source/stringcodec/utf8.cpp

   1 /* $Id$
   2
   3 This file is part of libmspstrings
   4 Copyright © 2006-2007 Mikko Rasa
   5 Distributed under the LGPL
   6 */
   7
   8 #include "utf8.h"
   9
  10 using namespace std;
  11
  12 namespace Msp {
  13 namespace Codecs {
  14
  15 void Utf8::Encoder::encode_char(UnicodeChar ch, string &buf)
  16 {
  17         if(ch<0 || ch>0x10FFFF)
  18                 return error(ch, buf, "Can't express character in UTF-8");
  19
  20         unsigned bytes = 1;
  21         if(ch>0xFFFF)
  22                 bytes = 4;
  23         else if(ch>0x7FF)
  24                 bytes = 3;
  25         else if(ch>0x7F)
  26                 bytes = 2;
  27
  28         if(bytes==1)
  29                 buf += ch;
  30         else
  31         {
  32                 char utf[4];
  33
  34                 utf[0] = 0xFF<<(8-bytes) | ch>>(bytes*6-6);
  35                 for(unsigned j=bytes-1; j>0; --j)
  36                 {
  37                         utf[j] = 0x80 | (ch&0x3F);
  38                         ch >>= 6;
  39                 }
  40
  41                 buf.append(utf, bytes);
  42         }
  43 }
  44
  45 void Utf8::Encoder::transliterate(UnicodeChar, string &buf)
  46 {
  47         buf.append("\357\277\275", 3);  // � U+FFFE Replacement Character
  48 }
  49
  50
  51 UnicodeChar Utf8::Decoder::decode_char(const string &str, string::const_iterator &i)
  52 {
  53         if(i==str.end())
  54                 return error("No input");
  55
  56         if((*i&0xC0)==0x80)
  57         {
  58                 UnicodeChar result = error("UTF-8 tail byte found when expecting head");
  59                 ++i;
  60                 return result;
  61         }
  62         else if(*i&0x80)
  63         {
  64                 unsigned bytes = 2;
  65                 unsigned mask = 0x20;
  66                 for(; *i&mask; mask>>=1)
  67                         ++bytes;
  68
  69                 string::const_iterator j = i;
  70
  71                 UnicodeChar result = (*j++)&(mask-1);
  72
  73                 unsigned k;
  74                 for(k=1; (k<bytes && j!=str.end() && (*j&0xC0)==0x80); ++k)
  75                         result = (result<<6) | ((*j++)&0x3F);
  76
  77                 if(k<bytes)
  78                         result = error("Incomplete UTF-8 character");
  79                 else if(!(result>>(bytes*5-4)) || !(result>>7))
  80                         result = error("Denormalized UTF-8 multibyte sequence");
  81                 else if(result>0x10FFFF || (result>=0xD800 && result<=0xDFFF))
  82                         result = error("Invalid Unicode code point");
  83
  84                 i = j;
  85                 return result;
  86         }
  87         else
  88                 return *i++;
  89 }
  90
  91 } // namespace Codecs
  92 } // namespace Msp