PaperForge

A comprehensive, pure-Rust PDF generation, editing, inspection, rendering, and document-processing engine.

Overview

PaperForge is a native Rust alternative to Python ReportLab and fragmented Rust PDF tooling. One coherent API for generating, editing, parsing, rendering, and extracting PDF documents.

Quick Start

Add to your Cargo.toml:

[dependencies]
paperforge = "0.1"

Generate a PDF:

use paperforge::prelude::*;

fn main() -> Result<(), PdfError> {
    let mut doc = Document::new();

    let page = doc.add_page(PageSize::A4);

    page.text()
        .at(50.mm(), 250.mm())
        .font_size(24.0)
        .write("Hello, Rust!");

    doc.save("hello.pdf")?;
    Ok(())
}

Features

Architecture

paperforge/
├── paperforge/          # Main public crate
├── paperforge-core/     # PDF-independent abstractions
├── paperforge-pdf/      # PDF serialization/parsing
├── paperforge-layout/   # High-level document layout
├── paperforge-fonts/    # Font handling
├── paperforge-images/   # Image support
├── paperforge-render/   # PDF rendering
├── paperforge-forms/    # AcroForm support
├── paperforge-annotations/ # Annotations
├── paperforge-security/ # Encryption/security
├── paperforge-extract/  # Text/image extraction
└── paperforge-cli/      # CLI tool

Benchmarks

Measured with criterion (30 samples, median) on an AMD Ryzen 5 7235HS / Windows 11 (2026-08-12). Workload: a 100-page A4 text document — one heading and one body line per page — built with each library's high-level (or object-level, for lopdf) API and then parsed from the same in-memory bytes. Every library builds the identical workload. Reproduce with cargo bench -p paperforge-bench.

¹ lopdf writes uncompressed streams by default; both columns use each library's default settings.

Notes:

• Generation beats printpdf by ~2.8×. paperforge (default: flate-compressed content, level 6) generates the 100-page document in 1.30 ms — 5× faster than the previous release (5.97 ms) and ~2.8× faster than printpdf (3.66 ms). lopdf remains the fastest because its default output skips compression entirely and is object-level; with compression disabled paperforge (363 µs) is within ~20% of lopdf (306 µs) on the same uncompressed workload.
• Parsing beats lopdf by ~33% — 254 µs vs 377 µs on the same compressed 100-page document.
• Output is 1.7× smaller than printpdf (35.9 KB vs 60.9 KB). paperforge's output is also 25% smaller than the previous release (47.7 KB) thanks to a single shared /Resources object referenced by every page. lopdf's 27.4 KB is uncompressed object-level output that omits required structure such as the /Parent link; paperforge keeps full spec compliance (ISO 32000-1).
• The speedups come from reusing zlib compressor state across content streams (4–6× faster flate on many small streams), building content streams into a reused buffer without per-command format! allocations, a fast hasher for dictionary keys, and manual integer/number parsing in the parser (no per-number UTF-8 validation or str::parse).
• The parse benchmark surfaced a real interop bug: xref table entries were 19 bytes instead of the spec-mandated 20 (ISO 32000-1 §7.5.4), which strict parsers such as lopdf reject. Entries are now spec-compliant, locked in by serializer::tests::xref_entries_are_exactly_20_bytes.
• Criterion also writes per-run reports and regression tracking to target/criterion/.

Full methodology and analysis are in docs/benchmarks.md.

Feature Flags

[dependencies]
paperforge = { version = "0.1", features = ["generate", "parse", "layout", "fonts", "images"] }

Available features: generate, parse, edit, layout, fonts, images, render, forms, annotations, security, extract, svg.

License

Dual-licensed under MIT OR Apache-2.0.