UTF-8 string character count in C/C++

The following code defines, tests and illustrates the use of utf8len() function. Which is a small piece of code for counting characters in UTF-8 (multibyte) string.

Compile this example with GCC by running: $ gcc utf8len.c -lrt -o utf8len

The RT library is used for the high precision clock only, you don't need to link it if you are using the function itself into your own code. This utf8len() function provides a portable (and small footprint) way of counting UTF-8 charactes in standard C or C++. This test source code has UTF-8 characters, you have to check the source file doesn't get corrupted when copy/pasting the code. Average time for functions have a small overhead from the for loop. But the overhead is the same for both tests.

#include <stdio.h>
#include <string.h>
#include <time.h>

#define STRINGS 8
#define TIMES 1000000

// Be sure your locale is set to "en_US.UTF-8" or whatever language with UTF-8

size_t utf8len(const char *s);
void print_diff(const char *func, int times, struct timespec *s, struct timespec *e);

const char * list[] = {"english", "español", "עברית", "Ελληνικά", "Українська", "한국어", "ﻑﺍﺮﺳی", "日本語"};

int main(int argc, char ** argv)
{
  int i;
  struct timespec start, end;

  for(i=0; i < STRINGS; ++i)
    printf("%s\nstrlen(): %zu \t utf8len(): %zu\n", list[i], strlen(list[i]), utf8len(list[i]));


  clock_gettime(CLOCK_REALTIME, &start);
  for(i=0; i<TIMES; ++i)
    strlen(list[i%STRINGS]);
  clock_gettime(CLOCK_REALTIME, &end);
  print_diff("strlen", TIMES, &start, &end);

  clock_gettime(CLOCK_REALTIME, &start);
  for(i=0; i<TIMES; ++i)
    utf8len(list[i%STRINGS]);
  clock_gettime(CLOCK_REALTIME, &end);
  print_diff("utf8len", TIMES, &start, &end);

  return 0;
}


size_t utf8len(const char *s)
{
  size_t len = 0;
  while(*s)
    len += (*(s++)&0xC0)!=0x80;
  return len;
}

// Aux

void print_diff(const char *func, int times, struct timespec *s, struct timespec *e)
{
  printf("Average %s time was: %6.3f nanoseconds\n", func, ((double)((e->tv_sec-s->tv_sec)*1000000000 ) + (e->tv_nsec-s->tv_nsec)) / times);
}