Module org.elasticsearch.server
Package org.elasticsearch.index.codec.vectors

Class VectorScoringUtils

java.lang.Object
org.elasticsearch.index.codec.vectors.VectorScoringUtils
public final class VectorScoringUtils extends Object
Utility methods for vector scoring and collection.

  • Method Summary

    Modifier and Type
    Method
    Description
    static org.apache.lucene.search.VectorScorer
    denseVectorScorer(org.apache.lucene.util.hnsw.RandomVectorScorer scorer, org.apache.lucene.index.KnnVectorValues.DocIndexIterator iterator)
    VectorScorer for dense vectors; bulk path uses VectorScorer.Bulk.fromRandomScorerDense(org.apache.lucene.util.hnsw.RandomVectorScorer, org.apache.lucene.index.KnnVectorValues.DocIndexIterator, org.apache.lucene.search.DocIdSetIterator).
    static void
    scoreAndCollectAll(org.apache.lucene.search.KnnCollector knnCollector, org.apache.lucene.search.AcceptDocs acceptDocs, org.apache.lucene.search.VectorScorer vectorScorer)
    Scores all vectors using the provided scorer and collects batches of documents when the batch's maxScore beats the collector's minCompetitiveSimilarity.
    static org.apache.lucene.search.VectorScorer
    sparseVectorScorer(org.apache.lucene.util.hnsw.RandomVectorScorer scorer, org.apache.lucene.index.KnnVectorValues.DocIndexIterator iterator)
    VectorScorer for sparse vectors; bulk path uses VectorScorer.Bulk.fromRandomScorerSparse(org.apache.lucene.util.hnsw.RandomVectorScorer, org.apache.lucene.index.KnnVectorValues.DocIndexIterator, org.apache.lucene.search.DocIdSetIterator).

    Methods inherited from class java.lang.Object

    clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

  • Method Details

    • denseVectorScorer

      public static org.apache.lucene.search.VectorScorer denseVectorScorer(org.apache.lucene.util.hnsw.RandomVectorScorer scorer, org.apache.lucene.index.KnnVectorValues.DocIndexIterator iterator)
      VectorScorer for dense vectors; bulk path uses VectorScorer.Bulk.fromRandomScorerDense(org.apache.lucene.util.hnsw.RandomVectorScorer, org.apache.lucene.index.KnnVectorValues.DocIndexIterator, org.apache.lucene.search.DocIdSetIterator).
      Parameters:
      scorer - scorer for the values behind iterator
      iterator - iterator over the same values instance as scorer

    • sparseVectorScorer

      public static org.apache.lucene.search.VectorScorer sparseVectorScorer(org.apache.lucene.util.hnsw.RandomVectorScorer scorer, org.apache.lucene.index.KnnVectorValues.DocIndexIterator iterator)
      VectorScorer for sparse vectors; bulk path uses VectorScorer.Bulk.fromRandomScorerSparse(org.apache.lucene.util.hnsw.RandomVectorScorer, org.apache.lucene.index.KnnVectorValues.DocIndexIterator, org.apache.lucene.search.DocIdSetIterator).
      Parameters:
      scorer - scorer for the values behind iterator
      iterator - iterator over the same values instance as scorer

    • scoreAndCollectAll

      public static void scoreAndCollectAll(org.apache.lucene.search.KnnCollector knnCollector, org.apache.lucene.search.AcceptDocs acceptDocs, org.apache.lucene.search.VectorScorer vectorScorer) throws IOException
      Scores all vectors using the provided scorer and collects batches of documents when the batch's maxScore beats the collector's minCompetitiveSimilarity.
      Parameters:
      knnCollector - the collector to collect scored vectors
      acceptDocs - the accept docs to filter vectors
      vectorScorer - the vector scorer
      Throws:
      IOException - if an I/O error occurs