utf decode

utf_decode.cxx

// https://en.wikipedia.org/wiki/UTF-8

#include <iostream>
#include <string>
#include <iomanip>

using u_char = unsigned char;

enum utf_mask: u_char
{
    char_part = 0b10000000,
    part_mask = 0b11000000,
    n1_value  = 0b00000000,
    n1_mask   = 0b10000000,
    n2_value  = 0b11000000,
    n2_mask   = 0b11100000,
    n3_value  = 0b11100000,
    n3_mask   = 0b11110000,
    n4_value  = 0b11110000,
    n4_mask   = 0b11111000,
};

bool eq(char c, u_char v, u_char m)
{
    return v == (c & m);
}

bool is_char_part(char c)
{
    return eq(c, char_part, part_mask);
}

int get_size(char c)
{
    if (is_char_part(c)) return -1;
    
    if (eq(c, n1_value, n1_mask)) return 1;
    if (eq(c, n2_value, n2_mask)) return 2;
    if (eq(c, n3_value, n3_mask)) return 3;
    if (eq(c, n4_value, n4_mask)) return 4;
    
    return -2;
}

int main()
{
    using namespace std::literals;
    
    std::string str = u8"абв";
    
    for (size_t i = 0; i < str.size();)
    {
        char c = str[i];
        int n = get_size(c);
        if (n <= 0)
        {
            std::cout << "<?>" << std::endl;
            ++i;
        }
        else
        {
            std::cout << str.substr(i, n) << std::endl;
            i += n;
        }
    }
}