v0.27.0 — Spot Colors, Streaming Parser, Builder Pattern, PDF/X, Plugins

What's New in v0.27.0

The biggest release yet — 29 roadmap items across patch, minor, and major tiers. 82 files changed, 19,668 insertions.

Major Features

• Spot Colors & DeviceN — Separation and DeviceN color spaces for professional print production (spotColor(), deviceNColor(), CMYK↔RGB conversion)
• Streaming PDF Parser — Process multi-GB PDFs without loading into memory via StreamingPdfParser (Web Streams API)
• PdfDocumentBuilder — Fluent builder pattern with method chaining: .create().setTitle().addPage().save()
• PDF/X Compliance — ISO 15930 validation and enforcement (validatePdfX, enforcePdfX, output intents)
• Plugin System — Lifecycle hooks (onBeforeAddPage, onAfterSave, etc.) with PdfPluginManager + 3 built-in plugins (timestamp, metadata, accessibility)
• Advanced Text Layout — Knuth-Plass line breaking, rich text spans, multi-column layout engine
• Unified Async API — embedPng() is now async (matching embedJpeg()) for consistent API
• PDF/UA Accessibility — Full validator with enforcement for accessible document creation

Minor Features

• Header/Footer Engine — Template variables ({page}, {pages}, {date}, {title}), Roman/alpha numbering
• Batch Processing — processBatch(), batchMerge(), batchFlatten() with error strategies (fail-fast/continue/collect)
• SVG Gradient Interpolation — Linear RGB interpolation with sRGB gamma correction
• Enhanced Redaction — Overlay alignment options for redaction annotations
• Page Labels & Bookmarks — Full API for document navigation
• Inline WASM Preloading — preloadInlineWasm() for eager initialization

Patch Fixes

• ES2025+ modernization audit — zero anti-patterns across entire codebase
• TypeScript strict mode compliance (exactOptionalPropertyTypes)
• Operator builder helpers with modern camelCase aliases

Stats

• 5,904 tests passing across 236 test suites
• 82 files changed, +19,668 / -675 lines
• 1 runtime dependency (fflate)

Full Changelog: https://github.com/ABCrimson/modern-pdf-lib/blob/master/CHANGELOG.md

v0.26.0

Added

• PDF form flattening (src/form/formFlatten.ts):

  • flattenForm() — burn all form fields into page content, remove AcroForm
  • flattenField() / flattenFields() — flatten specific fields by name
  • FlattenOptions with preserveReadOnly support

• Bookmarks / Outlines API (src/core/outlines.ts):

  • addBookmark() — add nested bookmarks with bold/italic/color/position
  • getBookmarks() — return the full bookmark tree
  • removeBookmark() / removeAllBookmarks() — bookmark management

• Page labels (src/core/pageLabels.ts):

  • setPageLabels() — set label ranges (decimal, roman, alpha, custom prefix)
  • getPageLabels() / removePageLabels() — label management
  • Catalog integration with /PageLabels number tree

• AES-256 encryption (PDF 2.0, V=5 R=6):

  • Full writer integration — doc.encrypt({ algorithm: 'aes-256' }) now works end-to-end
  • Algorithm 2.B key derivation (ISO 32000-2), SASLprep password normalization
  • %PDF-2.0 header for AES-256 encrypted documents
  • Trailer /Encrypt and /ID dictionary generation

• SVG-to-PDF vector conversion (enhanced):

  • Text rendering with standard font resolution (Helvetica/Times/Courier families)
  • Fill-rule support (nonzero/evenodd → f/f*/B/B*)
  • Stroke properties: linecap, linejoin, miterlimit, dasharray, dashoffset
  • 60 new SVG conversion tests

• Redaction application (src/annotation/applyRedactions.ts):

  • applyRedactions() — apply all redaction annotations (fill area, overlay text, remove annotation)
  • applyRedaction() — apply a single redaction by page/annotation index

• Batch processing (src/batch/batchProcessor.ts):

  • processBatch() — process multiple PDFs with bounded concurrency
  • batchMerge() / batchFlatten() — parallel merge and flatten operations
  • Progress callbacks, error isolation, runtime-aware concurrency

