docling 2.4.2__py3-none-any.whl → 2.5.1__py3-none-any.whl

docling/backend/html_backend.py

@@ -120,6 +120,8 @@ class HTMLDocumentBackend(DeclarativeDocumentBackend):
             self.handle_header(element, idx, doc)
         elif element.name in ["p"]:
             self.handle_paragraph(element, idx, doc)
+        elif element.name in ["pre"]:
+            self.handle_code(element, idx, doc)
         elif element.name in ["ul", "ol"]:
             self.handle_list(element, idx, doc)
         elif element.name in ["li"]:
@@ -205,6 +207,16 @@ class HTMLDocumentBackend(DeclarativeDocumentBackend):
                 level=hlevel,
             )
 
+    def handle_code(self, element, idx, doc):
+        """Handles monospace code snippets (pre)."""
+        if element.text is None:
+            return
+        text = element.text.strip()
+        label = DocItemLabel.CODE
+        if len(text) == 0:
+            return
+        doc.add_text(parent=self.parents[self.level], label=label, text=text)
+
     def handle_paragraph(self, element, idx, doc):
         """Handles paragraph tags (p)."""
         if element.text is None:

docling/backend/mspowerpoint_backend.py

@@ -358,41 +358,36 @@ class MsPowerpointDocumentBackend(DeclarativeDocumentBackend, PaginatedDocumentB
 
             size = Size(width=slide_width, height=slide_height)
             parent_page = doc.add_page(page_no=slide_ind + 1, size=size)
-            # parent_page = doc.add_page(page_no=slide_ind, size=size, hash=hash)
-
-            # Loop through each shape in the slide
-            for shape in slide.shapes:
 
+            def handle_shapes(shape, parent_slide, slide_ind, doc):
+                handle_groups(shape, parent_slide, slide_ind, doc)
                 if shape.has_table:
                     # Handle Tables
                     self.handle_tables(shape, parent_slide, slide_ind, doc)
-
                 if shape.shape_type == MSO_SHAPE_TYPE.PICTURE:
-                    # Handle Tables
+                    # Handle Pictures
                     self.handle_pictures(shape, parent_slide, slide_ind, doc)
-
                 # If shape doesn't have any text, move on to the next shape
                 if not hasattr(shape, "text"):
-                    continue
+                    return
                 if shape.text is None:
-                    continue
+                    return
                 if len(shape.text.strip()) == 0:
-                    continue
+                    return
                 if not shape.has_text_frame:
-                    _log.warn("Warning: shape has text but not text_frame")
-                    continue
-
-                # if shape.is_placeholder:
-                # Handle Titles (Headers) and Subtitles
-                # Check if the shape is a placeholder (titles are placeholders)
-                # self.handle_title(shape, parent_slide, slide_ind, doc)
-                # self.handle_text_elements(shape, parent_slide, slide_ind, doc)
-                # else:
-
+                    _log.warning("Warning: shape has text but not text_frame")
+                    return
                 # Handle other text elements, including lists (bullet lists, numbered lists)
                 self.handle_text_elements(shape, parent_slide, slide_ind, doc)
+                return
+
+            def handle_groups(shape, parent_slide, slide_ind, doc):
+                if shape.shape_type == MSO_SHAPE_TYPE.GROUP:
+                    for groupedshape in shape.shapes:
+                        handle_shapes(groupedshape, parent_slide, slide_ind, doc)
 
