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

Interface PartitionedDocValues

All Known Implementing Classes:
AbstractTSDBDocValuesProducer.BaseSortedDocValues
public interface PartitionedDocValues
An extension for doc-values which is the primary sort key, and values are grouped by some prefix bytes. One example is `_tsid` in time-series, where time-series can be grouped by the first few bytes, and the query engine can process a slice of time-series containing all "continuous" time-series sharing the same prefix. This is necessary instead of querying a slice of docs because some aggregations such as rate require all data points from a single time-series to be processed by the same operator.

  • Method Details

    • prefixPartitions

      @Nullable PartitionedDocValues.PrefixPartitions prefixPartitions(PartitionedDocValues.PrefixPartitions reused) throws IOException
      Returns the prefixed partition from the doc-values of this field if exists.
      Parameters:
      reused - an existing prefix partitions can be reused to avoid allocating memory
      Throws:
      IOException

    • prefixPartitionBits

      int prefixPartitionBits()
      The number of bits used in prefix partitions

    • canPartitionByTsidPrefix

      static boolean canPartitionByTsidPrefix(org.apache.lucene.search.IndexSearcher searcher) throws IOException
      Check if the given index searcher can be partitioned by tsid prefix.
      Parameters:
      searcher - the index searcher to check
      Returns:
      true if all non-empty segments support tsid prefix partitioning
      Throws:
      IOException