docling 1.6.2__py3-none-any.whl → 1.19.1__py3-none-any.whl

docling/backend/abstract_backend.py

@@ -1,14 +1,15 @@
 from abc import ABC, abstractmethod
 from io import BytesIO
 from pathlib import Path
-from typing import Any, Iterable, Optional, Union
+from typing import TYPE_CHECKING, Any, Iterable, Optional, Union
 
 from PIL import Image
 
+if TYPE_CHECKING:
+    from docling.datamodel.base_models import BoundingBox, Cell, PageSize
+
 
 class PdfPageBackend(ABC):
-    def __init__(self, page_obj: Any) -> object:
-        pass
 
     @abstractmethod
     def get_text_in_rect(self, bbox: "BoundingBox") -> str:
@@ -19,12 +20,12 @@ class PdfPageBackend(ABC):
         pass
 
     @abstractmethod
-    def get_bitmap_rects(self, scale: int = 1) -> Iterable["BoundingBox"]:
+    def get_bitmap_rects(self, float: int = 1) -> Iterable["BoundingBox"]:
         pass
 
     @abstractmethod
     def get_page_image(
-        self, scale: int = 1, cropbox: Optional["BoundingBox"] = None
+        self, scale: float = 1, cropbox: Optional["BoundingBox"] = None
     ) -> Image.Image:
         pass
 
@@ -32,6 +33,10 @@ class PdfPageBackend(ABC):
     def get_size(self) -> "PageSize":
         pass
 
+    @abstractmethod
+    def is_valid(self) -> bool:
+        pass
+
     @abstractmethod
     def unload(self):
         pass
@@ -39,8 +44,9 @@ class PdfPageBackend(ABC):
 
 class PdfDocumentBackend(ABC):
     @abstractmethod
-    def __init__(self, path_or_stream: Union[BytesIO, Path]):
-        pass
+    def __init__(self, path_or_stream: Union[BytesIO, Path], document_hash: str):
+        self.path_or_stream = path_or_stream
+        self.document_hash = document_hash
 
     @abstractmethod
     def load_page(self, page_no: int) -> PdfPageBackend:
@@ -56,4 +62,7 @@ class PdfDocumentBackend(ABC):
 
     @abstractmethod
     def unload(self):
-        pass
+        if isinstance(self.path_or_stream, BytesIO):
+            self.path_or_stream.close()
+
+        self.path_or_stream = None

docling/backend/docling_parse_backend.py

@@ -1,9 +1,8 @@
 import logging
 import random
-import time
 from io import BytesIO
 from pathlib import Path
-from typing import Iterable, Optional, Union
+from typing import Iterable, List, Optional, Union
 
 import pypdfium2 as pdfium
 from docling_parse.docling_parse import pdf_parser
@@ -17,13 +16,26 @@ _log = logging.getLogger(__name__)
 
 
 class DoclingParsePageBackend(PdfPageBackend):
-    def __init__(self, page_obj: PdfPage, docling_page_obj):
-        super().__init__(page_obj)
+    def __init__(
+        self, parser: pdf_parser, document_hash: str, page_no: int, page_obj: PdfPage
+    ):
         self._ppage = page_obj
-        self._dpage = docling_page_obj
-        self.text_page = None
+        parsed_page = parser.parse_pdf_from_key_on_page(document_hash, page_no)
+
+        self.valid = "pages" in parsed_page
+        if self.valid:
+            self._dpage = parsed_page["pages"][0]
+        else:
+            _log.info(
+                f"An error occured when loading page {page_no} of document {document_hash}."
+            )
+
+    def is_valid(self) -> bool:
+        return self.valid
 
     def get_text_in_rect(self, bbox: BoundingBox) -> str:
+        if not self.valid:
+            return ""
         # Find intersecting cells on the page
         text_piece = ""
         page_size = self.get_size()
@@ -55,9 +67,12 @@ class DoclingParsePageBackend(PdfPageBackend):
         return text_piece
 
     def get_text_cells(self) -> Iterable[Cell]:
-        cells = []
+        cells: List[Cell] = []
         cell_counter = 0
 
