fix(docx): merged table cells not properly converted (#857)

* fix(docx): merged cells not properly converted

Fix conversion issue of merged cells in Word tables leading to repeated text.
Simplify Word table conversion code.
Add docx file with several table formats for regression tests.

Signed-off-by: Cesar Berrospi Ramis <75900930+ceberam@users.noreply.github.com>

* chore: add type hinting to docx backend

Signed-off-by: Cesar Berrospi Ramis <75900930+ceberam@users.noreply.github.com>

---------

Signed-off-by: Cesar Berrospi Ramis <75900930+ceberam@users.noreply.github.com>

docling/backend/msword_backend.py

@@ -2,21 +2,28 @@ import logging
 import re
 from io import BytesIO
 from pathlib import Path
-from typing import Set, Union
+from typing import Any, Optional, Union
 
-import docx
 from docling_core.types.doc import (
     DocItemLabel,
     DoclingDocument,
     DocumentOrigin,
     GroupLabel,
     ImageRef,
+    NodeItem,
     TableCell,
     TableData,
 )
+from docx import Document
+from docx.document import Document as DocxDocument
+from docx.oxml.table import CT_Tc
+from docx.oxml.xmlchemy import BaseOxmlElement
+from docx.table import Table, _Cell
+from docx.text.paragraph import Paragraph
 from lxml import etree
 from lxml.etree import XPath
 from PIL import Image, UnidentifiedImageError
+from typing_extensions import override
 
 from docling.backend.abstract_backend import DeclarativeDocumentBackend
 from docling.datamodel.base_models import InputFormat
@@ -26,7 +33,10 @@ _log = logging.getLogger(__name__)
 
 
 class MsWordDocumentBackend(DeclarativeDocumentBackend):
-    def __init__(self, in_doc: "InputDocument", path_or_stream: Union[BytesIO, Path]):
+    @override
+    def __init__(
+        self, in_doc: "InputDocument", path_or_stream: Union[BytesIO, Path]
+    ) -> None:
         super().__init__(in_doc, path_or_stream)
         self.XML_KEY = (
             "{http://schemas.openxmlformats.org/wordprocessingml/2006/main}val"
@@ -36,19 +46,19 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
         }
         # self.initialise(path_or_stream)
         # Word file:
-        self.path_or_stream = path_or_stream
+        self.path_or_stream: Union[BytesIO, Path] = path_or_stream
-        self.valid = False
+        self.valid: bool = False
         # Initialise the parents for the hierarchy
-        self.max_levels = 10
+        self.max_levels: int = 10
-        self.level_at_new_list = None
+        self.level_at_new_list: Optional[int] = None
-        self.parents = {}  # type: ignore
+        self.parents: dict[int, Optional[NodeItem]] = {}
         for i in range(-1, self.max_levels):
             self.parents[i] = None
 
         self.level = 0
         self.listIter = 0
 
-        self.history = {
+        self.history: dict[str, Any] = {
             "names": [None],
             "levels": [None],
             "numids": [None],
@@ -58,9 +68,9 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
         self.docx_obj = None
         try:
             if isinstance(self.path_or_stream, BytesIO):
-                self.docx_obj = docx.Document(self.path_or_stream)
+                self.docx_obj = Document(self.path_or_stream)
             elif isinstance(self.path_or_stream, Path):
-                self.docx_obj = docx.Document(str(self.path_or_stream))
+                self.docx_obj = Document(str(self.path_or_stream))
 
             self.valid = True
         except Exception as e:
@@ -68,13 +78,16 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
                 f"MsPowerpointDocumentBackend could not load document with hash {self.document_hash}"
             ) from e
 
+    @override
     def is_valid(self) -> bool:
         return self.valid
 
     @classmethod
+    @override
     def supports_pagination(cls) -> bool:
         return False
 
+    @override
     def unload(self):
         if isinstance(self.path_or_stream, BytesIO):
             self.path_or_stream.close()
@@ -82,11 +95,17 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
         self.path_or_stream = None
 
