src/support/unicode.C

   1 /**
   2  * \file unicode.C
   3  * This file is part of LyX, the document processor.
   4  * Licence details can be found in the file COPYING.
   5  *
   6  * \author Lars Gullik Bjønnes
   7  *
   8  * Full author contact details are available in file CREDITS.
   9  *
  10  * A collection of unicode conversion functions, using iconv.
  11  */
  12
  13 #include <config.h>
  14
  15 #include "unicode.h"
  16
  17 #include "debug.h"
  18
  19 #include <iconv.h>
  20
  21 #include <cerrno>
  22 #include <iomanip>
  23 #include <map>
  24
  25
  26 namespace lyx {
  27
  28 using std::endl;
  29
  30 #ifdef WORDS_BIGENDIAN
  31         char const * ucs4_codeset = "UCS-4BE";
  32         char const * ucs2_codeset = "UCS-2BE";
  33 #else
  34         char const * ucs4_codeset = "UCS-4LE";
  35         char const * ucs2_codeset = "UCS-2LE";
  36 #endif
  37
  38 namespace {
  39
  40 template<typename RetType, typename InType>
  41 std::vector<RetType>
  42 iconv_convert(iconv_t * cd,
  43               char const * tocode,
  44               char const * fromcode,
  45               InType const * buf,
  46               size_t buflen)
  47 {
  48         if (buflen == 0)
  49                 return std::vector<RetType>();
  50
  51         if (*cd == (iconv_t)(-1)) {
  52                 *cd = iconv_open(tocode, fromcode);
  53                 if (*cd == (iconv_t)(-1)) {
  54                         lyxerr << "Error returned from iconv_open" << endl;
  55                         switch (errno) {
  56                         case EINVAL:
  57                                 lyxerr << "EINVAL The conversion from " << fromcode
  58                                        << " to " << tocode
  59                                        << " is not supported by the implementation."
  60                                        << endl;
  61                                 break;
  62                         default:
  63                                 lyxerr << "\tSome other error: " << errno << endl;
  64                                 break;
  65                         }
  66                 }
  67         }
  68
  69         char ICONV_CONST * inbuf = const_cast<char ICONV_CONST *>(reinterpret_cast<char const *>(buf));
  70         size_t inbytesleft = buflen * sizeof(InType);
  71         // The preamble of the user guide is more than 11.500 characters, so we go for 32kb
  72         size_t const outsize = 32768;
  73         static char out[outsize];
  74         char * outbuf = out;
  75         size_t outbytesleft = outsize;
  76
  77         size_t res = iconv(*cd, &inbuf, &inbytesleft, &outbuf, &outbytesleft);
  78
  79         if (res == (size_t)(-1)) {
  80                 lyxerr << "Error returned from iconv" << endl;
  81                 switch (errno) {
  82                 case E2BIG:
  83                         lyxerr << "E2BIG  There is not sufficient room at *outbuf." << endl;
  84                         break;
  85                 case EILSEQ:
  86                         lyxerr << "EILSEQ An invalid multibyte sequence"
  87                                << " has been encountered in the input.\n"
  88                                << "When converting from " << fromcode
  89                                << " to " << tocode << ".\n";
  90                         lyxerr << "Input: " << std::hex;
  91                         for (size_t i = 0; i < buflen; ++i) {
  92                                 boost::uint32_t const b = buf[i];
  93                                 lyxerr << "0x" << b << " ";
  94                         }
  95                         lyxerr << endl;
  96                         break;
  97                 case EINVAL:
  98                         lyxerr << "EINVAL An incomplete multibyte sequence"
  99                                << " has been encountered in the input.\n"
 100                                << "When converting from " << fromcode
 101                                << " to " << tocode << ".\n";
 102                         lyxerr << "Input: " << std::hex;
 103                         for (size_t i = 0; i < buflen; ++i) {
 104                                 boost::uint32_t const b = buf[i];
 105                                 lyxerr << "0x" << b << " ";
 106                         }
 107                         lyxerr << endl;
 108                         break;
 109                 default:
 110                         lyxerr << "\tSome other error: " << errno << endl;