• PDF linearization (fast web view) — complete rewrite:

  • linearizePdf() — full page classification, hint tables, xref streams
  • delinearizePdf() — convert linearized PDF back to normal
  • getLinearizationInfo() — extract linearization parameters
  • Page offset and shared object hint tables per PDF spec §F

• PDF/UA accessibility validation (src/accessibility/pdfUaValidator.ts):

  • validatePdfUa() — 12 checks: structure tree, language, title, headings, alt text, tables, lists, reading order, fonts, contrast, bookmarks, tab order
  • enforcePdfUa() — auto-fix language, title, structure tree, tab order
  • Page tab order API (setTabOrder())

• WASM-by-default infrastructure:

  • getInlineWasmBytes() — decode base64-embedded WASM at runtime
  • scripts/generate-inline-wasm.ts — build-time code generation for inline WASM
  • generate:wasm-inline npm script

• 586 new tests across 20+ test files:

  • Form flattening (26), outlines (24), page labels (31), AES-256 (55)
  • SVG conversion (60 new), redaction (17), batch processing (18)
  • Linearization (44), PDF/UA validation (51), inline WASM (17)
  • PDF/A enforcement (24), sRGB ICC profile (20), transparency flattener (18)
  • JPEG2000 tiles (25), JPEG2000 bit depth (41)
  • Object pool (26), PDF value helpers (24), validation (60)

Changed

• WASM loader — now resolves dist/wasm/ layout first (npm consumers), falls back to dev src/wasm/ layout
• tsdown config — removed duplicate IIFE config (esbuild script handles it)
• Coverage threshold — branch coverage raised from 80% to 85%
• Modernization audit — 32 changes across 14 source files:
  • 22× isNaN() → Number.isNaN(), isFinite() → Number.isFinite()
  • .indexOf() >= 0 → .includes() in generated Acrobat JS
  • Removed 4 as unknown casts via proper typing
  • .split('') → [...str] spread for string iteration
• PdfWriter — now async with full AES-256 encryption integration
• API audit — comprehensive audit doc at docs/plans/api-audit-results.md
• Test count: 5,135 tests passing across 227 test suites

Fixed

• IIFE test failures from tsdown config removal
• fieldVisibility test updated for .includes() modernization

Full Changelog: v0.25.0...v0.26.0

v0.25.0 — Test Coverage, Complete Exports, CI/Build Pipeline

Added

• 267 new tests — comprehensive coverage for previously untested security-critical and core modules:

  • signatureVerifier.test.ts (23 tests) — digital signature verification with tamper detection, multi-sig, hash algorithms
  • sha256.test.ts (28 tests) — NIST FIPS 180-4 test vectors for SHA-256/384/512
  • textShaper.test.ts (32 tests) — Unicode text shaping, RTL, glyph mapping, WASM fallback
  • jpegEmbed.test.ts (42 tests) — SOF parsing, color space detection, Adobe APP14, error handling
  • cffEmbed.test.ts (36 tests) — OpenType table directory parsing, tag matching, all standard tags
  • freeTextAnnotation.test.ts (16 tests) — creation, parsing, alignment, font size
  • inkAnnotation.test.ts (12 tests) — ink list management, multi-path, clear workflow
  • redactAnnotation.test.ts (13 tests) — overlay text, interior color, quad points
  • stampAnnotation.test.ts (13 tests) — all 14 standard stamp names, custom stamps
  • detect.test.ts (16 tests) — runtime detection for Node/Deno/Bun, caching, adapter creation
  • adapter.test.ts (18 tests) — runtime adapter interface, override, stream creation
  • cli.test.ts (18 tests) — help text, argument parsing, validation, optimize command

• Complete sub-entry point exports:

  • modern-pdf-lib/create — added ~30 missing exports: gradients, patterns, page manipulation, PDF embedding, QR code options, transparency, document metadata
  • modern-pdf-lib/parse — added Operand type, AnalyzeImagesOptions, PdfParseError, formatHexContext
  • modern-pdf-lib/forms — added 7 appearance generators + 9 option types for complete form appearance API

