A better compressed bitset
Roaring bitmaps are compressed bitmaps that combine the speed of uncompressed bitsets with the memory efficiency of compression — often hundreds of times faster than older formats.
// Java — RoaringBitmap
import org.roaringbitmap.RoaringBitmap;
RoaringBitmap a = RoaringBitmap.bitmapOf(1, 2, 3, 1000);
RoaringBitmap b = RoaringBitmap.bitmapOf(2, 3, 4, 1000);
RoaringBitmap c = RoaringBitmap.and(a, b);
System.out.println(c.getCardinality()); // 3
// C++ — CRoaring
#include "roaring/roaring.hh"
using roaring::Roaring;
Roaring a = Roaring::bitmapOf(4, 1, 2, 3, 1000);
Roaring b = Roaring::bitmapOf(4, 2, 3, 4, 1000);
Roaring c = a & b;
std::cout << c.cardinality() << std::endl;
// Go — roaring
import "github.com/RoaringBitmap/roaring/v2"
a := roaring.BitmapOf(1, 2, 3, 1000)
b := roaring.BitmapOf(2, 3, 4, 1000)
c := roaring.And(a, b)
fmt.Println(c.GetCardinality())
# Python — PyRoaringBitMap
from pyroaring import BitMap
a = BitMap([1, 2, 3, 1000])
b = BitMap([2, 3, 4, 1000])
print(len(a & b)) # 3
Compressed bitsets, without the trade-offs
Roaring bitmaps combine the speed of uncompressed bitsets with the memory efficiency of compression — and consistently outperform older formats like WAH, EWAH, and Concise.
Fast by design
Random access in constant time. Set operations are implemented as native bitwise AND/OR/ANDNOT on dense chunks — often hundreds of times faster than run-length-encoded alternatives.
Adaptive storage
Each 65 536-integer chunk picks the best container — array, bitmap, or run — so dense, sparse, and clustered data are all encoded efficiently and without manual tuning.
Portable format
A published serialization specification means bitmaps written by one implementation can be read by any other — across Java, C, Go, Rust, Python, and more.
Battle-tested at scale
Powering YouTube's SQL engine, Apache Lucene, Druid, Spark, ClickHouse, Elasticsearch, Pinot, and dozens more — handling billions of operations daily.
Backed by research
Published in Software: Practice and Experience and recommended by independent benchmarking studies. Read the papers.
Open community
Apache 2.0 licensed. Developed in the open on GitHub by 40+ contributors across academia and industry. Discussion happens on the user group.
Use Roaring for bitmap compression whenever possible. Do not use other bitmap compression methods.
Used by the systems you already know
Roaring is a building block in many of the world's most popular databases, search engines, and analytics platforms.
Available in every major language
Maintained reference libraries and community ports cover the entire ecosystem. All implementations are interoperable through the shared format specification.
Get started in seconds
Add Roaring to your favorite language — the API is small, the format is portable, and the performance is uncompromising.