docling 2.0.0__py3-none-any.whl → 2.2.0__py3-none-any.whl

docling/backend/md_backend.py

@@ -0,0 +1,293 @@
+import logging
+from io import BytesIO
+from pathlib import Path
+from typing import Set, Union
+
+import marko
+import marko.ext
+import marko.ext.gfm
+import marko.inline
+from docling_core.types.doc import (
+    DocItemLabel,
+    DoclingDocument,
+    DocumentOrigin,
+    GroupLabel,
+    TableCell,
+    TableData,
+)
+from marko import Markdown
+
+from docling.backend.abstract_backend import DeclarativeDocumentBackend
+from docling.datamodel.base_models import InputFormat
+from docling.datamodel.document import InputDocument
+
+_log = logging.getLogger(__name__)
+
+
+class MarkdownDocumentBackend(DeclarativeDocumentBackend):
+    def __init__(self, in_doc: "InputDocument", path_or_stream: Union[BytesIO, Path]):
+        super().__init__(in_doc, path_or_stream)
+
+        _log.debug("MD INIT!!!")
+
+        # Markdown file:
+        self.path_or_stream = path_or_stream
+        self.valid = True
+        self.markdown = ""  # To store original Markdown string
+
+        self.in_table = False
+        self.md_table_buffer: list[str] = []
+        self.inline_text_buffer = ""
+
+        try:
+            if isinstance(self.path_or_stream, BytesIO):
+                text_stream = self.path_or_stream.getvalue().decode("utf-8")
+                self.markdown = text_stream
+            if isinstance(self.path_or_stream, Path):
+                with open(self.path_or_stream, "r", encoding="utf-8") as f:
+                    md_content = f.read()
+                    self.markdown = md_content
+            self.valid = True
+
+            _log.debug(self.markdown)
+        except Exception as e:
+            raise RuntimeError(
+                f"Could not initialize MD backend for file with hash {self.document_hash}."
+            ) from e
+        return
+
+    def close_table(self, doc=None):
+        if self.in_table:
+            _log.debug("=== TABLE START ===")
+            for md_table_row in self.md_table_buffer:
+                _log.debug(md_table_row)
+            _log.debug("=== TABLE END ===")
+            tcells = []
+            result_table = []
+            for n, md_table_row in enumerate(self.md_table_buffer):
+                data = []
+                if n == 0:
+                    header = [t.strip() for t in md_table_row.split("|")[1:-1]]
+                    for value in header:
+                        data.append(value)
+                    result_table.append(data)
+                if n > 1:
+                    values = [t.strip() for t in md_table_row.split("|")[1:-1]]
+                    for value in values:
+                        data.append(value)
+                    result_table.append(data)
+
+            for trow_ind, trow in enumerate(result_table):
+                for tcol_ind, cellval in enumerate(trow):
+                    row_span = (
+                        1  # currently supporting just simple tables (without spans)
+                    )
+                    col_span = (
+                        1  # currently supporting just simple tables (without spans)
+                    )
+                    icell = TableCell(
+                        text=cellval.strip(),
+                        row_span=row_span,
+                        col_span=col_span,
+                        start_row_offset_idx=trow_ind,
+                        end_row_offset_idx=trow_ind + row_span,
+                        start_col_offset_idx=tcol_ind,
+                        end_col_offset_idx=tcol_ind + col_span,
+                        col_header=False,
+                        row_header=False,
+                    )
+                    tcells.append(icell)
+
+            num_rows = len(result_table)
+            num_cols = len(result_table[0])
+            self.in_table = False
+            self.md_table_buffer = []  # clean table markdown buffer
+            # Initialize Docling TableData
+            data = TableData(num_rows=num_rows, num_cols=num_cols, table_cells=tcells)
+            # Populate
+            for tcell in tcells:
+                data.table_cells.append(tcell)
+            if len(tcells) > 0:
+                doc.add_table(data=data)
+        return
+
+    def process_inline_text(self, parent_element, doc=None):
+        # self.inline_text_buffer += str(text_in)
+        txt = self.inline_text_buffer.strip()
+        if len(txt) > 0:
+            doc.add_text(
+                label=DocItemLabel.PARAGRAPH,
+                parent=parent_element,
+                text=txt,
+            )
+        self.inline_text_buffer = ""
+
+    def iterate_elements(self, element, depth=0, doc=None, parent_element=None):
+        # Iterates over all elements in the AST
+        # Check for different element types and process relevant details
+        if isinstance(element, marko.block.Heading):
+            self.close_table(doc)
+            self.process_inline_text(parent_element, doc)
+            _log.debug(
+                f" - Heading level {element.level}, content: {element.children[0].children}"
+            )
+            if element.level == 1:
+                doc_label = DocItemLabel.TITLE
+            else:
+                doc_label = DocItemLabel.SECTION_HEADER
+            snippet_text = element.children[0].children.strip()
+
+            parent_element = doc.add_text(
+                label=doc_label, parent=parent_element, text=snippet_text
+            )
+
+        elif isinstance(element, marko.block.List):
+            self.close_table(doc)
+            self.process_inline_text(parent_element, doc)
+            _log.debug(f" - List {'ordered' if element.ordered else 'unordered'}")
+            list_label = GroupLabel.LIST
+            if element.ordered:
+                list_label = GroupLabel.ORDERED_LIST
+            parent_element = doc.add_group(