Fixed

• CI/CD branch mismatch — .github/workflows/ci.yml triggered on main but repo uses master; CI was never running on pushes or PRs
• Missing sidebar entries — Image Formats and Multi-Party Signing guides were invisible in VitePress navigation
• CI WASM build missing crates — build-wasm-ci.sh was missing jbig2 and jpeg modules (only had 4/6)
• WASM binaries excluded from npm package — added dist/wasm/** to package.json files field
• No WASM copy step in build pipeline — added copy:wasm script to copy built WASM from src/wasm/*/pkg/ to dist/wasm/
• IIFE bundle not in build pipeline — added build:iife to build:all script chain

Changed

• Code deduplication — inlined .toHex() at 8 call sites, removed 3 identical wrapper functions from ocsp.ts, crl.ts, chainValidator.ts
• Test count: 4,549 tests passing across 211 test suites (up from 4,282 across 199)

Full Changelog: v0.24.10...v0.25.0

v0.24.10

Added

• Incremental save with signature preservation (src/signature/incrementalSave.ts):

  • saveIncrementalWithSignaturePreservation() — preserves all existing signature byte ranges
  • appendIncrementalUpdate() — pure append-only incremental updates
  • parseExistingTrailer() — parse trailer info from existing PDF
  • findExistingSignatures() / validateByteRangeIntegrity() — signature detection and validation

• Multi-signature chain validation (src/signature/multiSignatureValidator.ts):

  • validateSignatureChain() — validates ordered chain of signatures and byte range coverage

• MDP certification policy (src/signature/mdpPolicy.ts):

  • MdpPermission enum (NoChanges, FormFillAndSign, FormFillSignAnnotate)
  • setCertificationLevel() / getCertificationLevel() — DocMDP transform methods

• Modification detection (src/signature/modificationDetector.ts):

  • detectModifications() — checks if modifications exceed MDP-permitted level

• Signature field locking (src/signature/fieldLock.ts):

  • addFieldLock() / getFieldLocks() — lock specific fields after signing

• Document diff (src/signature/documentDiff.ts):

  • diffSignedContent() — compares signed version vs current content

• Counter-signatures (src/signature/counterSignature.ts):

  • addCounterSignature() / getCounterSignatures() — sign existing signatures

• LTV embedding (src/signature/ltvEmbed.ts):

  • embedLtvData() — embeds CRL + OCSP + certs in Document Security Store (DSS)
  • hasLtvData() / buildDssDictionary() — LTV data detection and building

• Incremental save optimization (src/signature/incrementalOptimizer.ts):

  • optimizeIncrementalSave() — FNV-1a object hashing, only append changed objects
  • findChangedObjects() / computeObjectHash() — object-level diff detection

• WebP image embedding (src/assets/image/webpDecode.ts):

  • decodeWebP() — pure TypeScript VP8 (lossy) and VP8L (lossless) decoder
  • Alpha channel support via ALPH chunk parsing
  • DCT coefficient decoding, YUV420→RGB, macroblock processing
  • LZ77, Huffman coding, spatial prediction (13 modes) for lossless

• TIFF image embedding (src/assets/image/tiffDecode.ts):

  • decodeTiff() / decodeTiffPage() / decodeTiffAll() — full TIFF decoder
  • Uncompressed, PackBits, LZW, Deflate, JPEG-in-TIFF compression support
  • Multi-page TIFF (IFD chain following)
  • getTiffPageCount() / parseTiffIfd() — TIFF introspection

• TIFF CMYK support (src/assets/image/tiffCmyk.ts):

  • embedTiffCmyk() — native DeviceCMYK embedding (no RGB conversion)
  • convertTiffCmykToRgb() — CMYK→RGB conversion when needed

• Image format auto-detection (src/assets/image/formatDetect.ts):

  • detectImageFormat() — detects PNG/JPEG/WebP/TIFF from magic bytes
  • Updated embedImage() to support all 4 formats automatically

• WebP optimization (src/assets/image/webpOptimize.ts):

  • webpToJpeg() / webpToPng() — format conversion
  • recompressWebP() — decode → JPEG/PNG re-encode pipeline

