CHANGELOG.md

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

Unreleased

0.2.6 - 2026-04-21

Dependencies

• Bump selectors 0.36 → 0.37
• Bump pyo3 0.28.2 → 0.28.3
• Bump rustls-webpki to 0.103.12
• Bump @napi-rs/cli 3.6.1 → 3.6.2
• Bump @emnapi/runtime (Node.js and WASM packages)
• Bump @biomejs/biome (Node.js and WASM packages)
• Bump pytest (Python package)
• Bump CI actions: lewagon/wait-on-check-action 1.6.1 → 1.7.0, actions/upload-pages-artifact 4 → 5, softprops/action-gh-release 2 → 3, pnpm/action-setup 5 → 6, dependabot/fetch-metadata 2 → 3, actions/deploy-pages 4 → 5, codecov/codecov-action 5 → 6

0.2.5 - 2026-03-17

Changed

• Replace markup5ever_rcdom with a custom DocBuilderSink (html5ever::TreeSink) that builds the internal Document directly, eliminating intermediate RcDom allocation and Rc/RefCell overhead

Dependencies

• Bump html5ever 0.38 → 0.39, markup5ever 0.38 → 0.39
• Remove markup5ever_rcdom (unmaintained, incompatible with markup5ever 0.39)

Fixed

• DocBuilderSink: depth tracking correctly updated after reparent_children and insert_before (foster parenting edge case)
• DocBuilderSink: text coalescing in append_before_sibling now checks the sibling node itself rather than its last child, matching html5ever RcDom semantics
• DocBuilderSink: is_mathml_annotation_xml_integration_point now respects ElementFlags instead of unconditionally returning false
• DocBuilderSink: create_pi returns Phantom handle instead of creating an empty text node
• Remove redundant strip = true from [tool.maturin] in pyproject.toml (conflicts with maturin develop in maturin >= 1.12)
• Add biome.json to exclude .d.ts files from biome linting (biome 2.4.7 internal panic on declaration files)

0.2.4 - 2026-02-20

Dependencies

• Bump workspace dependencies (clap, syn, serde, wasm-bindgen, pyo3)

0.2.3 - 2026-02-16

Dependencies

• Bump workspace Rust dependencies (rust-minor group)
• Bump @biomejs/biome in Node.js and WASM packages

0.2.2 - 2026-02-03

Fixed

• Switch CLI from native-tls to rustls for cross-compilation support (fixes linux-aarch64 build)

0.2.1 - 2026-02-03

Added

• Phase 18: Streaming & Advanced Performance

  • Full lol_html streaming integration with constant O(1) memory usage
  • StreamingParser with event-driven architecture for large documents
  • Memory-mapped file support via mmap feature
  • Streaming text extraction and element handlers

• Phase 19: Developer Experience

  • Enhanced error tracking and diagnostics
  • Improved error messages with source context
  • Better panic handling and recovery

• Phase 20: Benchmark Infrastructure

  • Comprehensive benchmark suite with criterion
  • Cross-platform performance comparison tools
  • Memory profiling infrastructure

• Documentation

  • mdBook-based documentation site
  • Getting Started guide
  • User Guide with examples
  • API reference documentation

Changed

• Updated PyO3 to 0.28.0 with skip_from_py_object compatibility fix
• Dependabot auto-merge workflow for patch updates

Fixed

• CI wait-on-check circular dependency issue
• Dependabot workflow auto-approve removal

Dependencies

• pyo3: 0.27.2 → 0.28.0
• bytes: 1.11.0 → 1.11.1
• clap: 4.5.54 → 4.5.56
• actions/checkout: 4 → 6
• actions/cache: 4 → 5
• actions/upload-pages-artifact: 3 → 4
• @biomejs/biome: updated in Node.js and WASM packages

0.2.0 - 2026-01-20

Added

• Phase 14: Performance Optimization

  • SIMD-accelerated class selector matching with automatic platform detection (SSE4.2/AVX2/NEON/SIMD128)
  • TagId enum for efficient HTML5 tag interning (113 tags with O(1) lookup)
  • DocumentIndex for optimized ID and class-based lookups (O(1) ID, O(k) class)
  • Rayon-powered batch parsing via parse_batch() for parallel document processing
  • Selector fast-paths for common patterns (tag only, class only, ID only)
  • Query optimization with compiled selector caching and pre-compiled selector support

• Phase 15: Core Utilities Extraction

  • New scrape-core/src/utils.rs module with shared HTML escaping utilities
  • New scrape-core/src/serialize.rs module with centralized HTML serialization
  • escape_text() and escape_attr() with zero-copy Cow optimization
  • serialize_node() and collect_text() functions for consistent HTML handling
  • Zero-copy optimization reduces allocations by 50-70% in typical HTML serialization

• Phase 16: Trait Abstractions and Iterator Extensions

  • HtmlSerializer trait for unified HTML serialization API on Tag type
  • ElementFilter iterator extensions (.elements()) for element-only iteration
  • 6 new iterator types: ElementChildrenIter, ElementDescendantsIter, ElementAncestorsIter, etc.
  • Simplified binding navigation code by 45% per method across all platforms (Python/Node.js/WASM)
  • Comprehensive benchmark suite validating zero-overhead abstractions