+                label=list_label, name=f"list", parent=parent_element
+            )
+
+        elif isinstance(element, marko.block.ListItem):
+            self.close_table(doc)
+            self.process_inline_text(parent_element, doc)
+            _log.debug(" - List item")
+
+            snippet_text = str(element.children[0].children[0].children)
+            is_numbered = False
+            if parent_element.label == GroupLabel.ORDERED_LIST:
+                is_numbered = True
+            doc.add_list_item(
+                enumerated=is_numbered, parent=parent_element, text=snippet_text
+            )
+
+        elif isinstance(element, marko.inline.Image):
+            self.close_table(doc)
+            self.process_inline_text(parent_element, doc)
+            _log.debug(f" - Image with alt: {element.title}, url: {element.dest}")
+            doc.add_picture(parent=parent_element, caption=element.title)
+
+        elif isinstance(element, marko.block.Paragraph):
+            self.process_inline_text(parent_element, doc)
+
+        elif isinstance(element, marko.inline.RawText):
+            _log.debug(f" - Paragraph (raw text): {element.children}")
+            snippet_text = str(element.children).strip()
+            # Detect start of the table:
+            if "|" in snippet_text:
+                # most likely part of the markdown table
+                self.in_table = True
+                if len(self.md_table_buffer) > 0:
+                    self.md_table_buffer[len(self.md_table_buffer) - 1] += str(
+                        snippet_text
+                    )
+                else:
+                    self.md_table_buffer.append(snippet_text)
+            else:
+                self.close_table(doc)
+                self.in_table = False
+                # most likely just inline text
+                self.inline_text_buffer += str(
+                    element.children
+                )  # do not strip an inline text, as it may contain important spaces
+
+        elif isinstance(element, marko.inline.CodeSpan):
+            self.close_table(doc)
+            self.process_inline_text(parent_element, doc)
+            _log.debug(f" - Code Span: {element.children}")
+            snippet_text = str(element.children).strip()
+            doc.add_text(
+                label=DocItemLabel.CODE, parent=parent_element, text=snippet_text
+            )
+
+        elif isinstance(element, marko.block.CodeBlock):
+            self.close_table(doc)
+            self.process_inline_text(parent_element, doc)
+            _log.debug(f" - Code Block: {element.children}")
+            snippet_text = str(element.children[0].children).strip()
+            doc.add_text(
+                label=DocItemLabel.CODE, parent=parent_element, text=snippet_text
+            )
+
+        elif isinstance(element, marko.block.FencedCode):
+            self.close_table(doc)
+            self.process_inline_text(parent_element, doc)
+            _log.debug(f" - Code Block: {element.children}")
+            snippet_text = str(element.children[0].children).strip()
+            doc.add_text(
+                label=DocItemLabel.CODE, parent=parent_element, text=snippet_text
+            )
+
+        elif isinstance(element, marko.inline.LineBreak):
+            self.process_inline_text(parent_element, doc)
+            if self.in_table:
+                _log.debug("Line break in a table")
+                self.md_table_buffer.append("")
+
+        elif isinstance(element, marko.block.HTMLBlock):
+            self.process_inline_text(parent_element, doc)
+            self.close_table(doc)
+            _log.debug("HTML Block: {}".format(element))
+            if (
+                len(element.children) > 0
+            ):  # If Marko doesn't return any content for HTML block, skip it
+                snippet_text = str(element.children).strip()
+                doc.add_text(
+                    label=DocItemLabel.CODE, parent=parent_element, text=snippet_text
+                )
+        else:
+            if not isinstance(element, str):
+                self.close_table(doc)
+                _log.debug("Some other element: {}".format(element))
+
+        # Iterate through the element's children (if any)
+        if not isinstance(element, marko.block.ListItem):
+            if not isinstance(element, marko.block.Heading):
+                if not isinstance(element, marko.block.FencedCode):
+                    # if not isinstance(element, marko.block.Paragraph):
+                    if hasattr(element, "children"):
+                        for child in element.children:
+                            self.iterate_elements(child, depth + 1, doc, parent_element)
+
+    def is_valid(self) -> bool:
+        return self.valid
+
+    def unload(self):
+        if isinstance(self.path_or_stream, BytesIO):
+            self.path_or_stream.close()
+        self.path_or_stream = None
+
+    @classmethod
+    def supports_pagination(cls) -> bool:
+        return False
+
+    @classmethod
+    def supported_formats(cls) -> Set[InputFormat]:
+        return {InputFormat.MD}
+
+    def convert(self) -> DoclingDocument:
+        _log.debug("converting Markdown...")
+
+        origin = DocumentOrigin(
+            filename=self.file.name or "file",
+            mimetype="text/markdown",
+            binary_hash=self.document_hash,
+        )
+
+        doc = DoclingDocument(name=self.file.stem or "file", origin=origin)
+
+        if self.is_valid():
+            # Parse the markdown into an abstract syntax tree (AST)
+            marko_parser = Markdown()
+            parsed_ast = marko_parser.parse(self.markdown)
+            # Start iterating from the root of the AST
+            self.iterate_elements(parsed_ast, 0, doc, None)
+        else:
+            raise RuntimeError(
+                f"Cannot convert md with {self.document_hash} because the backend failed to init."
+            )
+        return doc

