src/support/unicode.cpp

   1 /**
   2  * \file unicode.cpp
   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 <boost/cstdint.hpp>
  22
  23 #include <cerrno>
  24 #include <iomanip>
  25 #include <map>
  26
  27 using std::endl;
  28 using std::map;
  29 using std::make_pair;
  30 using std::string;
  31 using std::vector;
  32
  33 namespace {
  34
  35 #ifdef WORDS_BIGENDIAN
  36         char const * utf16_codeset = "UTF16-BE";
  37 #else
  38         char const * utf16_codeset = "UTF16-LE";
  39 #endif
  40
  41 }
  42
  43
  44 namespace lyx {
  45
  46 #ifdef WORDS_BIGENDIAN
  47         char const * ucs4_codeset = "UCS-4BE";
  48 #else
  49         char const * ucs4_codeset = "UCS-4LE";
  50 #endif
  51
  52 static const iconv_t invalid_cd = (iconv_t)(-1);
  53
  54
  55 struct IconvProcessor::Private {
  56         Private(): cd(invalid_cd) {}
  57         ~Private()
  58         {
  59                 if (cd != invalid_cd) {
  60                         if (iconv_close(cd) == -1) {
  61                                 lyxerr << "Error returned from iconv_close("
  62                                        << errno << ")" << endl;
  63                         }
  64                 }
  65         }
  66         iconv_t cd;
  67 };
  68
  69
  70 IconvProcessor::IconvProcessor(char const * tocode,
  71                 char const * fromcode): tocode_(tocode), fromcode_(fromcode),
  72                 pimpl_(new IconvProcessor::Private)
  73 {
  74 }
  75
  76
  77 IconvProcessor::IconvProcessor(IconvProcessor const & other)
  78         : tocode_(other.tocode_), fromcode_(other.fromcode_),
  79           pimpl_(new IconvProcessor::Private)
  80 {
  81 }
  82
  83
  84 IconvProcessor & IconvProcessor::operator=(IconvProcessor const & other)
  85 {
  86         if (&other == this)
  87                 return *this;
  88         tocode_ = other.tocode_;
  89         fromcode_ = other.fromcode_;
  90         pimpl_.reset(new Private);
  91         return *this;
  92 }
  93
  94
  95 IconvProcessor::~IconvProcessor() {}
  96
  97
  98 bool IconvProcessor::init()
  99 {
 100         if (pimpl_->cd != invalid_cd)
 101                 return true;
 102
 103         pimpl_->cd = iconv_open(tocode_.c_str(), fromcode_.c_str());
 104         if (pimpl_->cd != invalid_cd)
 105                 return true;
 106
 107         lyxerr << "Error returned from iconv_open" << endl;
 108         switch (errno) {
 109                 case EINVAL:
 110                         lyxerr << "EINVAL The conversion from " << fromcode_
 111                                 << " to " << tocode_
 112                                 << " is not supported by the implementation."
 113                                 << endl;
 114                         break;
 115                 default:
 116                         lyxerr << "\tSome other error: " << errno << endl;
 117                         break;
 118         }
 119         return false;
 120 }
 121
 122
 123 int IconvProcessor::convert(char const * buf, size_t buflen,
 124                 char * outbuf, size_t maxoutsize)
 125 {
 126         if (buflen == 0)
 127                 return 0;
 128
 129         if (pimpl_->cd == invalid_cd) {
 130                 if (!init())
 131                         return -1;
 132         }
 133
 134         char ICONV_CONST * inbuf = const_cast<char ICONV_CONST *>(buf);
 135         size_t inbytesleft = buflen;
 136         size_t outbytesleft = maxoutsize;
 137
 138         int res = iconv(pimpl_->cd, &inbuf, &inbytesleft, &outbuf, &outbytesleft);
 139
 140         //lyxerr << std::dec;
 141         //lyxerr << "Inbytesleft: " << inbytesleft << endl;
 142         //lyxerr << "Outbytesleft: " << outbytesleft << endl;
 143
 144         if (res != -1)
 145                 // Everything went well.
