This is a small benchmark measuring fmt.Sscanf() abilities to parse an integer, vs. strconv.ParseInt(). Run with "go test -bench=." (no quotes) and see for yourself; results are amazing... Sscanf() is 40 times slower parsing a simple integer than ParseInt().

result.txt

BenchmarkIntParse       50000000                57.1 ns/op
BenchmarkIntScanf        1000000              1911 ns/op
BenchmarkStrScanf        1000000              1675 ns/op

scanf_test.go

package scanf

import (
	"fmt"
	"strconv"
	"testing"
)

const str = "123423434"

func BenchmarkIntParse(b *testing.B) {
	for i := 0; i < b.N; i++ {
		bn, err := strconv.ParseInt(str, 10, 0)
		if err != nil {
			b.Fatal(err)
		}

		// We convert to int just to have exactly the same result in both
		// tests, thus benchmarking equivalent functions. I'm using the extra
		// assignment to get rid of the "var unused" compiler error. Note,
		// however, that only the first assignment generates an actual
		// assembler instruction. So, there's no need to worry about skewing
		// the test by the weird n = n.
		n := int(bn)
		n = n
	}
}

func BenchmarkIntScanf(b *testing.B) {
	for i := 0; i < b.N; i++ {
		var n int

		if _, err := fmt.Sscanf(str, "%d", &n); err != nil {
			b.Fatal(err)
		}
	}
}

func BenchmarkStrScanf(b *testing.B) {
	for i := 0; i < b.N; i++ {
		var s string

		if _, err := fmt.Sscanf(str, "%s", &s); err != nil {
			b.Fatal(err)
		}
	}
}