source/utf8.cpp

   1 #include "utf8.h"
   2
   3 using namespace std;
   4
   5 namespace Msp {
   6
   7 void Utf8::Encoder::encode_char(wchar_t c)
   8 {
   9         unsigned code=c;
  10         if(code>0x10FFFF)
  11                 throw CodecError("Can't express character in UTF-8");
  12
  13         unsigned bytes=1;
  14         if(code>0xFFFF)
  15                 bytes=4;
  16         else if(code>0x7FF)
  17                 bytes=3;
  18         else if(code>0x7F)
  19                 bytes=2;
  20
  21         if(bytes==1)
  22                 append(code);
  23         else
  24         {
  25                 char buf[4];
  26
  27                 buf[0]=0xFF<<(8-bytes) | code>>(bytes*6-6);
  28                 for(unsigned j=bytes-1; j>0; --j)
  29                 {
  30                         buf[j]=0x80 | code&0x3F;
  31                         code>>=6;
  32                 }
  33
  34                 append(buf, bytes);
  35         }
  36 }
  37
  38
  39 void Utf8::Decoder::decode_char(const string &str, string::const_iterator &i)
  40 {
  41         while(i!=str.end())
  42         {
  43                 if(bytes==0)
  44                 {
  45                         if((*i&0xC0)==0x80)
  46                                 throw CodecError("Invalid UTF-8 string (tail byte when expecting head)");
  47
  48                         else if(*i&0x80)
  49                         {
  50                                 unsigned mask=0x40;
  51                                 for(; *i&mask; mask>>=1)
  52                                         ++bytes;
  53
  54                                 if(bytes>3)
  55                                         throw CodecError("Invalid UTF-8 string (overlong multibyte sequence)");
  56
  57                                 code=(*i++)&(mask-1);
  58                                 if(!code)
  59                                         throw CodecError("Invalid UTF-8 string (denormalized multibyte sequence)");
  60                         }
  61                         else
  62                         {
  63                                 append(*i++);
  64                                 break;
  65                         }
  66                 }
  67                 else
  68                 {
  69                         if((*i&0xC0)!=0x80)
  70                                 throw CodecError("Invalid UTF-8 string (head byte when expecting tail)");
  71
  72                         code=code<<6 | (*i++)&0x3F;
  73                         --bytes;
  74
  75                         if(!bytes)
  76                         {
  77                                 if(code>0x10FFFF)
  78                                         throw CodecError("Invalid UTF-8 string (character out of range)");
  79                                 append(code);
  80                                 break;
  81                         }
  82                 }
  83         }
  84 }
  85
  86 void Utf8::Decoder::sync()
  87 {
  88         if(bytes)
  89                 throw CodecError("Sync in the middle of multibyte UTF-8 sequence");
  90 }
  91
  92 void Utf8::Decoder::reset()
  93 {
  94         bytes=0;
  95 }
  96
  97 } // namespace Msp