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