-                # figures...
-                # doc.add_figure(data=BaseFigureData(), parent=self.parents[self.level], caption=None)
+            # Loop through each shape in the slide
+            for shape in slide.shapes:
+                handle_shapes(shape, parent_slide, slide_ind, doc)
 
         return doc

docling/backend/msword_backend.py

@@ -130,7 +130,6 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
     def walk_linear(self, body, docx_obj, doc) -> DoclingDocument:
         for element in body:
             tag_name = etree.QName(element).localname
-
             # Check for Inline Images (drawings or blip elements)
             found_drawing = etree.ElementBase.xpath(
                 element, ".//w:drawing", namespaces=self.xml_namespaces
@@ -201,7 +200,6 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
             label_str = ""
             label_level = 0
             if parts[0] == "Heading":
-                # print("{} - {}".format(parts[0], parts[1]))
                 label_str = parts[0]
                 label_level = self.str_to_int(parts[1], default=None)
             if parts[1] == "Heading":
@@ -217,19 +215,16 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
         if paragraph.text is None:
             # _log.warn(f"paragraph has text==None")
             return
-
         text = paragraph.text.strip()
         # if len(text)==0 # keep empty paragraphs, they seperate adjacent lists!
 
         # Common styles for bullet and numbered lists.
         # "List Bullet", "List Number", "List Paragraph"
-        # TODO: reliably identify wether list is a numbered list or not
+        # Identify wether list is a numbered list or not
         # is_numbered = "List Bullet" not in paragraph.style.name
         is_numbered = False
-
         p_style_name, p_level = self.get_label_and_level(paragraph)
         numid, ilevel = self.get_numId_and_ilvl(paragraph)
-        # print("numid: {}, ilevel: {}, text: {}".format(numid, ilevel, text))
 
         if numid == 0:
             numid = None
@@ -450,8 +445,13 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
         for row in table.rows:
             # Calculate the max number of columns
             num_cols = max(num_cols, sum(get_colspan(cell) for cell in row.cells))
-            # if row.cells:
-            #     num_cols = max(num_cols, len(row.cells))
+
+        if num_rows == 1 and num_cols == 1:
+            cell_element = table.rows[0].cells[0]
+            # In case we have a table of only 1 cell, we consider it furniture
+            # And proceed processing the content of the cell as though it's in the document body
+            self.walk_linear(cell_element._element, docx_obj, doc)
+            return
 
         # Initialize the table grid
         table_grid = [[None for _ in range(num_cols)] for _ in range(num_rows)]

docling/cli/main.py

@@ -153,6 +153,13 @@ def convert(
             ..., help="If enabled, the bitmap content will be processed using OCR."
         ),
     ] = True,
+    force_ocr: Annotated[
+        bool,
+        typer.Option(
+            ...,
+            help="Replace any existing text with OCR generated text over the full content.",
+        ),
+    ] = False,
     ocr_engine: Annotated[
         OcrEngine, typer.Option(..., help="The OCR engine to use.")
     ] = OcrEngine.EASYOCR,
@@ -219,11 +226,11 @@ def convert(
 
     match ocr_engine:
         case OcrEngine.EASYOCR:
-            ocr_options: OcrOptions = EasyOcrOptions()
+            ocr_options: OcrOptions = EasyOcrOptions(force_full_page_ocr=force_ocr)
         case OcrEngine.TESSERACT_CLI:
-            ocr_options = TesseractCliOcrOptions()
+            ocr_options = TesseractCliOcrOptions(force_full_page_ocr=force_ocr)
         case OcrEngine.TESSERACT:
-            ocr_options = TesseractOcrOptions()
+            ocr_options = TesseractOcrOptions(force_full_page_ocr=force_ocr)
         case _:
             raise RuntimeError(f"Unexpected OCR engine type {ocr_engine}")
 

docling/datamodel/pipeline_options.py

@@ -22,6 +22,7 @@ class TableStructureOptions(BaseModel):
 
 class OcrOptions(BaseModel):
     kind: str
+    force_full_page_ocr: bool = False  # If enabled a full page OCR is always applied
     bitmap_area_threshold: float = (
         0.05  # percentage of the area for a bitmap to processed with OCR
     )

docling/datamodel/settings.py

@@ -2,7 +2,7 @@ import sys
 from pathlib import Path
 
 from pydantic import BaseModel
-from pydantic_settings import BaseSettings
+from pydantic_settings import BaseSettings, SettingsConfigDict
 
 
 class DocumentLimits(BaseModel):
@@ -40,6 +40,8 @@ class DebugSettings(BaseModel):
 
 
 class AppSettings(BaseSettings):
+    model_config = SettingsConfigDict(env_prefix="DOCLING_", env_nested_delimiter="_")
+
     perf: BatchConcurrencySettings
     debug: DebugSettings
 

docling/models/base_ocr_model.py

@@ -10,7 +10,7 @@ from PIL import Image, ImageDraw
 from rtree import index
 from scipy.ndimage import find_objects, label
 
