Cargo.toml

[workspace]
members = [".", "pdf_oxide_mcp", "pdf_oxide_cli"]

[package]
name = "pdf_oxide"
version = "0.3.17"
edition = "2021"
authors = ["Yury Fedoseev <yfedoseev@gmail.com>"]
license = "MIT OR Apache-2.0"
description = "The fastest Rust PDF library with text extraction: 0.8ms mean, 100% pass rate on 3,830 PDFs. 5× faster than pdf_extract, 17× faster than oxidize_pdf. Extract, create, and edit PDFs."
repository = "https://github.com/yfedoseev/pdf_oxide"
homepage = "https://oxide.fyi"
documentation = "https://pdf.oxide.fyi/docs/getting-started/rust"
readme = "README.md"
keywords = ["pdf", "text-extraction", "pdf-parser", "pdf-to-markdown", "document-parser"]
categories = ["parsing", "text-processing"]

[lib]
name = "pdf_oxide"
crate-type = ["cdylib", "rlib"]  # cdylib for Python/WASM, rlib for Rust

[dependencies]
# Core dependencies
thiserror = "2.0"
log = "0.4"
env_logger = "0.11"
chrono = "0.4"  # Timestamp generation for benchmark results

# Parsing
nom = "8.0"
regex = "1.12"
lazy_static = "1.4"

# Data structures
bytes = "1.5"
indexmap = "2.2"
memchr = "2"     # SIMD-accelerated byte/substring search for content stream scanning
byteorder = "1.5"  # For binary parsing (TrueType cmap tables)
bitflags = "2"     # For form field flags
smallvec = "1.13"  # Stack-allocated small vectors for content stream operands (#165)

# Compression/Decompression
flate2 = { version = "1.1", default-features = false, features = ["zlib-rs"] }
weezl = "0.1"  # LZW compression
inflate = "0.4"  # Alternative inflate for better error recovery
libflate = "2.1"  # Another inflate implementation for error recovery
brotli = "8"  # Brotli compression (PDF 2.0, ISO 32000-2:2020)

# Font parsing
ttf-parser = "0.25"  # TrueType font parser for embedded font cmap tables
phf = { version = "0.13", features = ["macros"] }  # Perfect hash maps for Adobe Glyph List
encoding_rs = "0.8"  # Shift-JIS, EUC-JP, and other CJK encodings for predefined CMaps

# Image processing
image = { version = "0.24", default-features = false, features = ["png", "jpeg", "tiff"] }
base64 = "0.22"  # For embedding images as data URIs
tiff = "0.11"  # TIFF support including CCITT decompression
fax = "0.2"   # CCITT Group 3/4 decompression for scanned PDFs (more lenient with malformed EOFB)

# Encryption support
aes = "0.8"          # AES encryption for PDF 1.6+
cbc = "0.1"          # CBC mode for AES
md-5 = "0.10"        # MD5 for key derivation
sha2 = "0.10"        # SHA-256 for newer encryption
stringprep = "0.1"   # SASLprep for Unicode PDF passwords (R>=5)

# Digital signatures (optional)
x509-parser = { version = "0.18", optional = true }  # X.509 certificate parsing
cms = { version = "0.2", optional = true }            # CMS/PKCS#7 signatures
rsa = { version = "0.9", optional = true }            # RSA signing
der = { version = "0.8", optional = true }            # DER encoding
spki = { version = "0.7", optional = true }           # SubjectPublicKeyInfo
pkcs1 = { version = "0.7", optional = true }          # PKCS#1 RSA format
pkcs8 = { version = "0.10", optional = true }         # PKCS#8 private keys
signature = { version = "2.2", optional = true }      # Signature traits
sha1 = { version = "0.10", optional = true }          # SHA-1 for legacy signatures

# Parallel extraction
rayon = { version = "1.10", optional = true }

# Math and algorithms
ndarray = { version = "0.17", optional = true, features = ["std"] }
linfa = { version = "0.7", optional = true }
linfa-clustering = { version = "0.8", optional = true }

# ML (CPU-only)
tract-onnx = { version = "0.22", optional = true }
tokenizers = { version = "0.15", optional = true, default-features = false, features = ["onig"] }

# OCR - PaddleOCR via ONNX Runtime (optional)
ort = { version = "2.0.0-rc.11", optional = true, default-features = false, features = ["ndarray"] }
imageproc = { version = "0.26", optional = true }

