Compiler Explorer

Source code

#include <uchar.h>
#include <stddef.h>
#include <stdint.h>

// UTF-8 ranges and tags for encoding characters
#define MAX_ONE_B 0x80
#define MAX_TWO_B 0x800
#define MAX_THREE_B 0x10000
#define TAG_CONT 0b10000000
#define TAG_TWO_B 0b11000000
#define TAG_THREE_B 0b11100000
#define TAG_FOUR_B 0b11110000

size_t len_utf8(const char32_t code) {
    if (code < MAX_ONE_B) return 1;
    else if (code < MAX_TWO_B) return 2;
    else if (code < MAX_THREE_B) return 3;
    return 4;
}

size_t encode_utf8(const char32_t code, unsigned char *dst, const size_t cap) {
    const size_t len = len_utf8(code);
    if (len > cap) {
        return 0;
    }

switch (len) {
    // 1-byte sequence (ASCII)
    case 1: {
        dst[0] = (unsigned char)code;
        break;
    }
    // 2-byte sequence
    case 2: {
        dst[0] = ((unsigned char)(code >> 6 & 0x1F) | TAG_TWO_B);
        dst[1] = ((unsigned char)(code & 0x3F) | TAG_CONT);
        break;
    }
    // 3-byte sequence
    case 3: {
        dst[0] = ((unsigned char)(code >> 12 & 0x0F) | TAG_THREE_B);
        dst[1] = ((unsigned char)(code >> 6 & 0x3F) | TAG_CONT);
        dst[2] = ((unsigned char)(code & 0x3F) | TAG_CONT);
        break;
    }
    // 4-byte sequence
    case 4: {
        dst[0] = ((unsigned char)(code >> 18 & 0x07) | TAG_FOUR_B);
        dst[1] = ((unsigned char)(code >> 12 & 0x3F) | TAG_CONT);
        dst[2] = ((unsigned char)(code >> 6 & 0x3F) | TAG_CONT);
        dst[3] = ((unsigned char)(code & 0x3F) | TAG_CONT);
        break;
    }
    default: {
        __builtin_unreachable();
    }
    }
    return len;
}