-from docling.datamodel.base_models import OcrCell, Page
+from docling.datamodel.base_models import Cell, OcrCell, Page
 from docling.datamodel.document import ConversionResult
 from docling.datamodel.pipeline_options import OcrOptions
 from docling.datamodel.settings import settings
@@ -73,7 +73,9 @@ class BaseOcrModel(BasePageModel):
         coverage, ocr_rects = find_ocr_rects(page.size, bitmap_rects)
 
         # return full-page rectangle if sufficiently covered with bitmaps
-        if coverage > max(BITMAP_COVERAGE_TRESHOLD, self.options.bitmap_area_threshold):
+        if self.options.force_full_page_ocr or coverage > max(
+            BITMAP_COVERAGE_TRESHOLD, self.options.bitmap_area_threshold
+        ):
             return [
                 BoundingBox(
                     l=0,
@@ -96,7 +98,7 @@ class BaseOcrModel(BasePageModel):
             return ocr_rects
 
     # Filters OCR cells by dropping any OCR cell that intersects with an existing programmatic cell.
-    def filter_ocr_cells(self, ocr_cells, programmatic_cells):
+    def _filter_ocr_cells(self, ocr_cells, programmatic_cells):
         # Create R-tree index for programmatic cells
         p = index.Property()
         p.dimension = 2
@@ -117,6 +119,23 @@ class BaseOcrModel(BasePageModel):
         ]
         return filtered_ocr_cells
 
+    def post_process_cells(self, ocr_cells, programmatic_cells):
+        r"""
+        Post-process the ocr and programmatic cells and return the final list of of cells
+        """
+        if self.options.force_full_page_ocr:
+            # If a full page OCR is forced, use only the OCR cells
+            cells = [
+                Cell(id=c_ocr.id, text=c_ocr.text, bbox=c_ocr.bbox)
+                for c_ocr in ocr_cells
+            ]
+            return cells
+
+        ## Remove OCR cells which overlap with programmatic cells.
+        filtered_ocr_cells = self._filter_ocr_cells(ocr_cells, programmatic_cells)
+        programmatic_cells.extend(filtered_ocr_cells)
+        return programmatic_cells
+
     def draw_ocr_rects_and_cells(self, conv_res, page, ocr_rects, show: bool = False):
         image = copy.deepcopy(page.image)
         draw = ImageDraw.Draw(image, "RGBA")

docling/models/easyocr_model.py

@@ -5,7 +5,7 @@ import numpy
 import torch
 from docling_core.types.doc import BoundingBox, CoordOrigin
 
-from docling.datamodel.base_models import OcrCell, Page
+from docling.datamodel.base_models import Cell, OcrCell, Page
 from docling.datamodel.document import ConversionResult
 from docling.datamodel.pipeline_options import EasyOcrOptions
 from docling.datamodel.settings import settings
@@ -31,12 +31,9 @@ class EasyOcrModel(BaseOcrModel):
                     "Alternatively, Docling has support for other OCR engines. See the documentation."
                 )
 
-            use_gpu = (
-                False if torch.backends.mps.is_available() else self.options.use_gpu
-            )
             self.reader = easyocr.Reader(
                 lang_list=self.options.lang,
-                gpu=use_gpu,
+                gpu=self.options.use_gpu,
                 model_storage_directory=self.options.model_storage_directory,
                 download_enabled=self.options.download_enabled,
             )
@@ -91,12 +88,8 @@ class EasyOcrModel(BaseOcrModel):
                         ]
                         all_ocr_cells.extend(cells)
 
