# tdunning/t-digest

A new data structure for accurate on-line accumulation of rank-based statistics such as quantiles and trimmed means

Repository: https://github.com/tdunning/t-digest
Canonical: https://ross.abutalabs.com/products/t-digest
Language: Java
License: Apache-2.0
License Family: permissive
Topics: quantile, accuracy, online-algorithms, t-digest
Last push: 2025-02-17T07:19:38+00:00

## Health v2 (maintenance only)
Score: 35/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 7, release rhythm 35, longevity 100
- inputs: {"age_days": 4671, "days_push": 562, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2166, forks 232 (observed 2026-08-28T04:06:21.355819+00:00)

## What it is
A Java library implementing the t-digest data structure for accurate online accumulation of rank-based statistics such as quantiles and trimmed means. Digests are compact, fast, and mergeable, making them suitable for map-reduce and parallel streaming workloads.

## Use cases
- estimate quantiles of a large data stream in constant memory
- compute median and percentile statistics in Spark or map-reduce jobs
- monitor latency percentiles for high-throughput services
- compute trimmed means over streaming data
- merge distributed sketch summaries into one accurate summary

## When to choose
- you need highly accurate extreme quantiles (tails) from streaming or distributed data
- you need compact, mergeable summaries for parallel processing
- you work on the JVM and want a dependency-free, well-tested implementation

## When to avoid
- you need exact quantile values, not approximations
- you need histograms or other non-rank-based statistics
- you are not on the JVM and cannot use a Java port

## Facets
- artifact type: library
- maturity: stable
- function: data-science, analytics, streaming, serialization, benchmarking
- domain: data-science, analytics, big-data, performance
- platform: jvm, cross-platform
- tags: t-digest, quantile-estimation, sketching, streaming-statistics, mergeable-summaries, approximate-algorithms, algorithms

## Member repositories
- tdunning/t-digest (main) score 35

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:21.355819+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T02:49:57.570755+00:00, confidence not recorded.
  - readme: https://github.com/tdunning/t-digest (fetched 2026-08-28T04:06:21.355819+00:00, sha 63d6eb56569c)
- Data as of 2026-08-30T08:39:29.467469+00:00.