional[BaseVisualizer] = None
     params: Params = Params()
@@ -45,7 +60,21 @@ class TableVisualizer(BaseVisualizer):
 
                 tl_bbox = cell.bbox.to_top_left_origin(page_height=page_height)
 
-                cell_color = (256, 0, 0, 32)  # Transparent black for cells
+                cell_color = self.params.cell_color  # Transparent black for cells
+                cell_outline = self.params.cell_outline
+                if cell.column_header:
+                    cell_color = (
+                        self.params.col_header_color
+                    )  # Transparent black for cells
+                    cell_outline = self.params.col_header_outline
+                if cell.row_header:
+                    cell_color = (
+                        self.params.row_header_color
+                    )  # Transparent black for cells
+                    cell_outline = self.params.row_header_outline
+                if cell.row_section:
+                    cell_color = self.params.row_header_color
+                    cell_outline = self.params.row_header_outline
 
                 cx0, cy0, cx1, cy1 = tl_bbox.as_tuple()
                 cx0 *= scale_x
@@ -55,10 +84,68 @@ class TableVisualizer(BaseVisualizer):
 
                 draw.rectangle(
                     [(cx0, cy0), (cx1, cy1)],
-                    outline=(256, 0, 0, 128),
+                    outline=cell_outline,
                     fill=cell_color,
                 )
 
+    def _draw_table_rows(
+        self,
+        table: TableItem,
+        page_image: Image,
+        page_height: float,
+        scale_x: float,
+        scale_y: float,
+    ):
+        """Draw individual table cells."""
+        draw = ImageDraw.Draw(page_image, "RGBA")
+
+        rows = table.data.get_row_bounding_boxes()
+
+        for rid, bbox in rows.items():
+
+            tl_bbox = bbox.to_top_left_origin(page_height=page_height)
+
+            cx0, cy0, cx1, cy1 = tl_bbox.as_tuple()
+            cx0 *= scale_x
+            cx1 *= scale_x
+            cy0 *= scale_y
+            cy1 *= scale_y
+
+            draw.rectangle(
+                [(cx0, cy0), (cx1, cy1)],
+                outline=self.params.row_outline,
+                fill=self.params.row_color,
+            )
+
+    def _draw_table_cols(
+        self,
+        table: TableItem,
+        page_image: Image,
+        page_height: float,
+        scale_x: float,
+        scale_y: float,
+    ):
+        """Draw individual table cells."""
+        draw = ImageDraw.Draw(page_image, "RGBA")
+
+        cols = table.data.get_column_bounding_boxes()
+
+        for cid, bbox in cols.items():
+
+            tl_bbox = bbox.to_top_left_origin(page_height=page_height)
+
+            cx0, cy0, cx1, cy1 = tl_bbox.as_tuple()
+            cx0 *= scale_x
+            cx1 *= scale_x
+            cy0 *= scale_y
+            cy1 *= scale_y
+
+            draw.rectangle(
+                [(cx0, cy0), (cx1, cy1)],
+                outline=self.params.col_outline,
+                fill=self.params.col_color,
+            )
+
     def _draw_doc_tables(
         self,
         doc: DoclingDocument,
@@ -108,6 +195,24 @@ class TableVisualizer(BaseVisualizer):
                             scale_y=image.height / doc.pages[page_nr].size.height,
                         )
 
+                    if self.params.show_rows:
+                        self._draw_table_rows(
+                            table=elem,
+                            page_height=doc.pages[page_nr].size.height,
+                            page_image=image,
+                            scale_x=image.width / doc.pages[page_nr].size.width,
+                            scale_y=image.height / doc.pages[page_nr].size.height,
+                        )
+
+                    if self.params.show_cols:
+                        self._draw_table_cols(
+                            table=elem,
+                            page_height=doc.pages[page_nr].size.height,
+                            page_image=image,
+                            scale_x=image.width / doc.pages[page_nr].size.width,
+                            scale_y=image.height / doc.pages[page_nr].size.height,
+                        )
+
                 else:
                     raise RuntimeError(f"Cannot visualize page-image for {page_nr}")
 