-                    ## Remove OCR cells which overlap with programmatic cells.
-                    filtered_ocr_cells = self.filter_ocr_cells(
-                        all_ocr_cells, page.cells
-                    )
-
-                    page.cells.extend(filtered_ocr_cells)
+                    # Post-process the cells
+                    page.cells = self.post_process_cells(all_ocr_cells, page.cells)
 
                 # DEBUG code:
                 if settings.debug.visualize_ocr:

docling/models/tesseract_ocr_cli_model.py

@@ -7,7 +7,7 @@ from typing import Iterable, Optional, Tuple
 import pandas as pd
 from docling_core.types.doc import BoundingBox, CoordOrigin
 
-from docling.datamodel.base_models import OcrCell, Page
+from docling.datamodel.base_models import Cell, OcrCell, Page
 from docling.datamodel.document import ConversionResult
 from docling.datamodel.pipeline_options import TesseractCliOcrOptions
 from docling.datamodel.settings import settings
@@ -170,12 +170,8 @@ class TesseractOcrCliModel(BaseOcrModel):
                             )
                             all_ocr_cells.append(cell)
 
-                    ## Remove OCR cells which overlap with programmatic cells.
-                    filtered_ocr_cells = self.filter_ocr_cells(
-                        all_ocr_cells, page.cells
-                    )
-
-                    page.cells.extend(filtered_ocr_cells)
+                    # Post-process the cells
+                    page.cells = self.post_process_cells(all_ocr_cells, page.cells)
 
                 # DEBUG code:
                 if settings.debug.visualize_ocr:

docling/models/tesseract_ocr_model.py

@@ -3,7 +3,7 @@ from typing import Iterable
 
 from docling_core.types.doc import BoundingBox, CoordOrigin
 
-from docling.datamodel.base_models import OcrCell, Page
+from docling.datamodel.base_models import Cell, OcrCell, Page
 from docling.datamodel.document import ConversionResult
 from docling.datamodel.pipeline_options import TesseractOcrOptions
 from docling.datamodel.settings import settings
@@ -140,12 +140,8 @@ class TesseractOcrModel(BaseOcrModel):
                         # del high_res_image
                         all_ocr_cells.extend(cells)
 
-                    ## Remove OCR cells which overlap with programmatic cells.
-                    filtered_ocr_cells = self.filter_ocr_cells(
-                        all_ocr_cells, page.cells
-                    )
-
-                    page.cells.extend(filtered_ocr_cells)
+                    # Post-process the cells
+                    page.cells = self.post_process_cells(all_ocr_cells, page.cells)
 
                 # DEBUG code:
                 if settings.debug.visualize_ocr:

{docling-2.4.2.dist-info → docling-2.5.1.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: docling
-Version: 2.4.2
+Version: 2.5.1
 Summary: SDK and CLI for parsing PDF, DOCX, HTML, and more, to a unified document representation for powering downstream workflows such as gen AI applications.
 Home-page: https://github.com/DS4SD/docling
 License: MIT