 146                 return maxoutsize - outbytesleft;
 147
 148         // There are some errors in the conversion
 149         lyxerr << "Error returned from iconv" << endl;
 150         switch (errno) {
 151                 case E2BIG:
 152                         lyxerr << "E2BIG  There is not sufficient room at *outbuf." << endl;
 153                         break;
 154                 case EILSEQ:
 155                         lyxerr << "EILSEQ An invalid multibyte sequence"
 156                                 << " has been encountered in the input.\n"
 157                                 << "When converting from " << fromcode_
 158                                 << " to " << tocode_ << ".\n";
 159                         lyxerr << "Input:" << std::hex;
 160                         for (size_t i = 0; i < buflen; ++i) {
 161                                 // char may be signed, avoid output of
 162                                 // something like 0xffffffc2
 163                                 boost::uint32_t const b =
 164                                         *reinterpret_cast<unsigned char const *>(buf + i);
 165                                 lyxerr << " 0x" << b;
 166                         }
 167                         lyxerr << std::dec << endl;
 168                         break;
 169                 case EINVAL:
 170                         lyxerr << "EINVAL An incomplete multibyte sequence"
 171                                 << " has been encountered in the input.\n"
 172                                 << "When converting from " << fromcode_
 173                                 << " to " << tocode_ << ".\n";
 174                         lyxerr << "Input:" << std::hex;
 175                         for (size_t i = 0; i < buflen; ++i) {
 176                                 // char may be signed, avoid output of
 177                                 // something like 0xffffffc2
 178                                 boost::uint32_t const b =
 179                                         *reinterpret_cast<unsigned char const *>(buf + i);
 180                                 lyxerr << " 0x" << b;
 181                         }
 182                         lyxerr << std::dec << endl;
 183                         break;
 184                 default:
 185                         lyxerr << "\tSome other error: " << errno << endl;
 186                         break;
 187         }
 188         // We got an error so we close down the conversion engine
 189         if (iconv_close(pimpl_->cd) == -1) {
 190                 lyxerr << "Error returned from iconv_close("
 191                         << errno << ")" << endl;
 192         }
 193         pimpl_->cd = invalid_cd;
 194         return -1;
 195 }
 196
 197
 198 namespace {
 199
 200
 201 template<typename RetType, typename InType>
 202 vector<RetType>
 203 iconv_convert(IconvProcessor & processor,
 204               InType const * buf,
 205               size_t buflen)
 206 {
 207         if (buflen == 0)
 208                 return vector<RetType>();
 209
 210         char const * inbuf = reinterpret_cast<char const *>(buf);
 211         size_t inbytesleft = buflen * sizeof(InType);
 212
 213         size_t const outsize = 32768;
 214         static char out[outsize];
 215         char * outbuf = out;
 216
 217         int bytes = processor.convert(inbuf, inbytesleft, outbuf, outsize);
 218         if (bytes <= 0)
 219                 // Conversion failed
 220                 // FIXME Maybe throw an exception and handle that in the caller?