{docling_core-2.35.0 → docling_core-2.37.0}/docling_core/types/doc/document.py

@@ -38,7 +38,7 @@ from typing_extensions import Annotated, Self, deprecated
 from docling_core.search.package import VERSION_PATTERN
 from docling_core.types.base import _JSON_POINTER_REGEX
 from docling_core.types.doc import BoundingBox, Size
-from docling_core.types.doc.base import ImageRefMode
+from docling_core.types.doc.base import CoordOrigin, ImageRefMode
 from docling_core.types.doc.labels import (
     CodeLanguageLabel,
     DocItemLabel,
@@ -372,6 +372,119 @@ class TableData(BaseModel):  # TBD
 
         return table_data
 
+    def get_row_bounding_boxes(self) -> dict[int, BoundingBox]:
+        """Get the minimal bounding box for each row in the table.
+
+        Returns:
+        List[Optional[BoundingBox]]: A list where each element is the minimal
+        bounding box that encompasses all cells in that row, or None if no
+        cells in the row have bounding boxes.
+        """
+        coords = []
+        for cell in self.table_cells:
+            if cell.bbox is not None:
+                coords.append(cell.bbox.coord_origin)
+
+        if len(set(coords)) > 1:
+            raise ValueError(
+                "All bounding boxes must have the same \
+                CoordOrigin to compute their union."
+            )
+
+        row_bboxes: dict[int, BoundingBox] = {}
+
+        for row_idx in range(self.num_rows):
+            row_cells_with_bbox: dict[int, list[BoundingBox]] = {}
+
+            # Collect all cells in this row that have bounding boxes
+            for cell in self.table_cells:
+
+                if (
+                    cell.bbox is not None
+                    and cell.start_row_offset_idx <= row_idx < cell.end_row_offset_idx
+                ):
+
+                    row_span = cell.end_row_offset_idx - cell.start_row_offset_idx
+                    if row_span in row_cells_with_bbox:
+                        row_cells_with_bbox[row_span].append(cell.bbox)
+                    else:
+                        row_cells_with_bbox[row_span] = [cell.bbox]
+
+            # Calculate the enclosing bounding box for this row
+            if len(row_cells_with_bbox) > 0:
+                min_row_span = min(row_cells_with_bbox.keys())
+                row_bbox: BoundingBox = BoundingBox.enclosing_bbox(
+                    row_cells_with_bbox[min_row_span]
+                )
+
+                for rspan, bboxs in row_cells_with_bbox.items():
+                    for bbox in bboxs:
+                        row_bbox.l = min(row_bbox.l, bbox.l)
+                        row_bbox.r = max(row_bbox.r, bbox.r)
+
+                row_bboxes[row_idx] = row_bbox
+
+        return row_bboxes
+
+    def get_column_bounding_boxes(self) -> dict[int, BoundingBox]:
+        """Get the minimal bounding box for each column in the table.
+
+        Returns:
+            List[Optional[BoundingBox]]: A list where each element is the minimal
+            bounding box that encompasses all cells in that column, or None if no
+            cells in the column have bounding boxes.
+        """
+        coords = []
+        for cell in self.table_cells:
+            if cell.bbox is not None:
+                coords.append(cell.bbox.coord_origin)
+
+        if len(set(coords)) > 1:
+            raise ValueError(
+                "All bounding boxes must have the same \
+                CoordOrigin to compute their union."
+            )
+
+        col_bboxes: dict[int, BoundingBox] = {}
+
+        for col_idx in range(self.num_cols):
+            col_cells_with_bbox: dict[int, list[BoundingBox]] = {}
+
+            # Collect all cells in this row that have bounding boxes
+            for cell in self.table_cells:
+
+                if (
+                    cell.bbox is not None
+                    and cell.start_col_offset_idx <= col_idx < cell.end_col_offset_idx
+                ):
+
+                    col_span = cell.end_col_offset_idx - cell.start_col_offset_idx
+                    if col_span in col_cells_with_bbox:
+                        col_cells_with_bbox[col_span].append(cell.bbox)
+                    else:
+                        col_cells_with_bbox[col_span] = [cell.bbox]
+
+            # Calculate the enclosing bounding box for this row
+            if len(col_cells_with_bbox) > 0:
+                min_col_span = min(col_cells_with_bbox.keys())
+                col_bbox: BoundingBox = BoundingBox.enclosing_bbox(
+                    col_cells_with_bbox[min_col_span]
+                )
+
+                for rspan, bboxs in col_cells_with_bbox.items():
+                    for bbox in bboxs:
+                        if bbox.coord_origin == CoordOrigin.TOPLEFT:
+                            col_bbox.b = max(col_bbox.b, bbox.b)
+                            col_bbox.t = min(col_bbox.t, bbox.t)
+
+                        elif bbox.coord_origin == CoordOrigin.BOTTOMLEFT:
+                            col_bbox.b = min(col_bbox.b, bbox.b)
+                            col_bbox.t = max(col_bbox.t, bbox.t)
+
+                col_bboxes[col_idx] = col_bbox
+
+        return col_bboxes
+
 
