utf8decode.go

package main

import (
	"fmt"
)

func main() {
	fmt.Printf("U+%04X\n", decode([]byte{0x61}))
	fmt.Printf("U+%04X\n", decode([]byte{0xc2, 0xa1}))
	fmt.Printf("U+%04X\n", decode([]byte{0xea, 0x9c, 0xb3}))
	fmt.Printf("U+%04X\n", decode([]byte{0xf0, 0x9f, 0x8e, 0x87}))
}

func decode(buf []byte) rune {
	// Find the first zero bit
	
	if buf[0] & 0x80 == 0 {
		// ASCII
		return rune(buf[0])
	}

	if buf[0]&(0x80>>1) == 0 {
		return 0xFFFD // Invalid as the first byte.
	}

	var size int
	for i := 2; i <= 4; i++ {
		if buf[0]&(0x80>>i) == 0 {
			size = i
			break
		}
	}

	if size == 0 {
		return 0xFFFD
	}

	r := rune(buf[0]) & (0x7f >> size)

	for _, b := range buf[1:size] {
		// Make sure the two high bits are 1 and 0 respectively.
		if b&(0x80|0x40) != 0x80 {
			return 0xFFFD
		}

		// There are 6 bits of the code point in this byte. So shift
		// everything to make room.
		r <<= 6

		// Put the low 6 bits of b on the low 6 bits of char.
		r |= rune(b & (0xff >> 2))
	}

	return r
}