• TIFF direct embedding (src/assets/image/tiffDirectEmbed.ts):

  • embedTiffDirect() — direct strip/tile mapping (skip decode/re-encode)
  • canDirectEmbed() — format compatibility check

• Documentation:

  • Multi-signature workflow guide (docs/guide/multi-sign.md)
  • Image format support guide (docs/guide/image-formats.md)

Stats

• 4,282 tests passing across 199 suites (+285 new tests)
• Zero TypeScript errors (strict mode + exactOptionalPropertyTypes)
• 15 new source files, 15 new test files, 2 new guides

Full Changelog: v0.22.9...v0.24.10

v0.22.9

Added

• CRL/OCSP revocation checking (src/signature/):

  • OCSP request building, response parsing, and certificate status checking
  • CRL parsing, download, and revocation checking with distribution points
  • Certificate chain building and validation with Web Crypto signature verification
  • TTL-based OCSP/CRL response caching with automatic expiry
  • OCSP stapling — embed/extract OCSP responses in PKCS#7 signatures
  • Delta CRL parsing and merging with base CRLs (RFC 5280)
  • Enhanced verification with structured results (chain, revocation, timestamps)
  • Offline revocation data extraction and verification without network
  • Custom trust store class for enterprise PKI certificate management
  • Key usage, extended key usage, and certificate policy validation
  • New guide: docs/guide/verification.md

• JPEG2000 (JPXDecode) decoder (src/parser/):

  • Full JP2/J2K decoder with MQ arithmetic coder and discrete wavelet transform
  • JPXDecode filter integration in the stream decoder pipeline
  • JP2 container vs J2K codestream detection and parsing
  • Alpha channel detection, separation, and premultiplication
  • JP2-to-JPEG transcoding for downstream JPEG workflows
  • 16-bit to 8-bit bit depth normalization
  • Tiled and region-of-interest decoding for large images
  • WASM bridge with automatic JS fallback
  • New guide: docs/guide/jpeg2000.md

• Form field JavaScript evaluation (src/form/):

  • Arithmetic expression evaluator and AFSimple_Calculate parser
  • Field calculation order with dependency graphs and topological sort
  • AFNumber_Format/Keystroke and AFDate_FormatEx/KeystrokeEx builtins
  • Field validation: email, phone, range, regex, and length constraints
  • AFPercent_Format, AFSpecial_Format (ZIP, SSN, phone number)
  • Field visibility toggle with condition-based show/hide
  • getField() cross-field references with field proxy objects
  • Document-level scripts: open, close, print, and save actions
  • Sandboxed script execution with reserved word filtering and strict mode
  • New guide: docs/guide/form-scripts.md

Changed

• Test count: 3,260 → 3,997 across 184 test suites (was 158).

Full Changelog: v0.19.9...v0.22.9

v0.19.9

Added

• Table layout engine (src/layout/): Full-featured table rendering with PDF operators.

  • renderTable() — single-page table rendering to content-stream operators
  • renderMultiPageTable() — automatic page breaks with header row repetition
  • Column width modes: fixed, percentage, flex (weighted), auto-fit (content-based)
  • Cell padding (per-cell or table-wide, four-sided or uniform)
  • Horizontal alignment (left, center, right) and vertical alignment (top, middle, bottom)
  • Colspan and rowspan with 2D occupation grid tracking
  • Alternating row colors, header background/text color
  • Nested tables as cell content (recursive rendering)
  • Rich cell content with TextRun[] (per-run font, color, size)
  • Text overflow modes: wrap, truncate, ellipsis, shrink-to-fit
  • Styling presets: minimalPreset(), stripedPreset(), borderedPreset(), professionalPreset()
  • applyPreset() and applyTablePreset() for named preset selection
  • PdfPage.drawTable() convenience method
  • Overflow utilities: estimateTextWidth(), wrapText(), truncateText(), ellipsisText(), shrinkFontSize()
  • 200 tests across 11 test files

• QR code & barcode engine (src/barcode/): 9 barcode formats with full encoding.