• Phase 17: Advanced Type Safety with Typestate Patterns

  • DocumentState sealed trait with three lifecycle states: Building, Queryable, Sealed
  • Compile-time enforced document lifecycle via PhantomData (zero runtime overhead)
  • Type-safe state transitions: Building → Queryable → Sealed
  • NodeType sealed trait preventing external implementations with private module pattern
  • Marker types (ElementMarker, TextMarker, CommentMarker) for enhanced type safety
  • Full backward compatibility via Document type alias (Document = DocumentImpl)

Changed

• Breaking Changes (Internal API)

  • Document is now DocumentImpl with generic state parameter
  • Parser APIs now use DocumentImpl during construction
  • Public Document type alias maintains backward compatibility
  • New types exported: DocumentImpl, DocumentState, Building, Queryable, Sealed, NodeType, markers

• Performance improvements across all binding libraries (Python/Node.js/WASM)

• Eliminated 308 lines of duplicated code across bindings

• More efficient DOM traversal with cached selector state

• Reduced memory allocations in text extraction and HTML serialization

Fixed

• Resolved 5 FIXME/TODO markers related to code duplication and unfinished implementations
• Fixed inconsistent text escaping patterns between core and bindings
• Improved consistency in HTML serialization across all platforms

Security

• Zero unsafe code in Phase 17 implementation
• All sealed traits prevent unintended implementations
• Type-safe state transitions prevent invalid operations at compile time

Performance

• Query Performance: SIMD-accelerated class matching (2-10x improvement on large documents)
• Memory: 50-70% reduction in serialization allocations via Cow
• Parallel: Batch parsing scales near-linearly with thread count (Rayon)
• Zero-cost abstractions: Phase 16-17 traits generate identical code to manual implementations

Test Coverage

• 1,121 comprehensive tests passing across all platforms
• 506 Rust unit tests + 116 doctests
• 180 Python tests, 201 Node.js tests, 118 WASM tests
• 100% coverage on all new modules (utils.rs, serialize.rs, state.rs, node_type.rs)
• Compile-time type safety verified via sealed traits and typestate patterns

0.1.6 - 2026-01-16

Fixed

• Use --ignore-scripts for npm main package publish (platform packages already published in loop)

0.1.5 - 2026-01-16

Fixed

• Fixed contents: write permission for npm publish job (napi prepublish needs to create releases)

0.1.4 - 2026-01-16

Fixed

• Fixed npm auth for platform packages publishing (explicit .npmrc setup)

0.1.3 - 2026-01-16

Fixed

• Added GITHUB_TOKEN for napi prepublish in npm release workflow

0.1.2 - 2026-01-16

Fixed

• Fixed doctests using private dom module path
• Fixed pnpm lockfile compatibility in CI (upgraded to pnpm 10)
• Fixed maturin cross-compilation with --find-interpreter flag
• Disabled linux-arm64-musl Node.js build temporarily (requires proper cross-compiler)

0.1.1 - 2026-01-16

Added

• CLI tool (scrape-cli): Command-line HTML extraction tool
  • CSS selector-based extraction with -s/--selector
  • Multiple output formats: text, JSON, CSV, HTML (-f/--format)
  • Batch processing with parallel execution via Rayon
  • Named extractions support (-n/--named)
  • Attribute extraction (-a/--attrs)
  • HTML fragment output (--html)
  • File and stdin input support
• CLI binary builds for 5 platforms in GitHub releases (linux-x64, linux-arm64, darwin-x64, darwin-arm64, windows-x64)

Changed

• Renamed Python package from scrape-rs to fast-scrape on PyPI
• Renamed Node.js package from scrape-rs to @fast-scrape/node on npm
• Renamed WASM package from @scrape-rs/wasm to @fast-scrape/wasm on npm

Fixed

• HTML comment escaping in CLI output to prevent injection via filenames

0.1.0 - 2026-01-16

Added

• Core HTML parsing engine based on html5ever for spec-compliant parsing
• Arena-based DOM tree for memory efficiency and cache-friendly traversal
• CSS selector support via selectors crate with full CSS3 selector syntax
• SIMD-accelerated byte scanning with memchr (SSE4.2/AVX2 on x86_64, NEON on ARM64)
• Python bindings via PyO3 (PyPI: scrape-rs)
• Node.js bindings via napi-rs (npm: scrape-rs)
• WASM bindings for browsers (npm: @scrape-rs/wasm)
• Cross-platform API consistency across all bindings
• BeautifulSoup-compatible API with Soup and Tag types
• find() and find_all() methods for element search
• Attribute access and text extraction utilities
• Parallel processing support via Rayon (native platforms only)
• Comprehensive test suite with shared test cases across bindings
• Fuzzing targets for parser and selector components
• Benchmarks comparing against BeautifulSoup and Cheerio

Performance

• 10-50x faster than BeautifulSoup for parsing and queries
• Zero-copy DOM navigation
• WASM bundle under 500KB gzipped