+        if not self.valid:
+            return cells
+
         page_size = self.get_size()
 
         parser_width = self._dpage["width"]
@@ -114,7 +129,7 @@ class DoclingParsePageBackend(PdfPageBackend):
 
         return cells
 
-    def get_bitmap_rects(self, scale: int = 1) -> Iterable[BoundingBox]:
+    def get_bitmap_rects(self, scale: float = 1) -> Iterable[BoundingBox]:
         AREA_THRESHOLD = 32 * 32
 
         for i in range(len(self._dpage["images"])):
@@ -129,7 +144,7 @@ class DoclingParsePageBackend(PdfPageBackend):
                 yield cropbox
 
     def get_page_image(
-        self, scale: int = 1, cropbox: Optional[BoundingBox] = None
+        self, scale: float = 1, cropbox: Optional[BoundingBox] = None
     ) -> Image.Image:
 
         page_size = self.get_size()
@@ -168,40 +183,41 @@ class DoclingParsePageBackend(PdfPageBackend):
     def unload(self):
         self._ppage = None
         self._dpage = None
-        self.text_page = None
 
 
 class DoclingParseDocumentBackend(PdfDocumentBackend):
-    def __init__(self, path_or_stream: Union[BytesIO, Path]):
-        super().__init__(path_or_stream)
-        self._pdoc = pdfium.PdfDocument(path_or_stream)
-        # Parsing cells with docling_parser call
-        parser = pdf_parser()
+    def __init__(self, path_or_stream: Union[BytesIO, Path], document_hash: str):
+        super().__init__(path_or_stream, document_hash)
 
-        start_pb_time = time.time()
+        self._pdoc = pdfium.PdfDocument(path_or_stream)
+        self.parser = pdf_parser()
 
+        success = False
         if isinstance(path_or_stream, BytesIO):
-            self._parser_doc = parser.find_cells_from_bytesio(path_or_stream)
-        else:
-            self._parser_doc = parser.find_cells(str(path_or_stream))
+            success = self.parser.load_document_from_bytesio(
+                document_hash, path_or_stream
+            )
+        elif isinstance(path_or_stream, Path):
+            success = self.parser.load_document(document_hash, str(path_or_stream))
 
-        end_pb_time = time.time() - start_pb_time
-        _log.info(
-            f"Time to parse {path_or_stream.name} with docling-parse: time={end_pb_time:.3f}"
-        )
+        if not success:
+            raise RuntimeError(
+                f"docling-parse could not load document {document_hash}."
+            )
 
     def page_count(self) -> int:
-        return len(self._parser_doc["pages"])
+        return len(self._pdoc)  # To be replaced with docling-parse API
 
     def load_page(self, page_no: int) -> DoclingParsePageBackend:
         return DoclingParsePageBackend(
-            self._pdoc[page_no], self._parser_doc["pages"][page_no]
+            self.parser, self.document_hash, page_no, self._pdoc[page_no]
         )
 
     def is_valid(self) -> bool:
         return self.page_count() > 0
 
     def unload(self):
+        super().unload()
+        self.parser.unload_document(self.document_hash)
         self._pdoc.close()
         self._pdoc = None
-        self._parser_doc = None

docling/backend/pypdfium2_backend.py

@@ -1,3 +1,4 @@
+import logging
 import random
 from io import BytesIO
 from pathlib import Path
@@ -6,19 +7,34 @@ from typing import Iterable, List, Optional, Union
 import pypdfium2 as pdfium
 import pypdfium2.raw as pdfium_c
 from PIL import Image, ImageDraw
-from pypdfium2 import PdfPage
+from pypdfium2 import PdfPage, PdfTextPage
+from pypdfium2._helpers.misc import PdfiumError
 
 from docling.backend.abstract_backend import PdfDocumentBackend, PdfPageBackend
 from docling.datamodel.base_models import BoundingBox, Cell, CoordOrigin, PageSize
 
