Skip to content

README.md

Latest commit

 

History

History
78 lines (60 loc) · 2.44 KB

README.md

File metadata and controls

78 lines (60 loc) · 2.44 KB

BBHash

BBHash is a fast Go implementation of a minimal perfect hash function for large key sets.

Installing the module for use in your own project

% go get github.com/relab/bbhash

How to use the package

import (
	"fmt"

	"github.com/relab/bbhash"
)

func ExampleBBHash_Find() {
	keys := []uint64{1, 2, 3, 4, 5, 6, 7, 8, 9, 10}
	bb, err := bbhash.New(keys, bbhash.Gamma(1.5))
	if err != nil {
		panic(err)
	}
	for _, key := range keys {
		hashIndex := bb.Find(key)
		fmt.Printf("%d, ", hashIndex)
	}
	fmt.Println()
	// Output:
	// 2, 6, 8, 3, 5, 7, 1, 9, 10, 4,
}

Advanced usage

The bbhash.New function takes a slice of keys as its first argument. The keys must be unique and of type uint64. New also takes zero or more bbhash.Option arguments. These are the available options:

Option Description
Gamma(float64) Set the gamma parameter of the BBHash algorithm. Default is 2.0.
InitialLevels(int) Set the initial number of levels in the BBHash algorithm. Default is 32.
Partitions(int) Set the number of partitions to split the keys into and compute parallel.
WithReverseMap() Create a reverse map that allows you to retrieve the key from the hash index.
Parallel() Use parallelism in the BBHash algorithm. Prefer the Partitions option instead.

The options can be combined like this:

bb, err := bbhash.New(keys)
bb, err := bbhash.New(keys, bbhash.Gamma(1.5), bbhash.InitialLevels(64))
bb, err := bbhash.New(keys, bbhash.InitialLevels(20))
bb, err := bbhash.New(keys, bbhash.Gamma(1.5), bbhash.Partitions(4))
bb, err := bbhash.New(keys, bbhash.Gamma(1.5), bbhash.Partitions(4), bbhash.WithReverseMap())

But the following combinations are not supported:

bb, err := bbhash.New(keys, bbhash.Parallel(), bbhash.Partitions(4))
bb, err := bbhash.New(keys, bbhash.Parallel(), bbhash.WithReverseMap())

Credits

Implemented by Hein Meling.

The implementation is mainly inspired by Sudhi Herle's Go implementation. Damian Gryski also has a Go implementation.

The algorithm is described in the paper: Fast and Scalable Minimal Perfect Hashing for Massive Key Sets Antoine Limasset, Guillaume Rizk, Rayan Chikhi, and Pierre Peterlongo. Their C++ implementation is available here.