rvben/rumdl v0.0.169

on GitHub

Performance

MASSIVE PERFORMANCE IMPROVEMENTS: This release delivers 7-53x faster linting through systematic elimination of O(n²) bottlenecks and algorithmic optimizations. rumdl is now 16-29x faster than markdownlint-cli2 on real-world repositories.

• Fix catastrophic O(n²) bottleneck in FilteredLinesIter ⭐ THE GAME CHANGER

  • Eliminated content.lines().collect() from inside Iterator::next() method
  • Single-file optimization delivered 7,600x speedup for affected rules
  • MD011: 6.00s → 0.785ms (7,644x faster)
  • MD012: 5.95s → 1.643ms (3,621x faster)
  • Impact: Stress test (10k lines): 9,987ms → 187ms (53x faster)
  • Impact: Rust Book (478 files): 2,646ms → 269ms (9.8x faster)
  • Impact: PyO3 (76 files): 2,004ms → 262ms (7.6x faster)

• Pre-compute LintContext data structures to eliminate redundant work

  • LineIndex: Eliminated 46× content cloning across all rules
  • Jinja template ranges: Pre-computed once instead of O(n×m) scanning
  • Table blocks: Computed once instead of 4× independent scans per document
  • HTML comment ranges: Pre-computed with 50-70x speedup
  • Code block line mapping: Optimized with binary search instead of linear scan
  • Autodoc blocks (MkDocs): Pre-computed to avoid O(n²) scaling

• Optimize link and image parsing with binary search

  • Replaced linear search with binary search for code span checks
  • Faster link/image parsing across all reference-based rules

• Replace regex with byte-level parsing for better performance

  • List detection: Manual byte scanning instead of regex
  • Blockquote prefix: Byte-level parsing instead of regex
  • Line info parsing: Eliminated redundant blockquote prefix parsing

• Streaming parser architecture

  • Replaced AST parsing with pulldown-cmark streaming for lower memory overhead
  • Better scalability for very large documents

• MD046 optimization

  • Pre-compute list/tab contexts
  • Use ctx.code_blocks instead of independent detection

Benchmarks

vs Previous Version (v0.0.168):

• Rust Book (478 files): 2,646ms → 269ms (9.8x faster)
• Stress Test (10,514 lines): 9,987ms → 187ms (53.4x faster)
• PyO3 (76 files): 2,004ms → 262ms (7.6x faster)

vs markdownlint-cli2 (industry standard):

• Rust Book: 10.3s vs 0.35s (29.4x faster)
• PyO3: 4.4s vs 0.26s (16.9x faster)

Current Performance:

• 0.7ms per file (Rust Book)
• 6.3x parallelization efficiency
• Zero O(n²) algorithmic bottlenecks remaining

Added

• MDX and Quarto flavor support
  • New flavors: MarkdownFlavor::MDX and MarkdownFlavor::Quarto
  • JSX component handling: Skip linting inside JSX elements
  • ESM import/export support: Ignore JavaScript import/export statements
  • Quarto code chunk support: Recognize {r}, {python} chunk delimiters
  • Essential for modern React-based documentation and data science workflows

Fixed

• MD011: Intelligent URL vs text detection in reversed links

  • Smarter classification of link components as URL or text
  • Reduces false positives in edge cases like (Generic<T>)[link]
  • Correctly handles (http://url)[text] vs (text)[#anchor]
  • Improved heuristics for ambiguous single-word patterns

• MD013: Preserve HTML blocks in list items during reflow

  • HTML blocks inside list items are now preserved during auto-fix
  • Prevents broken indentation and tag structure
  • Better handling of complex nested content

Changed

• Architecture improvements for pre-computation pattern
  • LintContext now owns shared data structures
  • Rules consume pre-computed data instead of duplicating work
  • Consistent binary search APIs for range lookups
  • Foundation for future optimizations

Downloads

Installation

Using uv (Recommended)

uv tool install rumdl

Using pip

pip install rumdl

Using pipx

pipx install rumdl

Direct Download

Download the appropriate binary for your platform from the table above, extract it, and add it to your PATH.