     @classmethod
-    def supported_formats(cls) -> Set[InputFormat]:
+    @override
+    def supported_formats(cls) -> set[InputFormat]:
         return {InputFormat.DOCX}
 
+    @override
     def convert(self) -> DoclingDocument:
-        # Parses the DOCX into a structured document model.
+        """Parses the DOCX into a structured document model.
+
+        Returns:
+            The parsed document.
+        """
 
         origin = DocumentOrigin(
             filename=self.file.name or "file",
@@ -104,23 +123,29 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
                 f"Cannot convert doc with {self.document_hash} because the backend failed to init."
             )
 
-    def update_history(self, name, level, numid, ilevel):
+    def update_history(
+        self,
+        name: str,
+        level: Optional[int],
+        numid: Optional[int],
+        ilevel: Optional[int],
+    ):
         self.history["names"].append(name)
         self.history["levels"].append(level)
 
         self.history["numids"].append(numid)
         self.history["indents"].append(ilevel)
 
-    def prev_name(self):
+    def prev_name(self) -> Optional[str]:
         return self.history["names"][-1]
 
-    def prev_level(self):
+    def prev_level(self) -> Optional[int]:
         return self.history["levels"][-1]
 
-    def prev_numid(self):
+    def prev_numid(self) -> Optional[int]:
         return self.history["numids"][-1]
 
-    def prev_indent(self):
+    def prev_indent(self) -> Optional[int]:
         return self.history["indents"][-1]
 
     def get_level(self) -> int:
@@ -130,7 +155,12 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
                 return k
         return 0
 
-    def walk_linear(self, body, docx_obj, doc) -> DoclingDocument:
+    def walk_linear(
+        self,
+        body: BaseOxmlElement,
+        docx_obj: DocxDocument,
+        doc: DoclingDocument,
+    ) -> DoclingDocument:
         for element in body:
             tag_name = etree.QName(element).localname
             # Check for Inline Images (blip elements)
@@ -150,7 +180,7 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
                     _log.debug("could not parse a table, broken docx table")
 
             elif drawing_blip:
-                self.handle_pictures(element, docx_obj, drawing_blip, doc)
+                self.handle_pictures(docx_obj, drawing_blip, doc)
             # Check for the sdt containers, like table of contents
             elif tag_name in ["sdt"]:
                 sdt_content = element.find(".//w:sdtContent", namespaces=namespaces)
@@ -167,7 +197,7 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
                 _log.debug(f"Ignoring element in DOCX with tag: {tag_name}")
         return doc
 
-    def str_to_int(self, s, default=0):
+    def str_to_int(self, s: Optional[str], default: Optional[int] = 0) -> Optional[int]:
         if s is None:
             return None
         try:
@@ -175,7 +205,7 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
         except ValueError:
             return default
 
-    def split_text_and_number(self, input_string):
+    def split_text_and_number(self, input_string: str) -> list[str]:
         match = re.match(r"(\D+)(\d+)$|^(\d+)(\D+)", input_string)
         if match:
             parts = list(filter(None, match.groups()))
@@ -183,7 +213,9 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
         else:
             return [input_string]
 
-    def get_numId_and_ilvl(self, paragraph):
+    def get_numId_and_ilvl(
+        self, paragraph: Paragraph
+    ) -> tuple[Optional[int], Optional[int]]:
         # Access the XML element of the paragraph
         numPr = paragraph._element.find(
             ".//w:numPr", namespaces=paragraph._element.nsmap
@@ -196,13 +228,11 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
             numId = numId_elem.get(self.XML_KEY) if numId_elem is not None else None
             ilvl = ilvl_elem.get(self.XML_KEY) if ilvl_elem is not None else None
 
