#collections #probabilistic


Implementation of HyperBitBit data structure

2 releases

0.0.1-alpha.2 Dec 24, 2020
0.0.1-alpha.1 Dec 19, 2020

#813 in Data structures


94 lines


ci-badge Crates.io Documentation

A native rust implementation of a HyperBitBit algorithm introduced by by Robert Sedgewick in AC11-Cardinality.pdf


  • HyperBitBit, for N < 2^64
  • Uses 128 + 8 bit of space
  • Estimated cardinality withing 10% or actuals for large N.

Consider HyperLogLog variants for productions usage, sine this data structure extensively studied, merge able and more accurate. HyperBitBit is extremely cheap and fast alternative in expense of accuracy.


This crate is on crates.io and can be used by adding hyperbitbit to the dependencies in your project's Cargo.toml.

hyperbitbit = "0.0.1-alpha.1"

If you want serde support, include the feature like this:

hyperbitbit = { version = "0.0.1-alpha.1", features = ["serde_support"] }

add this to your crate root:

extern crate hyperbitbit;


Create a HyperBitBit, add more data and estimate cardinality

use hyperbitbit::HyperBitBit;
use rand::distributions::Alphanumeric;
use rand::prelude::*;
use std::collections::HashSet;

fn main() {
    // create hbb for cardinality estimation
    let mut hbb = HyperBitBit::new();
    // hash set to measure exact cardinality
    let mut items = HashSet::new();
    // fixed seed for RNG for reproducibly results
    let mut rng = StdRng::seed_from_u64(42);
    // put bunch of random strings on hbb and hash set for comparison
    let maxn = 10000;
    for _ in 1..maxn {
        let s = (&mut rng).sample_iter(&Alphanumeric).take(4).collect::<String>();

    let expected: i64 = items.len() as i64;
    let rel: f64 = (100.0 * (expected - hbb.cardinality() as i64) as f64) / (expected as f64);

    println!("Actuals cardinality:   {:?}", expected);
    println!("Estimated cardinality: {:?}", hbb.cardinality());
    println!("Error % cardinality:   {:.2}", rel);


Licensed under the Apache License, Version 2.0