 class PictureTabularChartData(PictureChartData):
     """Base class for picture chart data.
@@ -623,6 +736,7 @@ class ContentLayer(str, Enum):
 
     BODY = "body"
     FURNITURE = "furniture"
+    BACKGROUND = "background"
 
 
 DEFAULT_CONTENT_LAYERS = {ContentLayer.BODY}
@@ -860,6 +974,7 @@ class TextItem(DocItem):
         DocItemLabel.PARAGRAPH,
         DocItemLabel.REFERENCE,
         DocItemLabel.TEXT,
+        DocItemLabel.EMPTY_VALUE,
     ]
 
     orig: str  # untreated representation
@@ -2867,23 +2982,46 @@ class DoclingDocument(BaseModel):
 
     def print_element_tree(self):
         """Print_element_tree."""
-        for ix, (item, level) in enumerate(self.iterate_items(with_groups=True)):
+        for ix, (item, level) in enumerate(
+            self.iterate_items(
+                with_groups=True,
+                traverse_pictures=True,
+                included_content_layers={cl for cl in ContentLayer},
+            )
+        ):
             if isinstance(item, GroupItem):
                 print(
                     " " * level,
                     f"{ix}: {item.label.value} with name={item.name}",
                 )
+            elif isinstance(item, TextItem):
+                print(
+                    " " * level,
+                    f"{ix}: {item.label.value}: {item.text[:min(len(item.text), 100)]}",
+                )
+
             elif isinstance(item, DocItem):
                 print(" " * level, f"{ix}: {item.label.value}")
 
     def export_to_element_tree(self) -> str:
         """Export_to_element_tree."""
         texts = []
-        for ix, (item, level) in enumerate(self.iterate_items(with_groups=True)):
+        for ix, (item, level) in enumerate(
+            self.iterate_items(
+                with_groups=True,
+                traverse_pictures=True,
+                included_content_layers={cl for cl in ContentLayer},
+            )
+        ):
             if isinstance(item, GroupItem):
                 texts.append(
                     " " * level + f"{ix}: {item.label.value} with name={item.name}"
                 )
+            elif isinstance(item, TextItem):
+                texts.append(
+                    " " * level
+                    + f"{ix}: {item.label.value}: {item.text[:min(len(item.text), 100)]}"
+                )
             elif isinstance(item, DocItem):
                 texts.append(" " * level + f"{ix}: {item.label.value}")
 

{docling_core-2.35.0 → docling_core-2.37.0}/docling_core/types/doc/labels.py

@@ -27,6 +27,9 @@ class DocItemLabel(str, Enum):
     KEY_VALUE_REGION = "key_value_region"
     GRADING_SCALE = "grading_scale"  # for elements in forms, questionaires representing a grading scale
     # e.g. [strongly disagree | ... | ... | strongly agree]
+    # e.g. ★★☆☆☆
+    HANDWRITTEN_TEXT = "handwritten_text"
+    EMPTY_VALUE = "empty_value"  # used for empty value fields in fillable forms
 