-            return self.str_to_int(numId, default=None), self.str_to_int(
+            return self.str_to_int(numId, None), self.str_to_int(ilvl, None)
-                ilvl, default=None
-            )
 
         return None, None  # If the paragraph is not part of a list
 
-    def get_label_and_level(self, paragraph):
+    def get_label_and_level(self, paragraph: Paragraph) -> tuple[str, Optional[int]]:
         if paragraph.style is None:
             return "Normal", None
         label = paragraph.style.style_id
@@ -218,20 +248,25 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
 
         if "Heading" in label and len(parts) == 2:
             parts.sort()
-            label_str = ""
+            label_str: str = ""
-            label_level = 0
+            label_level: Optional[int] = 0
             if parts[0] == "Heading":
                 label_str = parts[0]
-                label_level = self.str_to_int(parts[1], default=None)
+                label_level = self.str_to_int(parts[1], None)
             if parts[1] == "Heading":
                 label_str = parts[1]
-                label_level = self.str_to_int(parts[0], default=None)
+                label_level = self.str_to_int(parts[0], None)
             return label_str, label_level
         else:
             return label, None
 
-    def handle_text_elements(self, element, docx_obj, doc):
+    def handle_text_elements(
-        paragraph = docx.text.paragraph.Paragraph(element, docx_obj)
+        self,
+        element: BaseOxmlElement,
+        docx_obj: DocxDocument,
+        doc: DoclingDocument,
+    ) -> None:
+        paragraph = Paragraph(element, docx_obj)
 
         if paragraph.text is None:
             return
@@ -255,11 +290,7 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
             and p_style_id not in ["Title", "Heading"]
         ):
             self.add_listitem(
-                element,
-                docx_obj,
                 doc,
-                p_style_id,
-                p_level,
                 numid,
                 ilevel,
                 text,
@@ -284,13 +315,13 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
                 self.level = 0
 
         if p_style_id in ["Title"]:
-            for key, val in self.parents.items():
+            for key in range(len(self.parents)):
                 self.parents[key] = None
             self.parents[0] = doc.add_text(
                 parent=None, label=DocItemLabel.TITLE, text=text
             )
         elif "Heading" in p_style_id:
-            self.add_header(element, docx_obj, doc, p_style_id, p_level, text)
+            self.add_header(doc, p_level, text)
 
         elif p_style_id in [
             "Paragraph",
@@ -318,7 +349,9 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
         self.update_history(p_style_id, p_level, numid, ilevel)
         return
 
-    def add_header(self, element, docx_obj, doc, curr_name, curr_level, text: str):
+    def add_header(
+        self, doc: DoclingDocument, curr_level: Optional[int], text: str
+    ) -> None:
         level = self.get_level()
         if isinstance(curr_level, int):
             if curr_level > level:
@@ -331,7 +364,7 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
                     )
             elif curr_level < level:
                 # remove the tail
-                for key, val in self.parents.items():
+                for key in range(len(self.parents)):
                     if key >= curr_level:
                         self.parents[key] = None
 
@@ -350,22 +383,18 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
 
     def add_listitem(
         self,
-        element,
+        doc: DoclingDocument,
-        docx_obj,
+        numid: int,
-        doc,
+        ilevel: int,
-        p_style_id,
-        p_level,
-        numid,
-        ilevel,
         text: str,
-        is_numbered=False,
+        is_numbered: bool = False,
-    ):
+    ) -> None:
-        # is_numbered = is_numbered
         enum_marker = ""
 
         level = self.get_level()
+        prev_indent = self.prev_indent()
         if self.prev_numid() is None:  # Open new list
-            self.level_at_new_list = level  # type: ignore
+            self.level_at_new_list = level
 
             self.parents[level] = doc.add_group(
                 label=GroupLabel.LIST, name="list", parent=self.parents[level - 1]
@@ -384,10 +413,13 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
             )
 
         elif (
-            self.prev_numid() == numid and self.prev_indent() < ilevel
+            self.prev_numid() == numid
+            and self.level_at_new_list is not None
+            and prev_indent is not None
+            and prev_indent < ilevel
         ):  # Open indented list
             for i in range(
-                self.level_at_new_list + self.prev_indent() + 1,
+                self.level_at_new_list + prev_indent + 1,
                 self.level_at_new_list + ilevel + 1,
             ):
                 # Determine if this is an unordered list or an ordered list.
@@ -416,7 +448,12 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
                 text=text,
             )
 