docling/backend/mspowerpoint_backend.py

@@ -83,21 +83,14 @@ class MsPowerpointDocumentBackend(DeclarativeDocumentBackend, PaginatedDocumentB
         # Parses the PPTX into a structured document model.
         # origin = DocumentOrigin(filename=self.path_or_stream.name, mimetype=next(iter(FormatToMimeType.get(InputFormat.PPTX))), binary_hash=self.document_hash)
 
-        fname = ""
-        if isinstance(self.path_or_stream, Path):
-            fname = self.path_or_stream.name
-
         origin = DocumentOrigin(
-            filename=fname,
+            filename=self.file.name or "file",
             mimetype="application/vnd.ms-powerpoint",
             binary_hash=self.document_hash,
         )
-        if len(fname) > 0:
-            docname = Path(fname).stem
-        else:
-            docname = "stream"
+
         doc = DoclingDocument(
-            name=docname, origin=origin
+            name=self.file.stem or "file", origin=origin
         )  # must add origin information
         doc = self.walk_linear(self.pptx_obj, doc)
 
@@ -119,10 +112,16 @@ class MsPowerpointDocumentBackend(DeclarativeDocumentBackend, PaginatedDocumentB
 
     def handle_text_elements(self, shape, parent_slide, slide_ind, doc):
         is_a_list = False
+        is_list_group_created = False
         enum_list_item_value = 0
+        new_list = None
+        bullet_type = "None"
+        list_text = ""
+        list_label = GroupLabel.LIST
+        prov = self.generate_prov(shape, slide_ind, shape.text.strip())
+
+        # Identify if shape contains lists
         for paragraph in shape.text_frame.paragraphs:
-            enum_list_item_value += 1
-            bullet_type = "None"
             # Check if paragraph is a bullet point using the `element` XML
             p = paragraph._element
             if (
@@ -143,29 +142,32 @@ class MsPowerpointDocumentBackend(DeclarativeDocumentBackend, PaginatedDocumentB
             if paragraph.level > 0:
                 # Most likely a sub-list
                 is_a_list = True
-            list_text = paragraph.text.strip()
-
-            prov = self.generate_prov(shape, slide_ind, shape.text.strip())
 
             if is_a_list:
                 # Determine if this is an unordered list or an ordered list.
                 # Set GroupLabel.ORDERED_LIST when it fits.
-                list_label = GroupLabel.LIST
                 if bullet_type == "Numbered":
                     list_label = GroupLabel.ORDERED_LIST
 
