utf8.c

00001 /****************************************************************************
00002  *                               utf8.c
00003  * This file contains functions to decode utf8 characters in the input 
00004  * stream.
00005  *
00006  **************************************************************************/
00007 
00008 #include "utf8.h"
00009 #include <stdio.h>
00010 
00017 #define CONT(i) ((in[i]&0xc0) == 0x80)
00018 #define VAL(i, s) ((in[i]&0x3f) << s)
00019 
00020 char *
00021 __utf8_get_char(const char *in, int *chr)
00022 {                                       /* 2-byte, 0x80-0x7ff */
00023   if ( (in[0]&0xe0) == 0xc0 && CONT(1) )
00024   { *chr = ((in[0]&0x1f) << 6)|VAL(1,0);
00025     return (char *)in+2;
00026   }
00027                                         /* 3-byte, 0x800-0xffff */
00028   if ( (in[0]&0xf0) == 0xe0 && CONT(1) && CONT(2) )
00029   { *chr = ((in[0]&0xf) << 12)|VAL(1,6)|VAL(2,0);
00030     return (char *)in+3;
00031   }
00032                                         /* 4-byte, 0x10000-0x1FFFFF */
00033   if ( (in[0]&0xf8) == 0xf0 && CONT(1) && CONT(2) && CONT(3) )
00034   { *chr = ((in[0]&0x7) << 18)|VAL(1,12)|VAL(2,6)|VAL(3,0);
00035     return (char *)in+4;
00036   }
00037                                         /* 5-byte, 0x200000-0x3FFFFFF */
00038   if ( (in[0]&0xfc) == 0xf8 && CONT(1) && CONT(2) && CONT(3) && CONT(4) )
00039   { *chr = ((in[0]&0x3) << 24)|VAL(1,18)|VAL(2,12)|VAL(3,6)|VAL(4,0);
00040     return (char *)in+5;
00041   }
00042                                         /* 6-byte, 0x400000-0x7FFFFFF */
00043   if ( (in[0]&0xf8) == 0xf0 && CONT(1) && CONT(2) && CONT(3) && CONT(4) && CONT(5) )
00044   { *chr = ((in[0]&0x1) << 30)|VAL(1,24)|VAL(2,18)|VAL(3,12)|VAL(4,6)|VAL(5,0);
00045     return (char *)in+4;
00046   }
00047 
00048   *chr = *in;
00049 /*fprintf(stderr, "UTF-8 decoder: bad sequence\n");*/
00050   
00051   return (char *)in+1;
00052 }

Generated on Wed Jul 26 13:30:46 2006 for XSB by  doxygen 1.4.5