-        elif self.prev_numid() == numid and ilevel < self.prev_indent():  # Close list
+        elif (
+            self.prev_numid() == numid
+            and self.level_at_new_list is not None
+            and prev_indent is not None
+            and ilevel < prev_indent
+        ):  # Close list
             for k, v in self.parents.items():
                 if k > self.level_at_new_list + ilevel:
                     self.parents[k] = None
@@ -434,7 +471,7 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
             )
             self.listIter = 0
 
-        elif self.prev_numid() == numid or self.prev_indent() == ilevel:
+        elif self.prev_numid() == numid or prev_indent == ilevel:
             # TODO: Set marker and enumerated arguments if this is an enumeration element.
             self.listIter += 1
             if is_numbered:
@@ -448,31 +485,16 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
             )
         return
 
-    def handle_tables(self, element, docx_obj, doc):
+    def handle_tables(
-
+        self,
-        # Function to check if a cell has a colspan (gridSpan)
+        element: BaseOxmlElement,
-        def get_colspan(cell):
+        docx_obj: DocxDocument,
-            grid_span = cell._element.xpath("@w:gridSpan")
+        doc: DoclingDocument,
-            if grid_span:
+    ) -> None:
-                return int(grid_span[0])  # Return the number of columns spanned
+        table: Table = Table(element, docx_obj)
-            return 1  # Default is 1 (no colspan)
-
-        # Function to check if a cell has a rowspan (vMerge)
-        def get_rowspan(cell):
-            v_merge = cell._element.xpath("@w:vMerge")
-            if v_merge:
-                return v_merge[
-                    0
-                ]  # 'restart' indicates the beginning of a rowspan, others are continuation
-            return 1
-
-        table = docx.table.Table(element, docx_obj)
-
         num_rows = len(table.rows)
-        num_cols = 0
+        num_cols = len(table.columns)
-        for row in table.rows:
+        _log.debug(f"Table grid with {num_rows} rows and {num_cols} columns")
-            # Calculate the max number of columns
-            num_cols = max(num_cols, sum(get_colspan(cell) for cell in row.cells))
 
         if num_rows == 1 and num_cols == 1:
             cell_element = table.rows[0].cells[0]
@@ -481,59 +503,56 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
             self.walk_linear(cell_element._element, docx_obj, doc)
             return
 
-        # Initialize the table grid
+        data = TableData(num_rows=num_rows, num_cols=num_cols)
-        table_grid = [[None for _ in range(num_cols)] for _ in range(num_rows)]
+        cell_set: set[CT_Tc] = set()
-
-        data = TableData(num_rows=num_rows, num_cols=num_cols, table_cells=[])
-
         for row_idx, row in enumerate(table.rows):
+            _log.debug(f"Row index {row_idx} with {len(row.cells)} populated cells")
             col_idx = 0
-            for c, cell in enumerate(row.cells):
+            while col_idx < num_cols:
-                row_span = get_rowspan(cell)
+                cell: _Cell = row.cells[col_idx]
-                col_span = get_colspan(cell)
+                _log.debug(
+                    f" col {col_idx} grid_span {cell.grid_span} grid_cols_before {row.grid_cols_before}"
+                )
+                if cell is None or cell._tc in cell_set:
+                    _log.debug(f"  skipped since repeated content")
+                    col_idx += cell.grid_span
+                    continue
+                else:
+                    cell_set.add(cell._tc)
 
-                cell_text = cell.text
+                spanned_idx = row_idx
-                # In case cell doesn't return text via docx library:
+                spanned_tc: Optional[CT_Tc] = cell._tc
-                if len(cell_text) == 0:
+                while spanned_tc == cell._tc:
-                    cell_xml = cell._element
+                    spanned_idx += 1
+                    spanned_tc = (
+                        table.rows[spanned_idx].cells[col_idx]._tc
+                        if spanned_idx < num_rows
+                        else None
+                    )
+                _log.debug(f"  spanned before row {spanned_idx}")
 