 111                         break;
 112                 }
 113                 // We got an error so we close down the conversion engine
 114                 if (iconv_close(*cd) == -1) {
 115                         lyxerr << "Error returned from iconv_close("
 116                                << errno << ")" << endl;
 117                 }
 118                 *cd = (iconv_t)(-1);
 119         }
 120
 121         //lyxerr << std::dec;
 122         //lyxerr << "Inbytesleft: " << inbytesleft << endl;
 123         //lyxerr << "Outbytesleft: " << outbytesleft << endl;
 124         int bytes = outsize - outbytesleft;
 125
 126         RetType const * tmp = reinterpret_cast<RetType const *>(out);
 127         return std::vector<RetType>(tmp, tmp + bytes / sizeof(RetType));
 128 }
 129
 130 } // anon namespace
 131
 132
 133 std::vector<lyx::char_type> utf8_to_ucs4(std::vector<char> const & utf8str)
 134 {
 135         if (utf8str.empty())
 136                 return std::vector<lyx::char_type>();
 137
 138         return utf8_to_ucs4(&utf8str[0], utf8str.size());
 139 }
 140
 141
 142 std::vector<lyx::char_type>
 143 utf8_to_ucs4(char const * utf8str, size_t ls)
 144 {
 145         static iconv_t cd = (iconv_t)(-1);
 146         return iconv_convert<lyx::char_type>(&cd, ucs4_codeset, "UTF-8",
 147                                               utf8str, ls);
 148 }
 149
 150
 151 lyx::char_type
 152 ucs2_to_ucs4(unsigned short c)
 153 {
 154         return ucs2_to_ucs4(&c, 1)[0];
 155 }
 156
 157
 158 std::vector<lyx::char_type>
 159 ucs2_to_ucs4(std::vector<unsigned short> const & ucs2str)
 160 {
 161         if (ucs2str.empty())
 162                 return std::vector<lyx::char_type>();
 163
 164         return ucs2_to_ucs4(&ucs2str[0], ucs2str.size());
 165 }
 166
 167
 168 std::vector<lyx::char_type>
 169 ucs2_to_ucs4(unsigned short const * ucs2str, size_t ls)
 170 {
 171         static iconv_t cd = (iconv_t)(-1);
 172         return iconv_convert<lyx::char_type>(&cd, ucs4_codeset, ucs2_codeset,
 173                                               ucs2str, ls);
 174 }
 175
 176
 177 unsigned short
 178 ucs4_to_ucs2(lyx::char_type c)
 179 {
 180         return ucs4_to_ucs2(&c, 1)[0];
 181 }
 182
 183
 184 std::vector<unsigned short>
 185 ucs4_to_ucs2(std::vector<lyx::char_type> const & ucs4str)
 186 {
 187         if (ucs4str.empty())
 188                 return std::vector<unsigned short>();
 189
 190         return ucs4_to_ucs2(&ucs4str[0], ucs4str.size());
 191 }
 192
 193
 194 std::vector<unsigned short>
 195 ucs4_to_ucs2(lyx::char_type const * s, size_t ls)
 196 {
 197         static iconv_t cd = (iconv_t)(-1);
 198         return iconv_convert<unsigned short>(&cd, ucs2_codeset, ucs4_codeset,
 199                                              s, ls);
 200 }
 201
 202
 203 std::vector<char>
 204 ucs4_to_utf8(lyx::char_type c)
 205 {
 206         static iconv_t cd = (iconv_t)(-1);
 207         return iconv_convert<char>(&cd, "UTF-8", ucs4_codeset, &c, 1);
 208 }
 209
 210
 211 std::vector<char>
 212 ucs4_to_utf8(std::vector<lyx::char_type> const & ucs4str)
 213 {
 214         if (ucs4str.empty())
 215                 return std::vector<char>();
 216
 217         return ucs4_to_utf8(&ucs4str[0], ucs4str.size());
 218 }
 219
 220
 221 std::vector<char>
 222 ucs4_to_utf8(lyx::char_type const * ucs4str, size_t ls)
 223 {
 224         static iconv_t cd = (iconv_t)(-1);
 225         return iconv_convert<char>(&cd, "UTF-8", ucs4_codeset,
 226                                    ucs4str, ls);
 227 }
 228
 229
 230 std::vector<lyx::char_type>
 231 eightbit_to_ucs4(char const * s, size_t ls, std::string const & encoding)
 232 {
 233         static std::map<std::string, iconv_t> cd;
 234         if (cd.find(encoding) == cd.end())
 235                 cd[encoding] = (iconv_t)(-1);
 236         return iconv_convert<char_type>(&cd[encoding], ucs4_codeset,
 237                                         encoding.c_str(), s, ls);
 238 }
 239
 240
 241 std::vector<char>
 242 ucs4_to_eightbit(lyx::char_type const * ucs4str, size_t ls, std::string const & encoding)
 243 {
 244         static std::map<std::string, iconv_t> cd;
 245         if (cd.find(encoding) == cd.end())
 246                 cd[encoding] = (iconv_t)(-1);
 247         return iconv_convert<char>(&cd[encoding], encoding.c_str(),
 248                                    ucs4_codeset, ucs4str, ls);
 249 }
 250
 251 } // namespace lyx