GoLLRB is a Left-Leaning Red-Black (LLRB) implementation of 2-3 balanced binary search trees in Go Language.

As of this writing and to the best of the author's knowledge, Go still does not have a balanced binary search tree (BBST) data structure. These data structures are quite useful in a variety of cases. A BBST maintains elements in sorted order under dynamic updates (inserts and deletes) and can support various order-specific queries. Furthermore, in practice one often implements other common data structures like Priority Queues, using BBST's.

2-3 trees (a type of BBST's), as well as the runtime-similar 2-3-4 trees, are the de facto standard BBST algoritms found in implementations of Python, Java, and other libraries. The LLRB method of implementing 2-3 trees is a recent improvement over the traditional implementation. The LLRB approach was discovered relatively recently (in 2008) by Robert Sedgewick of Princeton University.

GoLLRB is a Go implementation of LLRB 2-3 trees.

GoLLRB has been used in some pretty heavy-weight machine learning tasks over many gigabytes of data. I consider it to be in stable, perhaps even production, shape. There are no known bugs.

With a healthy Go Language installed, simply run go get github.com/petar/GoLLRB/llrb

package main

import (
	"fmt"
	"github.com/petar/GoLLRB/llrb"
)

func lessInt(a, b interface{}) bool { return a.(int) < b.(int) }

func main() {
	tree := llrb.New(lessInt)
	tree.ReplaceOrInsert(1)
	tree.ReplaceOrInsert(2)
	tree.ReplaceOrInsert(3)
	tree.ReplaceOrInsert(4)
	tree.DeleteMin()
	tree.Delete(4)
	c := tree.IterAscend()
	for {
		u := <-c
		if u == nil {
			break
		}
		fmt.Printf("%d\n", int(u.(int)))
	}
}

GoLLRB was written by Petar Maymounkov.

Follow me on Twitter @maymounkov!