-                    texts = [""]
+                table_cell = TableCell(
-                    for elem in cell_xml.iter():
+                    text=cell.text,
-                        if elem.tag.endswith("t"):  # <w:t> tags that contain text
+                    row_span=spanned_idx - row_idx,
-                            if elem.text:
+                    col_span=cell.grid_span,
-                                texts.append(elem.text)
+                    start_row_offset_idx=row.grid_cols_before + row_idx,
-                    # Join the collected text
+                    end_row_offset_idx=row.grid_cols_before + spanned_idx,
-                    cell_text = " ".join(texts).strip()
-
-                # Find the next available column in the grid
-                while table_grid[row_idx][col_idx] is not None:
-                    col_idx += 1
-
-                # Fill the grid with the cell value, considering rowspan and colspan
-                for i in range(row_span if row_span == "restart" else 1):
-                    for j in range(col_span):
-                        table_grid[row_idx + i][col_idx + j] = ""
-
-                cell = TableCell(
-                    text=cell_text,
-                    row_span=row_span,
-                    col_span=col_span,
-                    start_row_offset_idx=row_idx,
-                    end_row_offset_idx=row_idx + row_span,
                     start_col_offset_idx=col_idx,
-                    end_col_offset_idx=col_idx + col_span,
+                    end_col_offset_idx=col_idx + cell.grid_span,
                     col_header=False,
                     row_header=False,
                 )
-
+                data.table_cells.append(table_cell)
-                data.table_cells.append(cell)
+                col_idx += cell.grid_span
 
         level = self.get_level()
         doc.add_table(data=data, parent=self.parents[level - 1])
         return
 
-    def handle_pictures(self, element, docx_obj, drawing_blip, doc):
+    def handle_pictures(
-        def get_docx_image(element, drawing_blip):
+        self, docx_obj: DocxDocument, drawing_blip: Any, doc: DoclingDocument
+    ) -> None:
+        def get_docx_image(drawing_blip):
             rId = drawing_blip[0].get(
                 "{http://schemas.openxmlformats.org/officeDocument/2006/relationships}embed"
             )
@@ -546,7 +565,7 @@ class MsWordDocumentBackend(DeclarativeDocumentBackend):
         level = self.get_level()
         # Open the BytesIO object with PIL to create an Image
         try:
-            image_data = get_docx_image(element, drawing_blip)
+            image_data = get_docx_image(drawing_blip)
             image_bytes = BytesIO(image_data)
             pil_image = Image.open(image_bytes)
             doc.add_picture(

