Raw DOCX

A Python library that extends python-docx to convert Word documents into structured, traversable Python objects with export to dictionary, HTML, and plain text formats.

Installation

pip install raw_docx

Features

• Document hierarchy - Multi-level section numbering that mirrors Word: only headings Word actually numbers get a number; unnumbered headings (front matter, bullets, "no number" styles) are kept as sections without one
• Rich text - Colors, highlighting, bold, italic, superscript, and subscript
• Tables - Full support for merged cells (row and column spans)
• Nested lists - Arbitrary nesting depth with level tracking and numId boundary detection
• Indentation hierarchy - Infers nesting from indentation when all items share the same level
• Bookmarks and cross-references - Bookmark anchors and field-based references
• Image extraction - Extracts embedded images with base64 HTML embedding
• Multiple export formats - Dictionary, HTML, and plain text
• Search - Find text across sections, tables, and the full document
• Error tracking - Integrated logging via simple_error_log

Quick Start

from raw_docx import RawDocx

# Load and process a document
docx = RawDocx("path/to/document.docx", work_dir="/tmp/output")

# Disable indentation-based hierarchy inference if needed
# docx = RawDocx("path/to/document.docx", infer_indent_hierarchy=False)

# Access the structured document
document = docx.target_document

# Export to dictionary
data = docx.to_dict()

# Work with sections
section = document.section_by_title("Introduction")
paragraphs = section.paragraphs()
tables = section.tables()
lists = section.lists()

# Search for content
results = section.find("keyword")

# Generate HTML
html = section.to_html()

Key Classes

Requirements

• Python >= 3.12
• python-docx >= 1.1.2
• simple_error_log >= 0.6.0

License

MIT - see LICENSE for details.

Build and Release

pytest
ruff format
ruff check
python3 -m build --sdist --wheel
twine upload dist/*