ripdoc 0.2.1

Rust-native PDF extraction. 73x faster than pdfplumber.

ripdoc

Rust-native PDF extraction. 73x faster than pdfplumber.

ripdoc is a drop-in replacement for pdfplumber built entirely in Rust with Python bindings via PyO3. Extract text, tables, words, and more from PDFs at speeds that feel instant.

$ python bench.py report.pdf (200 pages)

ripdoc       0.16s  ██
pymupdf      5.97s  ████████████████████████████████████████
pdfplumber  11.63s  █████████████████████████████████████████████████████████████████████████
pdfminer    15.55s  ██████████████████████████████████████████████████████████████████████████████████████████████████

Install

pip install ripdoc

Requires Python 3.8+. Pre-built wheels for macOS (arm64). Other platforms build from source (requires Rust toolchain).

Quick start

import ripdoc

pdf = ripdoc.open("report.pdf")

for page in pdf.pages:
    # Extract text
    text = page.extract_text()

    # Extract with layout preservation
    text = page.extract_text(layout=True)

    # Extract words with bounding boxes
    words = page.extract_words()

    # Extract tables
    tables = page.extract_tables()

    # Search for text
    results = page.search("revenue")

Drop-in pdfplumber replacement

Swap one import — everything else stays the same:

# Before
import pdfplumber
pdf = pdfplumber.open("report.pdf")

# After
import ripdoc as pdfplumber
pdf = pdfplumber.open("report.pdf")

Or use the explicit compat module:

import ripdoc.compat as pdfplumber

API

ripdoc.open(path) -> PDF

Open a PDF file. Also supports PDF.from_bytes(bytes).

PDF

Property / Method	Description
pdf.pages	List of Page objects
pdf.page_count	Number of pages
pdf.metadata	Document metadata dict
pdf.page(n)	Get page by number (1-indexed)

Page

Property / Method	Description
page.extract_text(layout=False)	Extract text, optionally preserving spatial layout
page.extract_words()	Words with bounding boxes (x0, top, x1, bottom)
page.extract_tables()	Tables as list of row lists
page.extract_table()	Largest table on the page
page.find_tables()	Table objects with metadata
page.search(query)	Find text matches with positions
page.chars	Individual characters with font info
page.lines	Line segments
page.rects	Rectangles
page.edges	Edges (used for table detection)
page.crop(bbox)	Crop to bounding box (x0, top, x1, bottom)
page.within_bbox(bbox)	Filter objects within bounding box
page.width / page.height	Page dimensions in points
page.page_number	1-indexed page number

Architecture

ripdoc
├── ripdoc-core     Pure Rust library (~5500 LOC)
│   ├── content_stream    PDF operator interpreter
│   ├── fonts/            Encoding, CMap, metrics
│   ├── geometry/         BBox, CTM, clustering
│   ├── text/             Word grouping, layout, search
│   ├── table/            Nurminen/Tabula algorithm
│   └── output/           Markdown, JSON, HTML, CSV
└── ripdoc-python   PyO3 bindings (~450 LOC)

Built on lopdf for low-level PDF structure parsing. All text extraction, table detection, and layout analysis is implemented from scratch in Rust.

Features

• Text extraction — simple and layout-preserving modes
• Table detection — Nurminen/Tabula algorithm with merged cell support
• Search — full-text search with bounding box positions
• Reading order — XY-cut algorithm + tagged PDF structure tree
• Output formats — Markdown, JSON, HTML, CSV
• Spatial queries — crop, within_bbox, character-level access
• pdfplumber compatible — same API, same patterns

Development

# Build and install locally
cd crates/ripdoc-python
maturin develop --release

# Run tests
cargo test

# Type check the visualizer frontend
cd visualizer/frontend && npx tsc --noEmit

License

MIT OR Apache-2.0