tests/data/groundtruth/docling_v2/word_tables.docx.html

@@ -0,0 +1,75 @@
+<!DOCTYPE html>
+<html lang="en">
+<head>
+    <link rel="icon" type="image/png"
+    href="https://ds4sd.github.io/docling/assets/logo.png"/>
+    <meta charset="UTF-8">
+    <title>
+    Powered by Docling
+    </title>
+    <style>
+    html {
+    background-color: LightGray;
+    }
+    body {
+    margin: 0 auto;
+    width:800px;
+    padding: 30px;
+    background-color: White;
+    font-family: Arial, sans-serif;
+    box-shadow: 10px 10px 10px grey;
+    }
+    figure{
+    display: block;
+    width: 100%;
+    margin: 0px;
+    margin-top: 10px;
+    margin-bottom: 10px;
+    }
+    img {
+    display: block;
+    margin: auto;
+    margin-top: 10px;
+    margin-bottom: 10px;
+    max-width: 640px;
+    max-height: 640px;
+    }
+    table {
+    min-width:500px;
+    background-color: White;
+    border-collapse: collapse;
+    cell-padding: 5px;
+    margin: auto;
+    margin-top: 10px;
+    margin-bottom: 10px;
+    }
+    th, td {
+    border: 1px solid black;
+    padding: 8px;
+    }
+    th {
+    font-weight: bold;
+    }
+    table tr:nth-child(even) td{
+    background-color: LightGray;
+    }
+    </style>
+    </head>
+<h2>Test with tables</h2>
+<p>A uniform table</p>
+<table><tbody><tr><td>Header 0.0</td><td>Header 0.1</td><td>Header 0.2</td></tr><tr><td>Cell 1.0</td><td>Cell 1.1</td><td>Cell 1.2</td></tr><tr><td>Cell 2.0</td><td>Cell 2.1</td><td>Cell 2.2</td></tr></tbody></table>
+<p></p>
+<p>A non-uniform table with horizontal spans</p>
+<table><tbody><tr><td>Header 0.0</td><td>Header 0.1</td><td>Header 0.2</td></tr><tr><td>Cell 1.0</td><td colspan="2">Merged Cell 1.1 1.2</td></tr><tr><td>Cell 2.0</td><td colspan="2">Merged Cell 2.1 2.2</td></tr></tbody></table>
+<p></p>
+<p>A non-uniform table with horizontal spans in inner columns</p>
+<table><tbody><tr><td>Header 0.0</td><td>Header 0.1</td><td>Header 0.2</td><td>Header 0.3</td></tr><tr><td>Cell 1.0</td><td colspan="2">Merged Cell 1.1 1.2</td><td>Cell 1.3</td></tr><tr><td>Cell 2.0</td><td colspan="2">Merged Cell 2.1 2.2</td><td>Cell 2.3</td></tr></tbody></table>
+<p></p>
+<p>A non-uniform table with vertical spans</p>
+<table><tbody><tr><td>Header 0.0</td><td>Header 0.1</td><td>Header 0.2</td></tr><tr><td>Cell 1.0</td><td rowspan="2">Merged Cell 1.1 2.1</td><td>Cell 1.2</td></tr><tr><td>Cell 2.0</td><td>Cell 2.2</td></tr><tr><td>Cell 3.0</td><td rowspan="2">Merged Cell 3.1 4.1</td><td>Cell 3.2</td></tr><tr><td>Cell 4.0</td><td>Cell 4.2</td></tr></tbody></table>
+<p></p>
+<p>A non-uniform table with all kinds of spans and empty cells</p>
+<table><tbody><tr><td>Header 0.0</td><td>Header 0.1</td><td>Header 0.2</td><td></td><td></td></tr><tr><td>Cell 1.0</td><td rowspan="2">Merged Cell 1.1 2.1</td><td>Cell 1.2</td><td></td><td></td></tr><tr><td>Cell 2.0</td><td>Cell 2.2</td><td></td><td></td></tr><tr><td>Cell 3.0</td><td rowspan="2">Merged Cell 3.1 4.1</td><td>Cell 3.2</td><td rowspan="3"></td><td></td></tr><tr><td>Cell 4.0</td><td>Cell 4.2</td><td rowspan="2">Merged Cell 4.4 5.4</td></tr><tr><td></td><td></td><td></td></tr><tr><td></td><td></td><td></td><td></td><td></td></tr><tr><td colspan="5"></td></tr><tr><td></td><td></td><td></td><td></td><td>Cell 8.4</td></tr></tbody></table>
+<p></p>
+<p></p>
+</html>

tests/data/groundtruth/docling_v2/word_tables.docx.itxt

@@ -0,0 +1,19 @@
+item-0 at level 0: unspecified: group _root_
+  item-1 at level 1: section: group header-0
+    item-2 at level 2: section_header: Test with tables
+      item-3 at level 3: paragraph: A uniform table
+      item-4 at level 3: table with [3x3]
+      item-5 at level 3: paragraph: 
+      item-6 at level 3: paragraph: A non-uniform table with horizontal spans
+      item-7 at level 3: table with [3x3]
+      item-8 at level 3: paragraph: 
+      item-9 at level 3: paragraph: A non-uniform table with horizontal spans in inner columns
+      item-10 at level 3: table with [3x4]
+      item-11 at level 3: paragraph: 
+      item-12 at level 3: paragraph: A non-uniform table with vertical spans
+      item-13 at level 3: table with [5x3]
+      item-14 at level 3: paragraph: 
+      item-15 at level 3: paragraph: A non-uniform table with all kinds of spans and empty cells
+      item-16 at level 3: table with [9x5]
+      item-17 at level 3: paragraph: 
+      item-18 at level 3: paragraph: 