-                new_list = doc.add_group(
-                    label=list_label, name=f"list", parent=parent_slide
-                )
-            else:
-                new_list = None
-
             if is_a_list:
                 _log.debug("LIST DETECTED!")
             else:
                 _log.debug("No List")
 
-            # for e in p.iter():
+        # If there is a list inside of the shape, create a new docling list to assign list items to
+        # if is_a_list:
+        #     new_list = doc.add_group(
+        #         label=list_label, name=f"list", parent=parent_slide
+        #     )
+
+        # Iterate through paragraphs to build up text
+        for paragraph in shape.text_frame.paragraphs:
+            # p_text = paragraph.text.strip()
+            p = paragraph._element
+            enum_list_item_value += 1
+            inline_paragraph_text = ""
+            inline_list_item_text = ""
+
             for e in p.iterfind(".//a:r", namespaces={"a": self.namespaces["a"]}):
                 if len(e.text.strip()) > 0:
                     e_is_a_list_item = False
@@ -187,15 +189,17 @@ class MsPowerpointDocumentBackend(DeclarativeDocumentBackend, PaginatedDocumentB
                         e_is_a_list_item = False
 
                     if e_is_a_list_item:
+                        if len(inline_paragraph_text) > 0:
+                            # output accumulated inline text:
+                            doc.add_text(
+                                label=doc_label,
+                                parent=parent_slide,
+                                text=inline_paragraph_text,
+                                prov=prov,
+                            )
                         # Set marker and enumerated arguments if this is an enumeration element.
-                        enum_marker = str(enum_list_item_value) + "."
-                        doc.add_list_item(
-                            marker=enum_marker,
-                            enumerated=is_numbered,
-                            parent=new_list,
-                            text=list_text,
-                            prov=prov,
-                        )
+                        inline_list_item_text += e.text
+                        # print(e.text)
                     else:
                         # Assign proper label to the text, depending if it's a Title or Section Header
                         # For other types of text, assign - PARAGRAPH
@@ -210,15 +214,34 @@ class MsPowerpointDocumentBackend(DeclarativeDocumentBackend, PaginatedDocumentB
                                 doc_label = DocItemLabel.TITLE
                             elif placeholder_type == PP_PLACEHOLDER.SUBTITLE:
                                 DocItemLabel.SECTION_HEADER
-
                         enum_list_item_value = 0
+                        inline_paragraph_text += e.text
 
-                        doc.add_text(
-                            label=doc_label,
-                            parent=parent_slide,
-                            text=list_text,
-                            prov=prov,
-                        )
+            if len(inline_paragraph_text) > 0:
+                # output accumulated inline text:
+                doc.add_text(
+                    label=doc_label,
+                    parent=parent_slide,
+                    text=inline_paragraph_text,
+                    prov=prov,
+                )
+
+            if len(inline_list_item_text) > 0:
+                enum_marker = ""
+                if is_numbered:
+                    enum_marker = str(enum_list_item_value) + "."
+                if not is_list_group_created:
+                    new_list = doc.add_group(
+                        label=list_label, name=f"list", parent=parent_slide
+                    )
+                    is_list_group_created = True
+                doc.add_list_item(
+                    marker=enum_marker,
+                    enumerated=is_numbered,
+                    parent=new_list,
+                    text=inline_list_item_text,
+                    prov=prov,
+                )
         return
 
     def handle_title(self, shape, parent_slide, slide_ind, doc):
@@ -311,7 +334,7 @@ class MsPowerpointDocumentBackend(DeclarativeDocumentBackend, PaginatedDocumentB
             if len(tcells) > 0:
                 # If table is not fully empty...
                 # Create Docling table
-                doc.add_table(data=data, prov=prov)
+                doc.add_table(parent=parent_slide, data=data, prov=prov)
         return
 
     def walk_linear(self, pptx_obj, doc) -> DoclingDocument:

docling/backend/msword_backend.py

@@ -85,20 +85,13 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
     def convert(self) -> DoclingDocument:
         # Parses the DOCX into a structured document model.
 