     # Additional labels for markup-based formats (e.g. HTML, Word)
     PARAGRAPH = "paragraph"
@@ -60,6 +63,9 @@ class DocItemLabel(str, Enum):
             DocItemLabel.KEY_VALUE_REGION: (183, 65, 14),
             DocItemLabel.PARAGRAPH: (255, 255, 153),
             DocItemLabel.REFERENCE: (176, 224, 230),
+            DocItemLabel.GRADING_SCALE: (255, 204, 204),
+            DocItemLabel.HANDWRITTEN_TEXT: (204, 255, 204),
+            DocItemLabel.EMPTY_VALUE: (220, 220, 220),
         }
         return color_map.get(label, (0, 0, 0))
 
@@ -166,7 +172,6 @@ class GraphCellLabel(str, Enum):
     KEY = "key"  # used to designate a key (label) of a key-value element
     VALUE = "value"  # Data value with or without explicit Key, but filled in,
     # e.g. telephone number, address, quantity, name, date
-    EMPTY_VALUE = "empty_value"  # used for empty value fields in fillable forms
     CHECKBOX = "checkbox"
 
     def __str__(self):
@@ -179,7 +184,6 @@ class GraphCellLabel(str, Enum):
         color_map = {
             GraphCellLabel.KEY: (255, 0, 0),
             GraphCellLabel.VALUE: (0, 255, 0),
-            GraphCellLabel.EMPTY_VALUE: (0, 0, 255),
         }
         return color_map.get(label, (0, 0, 0))
 

{docling_core-2.35.0 → docling_core-2.37.0}/docling_core.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: docling-core
-Version: 2.35.0
+Version: 2.37.0
 Summary: A python library to define and validate data types in Docling.
 Author-email: Cesar Berrospi Ramis <ceb@zurich.ibm.com>, Panos Vagenas <pva@zurich.ibm.com>, Michele Dolfi <dol@zurich.ibm.com>, Christoph Auer <cau@zurich.ibm.com>, Peter Staar <taa@zurich.ibm.com>
 Maintainer-email: Panos Vagenas <pva@zurich.ibm.com>, Michele Dolfi <dol@zurich.ibm.com>, Christoph Auer <cau@zurich.ibm.com>, Peter Staar <taa@zurich.ibm.com>, Cesar Berrospi Ramis <ceb@zurich.ibm.com>

{docling_core-2.35.0 → docling_core-2.37.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "docling-core"
-version = "2.35.0"  # DO NOT EDIT, updated automatically
+version = "2.37.0"  # DO NOT EDIT, updated automatically
 description = "A python library to define and validate data types in Docling."
 license = "MIT"
 license-files = ["LICENSE"]

{docling_core-2.35.0 → docling_core-2.37.0}/test/test_visualization.py

@@ -55,7 +55,7 @@ def test_doc_visualization_no_label():
             )
 
 
-def test_table_visualization_no_label():
+def test_table_visualization_for_cells():
     src = Path("./test/data/doc/2408.09869v3_enriched.json")
     doc = DoclingDocument.load_from_json(src)
 
@@ -66,3 +66,19 @@ def test_table_visualization_no_label():
         exp_file=VIZ_TEST_DATA_PATH / f"{src.stem}_table_viz_wout_lbl_p5.png",
         actual=viz_pages[5],
     )
+
+
+def test_table_visualization_for_rows_and_cols():
+    src = Path("./test/data/doc/2408.09869v3_enriched.json")
+    doc = DoclingDocument.load_from_json(src)
+
+    visualizer = TableVisualizer(
+        params=TableVisualizer.Params(show_cells=False, show_rows=True, show_cols=True)
+    )
+    viz_pages = visualizer.get_visualization(doc=doc)
+
+    verify(
+        exp_file=VIZ_TEST_DATA_PATH
+        / f"{src.stem}_table_viz_wout_lbl_p5_rows_and_cols.png",
+        actual=viz_pages[5],
+    )