tests/data/groundtruth/docling_v2/word_tables.docx.md

@@ -0,0 +1,44 @@
+## Test with tables
+
+A uniform table
+
+| Header 0.0   | Header 0.1   | Header 0.2   |
+|--------------|--------------|--------------|
+| Cell 1.0     | Cell 1.1     | Cell 1.2     |
+| Cell 2.0     | Cell 2.1     | Cell 2.2     |
+
+A non-uniform table with horizontal spans
+
+| Header 0.0   | Header 0.1          | Header 0.2          |
+|--------------|---------------------|---------------------|
+| Cell 1.0     | Merged Cell 1.1 1.2 | Merged Cell 1.1 1.2 |
+| Cell 2.0     | Merged Cell 2.1 2.2 | Merged Cell 2.1 2.2 |
+
+A non-uniform table with horizontal spans in inner columns
+
+| Header 0.0   | Header 0.1          | Header 0.2          | Header 0.3   |
+|--------------|---------------------|---------------------|--------------|
+| Cell 1.0     | Merged Cell 1.1 1.2 | Merged Cell 1.1 1.2 | Cell 1.3     |
+| Cell 2.0     | Merged Cell 2.1 2.2 | Merged Cell 2.1 2.2 | Cell 2.3     |
+
+A non-uniform table with vertical spans
+
+| Header 0.0   | Header 0.1          | Header 0.2   |
+|--------------|---------------------|--------------|
+| Cell 1.0     | Merged Cell 1.1 2.1 | Cell 1.2     |
+| Cell 2.0     | Merged Cell 1.1 2.1 | Cell 2.2     |
+| Cell 3.0     | Merged Cell 3.1 4.1 | Cell 3.2     |
+| Cell 4.0     | Merged Cell 3.1 4.1 | Cell 4.2     |
+
+A non-uniform table with all kinds of spans and empty cells
+
+| Header 0.0   | Header 0.1          | Header 0.2   |    |                     |
+|--------------|---------------------|--------------|----|---------------------|
+| Cell 1.0     | Merged Cell 1.1 2.1 | Cell 1.2     |    |                     |
+| Cell 2.0     | Merged Cell 1.1 2.1 | Cell 2.2     |    |                     |
+| Cell 3.0     | Merged Cell 3.1 4.1 | Cell 3.2     |    |                     |
+| Cell 4.0     | Merged Cell 3.1 4.1 | Cell 4.2     |    | Merged Cell 4.4 5.4 |
+|              |                     |              |    | Merged Cell 4.4 5.4 |
+|              |                     |              |    |                     |
+|              |                     |              |    |                     |
+|              |                     |              |    | Cell 8.4            |

tests/test_backend_msword.py

@@ -69,7 +69,6 @@ def verify_export(pred_text: str, gtfile: str):
         with open(gtfile, "r") as fr:
             true_text = fr.read()
 
-        assert pred_text == true_text, "pred_itxt==true_itxt"
         return pred_text == true_text
 
 
@@ -101,3 +100,7 @@ def test_e2e_docx_conversions():
 
         pred_json: str = json.dumps(doc.export_to_dict(), indent=2)
         assert verify_export(pred_json, str(gt_path) + ".json"), "export to json"
+
+        if docx_path.name == "word_tables.docx":
+            pred_html: str = doc.export_to_html()
+            assert verify_export(pred_html, str(gt_path) + ".html"), "export to html"