-        fname = ""
-        if isinstance(self.path_or_stream, Path):
-            fname = self.path_or_stream.name
-
         origin = DocumentOrigin(
-            filename=fname,
+            filename=self.file.name or "file",
             mimetype="application/vnd.openxmlformats-officedocument.wordprocessingml.document",
             binary_hash=self.document_hash,
         )
-        if len(fname) > 0:
-            docname = Path(fname).stem
-        else:
-            docname = "stream"
-        doc = DoclingDocument(name=docname, origin=origin)
+
+        doc = DoclingDocument(name=self.file.stem or "file", origin=origin)
         if self.is_valid():
             assert self.docx_obj is not None
             doc = self.walk_linear(self.docx_obj.element.body, self.docx_obj, doc)

docling/datamodel/base_models.py

@@ -30,6 +30,8 @@ class InputFormat(str, Enum):
     HTML = "html"
     IMAGE = "image"
     PDF = "pdf"
+    ASCIIDOC = "asciidoc"
+    MD = "md"
 
 
 class OutputFormat(str, Enum):
@@ -43,29 +45,33 @@ FormatToExtensions: Dict[InputFormat, List[str]] = {
     InputFormat.DOCX: ["docx", "dotx", "docm", "dotm"],
     InputFormat.PPTX: ["pptx", "potx", "ppsx", "pptm", "potm", "ppsm"],
     InputFormat.PDF: ["pdf"],
+    InputFormat.MD: ["md"],
     InputFormat.HTML: ["html", "htm", "xhtml"],
     InputFormat.IMAGE: ["jpg", "jpeg", "png", "tif", "tiff", "bmp"],
+    InputFormat.ASCIIDOC: ["adoc", "asciidoc", "asc"],
 }
 
-FormatToMimeType: Dict[InputFormat, Set[str]] = {
-    InputFormat.DOCX: {
+FormatToMimeType: Dict[InputFormat, List[str]] = {
+    InputFormat.DOCX: [
         "application/vnd.openxmlformats-officedocument.wordprocessingml.document",
         "application/vnd.openxmlformats-officedocument.wordprocessingml.template",
-    },
-    InputFormat.PPTX: {
+    ],
+    InputFormat.PPTX: [
         "application/vnd.openxmlformats-officedocument.presentationml.template",
         "application/vnd.openxmlformats-officedocument.presentationml.slideshow",
         "application/vnd.openxmlformats-officedocument.presentationml.presentation",
-    },
-    InputFormat.HTML: {"text/html", "application/xhtml+xml"},
-    InputFormat.IMAGE: {
+    ],
+    InputFormat.HTML: ["text/html", "application/xhtml+xml"],
+    InputFormat.IMAGE: [
         "image/png",
         "image/jpeg",
         "image/tiff",
         "image/gif",
         "image/bmp",
-    },
-    InputFormat.PDF: {"application/pdf"},
+    ],
+    InputFormat.PDF: ["application/pdf"],
+    InputFormat.ASCIIDOC: ["text/asciidoc"],
+    InputFormat.MD: ["text/markdown", "text/x-markdown"],
 }
 MimeTypeToFormat = {
     mime: fmt for fmt, mimes in FormatToMimeType.items() for mime in mimes

docling/datamodel/document.py

@@ -6,12 +6,6 @@ from pathlib import Path, PurePath
 from typing import TYPE_CHECKING, Dict, Iterable, List, Optional, Tuple, Type, Union
 
