source/stringcodec/codec.cpp

   1 #include "ascii.h"
   2 #include "codec.h"
   3 #include "iso2022jp.h"
   4 #include "iso646fi.h"
   5 #include "iso88591.h"
   6 #include "iso885915.h"
   7 #include "jisx0201.h"
   8 #include "jisx0208.h"
   9 #include "utf8.h"
  10 #include "windows1252.h"
  11
  12 using namespace std;
  13
  14 namespace Msp {
  15 namespace StringCodec {
  16
  17 bool Codec::detect(const string &str) const
  18 {
  19         Decoder *dec = create_decoder(IGNORE_ERRORS);
  20
  21         bool result = true;
  22         for(string::const_iterator i=str.begin(); (result && i!=str.end()); )
  23                 result = (dec->decode_char(str, i)!=-1);
  24
  25         delete dec;
  26
  27         return result;
  28 }
  29
  30 void Codec::Encoder::encode(const ustring &str, string &buf)
  31 {
  32         for(ustring::const_iterator i=str.begin(); i!=str.end(); ++i)
  33                 encode_char(*i, buf);
  34 }
  35
  36 string Codec::Encoder::encode(const ustring &str)
  37 {
  38         string buf;
  39         encode(str, buf);
  40         sync(buf);
  41         return buf;
  42 }
  43
  44
  45
  46 void Codec::Decoder::decode(const string &str, ustring &buf)
  47 {
  48         for(string::const_iterator i=str.begin(); i!=str.end();)
  49         {
  50                 unichar c = decode_char(str, i);
  51                 if(c!=-1)
  52                         buf += c;
  53         }
  54 }
  55
  56 ustring Codec::Decoder::decode(const string &str)
  57 {
  58         ustring buf;
  59         decode(str, buf);
  60         return buf;
  61 }
  62
  63 Codec *create_codec(const string &n)
  64 {
  65         string name;
  66         string::const_iterator i;
  67         for(i=n.begin(); i!=n.end(); ++i)
  68         {
  69                 if(*i==':')
  70                         break;
  71                 else if(isupper(*i))
  72                         name += tolower(*i);
  73                 else if(islower(*i) || isdigit(*i))
  74                         name += *i;
  75         }
  76
  77         ErrorMode em = DEFAULT;
  78         if(i!=n.end() && *i==':')
  79         {
  80                 string em_str(i+1, n.end());
  81                 if(em_str=="throw")
  82                         em = THROW_ON_ERROR;
  83                 else if(em_str=="ignore")
  84                         em = IGNORE_ERRORS;
  85                 else if(em_str=="trans" || em_str=="transliterate")
  86                         em = TRANSLITERATE;
  87                 else
  88                         throw invalid_argument("invalid error mode");
  89         }
  90
  91         if(name=="ascii") return new Ascii(em);
  92         if(name=="iso2022jp") return new Iso2022Jp(em);
  93         if(name=="iso646fi") return new Iso646Fi(em);
  94         if(name=="iso88591" || name=="latin1") return new Iso88591(em);
  95         if(name=="iso885915" || name=="latin9") return new Iso885915(em);
  96         if(name=="jisx0201") return new JisX0201(em);
  97         if(name=="jisx0208") return new JisX0208(em);
  98         if(name=="utf8") return new Utf8(em);
  99         if(name=="windows1252" || name=="cp1252") return new Windows1252(em);
 100         throw invalid_argument("unknown string codec");
 101 }
 102
 103 Codec *detect_codec(const string &str)
 104 {
 105         bool is_utf8 = true;
 106         bool is_ascii = true;
 107         bool is_latin1 = true;
 108         unsigned utf8_mb = 0;
 109
 110         for(string::const_iterator i=str.begin(); i!=str.end(); ++i)
 111         {
 112                 unsigned char c = *i;
 113                 if(c&0x80)
 114                 {
 115                         is_ascii = false;
 116                         if((c&0xC0)==0x80)
 117                         {
 118                                 if((c&0xE0)==0x80)
 119                                         is_latin1 = false;
 120                                 if(utf8_mb)
 121                                         --utf8_mb;
 122                                 else
 123                                         is_utf8 = false;
 124                         }
 125                         else if((c&0xC0)==0xC0)
 126                         {
 127                                 if(utf8_mb)
 128                                 {
 129                                         is_utf8 = false;
 130                                         utf8_mb = 0;
 131                                 }
 132                                 else
 133                                 {
 134                                         for(utf8_mb=1; (c>>(6-utf8_mb))&1; ++utf8_mb) ;
 135                                 }
 136                         }
 137                 }
 138                 else if(utf8_mb)
 139                 {
 140                         is_utf8 = false;
 141                         utf8_mb = 0;
 142                 }
 143         }
 144
 145         if(is_ascii)
 146                 return new Ascii;
 147         else if(is_utf8)
 148                 return new Utf8;
 149         else if(is_latin1)
 150                 return new Iso88591;
 151         else
 152                 return new Windows1252;
 153 }
 154
 155 } // namespace StringCodec
 156 } // namespace Msp