# Serialization
serde = { version = "1.0", features = ["derive"] }
serde_json = "1.0"

# XML parsing (for XMP metadata)
quick-xml = "0.39"

# Python bindings (optional)
# Using latest version to support Python 3.14+
pyo3 = { version = "0.27", features = ["abi3-py38"], optional = true }

# WASM support (optional)
wasm-bindgen = { version = "0.2", optional = true }
web-sys = { version = "0.3", optional = true, features = ["console"] }
js-sys = { version = "0.3", optional = true }
console_error_panic_hook = { version = "0.1", optional = true }
getrandom = { version = "0.4", optional = true, features = ["wasm_js"] }
serde-wasm-bindgen = { version = "0.6", optional = true }

# Table detection - page rendering (optional)
pdfium-render = { version = "0.8", optional = true }

# Rendering (optional - for page to image conversion)
tiny-skia = { version = "0.12", optional = true }
fontdb = { version = "0.23", optional = true }
rustybuzz = { version = "0.20", optional = true }

lzw = "0.10.0"
uuid = { version = "1.0", features = ["v4", "js"] }

# Barcode/QR code generation (optional)
qrcode = { version = "0.14", optional = true }
barcoders = { version = "2.0", optional = true, features = ["image"] }

# Office document conversion (optional)
calamine = { version = "0.33", optional = true }
zip = { version = "8.2", optional = true, default-features = false, features = ["deflate"] }

[dev-dependencies]
criterion = "0.8"
proptest = "1.4"
regex = "1.12"
tempfile = "3.10"
crc32fast = "1.3"
wasm-bindgen-test = "0.3"

[features]
default = []

# Enable logging for the extraction pipeline
logging = []

# Python bindings (v0.1.0)
python = ["pyo3"]

# The following features are planned for v1.0 and are not yet production-ready.
# We define them here to avoid check-cfg warnings, but their dependencies are commented out.
ml = ["dep:tract-onnx", "dep:ndarray", "dep:linfa", "dep:linfa-clustering", "dep:tokenizers"]
table-ml = ["ml", "dep:pdfium-render"]
ocr = ["dep:ort", "dep:imageproc", "dep:ndarray"]
gpu = ["dep:ort", "ml"]
wasm = ["dep:wasm-bindgen", "dep:web-sys", "dep:js-sys", "dep:console_error_panic_hook", "dep:getrandom", "dep:serde-wasm-bindgen"]
wasm-ml = ["wasm", "ml"]

# Debug feature for span merging analysis
debug-span-merging = []

# Page rendering to images (pure Rust via tiny-skia)
rendering = ["dep:tiny-skia", "dep:fontdb", "dep:rustybuzz"]

# Digital signatures (create and verify PDF signatures)
signatures = ["dep:x509-parser", "dep:cms", "dep:rsa", "dep:der", "dep:spki", "dep:pkcs1", "dep:pkcs8", "dep:signature", "dep:sha1"]

# Barcode/QR code generation
barcodes = ["dep:qrcode", "dep:barcoders"]

# Office document conversion (DOCX, XLSX, PPTX)
office = ["dep:calamine", "dep:zip"]

# Parallel page extraction via rayon
parallel = ["dep:rayon"]

[profile.release]
opt-level = 3
lto = true
codegen-units = 1
strip = true

# Size-optimized release (for WASM)
[profile.release-small]
inherits = "release"
opt-level = "z"
lto = true
codegen-units = 1
strip = true
panic = "abort"

[profile.bench]
opt-level = 3
lto = true

# Benchmarks
[[bench]]
name = "pdf_extraction_performance"
harness = false

[[bench]]
name = "word_boundary_benchmarks"
harness = false

[[bench]]
name = "script_detection_benchmarks"
harness = false

[[bench]]
name = "ligature_benchmarks"
harness = false

[[bench]]
name = "cjk_benchmarks"
harness = false

[[bench]]
name = "rtl_benchmarks"
harness = false

[[bench]]
name = "complex_script_benchmarks"
harness = false

[[bench]]
name = "full_pipeline_benchmarks"
harness = false

[package.metadata.docs.rs]
# Document with safe features only (tesseract-rs fails on docs.rs sandbox)
features = ["python", "logging", "debug-span-merging", "parallel"]
rustdoc-args = ["--cfg", "docsrs"]

# Python packaging (Maturin)
[package.metadata.maturin]
name = "pdf_oxide"
python-source = "python"