#include "stdio.h"
#include "unistd.h"

/*
 *  utf8conv.c - Program to convert UTF-8 encoded files to use
 *     XML entity references to the Unicode characters instead.
 *
 *  Reads from STDIN, outputs to STDOUT. 
 *
 *  To make: make utf8conv (yes, that simple)
 *
 *  TODO:  Add error checking
 *         Allow it to read/write (multiple)? files
 *         -h usage message
 * 
 *
 * DWS - 10/10/01
 * */


#define BUFSIZE 2048

char * buf;

/* int bget(int handle) 
 *
 * Reads a byte from the file handle, buffering things so it is reasonably 
 * fast. Should handle most cases of input underrun, but I cannot 
 * gaurantee it.
 * 
 */

int bget(in) {
  
  static i,size;
  if (!buf) {
    if ((buf = (char *)malloc (BUFSIZE))==NULL) {
      fprintf(stderr,"Unable to malloc %d bytes.\n\n", BUFSIZE);
      exit(1);
    }
    
    i=size=0;
  }

  if (i==size) {
    if ((size = read(in, buf, BUFSIZE)) < 0) {
      perror("Unable to read: " );
      exit(1);
    }
    
    i=0;
  }
  
  if (i<size) {
    return (*((unsigned char *)(buf+(i++))));
  } 
  else {
    return -1;
  }
}


int main (int argc, char **argv){ 
  int in, byte;
  
  in =0;  

  while ((byte = bget(in)) >= 0) {
    if (byte < 128) {
      /* 7 bit ASCII */
      putchar(byte);
    } 
    else {
      if ((byte >= 192) && (byte <224)) {
        /* Two byte character */
        int word;
        word = (byte & 31) * 64;
        word +=( bget (in) & 63);
        printf ("&#x%0*x;",(word < 0x100) ? 2 : 4 , word);
      } 
      else if (byte >= 224) {
        /* Three byte character */
        int word;
        word = (byte & 15) * 4096;
        word +=( bget (in) & 63) * 64;
        word +=( bget (in) & 63) ;
        printf ("&#x%0*x;", (word < 0x100) ? 2 : 4 ,word);
      }
    }
  }
  return 0;
}