v1.3.0 - Performance & API Enhancement (2026-03-03)

Added

• LinkFormat configuration for inline link formatting (markdown, html, none)
• CacheCleanup configuration with automatic background cleanup for TTL entries
• StartCleanup() and StopCleanup() methods for proactive cache management
• SetPoolLogger() function for pool corruption debug logging
• Optional variadic configuration parameters for New() constructor
• Smart config merging for ExtractConfig and LinkExtractionConfig
• Automatic cache goroutine cleanup via runtime.SetFinalizer
• Len() method to get current cache entry count
• FromFile variant methods to Extractor and LinkExtractor interfaces

Changed

• WalkNodes converted from recursive to iterative (prevents stack overflow on deep DOM)
• isPureASCII optimized with 64-bit batch processing (16% CPU hotspot reduced)
• Cache key hash length increased from 8 to 16 bytes (better collision resistance)
• Cache.Get() uses read-write lock separation for better concurrent performance
• ExtractToMarkdown() now uses DefaultConfig() for API consistency
• DefaultScorer uses lazy initialization with sync.Once
• Examples restructured from 9 to 8 focused files

Fixed

• Potential cache goroutine leak when Cache is garbage collected
• TOCTOU race condition in Cache.Get() method
• Potential nil pointer dereference in NewDefaultScorerWithConfig()
• Goroutine leak in withTimeout() with maximum limit protection
• Test error handling issues (unchecked errors, nil pointer access)

Performance

• Extract: ~26% faster
• ExtractWithCache: ~34% faster
• ExtractLargeDocument: ~22% faster
• CleanText: ~68% faster (replaced regex with manual scanning)
• ConcurrentExtract: ~29% faster
• Memory allocations reduced by 50-65% in key benchmarks

Security

• Library confirmed fully thread-safe (100+ race detection iterations)
• All shared state properly synchronized with appropriate primitives

Breaking Changes

• Destructive update!