+_log = logging.getLogger(__name__)
+
 
 class PyPdfiumPageBackend(PdfPageBackend):
-    def __init__(self, page_obj: PdfPage):
-        super().__init__(page_obj)
-        self._ppage = page_obj
-        self.text_page = None
+    def __init__(
+        self, pdfium_doc: pdfium.PdfDocument, document_hash: str, page_no: int
+    ):
+        self.valid = True  # No better way to tell from pypdfium.
+        try:
+            self._ppage: pdfium.PdfPage = pdfium_doc[page_no]
+        except PdfiumError as e:
+            _log.info(
+                f"An exception occured when loading page {page_no} of document {document_hash}.",
+                exc_info=True,
+            )
+            self.valid = False
+        self.text_page: Optional[PdfTextPage] = None
+
+    def is_valid(self) -> bool:
+        return self.valid
 
-    def get_bitmap_rects(self, scale: int = 1) -> Iterable[BoundingBox]:
+    def get_bitmap_rects(self, scale: float = 1) -> Iterable[BoundingBox]:
         AREA_THRESHOLD = 32 * 32
         for obj in self._ppage.get_objects(filter=[pdfium_c.FPDF_PAGEOBJ_IMAGE]):
             pos = obj.get_pos()
@@ -173,7 +189,7 @@ class PyPdfiumPageBackend(PdfPageBackend):
         return cells
 
     def get_page_image(
-        self, scale: int = 1, cropbox: Optional[BoundingBox] = None
+        self, scale: float = 1, cropbox: Optional[BoundingBox] = None
     ) -> Image.Image:
 
         page_size = self.get_size()
@@ -215,19 +231,25 @@ class PyPdfiumPageBackend(PdfPageBackend):
 
 
 class PyPdfiumDocumentBackend(PdfDocumentBackend):
-    def __init__(self, path_or_stream: Union[BytesIO, Path]):
-        super().__init__(path_or_stream)
-        self._pdoc = pdfium.PdfDocument(path_or_stream)
+    def __init__(self, path_or_stream: Union[BytesIO, Path], document_hash: str):
+        super().__init__(path_or_stream, document_hash)
+        try:
+            self._pdoc = pdfium.PdfDocument(path_or_stream)
+        except PdfiumError as e:
+            raise RuntimeError(
+                f"pypdfium could not load document {document_hash}"
+            ) from e
 
     def page_count(self) -> int:
         return len(self._pdoc)
 
     def load_page(self, page_no: int) -> PyPdfiumPageBackend:
-        return PyPdfiumPageBackend(self._pdoc[page_no])
+        return PyPdfiumPageBackend(self._pdoc, self.document_hash, page_no)
 
     def is_valid(self) -> bool:
         return self.page_count() > 0
 
     def unload(self):
+        super().unload()
         self._pdoc.close()
         self._pdoc = None