 import filetype
-from docling_core.types import BaseText
-from docling_core.types import Document as DsDocument
-from docling_core.types import DocumentDescription as DsDocumentDescription
-from docling_core.types import FileInfoObject as DsFileInfoObject
-from docling_core.types import PageDimensions, PageReference, Prov, Ref
-from docling_core.types import Table as DsSchemaTable
 from docling_core.types.doc import (
     DocItem,
     DocItemLabel,
@@ -22,7 +16,22 @@ from docling_core.types.doc import (
     TextItem,
 )
 from docling_core.types.doc.document import ListItem
-from docling_core.types.legacy_doc.base import Figure, GlmTableCell, TableCell
+from docling_core.types.legacy_doc.base import (
+    BaseText,
+    Figure,
+    GlmTableCell,
+    PageDimensions,
+    PageReference,
+    Prov,
+    Ref,
+)
+from docling_core.types.legacy_doc.base import Table as DsSchemaTable
+from docling_core.types.legacy_doc.base import TableCell
+from docling_core.types.legacy_doc.document import (
+    CCSDocumentDescription as DsDocumentDescription,
+)
+from docling_core.types.legacy_doc.document import CCSFileInfoObject as DsFileInfoObject
+from docling_core.types.legacy_doc.document import ExportedCCSDocument as DsDocument
 from docling_core.utils.file import resolve_file_source
 from pydantic import BaseModel
 from typing_extensions import deprecated
@@ -36,6 +45,8 @@ from docling.datamodel.base_models import (
     ConversionStatus,
     DocumentStream,
     ErrorItem,
+    FormatToExtensions,
+    FormatToMimeType,
     InputFormat,
     MimeTypeToFormat,
     Page,
@@ -134,11 +145,13 @@ class InputDocument(BaseModel):
                         self.valid = False
 
         except (FileNotFoundError, OSError) as e:
+            self.valid = False
             _log.exception(
                 f"File {self.file.name} not found or cannot be opened.", exc_info=e
             )
             # raise
         except RuntimeError as e:
+            self.valid = False
             _log.exception(
                 f"An unexpected error occurred while opening the document {self.file.name}",
                 exc_info=e,
@@ -157,6 +170,8 @@ class InputDocument(BaseModel):
             )
 
         self._backend = backend(self, path_or_stream=path_or_stream)
+        if not self._backend.is_valid():
+            self.valid = False
 
 
 class DocumentFormat(str, Enum):
@@ -471,26 +486,48 @@ class _DocumentConversionInput(BaseModel):
             else:
                 raise RuntimeError(f"Unexpected obj type in iterator: {type(obj)}")
 
-    def _guess_format(self, obj):
-        content = None
+    def _guess_format(self, obj: Union[Path, DocumentStream]):
+        content = b""  # empty binary blob
+        format = None
+
         if isinstance(obj, Path):
             mime = filetype.guess_mime(str(obj))
             if mime is None:
+                ext = obj.suffix[1:]
+                mime = self._mime_from_extension(ext)
+            if mime is None:  # must guess from
                 with obj.open("rb") as f:
                     content = f.read(1024)  # Read first 1KB
 
         elif isinstance(obj, DocumentStream):
-            obj.stream.seek(0)
             content = obj.stream.read(8192)
             obj.stream.seek(0)
             mime = filetype.guess_mime(content)
+            if mime is None:
+                ext = (
+                    obj.name.rsplit(".", 1)[-1]
+                    if ("." in obj.name and not obj.name.startswith("."))
+                    else ""
+                )
+                mime = self._mime_from_extension(ext)
 
-        if mime is None:
-            mime = self._detect_html_xhtml(content)
+        mime = mime or self._detect_html_xhtml(content)
+        mime = mime or "text/plain"
 
         format = MimeTypeToFormat.get(mime)
         return format
 
+    def _mime_from_extension(self, ext):
+        mime = None
+        if ext in FormatToExtensions[InputFormat.ASCIIDOC]:
+            mime = FormatToMimeType[InputFormat.ASCIIDOC][0]
+        elif ext in FormatToExtensions[InputFormat.HTML]:
+            mime = FormatToMimeType[InputFormat.HTML][0]
+        elif ext in FormatToExtensions[InputFormat.MD]:
+            mime = FormatToMimeType[InputFormat.MD][0]
+
+        return mime
+
     def _detect_html_xhtml(self, content):
         content_str = content.decode("ascii", errors="ignore").lower()
         # Remove XML comments

docling/datamodel/pipeline_options.py

@@ -22,6 +22,9 @@ class TableStructureOptions(BaseModel):
 
 class OcrOptions(BaseModel):
     kind: str
+    bitmap_area_threshold: float = (
+        0.05  # percentage of the area for a bitmap to processed with OCR
+    )
 
 
 class EasyOcrOptions(OcrOptions):