{docling-2.4.2.dist-info → docling-2.5.1.dist-info}/RECORD

@@ -4,31 +4,31 @@ docling/backend/abstract_backend.py,sha256=-or6kWVV7egQeyIuN-vI0Tr7Q1htalBZSlhgq
 docling/backend/asciidoc_backend.py,sha256=kXZxOLk_LvLFVZwnJVVwjmvc3QWZ0iiG7VnwjgtC3hI,14051
 docling/backend/docling_parse_backend.py,sha256=csWy6ZGxDuZfNr0YTrUU40DXqelN_TJksWIYoXxZMjU,7633
 docling/backend/docling_parse_v2_backend.py,sha256=gUr9_fwHbkj238oYQPJ9AxpjFL2jGvhjBlBQPblmSAg,8589
-docling/backend/html_backend.py,sha256=p3WlYta1f3e4osmvVR12KIUYLJimveTX8UwEkyPt7_g,15161
+docling/backend/html_backend.py,sha256=qbu1W8xoTGnXMuZPRPLq68hDbCEj6ygnpxP5gYaodAQ,15593
 docling/backend/md_backend.py,sha256=tmuSCghjor9PqKIiVieCuZ4_t5JEjZMy3cq7u3yTgyU,14032
-docling/backend/mspowerpoint_backend.py,sha256=J472AIH_IXvGg3D0FDmXhue1At_VSBD6n15c64Kxttw,15446
-docling/backend/msword_backend.py,sha256=FAUdP74QxGKo2xMZQ4WQGYwtpIBCTJ_FG17PBpRwhxI,17230
+docling/backend/mspowerpoint_backend.py,sha256=YaVJc6RXWmM1EPTp0TzAiXpGxu6K-MZdPNsmR_64LSg,15358
+docling/backend/msword_backend.py,sha256=IEqGz-lUrQw0tgBly_gv_mYGC0X0iNnGhkwnDWaDtBY,17341
 docling/backend/pdf_backend.py,sha256=unnw7QiRE1VXg6Pj-eYrtnFGrp5SSYiI324OlFxyv6c,2050
 docling/backend/pypdfium2_backend.py,sha256=B4bfv-dfzlWiKTfF8LN5fto_99YBu8A2c1_XIVwRUWI,8996
 docling/cli/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-docling/cli/main.py,sha256=IOeIpGoK_5AeE_6LYTU_nfZjqpZ5xeGaTCB8Vfsama0,9334
+docling/cli/main.py,sha256=7stF4dMjGVp5R0Gvcawm21rff5RbEQnWj8ZzoAHvV9k,9619
 docling/datamodel/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 docling/datamodel/base_models.py,sha256=fmkS6iTxGZCTtNCo2zsgMmBC11Ogf2Ht-mNIlZ9GP-o,5375
 docling/datamodel/document.py,sha256=9dQf_J18X_MEWs-Mg3Ed6BykFPJ79ETmkkxcssY-vYo,20698
-docling/datamodel/pipeline_options.py,sha256=PqQ4VjMDN16oWZSUYtskQEH366504OZmnjinCaOWmMc,2444
-docling/datamodel/settings.py,sha256=2-sYEnKLV_giGygUlBtiBd4CJYN5T9-3BdL6NpWkUYw,1155
+docling/datamodel/pipeline_options.py,sha256=-PXwqkdwSpWjIMCxyqwB8Q453szVNR1zVM-7d0PAOWQ,2530
+docling/datamodel/settings.py,sha256=JK8lZPBjUx2kD2q-Qpg-o3vOElADMcyQbRUL0EHZ7us,1263
 docling/document_converter.py,sha256=U52_rZQDm2wzrnsuUrvsfX2MnmOWFFhjBzfS8tEvt6Y,10595
 docling/models/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 docling/models/base_model.py,sha256=Yq_-FmUhqhE20vXYG3WiQXDRTIPjik1CyuEZ8iYTGAY,701
-docling/models/base_ocr_model.py,sha256=Ti0glL-_DVRfmP3MpywYVmkNf5RP6qhRg_UKzJuV1Dc,5663
