Refactor k-mer matching to use a pipeline architecture with improved concurrency and memory management:
- Replace sort.Slice with slices.SortFunc and cmp.Compare for better performance
- Introduce PreparedQueries struct to encapsulate query buckets with metadata
- Implement MergeQueries function to merge query buckets from multiple batches
- Rewrite MatchBatch to use pre-allocated results and mutexes instead of map-based accumulation
- Add seek optimization in matchPartition to reduce linear scanning
- Refactor match command to use a multi-stage pipeline with proper batching and merging
- Add index directory option for match command
- Improve parallel processing of sequence batches
This refactoring improves performance by reducing memory allocations, optimizing k-mer lookup, and implementing a more efficient pipeline for large-scale k-mer matching operations.
This commit introduces sparse index support for KDI files to enable fast random access during k-mer matching. It adds a new .kdx index file format and updates the KDI reader and writer to handle index creation and seeking. The changes include:
- New KdxIndex struct and related functions for loading, searching, and writing .kdx files
- Modified KdiReader to support seeking with the new index
- Updated KdiWriter to create .kdx index files during writing
- Enhanced KmerSetGroup.Contains to use the new index for faster lookups
- Added a new 'match' command to annotate sequences with k-mer match positions
The index is created automatically during KDI file creation and allows for O(log N / stride) binary search followed by at most stride linear scan steps, significantly improving performance for large datasets.
