trie: The fast and flexible Trie Tree implementation

The Trie Tree implementation in Go. It has flexible interface and works fast as Radix Tree implementation.

This is basically an implementation of the algorithm described in 簡単なトライ - LINE ENGINEERING. I really appreciate the amazing ideas and the clear and easy-to-understand explanation.

Import Path

github.com/acomagu/trie/v2

Benchmark

Run make bench to run it locally. The latest benchmark result is here.

Exact Matching

The task is to determine whether a string matches one of all Wikipedia titles.

Package	Time	Objects Allocated
acomagu/trie	1090 ns/op	0 allocs/op
sauerbraten/radix	2445 ns/op	0 allocs/op
dghubble/trie	2576 ns/op	0 allocs/op
hashicorp/go-immutable-radix	3660 ns/op	0 allocs/op
derekparker/trie	4010 ns/op	0 allocs/op
armon/go-radix	11745 ns/op	0 allocs/op
kkdai/radix	18809 ns/op	0 allocs/op
tchap/go-patricia/patricia	21498 ns/op	0 allocs/op

Longest Prefix

The task is to answer which of all Wikipedia titles can be the longest prefix of a string.

Package	Time	Objects Allocated
acomagu/trie	140 ns/op	0 allocs/op
hashicorp/go-immutable-radix	159 ns/op	0 allocs/op
tchap/go-patricia/patricia	252 ns/op	0 allocs/op
derekparker/trie	2374 ns/op	0 allocs/op
sauerbraten/radix	3264938 ns/op	0 allocs/op
armon/go-radix	22129827 ns/op	1 allocs/op

(dghubble/trie and kkdai/radix don't have way to do.)

Build

The task is to prepare Trie/Radix Tree with all of the Wikipedia titles.

Package	Time	Objects Allocated
sauerbraten/radix	118959250 ns/op	408564 allocs/op
acomagu/trie	542902000 ns/op	421906 allocs/op
dghubble/trie	609406300 ns/op	1136281 allocs/op
derekparker/trie	1046705400 ns/op	1801539 allocs/op
armon/go-radix	1750312500 ns/op	1446050 allocs/op
kkdai/radix	2280362300 ns/op	1742841 allocs/op
tchap/go-patricia/patricia	2898335700 ns/op	1150947 allocs/op
hashicorp/go-immutable-radix	7614342400 ns/op	45097986 allocs/op

Examples

The common preparation for each example:

import (
	"fmt"

	"github.com/acomagu/trie/v2"
)

keys := [][]byte{
	[]byte("ab"),
	[]byte("abc"),
	[]byte("abd"),
}
values := []int{1, 2, 3} // The type of value doesn't have to be int. Can be anything.
t := trie.New(keys, values)

New() takes keys and values as the arguments. values[i] is the value of the corresponding key, keys[i].

Exact Match

v, ok := t.Trace([]byte("abc")).Terminal()
fmt.Println(v, ok) // => 2 true

Playground

Longest Prefix

var v int
var match bool
for _, c := range []byte("abcxxx") {
	if t = t.TraceByte(c); t == nil {
		break
	}
	if vv, ok := t.Terminal(); ok {
		v = vv
		match = true
	}
}

fmt.Println(v, match) // => 2 true

Playground

No special function to get longest prefix because it can be implemented yourself easily using the existing methods.