Which is better to get the last X character of a Golang String?

Question

When I have string "hogemogehogemogehogemoge世界世界世界" which code is better to get last rune with avoiding memory allocation?

There are similar question about to get last X character of Golang String.

How to get the last X Characters of a Golang String?

I want to make sure which is prefered if I just want to get last rune, without any additional operation.

package main

import (
    "fmt"
    "unicode/utf8"
)

func main() {
    // which is more better for memory allocation?
    s := "hogemogehogemogehogemoge世界世界世界a"
    getLastRune(s, 3)
    getLastRune2(s, 3)
}

func getLastRune(s string, c int) {
    // DecodeLastRuneInString
    j := len(s)
    for i := 0; i < c && j > 0; i++ {
        _, size := utf8.DecodeLastRuneInString(s[:j])
        j -= size
    }
    lastByRune := s[j:]
    fmt.Println(lastByRune)
}

func getLastRune2(s string, c int) {
    // string -> []rune
    r := []rune(s)
    lastByRune := string(r[len(r)-c:])
    fmt.Println(lastByRune)
}

世界a

世界a

Answer 1

Whenever performance and allocations are the question, you should run benchmarks.

First let's modify your functions to not print but rather return the result:

func getLastRune(s string, c int) string {
    j := len(s)
    for i := 0; i < c && j > 0; i++ {
        _, size := utf8.DecodeLastRuneInString(s[:j])
        j -= size
    }
    return s[j:]
}

func getLastRune2(s string, c int) string {
    r := []rune(s)
    if c > len(r) {
        c = len(r)
    }
    return string(r[len(r)-c:])
}

And the benchmark functions:

var s = "hogemogehogemogehogemoge世界世界世界a"

func BenchmarkGetLastRune(b *testing.B) {
    for i := 0; i < b.N; i++ {
        getLastRune(s, 3)
    }
}

func BenchmarkGetLastRune2(b *testing.B) {
    for i := 0; i < b.N; i++ {
        getLastRune2(s, 3)
    }
}

Running them:

go test -bench . -benchmem

Results:

BenchmarkGetLastRune-4     30000000     36.9 ns/op     0 B/op    0 allocs/op
BenchmarkGetLastRune2-4    10000000    165 ns/op       0 B/op    0 allocs/op

getLastRune() is more than 4 times faster. Neither of them is making any allocations, but this is due to a compiler optimization (converting a string to []rune and back generally requires allocation).

If we run the benchmarks with optimizations disabled:

go test -gcflags '-N -l' -bench . -benchmem

Results:

BenchmarkGetLastRune-4     30000000    46.2 ns/op      0 B/op    0 allocs/op
BenchmarkGetLastRune2-4    10000000   197 ns/op       16 B/op    1 allocs/op

Compiler optimizations or not, getLastRune() is the clear winner.