UTF8 to Unicode codepoint

utf8_to_unicode_cp.cpp

#include <iostream>
#include <string>

using namespace std;

uint32_t to_unicode(size_t *bytes, char const *utf8)
{
  if (!utf8 || !*utf8) {
    *bytes = 0;
    return 0;
  }
  uint32_t unicode = 0;
  uint8_t first_byte = utf8[0];
  uint8_t len = 
    (first_byte >> 7) == 0 ? 1 :
    (first_byte & 0xf0) == 0xf0 ? 4 :
    (first_byte & 0xe0) == 0xe0 ? 3 :
    (first_byte & 0xc0) == 0xc0 ? 2 : 0
    ;
  unicode += (uint8_t)(first_byte << len) >> len;
  for(auto i = 1; utf8[i] && i < len; ++i) {
    unicode <<= 6;
    unicode += utf8[i] & 0x3F;
  }
  *bytes = len;
  return unicode;
}

int main() {
  char const *s = "中文測試a";
  size_t bytes = 0;
  
  while(*s) {
    uint32_t code = to_unicode(&bytes, s);
    cout << code << " " << bytes << endl;
    s += bytes;
  }
  return 0;
}