  • QR code (ISO 18004): versions 1–40, all EC levels (L/M/Q/H), 8 mask patterns, GF(256) Reed-Solomon
  • Code 128 (A/B/C auto-switching, mod-103 check digit)
  • EAN-13 / EAN-8 with L/G/R patterns and parity tables
  • UPC-A (delegates to EAN-13 with leading zero)
  • Code 39 with optional mod-43 check digit
  • ITF (Interleaved 2 of 5) with bearer bars
  • PDF417 (text/byte compaction, RS over GF(929), clusters 0/3/6)
  • Data Matrix ECC200 (RS over GF(256), Utah placement algorithm)
  • renderStyledBarcode() for styled rendering with text, borders, colors
  • readBarcode() / readCode128() / readEan13() for round-trip verification
  • PdfPage.drawQrCode() convenience method
  • 204 tests across 10 test files

• Browser utilities (src/browser/):

  • saveAsDownload(), saveAsBlob(), saveAsDataUrl(), openInNewTab()
  • Service Worker helpers: handlePdfRequest(), createPdfResponse(), isCacheAvailable()
  • PdfWorker class for Web Worker PDF generation
  • CSP compatibility: disableWasm option, isWasmDisabled(), runtime detection
  • WASM streaming: loadWasmModuleStreaming(), instantiateWasmModuleStreaming()

• PDF/A full enforcement (src/compliance/):

  • enforcePdfAFull() — complete pipeline chaining all PDF/A fixes
  • sRGB ICC profile generation (~3KB ICC v2 profile)
  • OutputIntent builder with OutputIntentOptions
  • ToUnicode CMap generation (WinAnsi, Symbol, ZapfDingbats)
  • Transparency detection and flattening
  • XMP metadata validation and generation for PDF/A conformance
  • PDF/A profile definitions (1a/1b, 2a/2b/2u, 3a/3b/3u)
  • Associated files for PDF/A-3 (/AF key)
  • Prohibited feature stripping (JavaScript, Launch, Sound, Movie, RichMedia)
  • veraPDF CLI wrapper for CI validation

• Documentation: New guides for tables, barcodes, PDF/A, CSP, and browser integration.

Changed

• Test count: 2,323 → 3,260 across 158 test suites (was 110).

Full Changelog: v0.15.1...v0.19.9

v0.15.1

Fixed

• CLI build: Added src/cli/index.ts as a tsdown entry point so dist/cli/index.mjs is produced during build. Previously, npx modern-pdf optimize would fail because the CLI file was never emitted.
• CJS compatibility: Replaced top-level await in CLI entry with an async main() wrapper to avoid Top-level await is not supported with CJS output build errors.

Full Changelog: v0.15.0...v0.15.1

v0.15.0 — JPEG WASM & Image Optimization

Added

• JPEG WASM module: New Rust WASM crate (src/wasm/jpeg/) using jpeg-encoder 0.7 + jpeg-decoder 0.3 for high-performance JPEG encoding and decoding. TypeScript bridge (initJpegWasm(), encodeJpegWasm(), decodeJpegWasm()) with graceful JS fallback when WASM is unavailable.
• JPEG quality auto-detection: estimateJpegQuality(jpegBytes) analyzes DQT (quantization table) markers to reverse-engineer the original JPEG quality level (1–100). Pure TypeScript — no WASM required.
• Progressive JPEG support: progressive option for JPEG encoding produces progressive scan-order JPEGs (better for web delivery).
• Chroma subsampling control: chromaSubsampling option ('4:4:4' | '4:2:2' | '4:2:0') for JPEG encoding. Default '4:2:0' matches industry standard for smallest files.
• CMYK JPEG handling: Automatic CMYK→RGB conversion before JPEG encoding using the standard formula R = 255 × (1 − C/255) × (1 − K/255).
• Batch image optimization API: optimizeAllImages(doc, options) walks all image XObjects in a parsed PDF, recompresses them as JPEG, and replaces stream data in-place. Returns detailed OptimizationReport with per-image breakdown. Options: quality, maxDpi, progressive, chromaSubsampling, skipSmallImages, minSavingsPercent, autoGrayscale.
• Image extraction API: extractImages(doc) collects metadata for all image XObjects across all pages. decodeImageStream(imageInfo) decodes stream data for further processing.
• Image deduplication: deduplicateImages(doc) detects identical images by FNV-1a hashing and replaces duplicate references with a single canonical copy. Returns DeduplicationReport with bytes-saved statistics.
• Grayscale auto-detection: isGrayscaleImage() detects RGB images where all pixels are effectively grayscale (R ≈ G ≈ B within tolerance). convertToGrayscale() converts using ITU-R BT.601 luma formula, saving ~66% for RGB→grayscale.
• DPI-aware downscaling: computeImageDpi() and computeTargetDimensions() calculate effective DPI from pixel dimensions and display size in points. Used by batch optimizer for automatic DPI-based downscaling.
• CLI tool: npx modern-pdf optimize input.pdf output.pdf [options] — command-line interface for image optimization with --quality, --progressive, --grayscale, --dedup, --chroma, --verbose options.
• Image optimization VitePress guide (docs/guide/image-optimization.md): Comprehensive guide covering the full image optimization API, CLI usage, and options reference.

