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 "utf16.h"
  10 #include "utf8.h"
  11 #include "windows1252.h"
  12
  13 using namespace std;
  14
  15 namespace Msp {
  16 namespace StringCodec {
  17
  18 bool Codec::detect(const string &str) const
  19 {
  20         Decoder *dec = create_decoder(IGNORE_ERRORS);
  21
  22         bool result = true;
  23         for(auto i=str.begin(); (result && i!=str.end()); )
  24                 result = (dec->decode_char(str, i)!=-1);
  25
  26         delete dec;
  27
  28         return result;
  29 }
  30
  31 void Codec::Encoder::encode(const ustring &str, string &buf)
  32 {
  33         for(unichar c: str)
  34                 encode_char(c, buf);
  35 }
  36
  37 string Codec::Encoder::encode(const ustring &str)
  38 {
  39         string buf;
  40         encode(str, buf);
  41         sync(buf);
  42         return buf;
  43 }
  44
  45
  46
  47 void Codec::Decoder::decode(const string &str, ustring &buf)
  48 {
  49         for(auto i=str.begin(); i!=str.end();)
  50         {
  51                 unichar c = decode_char(str, i);
  52                 if(c!=-1)
  53                         buf += c;
  54         }
  55 }
  56
  57 ustring Codec::Decoder::decode(const string &str)
  58 {
  59         ustring buf;
  60         decode(str, buf);
  61         return buf;
  62 }
  63
  64 Codec *create_codec(const string &n)
  65 {
  66         string name;
  67         string::const_iterator i;
  68         for(i=n.begin(); i!=n.end(); ++i)
  69         {
  70                 if(*i==':')
  71                         break;
  72                 else if(isupper(*i))
  73                         name += tolower(*i);
  74                 else if(islower(*i) || isdigit(*i))
  75                         name += *i;
  76         }
  77
  78         ErrorMode em = DEFAULT;
  79         if(i!=n.end() && *i==':')
  80         {
  81                 string em_str(i+1, n.end());
  82                 if(em_str=="throw")
  83                         em = THROW_ON_ERROR;
  84                 else if(em_str=="ignore")
  85                         em = IGNORE_ERRORS;
  86                 else if(em_str=="trans" || em_str=="transliterate")
  87                         em = TRANSLITERATE;
  88                 else
  89                         throw invalid_argument("invalid error mode");
  90         }
  91
  92         if(name=="ascii") return new Ascii(em);
  93         if(name=="iso2022jp") return new Iso2022Jp(em);
  94         if(name=="iso646fi") return new Iso646Fi(em);
  95         if(name=="iso88591" || name=="latin1") return new Iso88591(em);
  96         if(name=="iso885915" || name=="latin9") return new Iso885915(em);
  97         if(name=="jisx0201") return new JisX0201(em);
  98         if(name=="jisx0208") return new JisX0208(em);
  99         if(name=="utf8") return new Utf8(em);
 100         if(name=="utf16") return new Utf16(em, Utf16::AUTO);
 101         if(name=="utf16be") return new Utf16(em, Utf16::BIG);
 102         if(name=="utf16le") return new Utf16(em, Utf16::LITTLE);
 103         if(name=="windows1252" || name=="cp1252") return new Windows1252(em);
 104         throw invalid_argument("unknown string codec");
 105 }
 106
 107 Codec *detect_codec(const string &str)
 108 {
 109         bool is_utf8 = true;
 110         bool is_ascii = true;
 111         bool is_latin1 = true;
 112         unsigned utf8_mb = 0;
 113
 114         for(char c: str)
 115         {
 116                 if(c&0x80)
 117                 {
 118                         is_ascii = false;
 119                         if((c&0xC0)==0x80)
 120                         {
 121                                 if((c&0xE0)==0x80)
 122                                         is_latin1 = false;
 123                                 if(utf8_mb)
 124                                         --utf8_mb;
 125                                 else
 126                                         is_utf8 = false;
 127                         }
 128                         else if((c&0xC0)==0xC0)
 129                         {
 130                                 if(utf8_mb)
 131                                 {
 132                                         is_utf8 = false;
 133                                         utf8_mb = 0;
 134                                 }
 135                                 else
 136                                 {
 137                                         for(utf8_mb=1; (c>>(6-utf8_mb))&1; ++utf8_mb) ;
 138                                 }
 139                         }
 140                 }
 141                 else if(utf8_mb)
 142                 {
 143                         is_utf8 = false;
 144                         utf8_mb = 0;
 145                 }
 146         }
 147
 148         if(is_ascii)
 149                 return new Ascii;
 150         else if(is_utf8)
 151                 return new Utf8;
 152         else if(is_latin1)
 153                 return new Iso88591;
 154         else
 155                 return new Windows1252;
 156 }
 157
 158 } // namespace StringCodec
 159 } // namespace Msp