 221                 return vector<RetType>();
 222
 223         RetType const * tmp = reinterpret_cast<RetType const *>(out);
 224         return vector<RetType>(tmp, tmp + bytes / sizeof(RetType));
 225 }
 226
 227 } // anon namespace
 228
 229
 230 vector<char_type> utf8_to_ucs4(vector<char> const & utf8str)
 231 {
 232         if (utf8str.empty())
 233                 return vector<char_type>();
 234
 235         return utf8_to_ucs4(&utf8str[0], utf8str.size());
 236 }
 237
 238
 239 vector<char_type>
 240 utf8_to_ucs4(char const * utf8str, size_t ls)
 241 {
 242         static IconvProcessor processor(ucs4_codeset, "UTF-8");
 243         return iconv_convert<char_type>(processor, utf8str, ls);
 244 }
 245
 246
 247 vector<char_type>
 248 utf16_to_ucs4(unsigned short const * s, size_t ls)
 249 {
 250         static IconvProcessor processor(ucs4_codeset, utf16_codeset);
 251         return iconv_convert<char_type>(processor, s, ls);
 252 }
 253
 254
 255 vector<unsigned short>
 256 ucs4_to_utf16(char_type const * s, size_t ls)
 257 {
 258         static IconvProcessor processor(utf16_codeset, ucs4_codeset);
 259         return iconv_convert<unsigned short>(processor, s, ls);
 260 }
 261
 262
 263 vector<char>
 264 ucs4_to_utf8(char_type c)
 265 {
 266         static IconvProcessor processor("UTF-8", ucs4_codeset);
 267         return iconv_convert<char>(processor, &c, 1);
 268 }
 269
 270
 271 vector<char>
 272 ucs4_to_utf8(vector<char_type> const & ucs4str)
 273 {
 274         if (ucs4str.empty())
 275                 return vector<char>();
 276
 277         return ucs4_to_utf8(&ucs4str[0], ucs4str.size());
 278 }
 279
 280
 281 vector<char>
 282 ucs4_to_utf8(char_type const * ucs4str, size_t ls)
 283 {
 284         static IconvProcessor processor("UTF-8", ucs4_codeset);
 285         return iconv_convert<char>(processor, ucs4str, ls);
 286 }
 287
 288
 289 vector<char_type>
 290 eightbit_to_ucs4(char const * s, size_t ls, string const & encoding)
 291 {
 292         static map<string, IconvProcessor> processors;
 293         if (processors.find(encoding) == processors.end()) {
 294                 IconvProcessor processor(ucs4_codeset, encoding.c_str());
 295                 processors.insert(make_pair(encoding, processor));
 296         }
 297         return iconv_convert<char_type>(processors[encoding], s, ls);
 298 }
 299
 300
 301 vector<char>
 302 ucs4_to_eightbit(char_type const * ucs4str, size_t ls, string const & encoding)
 303 {
 304         static map<string, IconvProcessor> processors;
 305         if (processors.find(encoding) == processors.end()) {
 306                 IconvProcessor processor(encoding.c_str(), ucs4_codeset);
 307                 processors.insert(make_pair(encoding, processor));
 308         }
 309         return iconv_convert<char>(processors[encoding], ucs4str, ls);
 310 }
 311
 312
 313 char ucs4_to_eightbit(char_type ucs4, string const & encoding)
 314 {
 315         static map<string, IconvProcessor> processors;
 316         map<string, IconvProcessor>::iterator it = processors.find(encoding);
 317         if (it == processors.end()) {
 318                 IconvProcessor processor(encoding.c_str(), ucs4_codeset);
 319                 it = processors.insert(make_pair(encoding, processor)).first;
 320         }
 321
 322         char out;
 323         int const bytes = it->second.convert((char *)(&ucs4), 4, &out, 1);
 324         if (bytes > 0)
 325                 return out;
 326         return 0;
 327 }
 328
 329
 330 void ucs4_to_multibytes(char_type ucs4, vector<char> & out,
 331         string const & encoding)
 332 {
 333         static map<string, IconvProcessor> processors;
 334         map<string, IconvProcessor>::iterator it = processors.find(encoding);
 335         if (it == processors.end()) {
 336                 IconvProcessor processor(encoding.c_str(), ucs4_codeset);
 337                 it = processors.insert(make_pair(encoding, processor)).first;
 338         }
 339
 340         out.resize(4);
 341         int bytes = it->second.convert((char *)(&ucs4), 4, &out[0], 4);
 342         if (bytes > 0)
 343                 out.resize(bytes);
 344         else
 345                 out.clear();
 346 }
 347
 348 } // namespace lyx