utf8.c

#include "utf8.h"
#include <errno.h>

uint8_t
isutf8(const unsigned char *string)
{
  int nbytes;
  unsigned char *ptr = (unsigned char *) string;
  while(*ptr) {
    nbytes = 1;
#define SEQ(shift, result, increment) \
  if(((*ptr) >> (shift)) == (result)) {\
    while(nbytes < (increment)) {\
      if((*(ptr + nbytes) >> 6) != 2) {\
        return 0;\
      }\
      nbytes++;\
    }\
    ptr += (increment);\
    continue;\
  }

    SEQ(7, 0x00, 1) /* U+0000 U+007F */
    SEQ(5, 0x06, 2) /* U+0080 U+07FF */
    SEQ(4, 0x0e, 3) /* U+0800 U+FFFF */

    /* U+10000 U+10FFFF */
    /* check four byte format */
    if(((*ptr) >> (3)) == (0x1e)) {
      /* handle upper limit *
       * from RFC3629       */
      if(*(ptr) >= 0xf4) {
        return 0;
      }

      /* for each continuation byte ... */
      while(nbytes < (4)) {
        /* does it start correctly? */
        if((*(ptr + nbytes) >> 6) != 2) {
          return 0;
        }

        nbytes++; /* next byte */
      }

      ptr += (4); /* next character */
      continue;
    }
#undef SEQ

    errno = EINVAL;
    return 0; /* catch invalid byte */
  }

  return 1;
}

int32_t
utf8len(const unsigned char *string)
{
  int len = 0, nbytes;
  unsigned char *ptr = (unsigned char*) string;
  while(*ptr) {
    nbytes = 1;
#define SEQ(shift, result, increment) \
  if(((*ptr) >> (shift)) == (result)) {\
    while(nbytes < (increment)) {\
      if((*(ptr + nbytes) >> 6) != 2) {\
        errno = EINVAL;\
        return -1;\
      }\
      nbytes++;\
    }\
    len++; ptr += (increment);\
    continue;\
  }

    SEQ(7, 0x00, 1) /* U+0000 U+007F */
    SEQ(5, 0x06, 2) /* U+0080 U+07FF */
    SEQ(4, 0x0e, 3) /* U+0800 U+FFFF */

    /* U+10000 U+10FFFF */
    /* check four byte format */
    if(((*ptr) >> (3)) == (0x1e)) {
      /* handle upper limit *
       * from RFC3629       */
      if(*(ptr) >= 0xf4) {
        return 0;
      }

      /* for each continuation byte ... */
      while(nbytes < (4)) {
        /* does it start correctly? */
        if((*(ptr + nbytes) >> 6) != 2) {
          errno = EINVAL;
          return -1;
        }

        nbytes++; /* next byte */
      }

      len++;
      ptr += (4); /* next character */
      continue;
    }
#undef SEQ

    errno = EINVAL;
    return -1;
  }

  return len;
}

utf8.h

#ifndef _UTF_8_H
#define _UTF_8_H

#include <stdint.h>

/**
 *
 */
uint8_t isutf8(const unsigned char*);

/**
 *
 */
int32_t utf8len(const unsigned char*);

#endif