yfedoseev/pdf_oxide v0.3.10

v0.3.10 | 21 Issues Resolved — Parallel Extraction, WASM, Batch API, Text Quality, Table Improvements

on GitHub

New Features

• WASM build support (#151) — WebAssembly bindings via wasm-bindgen. New PdfDocument::open_from_bytes() constructor enables browser-side PDF extraction. Internal reader changed from BufReader<File> to BufReader<Cursor<Vec<u8>>> for portability.

• Parallel page extraction (#168) — New parallel feature flag with rayon-based multi-threaded extraction. ParallelExtractor distributes pages across worker threads, each opening its own PdfDocument instance. Global font cache ensures fonts are parsed only once.

• Batch processing API (#167) — New BatchProcessor for multi-PDF workflows with progress callbacks and error collection. Supports both sequential and parallel (with parallel feature) processing. BatchResult and BatchSummary for per-file results and aggregate statistics.

• OCR hybrid detection (#158) — New PageType enum (NativeText, ScannedPage, HybridPage) with multi-heuristic detection: text density, replacement character ratio, image coverage. extract_text_with_ocr() now intelligently selects between native text and OCR output.

• Full WASM/Python API parity (#172, #173) — 10 new method groups across WASM and Python bindings: form field get/set, image bytes extraction, PDF-from-images, form flattening, PDF merging, file embedding, page labels, XMP metadata. Python gains from_image(), from_images(), from_image_bytes() constructors.

Bug Fixes

• Circular XObject segfault (#163) — Fixed segfault from circular Form XObject references (X0→X1→X0) during image extraction. Cycle-detection stack now shared across all recursive Do calls.

• XRef /Prev chain overflow (#170) — XRef /Prev chain parsing rewritten from recursive (depth limit 100) to iterative with HashSet cycle detection. Supports 177+ incremental saves.

• Broken ligature text (#154) — repair_ligatures() post-processor fixes corrupted text from LaTeX PDFs with broken ToUnicode CMaps (fi→#, fl→$, ff→!, ffi→", ffl→%).

• Text extraction quality (#104) — Three sub-categories improved: markup/link/popup annotation text extraction with /RC fallback, leader dot normalization for TOC output, and Priority 3 predefined CMap support for CJK fonts (Japan1, GB1, CNS1, Korea1).

• Reduce .unwrap() usage (#155) — Replaced risky .unwrap() calls in core library paths with safe alternatives (match, expect with invariant comments, proper error propagation).

• Table extraction (#157) — Spatial table detector now detects merged cells (colspan/rowspan) via bounding box analysis, supports multi-line cell content, and uses font property analysis for header row detection.

• Form field persistence in incremental save (#174) — DocumentEditor.save() now correctly persists form field value changes when using incremental save mode.

• JPEG multi-filter extraction — FlateDecode+DCTDecode filter chains now decode correctly, fixing garbled output from PDFs that apply multiple filters to JPEG image streams.

• OCR CTC blank token & recursion — Fixed CTC decoding blank token handling and infinite recursion in OCR V5 strategy.

Code Quality

• 85% test coverage enforced (4,200+ library tests), clippy enforcement with zero warnings, 31 lint fixes across the codebase.
• Renamed version-referenced test files by topic for better discoverability.

Performance

• Image-only page skip (#133, #169) — page_cannot_have_text() pre-check skips content stream decompression for pages with no /Font resources and no Form XObjects.

• SmallVec operator operands (#165) — Stack-allocated SmallVec<[Object; 6]> replaces heap Vec<Object> for operator parsing. All standard PDF operators have ≤6 operands, eliminating per-operator allocation on graphics-heavy pages.

• Cross-document font cache (#166) — Process-level LRU font cache (Layer 6) with 1024-entry default capacity. Fonts parsed by one PdfDocument are available to all subsequent instances via OnceLock<Mutex<LruFontCache>>.

• Large document performance (#132) — Parallel extraction + global font cache + SmallVec + image-only skip combine to bring 1000-page PDFs well under 1 second on multi-core hardware.

Community Contributors

Thanks to everyone who reported issues and contributed to v0.3.10:

• @SeanPedersen — Reported corrupted ligature text in post-processing (#154)
• @dhdaines — Reported circular XObject segfault (#163) and outdated dataset URLs in benchmark docs (#162)
• @bhaswata08 — Reported missing extract_text_ocr in Python bindings (#171)

📥 Installation

Rust (crates.io)

cargo add pdf_oxide

Python (PyPI)

pip install pdf_oxide

JavaScript/WASM (npm)

npm install pdf-oxide-wasm

Pre-built Binaries
Download archives for Linux, macOS, and Windows from the assets below.

Platform Support

Changelog

See CHANGELOG.md for full details.