·
1836 commits
to main
since this release
Lucene 10.5.1 Release Highlights:
• Fixed DocValuesRangeIterator.docIDRunEnd() returning wrong run boundaries, which could cause bulk-scoring toskip or mis-score documents in range and ordinal set queries
• Restored HNSW graph reuse during merges with per-field vector formats (including the default codec).
• HNSW graph construction now checks periodically whether the surrounding merge was aborted, so IndexWriter#rollback() and abortMerges() no longer block until the full graph is built.
• Fixed scoring on scalar-quantized indexes after raw float vectors were dropped; FloatVectorValues#scorer now scores against quantized vectors, matching behavior when raw vectors are present.