word_tables.html

@@ -0,0 +1,75 @@
+<!DOCTYPE html>
+<html lang="en">
+<head>
+    <link rel="icon" type="image/png"
+    href="https://ds4sd.github.io/docling/assets/logo.png"/>
+    <meta charset="UTF-8">
+    <title>
+    Powered by Docling
+    </title>
+    <style>
+    html {
+    background-color: LightGray;
+    }
+    body {
+    margin: 0 auto;
+    width:800px;
+    padding: 30px;
+    background-color: White;
+    font-family: Arial, sans-serif;
+    box-shadow: 10px 10px 10px grey;
+    }
+    figure{
+    display: block;
+    width: 100%;
+    margin: 0px;
+    margin-top: 10px;
+    margin-bottom: 10px;
+    }
+    img {
+    display: block;
+    margin: auto;
+    margin-top: 10px;
+    margin-bottom: 10px;
+    max-width: 640px;
+    max-height: 640px;
+    }
+    table {
+    min-width:500px;
+    background-color: White;
+    border-collapse: collapse;
+    cell-padding: 5px;
+    margin: auto;
+    margin-top: 10px;
+    margin-bottom: 10px;
+    }
+    th, td {
+    border: 1px solid black;
+    padding: 8px;
+    }
+    th {
+    font-weight: bold;
+    }
+    table tr:nth-child(even) td{
+    background-color: LightGray;
+    }
+    </style>
+    </head>
+<h2>Test with tables</h2>
+<p>A uniform table</p>
+<table><tbody><tr><td>Header 0.0</td><td>Header 0.1</td><td>Header 0.2</td></tr><tr><td>Cell 1.0</td><td>Cell 1.1</td><td>Cell 1.2</td></tr><tr><td>Cell 2.0</td><td>Cell 2.1</td><td>Cell 2.2</td></tr></tbody></table>
+<p></p>
+<p>A non-uniform table with horizontal spans</p>
+<table><tbody><tr><td>Header 0.0</td><td>Header 0.1</td><td>Header 0.2</td></tr><tr><td>Cell 1.0</td><td colspan="2">Merged Cell 1.1 1.2</td></tr><tr><td>Cell 2.0</td><td colspan="2">Merged Cell 2.1 2.2</td></tr></tbody></table>
+<p></p>
+<p>A non-uniform table with horizontal spans in inner columns</p>
+<table><tbody><tr><td>Header 0.0</td><td>Header 0.1</td><td>Header 0.2</td><td>Header 0.3</td></tr><tr><td>Cell 1.0</td><td colspan="2">Merged Cell 1.1 1.2</td><td>Cell 1.3</td></tr><tr><td>Cell 2.0</td><td colspan="2">Merged Cell 2.1 2.2</td><td>Cell 2.3</td></tr></tbody></table>
+<p></p>
+<p>A non-uniform table with vertical spans</p>
+<table><tbody><tr><td>Header 0.0</td><td>Header 0.1</td><td>Header 0.2</td></tr><tr><td>Cell 1.0</td><td rowspan="2">Merged Cell 1.1 2.1</td><td>Cell 1.2</td></tr><tr><td>Cell 2.0</td><td>Cell 2.2</td></tr><tr><td>Cell 3.0</td><td rowspan="2">Merged Cell 3.1 4.1</td><td>Cell 3.2</td></tr><tr><td>Cell 4.0</td><td>Cell 4.2</td></tr></tbody></table>
+<p></p>
+<p>A non-uniform table with all kinds of spans and empty cells</p>
+<table><tbody><tr><td>Header 0.0</td><td>Header 0.1</td><td>Header 0.2</td><td></td><td></td></tr><tr><td>Cell 1.0</td><td rowspan="2">Merged Cell 1.1 2.1</td><td>Cell 1.2</td><td></td><td></td></tr><tr><td>Cell 2.0</td><td>Cell 2.2</td><td></td><td></td></tr><tr><td>Cell 3.0</td><td rowspan="2">Merged Cell 3.1 4.1</td><td>Cell 3.2</td><td rowspan="3"></td><td></td></tr><tr><td>Cell 4.0</td><td>Cell 4.2</td><td rowspan="2">Merged Cell 4.4 5.4</td></tr><tr><td></td><td></td><td></td></tr><tr><td></td><td></td><td></td><td></td><td></td></tr><tr><td colspan="5"></td></tr><tr><td></td><td></td><td></td><td></td><td>Cell 8.4</td></tr></tbody></table>
+<p></p>
+<p></p>
+</html>