docling/cli/main.py

@@ -0,0 +1,253 @@
+import importlib
+import json
+import logging
+import time
+import warnings
+from enum import Enum
+from pathlib import Path
+from typing import Annotated, Iterable, List, Optional
+
+import typer
+from docling_core.utils.file import resolve_file_source
+
+from docling.backend.docling_parse_backend import DoclingParseDocumentBackend
+from docling.backend.pypdfium2_backend import PyPdfiumDocumentBackend
+from docling.datamodel.base_models import ConversionStatus
+from docling.datamodel.document import ConversionResult, DocumentConversionInput
+from docling.datamodel.pipeline_options import (
+    EasyOcrOptions,
+    PipelineOptions,
+    TesseractCliOcrOptions,
+    TesseractOcrOptions,
+)
+from docling.document_converter import DocumentConverter
+
+warnings.filterwarnings(action="ignore", category=UserWarning, module="pydantic|torch")
+warnings.filterwarnings(action="ignore", category=FutureWarning, module="easyocr")
+
+_log = logging.getLogger(__name__)
+from rich.console import Console
+
+err_console = Console(stderr=True)
+
+
+app = typer.Typer(
+    name="Docling",
+    no_args_is_help=True,
+    add_completion=False,
+    pretty_exceptions_enable=False,
+)
+
+
+def version_callback(value: bool):
+    if value:
+        docling_version = importlib.metadata.version("docling")
+        docling_core_version = importlib.metadata.version("docling-core")
+        docling_ibm_models_version = importlib.metadata.version("docling-ibm-models")
+        docling_parse_version = importlib.metadata.version("docling-parse")
+        print(f"Docling version: {docling_version}")
+        print(f"Docling Core version: {docling_core_version}")
+        print(f"Docling IBM Models version: {docling_ibm_models_version}")
+        print(f"Docling Parse version: {docling_parse_version}")
+        raise typer.Exit()
+
+
+# Define an enum for the backend options
+class Backend(str, Enum):
+    PYPDFIUM2 = "pypdfium2"
+    DOCLING = "docling"
+
+
+# Define an enum for the ocr engines
+class OcrEngine(str, Enum):
+    EASYOCR = "easyocr"
+    TESSERACT_CLI = "tesseract_cli"
+    TESSERACT = "tesseract"
+
+
+def export_documents(
+    conv_results: Iterable[ConversionResult],
+    output_dir: Path,
+    export_json: bool,
+    export_md: bool,
+    export_txt: bool,
+    export_doctags: bool,
+):
+
+    success_count = 0
+    failure_count = 0
+
+    for conv_res in conv_results:
+        if conv_res.status == ConversionStatus.SUCCESS:
+            success_count += 1
+            doc_filename = conv_res.input.file.stem
+
+            # Export Deep Search document JSON format:
+            if export_json:
+                fname = output_dir / f"{doc_filename}.json"
+                with fname.open("w") as fp:
+                    _log.info(f"writing JSON output to {fname}")
+                    fp.write(json.dumps(conv_res.render_as_dict()))
+
+            # Export Text format:
+            if export_txt:
+                fname = output_dir / f"{doc_filename}.txt"
+                with fname.open("w") as fp:
+                    _log.info(f"writing Text output to {fname}")
+                    fp.write(conv_res.render_as_text())
+
+            # Export Markdown format:
+            if export_md:
+                fname = output_dir / f"{doc_filename}.md"
+                with fname.open("w") as fp:
+                    _log.info(f"writing Markdown output to {fname}")
+                    fp.write(conv_res.render_as_markdown())
+
+            # Export Document Tags format:
+            if export_doctags:
+                fname = output_dir / f"{doc_filename}.doctags"
+                with fname.open("w") as fp:
+                    _log.info(f"writing Doc Tags output to {fname}")
+                    fp.write(conv_res.render_as_doctags())
+
+        else:
+            _log.warning(f"Document {conv_res.input.file} failed to convert.")
+            failure_count += 1
+
+    _log.info(
+        f"Processed {success_count + failure_count} docs, of which {failure_count} failed"
+    )
+
+
+@app.command(no_args_is_help=True)
