結局Goの文字列結合はどれが一番早いのか

benchmark_bytebuffer_test.go

package main

import (
        "bytes"
        "fmt"
        "testing"
)

var m = [...]string{
        "AAAAAAAAA",
        "AAAAAAAAA",
        "AAAAAAAAA",
        "AAAAAAAAA",
        "AAAAAAAAA",
        "AAAAAAAAA",
        "AAAAAAAAA",
}

func BenchmarkSprintf(b *testing.B) {
        for i := 0; i < b.N; i++ {
                _ = fmt.Sprintf("%s", m)
        }

}

func BenchmarkCapBytesBuffer(b *testing.B) {
        for i := 0; i < b.N; i++ {
                var m2 = bytes.NewBuffer(make([]byte, 0, 100))
                for _, v := range m {
                        m2.WriteString(v)
                }
                _ = m2.String()
        }
}

func BenchmarkCapByteArray(b *testing.B) {
        for i := 0; i < b.N; i++ {
                var m2 = make([]byte, 0, 100)
                for _, v := range m {
                        m2 = append(m2, v...)
                }
                _ = string(m2)
        }
}

手元のMacBookProでとりあえず計測

BenchmarkSprintf         1000000              1479 ns/op
BenchmarkCapBytesBuffer  3000000               420 ns/op
BenchmarkCapByteArray   10000000               188 ns/op
ok      _/tmp/performance-sprintf-vs-bytebuffer 5.248s

BenchmarkSprintf         1000000              1433 ns/op
BenchmarkCapBytesBuffer  3000000               415 ns/op
BenchmarkCapByteArray   10000000               184 ns/op
ok      _/tmp/performance-sprintf-vs-bytebuffer 5.160s

BenchmarkSprintf         1000000              1476 ns/op
BenchmarkCapBytesBuffer  3000000               443 ns/op
BenchmarkCapByteArray   10000000               179 ns/op
ok      _/tmp/performance-sprintf-vs-bytebuffer 5.248s

結果、やっぱりキャパシティ付きのByte型appendが最速