Module org.elasticsearch.tdigest

Package org.elasticsearch.tdigest

package org.elasticsearch.tdigest

T-Digest library

This package contains a fork for the [T-Digest](https://github.com/tdunning/t-digest) library that's used for percentile calculations. Forking the library allows addressing bugs and inaccuracies around both the AVL- and the merging-based implementations that unblocks switching from the former to the latter, with substantial performance gains. It also unlocks the use of BigArrays and other ES-specific functionality to account for resources used in the digest data structures.
  • Class
    Description
    AbstractTDigest
     
    AVLTreeDigest
     
    Centroid
    A single centroid which represents a number of data points.
    Dist
    Reference implementations for cdf and quantile if we have all data sorted.
    HybridDigest
    Uses a SortingDigest implementation under the covers for small sample populations, then switches to MergingDigest.
    MergingDigest
    Maintains a t-digest by collecting new points in a buffer that is then sorted occasionally and merged into a sorted array that contains previously computed centroids.
    ScaleFunction
    Encodes the various scale functions for t-digests.
    Sort
    Static sorting methods
    SortingDigest
    Simple implementation of the TDigest interface that stores internally and sorts all samples to calculate quantiles and CDFs.
    TDigest
    Adaptive histogram based on something like streaming k-means crossed with Q-digest.