+def convert(
+    input_sources: Annotated[
+        List[str],
+        typer.Argument(
+            ...,
+            metavar="source",
+            help="PDF files to convert. Can be local file / directory paths or URL.",
+        ),
+    ],
+    export_json: Annotated[
+        bool,
+        typer.Option(
+            ..., "--json/--no-json", help="If enabled the document is exported as JSON."
+        ),
+    ] = False,
+    export_md: Annotated[
+        bool,
+        typer.Option(
+            ..., "--md/--no-md", help="If enabled the document is exported as Markdown."
+        ),
+    ] = True,
+    export_txt: Annotated[
+        bool,
+        typer.Option(
+            ..., "--txt/--no-txt", help="If enabled the document is exported as Text."
+        ),
+    ] = False,
+    export_doctags: Annotated[
+        bool,
+        typer.Option(
+            ...,
+            "--doctags/--no-doctags",
+            help="If enabled the document is exported as Doc Tags.",
+        ),
+    ] = False,
+    ocr: Annotated[
+        bool,
+        typer.Option(
+            ..., help="If enabled, the bitmap content will be processed using OCR."
+        ),
+    ] = True,
+    backend: Annotated[
+        Backend, typer.Option(..., help="The PDF backend to use.")
+    ] = Backend.DOCLING,
+    ocr_engine: Annotated[
+        OcrEngine, typer.Option(..., help="The OCR engine to use.")
+    ] = OcrEngine.EASYOCR,
+    output: Annotated[
+        Path, typer.Option(..., help="Output directory where results are saved.")
+    ] = Path("."),
+    version: Annotated[
+        Optional[bool],
+        typer.Option(
+            "--version",
+            callback=version_callback,
+            is_eager=True,
+            help="Show version information.",
+        ),
+    ] = None,
+):
+    logging.basicConfig(level=logging.INFO)
+
+    input_doc_paths: List[Path] = []
+    for src in input_sources:
+        source = resolve_file_source(source=src)
+        if not source.exists():
+            err_console.print(
+                f"[red]Error: The input file {source} does not exist.[/red]"
+            )
+            raise typer.Abort()
+        elif source.is_dir():
+            input_doc_paths.extend(list(source.glob("**/*.pdf")))
+            input_doc_paths.extend(list(source.glob("**/*.PDF")))
+        else:
+            input_doc_paths.append(source)
+
+    match backend:
+        case Backend.PYPDFIUM2:
+            do_cell_matching = ocr  # only do cell matching when OCR enabled
+            pdf_backend = PyPdfiumDocumentBackend
+        case Backend.DOCLING:
+            do_cell_matching = True
+            pdf_backend = DoclingParseDocumentBackend
+        case _:
+            raise RuntimeError(f"Unexpected backend type {backend}")
+
+    match ocr_engine:
+        case OcrEngine.EASYOCR:
+            ocr_options = EasyOcrOptions()
+        case OcrEngine.TESSERACT_CLI:
+            ocr_options = TesseractCliOcrOptions()
+        case OcrEngine.TESSERACT:
+            ocr_options = TesseractOcrOptions()
+        case _:
+            raise RuntimeError(f"Unexpected backend type {backend}")
+
+    pipeline_options = PipelineOptions(
+        do_ocr=ocr,
+        ocr_options=ocr_options,
+        do_table_structure=True,
+    )
+    pipeline_options.table_structure_options.do_cell_matching = do_cell_matching
+    doc_converter = DocumentConverter(
+        pipeline_options=pipeline_options,
+        pdf_backend=pdf_backend,
+    )
+
+    # Define input files
+    input = DocumentConversionInput.from_paths(input_doc_paths)
+
+    start_time = time.time()
+
+    conv_results = doc_converter.convert(input)
+
+    output.mkdir(parents=True, exist_ok=True)
+    export_documents(
+        conv_results,
+        output_dir=output,
+        export_json=export_json,
+        export_md=export_md,
+        export_txt=export_txt,
+        export_doctags=export_doctags,
+    )
+
+    end_time = time.time() - start_time
+
+    _log.info(f"All documents were converted in {end_time:.2f} seconds.")
+
+
+if __name__ == "__main__":
+    app()