Changed

• WASM modules: 5 → 6 (added jpeg for JPEG encode/decode).
• Test count: 2,243 → 2,323 across 110 test suites (was 103).
• package.json: Added "bin": { "modern-pdf": "./dist/cli/index.mjs" } for CLI support.

Full Changelog: v0.14.1...v0.15.0

v0.14.1

Added

• Unified embedImage() method: Auto-detects PNG (magic bytes 89 50 4E 47) vs JPEG (FF D8 FF) from the raw file data — no need to call embedPng() or embedJpeg() separately. Accepts Uint8Array or ArrayBuffer, throws descriptive errors for unsupported formats.
• Image optimization API exports: downscaleImage(), recompressImage(), and optimizeImage() are now exported from the main entry point along with their option types (DownscaleOptions, RecompressOptions, ImageOptimizeOptions, RawImageData, OptimizeResult).
• SASLprep password normalization (RFC 4013): V=5/R=6 (AES-256) password preparation now follows the full SASLprep profile — B.1 "mapped to nothing" characters are stripped, non-ASCII spaces are normalized to U+0020, NFKC normalization is applied, and prohibited characters (control chars, private use, surrogates, tagging) are rejected. This ensures correct password handling for internationalized passwords per ISO 32000-2.
• Visible signature appearances: signPdf() now accepts an appearance option with rect, text, fontSize, backgroundColor, borderColor, and borderWidth. When provided, the signature renders as a visible box on the page with auto-generated text (signer name from certificate CN, reason, location, date) or custom text lines. Uses a PDF Form XObject appearance stream with Helvetica.
• Popup annotation type (PdfPopupAnnotation): Floating window annotation that displays parent annotation text. Supports isOpen()/setOpen() and setParent()/getParent() for linking to parent annotations. Reference: PDF 1.7 §12.5.6.14.
• Caret annotation type (PdfCaretAnnotation): Marks text insertion points in review workflows. Supports getSymbol()/setSymbol() ('None' | 'P' for paragraph) and getCaretRect()/setCaretRect() for inner rectangle (RD) insets. Reference: PDF 1.7 §12.5.6.11.
• File attachment annotation type (PdfFileAttachmentAnnotation): Embeds a file as a clickable icon on a page. Supports getIcon()/setIcon() ('GraphPushPin' | 'PaperclipTag' | 'Paperclip' | 'Tag'), getFileName(), and buildFileSpec(registry) for building the embedded file stream, EF dictionary, and file specification. Reference: PDF 1.7 §12.5.6.15.
• Digital signatures guide (docs/guide/signatures.md): Comprehensive guide covering invisible and visible signing, verification, low-level ByteRange API, external signers, RFC 3161 timestamping, and key preparation with OpenSSL.
• Accessibility & tagged PDF guide (docs/guide/accessibility.md): Guide covering document language, title metadata, font embedding for Unicode, color contrast, multilingual content, PDF/UA compliance checklist, and XMP metadata.
• Annotations guide (docs/guide/annotations.md): Complete reference for all 18 annotation types with code examples — Text, Link, FreeText, Highlight, Underline, Squiggly, StrikeOut, Line, Square, Circle, Polygon, PolyLine, Stamp, Ink, Redact, Popup, Caret, FileAttachment. Includes annotation flags, appearance generation, and parsing from existing PDFs.