+docling/models/base_ocr_model.py,sha256=rGSpBF4dByITcsBaRIgvFKpiu0CrhmZS_PHIo686Dw0,6428
 docling/models/ds_glm_model.py,sha256=2OpWW8MMzCIshrtP36gDSRPYOCjv1ex34FqxD2nYjP4,11986
-docling/models/easyocr_model.py,sha256=zl-B3OR81r0AHt5zFP_A5TqJUSjIjo1-x4GsMZyIzVk,3983
+docling/models/easyocr_model.py,sha256=c2m4x9dZpSc-cMgeEdFBRVBlB78uMGlYD8Q_2gzRuMU,3734
 docling/models/layout_model.py,sha256=ZvbTSyxvXB5yLHNEti0Wv3trz0vwGuHySI5TCdApb0U,14011
 docling/models/page_assemble_model.py,sha256=kSGNiRKhmzkpFH7xCiT3rulMsgJmUXFa6Th_eB-cLEk,7103
 docling/models/page_preprocessing_model.py,sha256=1gVrZjObKxAvXkkKvXnIFApPOggzgiTFPtt1CGbMbSs,2763
 docling/models/table_structure_model.py,sha256=-ANSQpiN2avt3B9sbi7dHcoULUJbMBalAR5xxlrM7To,8421
-docling/models/tesseract_ocr_cli_model.py,sha256=ZflwQcD7YjhPqEB8bbgNgP14OBD4NNEJefUS8Lbr5X0,6511
-docling/models/tesseract_ocr_model.py,sha256=X9qlzwaTZLtSGXFIZuD7MO6EzFmHl1D-FjktUBko6us,6234
+docling/models/tesseract_ocr_cli_model.py,sha256=OfopQnt2FGwtLJTMtW9jbJZ9EN2G2QFkA_aACjuUuDs,6372
+docling/models/tesseract_ocr_model.py,sha256=RDf6iV1q-oXaGfZXv0bW6SqjHNKQvBUDlUsOkuz0neY,6095
 docling/pipeline/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 docling/pipeline/base_pipeline.py,sha256=IF1XWYgUGbdB4-teLkmM4Hvg_UNEfPrGuhExMRTUsk8,7168
 docling/pipeline/simple_pipeline.py,sha256=mZqANqUtAOFAyqQEShErQnAUz6tJFOl6zVoazEDJ_wE,2254
@@ -38,8 +38,8 @@ docling/utils/export.py,sha256=KyGF1BVDHPFfHVXZc8vegsWlFfOgGPP2YckWpTadyI8,4694
 docling/utils/layout_utils.py,sha256=vlN0rc8i0ayRGn3WnaG-pdmqEL00KKGl2zez3Gj-hrk,32074
 docling/utils/profiling.py,sha256=YaMGoB9MMZpagF9mb5ndoHj8Lpb9aIdb7El-Pl7IcFs,1753
 docling/utils/utils.py,sha256=llhXSbIDNZ1MHOwBEfLHBAoJIAYI7QlPIonlI1jLUJ0,1208
-docling-2.4.2.dist-info/LICENSE,sha256=mBb7ErEcM8VS9OhiGHnQ2kk75HwPhr54W1Oiz3965MY,1088
-docling-2.4.2.dist-info/METADATA,sha256=WCTdN8xVORJ5Nxt-1CQSmNEe4j5kNvgYu2BK25O0We4,6530
-docling-2.4.2.dist-info/WHEEL,sha256=sP946D7jFCHeNz5Iq4fL4Lu-PrWrFsgfLXbbkciIZwg,88
-docling-2.4.2.dist-info/entry_points.txt,sha256=VOSzV77znM52dz5ysaDuJ0ijl1cnfrh1ZPg8od5OcTs,48
-docling-2.4.2.dist-info/RECORD,,
+docling-2.5.1.dist-info/LICENSE,sha256=mBb7ErEcM8VS9OhiGHnQ2kk75HwPhr54W1Oiz3965MY,1088
+docling-2.5.1.dist-info/METADATA,sha256=qOFYM-E7GjYUIaHtwPoef22zJEWAhIZW8tlIALD17u0,6530
+docling-2.5.1.dist-info/WHEEL,sha256=sP946D7jFCHeNz5Iq4fL4Lu-PrWrFsgfLXbbkciIZwg,88
+docling-2.5.1.dist-info/entry_points.txt,sha256=VOSzV77znM52dz5ysaDuJ0ijl1cnfrh1ZPg8od5OcTs,48
+docling-2.5.1.dist-info/RECORD,,