docling/datamodel/base_models.py

@@ -9,6 +9,10 @@ from pydantic import BaseModel, ConfigDict, Field, model_validator
 from typing_extensions import Self
 
 from docling.backend.abstract_backend import PdfPageBackend
+from docling.datamodel.pipeline_options import (  # Must be imported here for backward compatibility.
+    PipelineOptions,
+    TableStructureOptions,
+)
 
 
 class ConversionStatus(str, Enum):
@@ -16,7 +20,7 @@ class ConversionStatus(str, Enum):
     STARTED = auto()
     FAILURE = auto()
     SUCCESS = auto()
-    SUCCESS_WITH_ERRORS = auto()
+    PARTIAL_SUCCESS = auto()
 
 
 class DocInputType(str, Enum):
@@ -29,6 +33,18 @@ class CoordOrigin(str, Enum):
     BOTTOMLEFT = auto()
 
 
+class DoclingComponentType(str, Enum):
+    PDF_BACKEND = auto()
+    MODEL = auto()
+    DOC_ASSEMBLER = auto()
+
+
+class ErrorItem(BaseModel):
+    component_type: DoclingComponentType
+    module_name: str
+    error_message: str
+
+
 class PageSize(BaseModel):
     width: float = 0.0
     height: float = 0.0
@@ -59,6 +75,15 @@ class BoundingBox(BaseModel):
 
         return out_bbox
 
+    def normalized(self, page_size: PageSize) -> "BoundingBox":
+        out_bbox = copy.deepcopy(self)
+        out_bbox.l /= page_size.width
+        out_bbox.r /= page_size.width
+        out_bbox.t /= page_size.height
+        out_bbox.b /= page_size.height
+
+        return out_bbox
+
     def as_tuple(self):
         if self.coord_origin == CoordOrigin.TOPLEFT:
             return (self.l, self.t, self.r, self.b)
@@ -66,7 +91,7 @@ class BoundingBox(BaseModel):
             return (self.l, self.b, self.r, self.t)
 
     @classmethod
-    def from_tuple(cls, coord: Tuple[float], origin: CoordOrigin):
+    def from_tuple(cls, coord: Tuple[float, ...], origin: CoordOrigin):
         if origin == CoordOrigin.TOPLEFT:
             l, t, r, b = coord[0], coord[1], coord[2], coord[3]
             if r < l:
@@ -85,7 +110,10 @@ class BoundingBox(BaseModel):
             return BoundingBox(l=l, t=t, r=r, b=b, coord_origin=origin)
 
     def area(self) -> float:
-        return (self.r - self.l) * (self.b - self.t)
+        area = (self.r - self.l) * (self.b - self.t)
+        if self.coord_origin == CoordOrigin.BOTTOMLEFT:
+            area = -area
+        return area
 
     def intersection_area_with(self, other: "BoundingBox") -> float:
         # Calculate intersection coordinates
@@ -225,19 +253,19 @@ class EquationPrediction(BaseModel):
 
 
 class PagePredictions(BaseModel):
-    layout: LayoutPrediction = None
-    tablestructure: TableStructurePrediction = None
-    figures_classification: FigureClassificationPrediction = None
-    equations_prediction: EquationPrediction = None
+    layout: Optional[LayoutPrediction] = None
+    tablestructure: Optional[TableStructurePrediction] = None
+    figures_classification: Optional[FigureClassificationPrediction] = None
+    equations_prediction: Optional[EquationPrediction] = None
 
 
 PageElement = Union[TextElement, TableElement, FigureElement]
 
 
 class AssembledUnit(BaseModel):
-    elements: List[PageElement]
-    body: List[PageElement]
-    headers: List[PageElement]
+    elements: List[PageElement] = []
+    body: List[PageElement] = []
+    headers: List[PageElement] = []
 
 
 class Page(BaseModel):
@@ -246,7 +274,7 @@ class Page(BaseModel):
     page_no: int
     page_hash: Optional[str] = None
     size: Optional[PageSize] = None
-    cells: List[Cell] = None
+    cells: List[Cell] = []
     predictions: PagePredictions = PagePredictions()
     assembled: Optional[AssembledUnit] = None
 
@@ -277,22 +305,6 @@ class DocumentStream(BaseModel):
     stream: BytesIO
 
 
-class TableStructureOptions(BaseModel):
-    do_cell_matching: bool = (
-        True
-        # True:  Matches predictions back to PDF cells. Can break table output if PDF cells
-        #        are merged across table columns.
-        # False: Let table structure model define the text cells, ignore PDF cells.
-    )
-
-
-class PipelineOptions(BaseModel):
-    do_table_structure: bool = True  # True: perform table structure extraction
-    do_ocr: bool = True  # True: perform OCR, replace programmatic PDF text
-
-    table_structure_options: TableStructureOptions = TableStructureOptions()
-
-
 class AssembleOptions(BaseModel):
     keep_page_images: Annotated[
         bool,