Performance

• Per-object encryption key caching: deriveObjectKey() in PdfEncryptionHandler now caches computed keys in a Map<number, Uint8Array> keyed on (objNum << 16) | genNum. For V=1-4 encrypted PDFs, this eliminates redundant MD5 computations when decrypting multiple strings/streams in the same object.
• File-level encryption key caching: computeFileEncryptionKey() now maintains an LRU cache (max 32 entries) keyed on password + encryption dictionary parameters. Re-opening the same PDF with the same password skips the expensive key derivation (especially Algorithm 2.B for R=6 which runs 64+ rounds of AES+SHA).

Changed

• Annotation count: 15 → 18 annotation types (added Popup, Caret, FileAttachment).
• Test count: 2,199 → 2,243 across 103 test suites (was 100).

Full Changelog: v0.14.0...v0.14.1

v0.14.0 — Performance Audit & Bug Fixes

Performance

Comprehensive internal performance audit across the entire codebase. All changes are internal hot-path optimizations — zero API surface changes.

Lexer & Content Stream Parser:

• String concatenation in readLiteralString(), readHexString(), readName() replaced with parts[] + .join('') pattern
• bytesToAscii() / decodeAscii() replaced with batch String.fromCharCode.apply()
• Hex string parsing rewritten to single-pass direct byte decoding with hexVal lookup table (eliminates intermediate string + parseInt)

LZW Decompression:

• Complete rewrite with pooled flat buffer (Uint8Array + Int32Array index pairs) replacing per-entry Uint8Array[] allocations
• Identity entries (0-255) initialized once and persist across table resets
• Pre-allocated output buffer with manual growth instead of number[] + .push()

XRef Recovery & Parsing:

• rebuildXrefFromScan() rewritten to scan raw Uint8Array bytes directly for obj pattern instead of TextDecoder.decode() + regex on the entire file
• Standard xref entries parsed directly from bytes (fixed 20-byte format) without TextDecoder
• Keyword checks (xref, trailer) replaced with direct byte comparison

PDF Object Serialization:

• escapeLiteralString(): 5 chained .replace() calls replaced with single-pass character loop
• PdfName.serialize(): String concatenation replaced with array + join
• formatNumber(): Regex trailing-zero trim replaced with manual digit loop

Cryptographic Key Derivation:

• Pre-allocated modKey buffer outside 19x RC4 iteration loops in owner/user password verification
• Direct K1 buffer construction in Algorithm 2.B (eliminates intermediate concatenation)

Other:

• ASCII85 decoder: Fixed-size group buffer + pre-allocated output
• SVG color parser: Module-level result cache (Map<string, ParsedColor>)
• XMP escapeXml(): Single-pass character loop replacing chained .replace()
• Inline image EI scanning: data.indexOf(0x45) jump instead of byte-by-byte scan
• Object stream header: Array + join replacing string concatenation

Fixed

• CCITT Group 3/4 2D decode bug: read2DMode() returned HORIZONTAL for bit pattern 011 instead of VERTICAL_PLUS_1 — correct logic existed but was unreachable due to premature return. This could cause incorrect rendering of CCITT Group 4 and Group 3 2D compressed images (scanned documents).
• customName font option ignored for empty strings: || operator treated empty string as falsy, falling through to postScriptName. Changed to ?? (nullish coalescing).
• embedPages() unnecessarily async: Method was declared async but contained only synchronous code, wrapping return in an unnecessary Promise. Now returns EmbeddedPdfPage[] directly.
• Duplicate hash computation in document merge: hashBytes() was called twice on the same stream data during cross-document page copy. Now computed once and reused.

Removed

• Dead PdfArr import alias in pdfWriter.ts
• Unused objectBuf variable allocation in object stream serialization
• Unused objectContainsPageRef() function in linearization module

Full Changelog: v0.13.0...v0.14.0