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 "support/unicode.h"
  16 #include "support/debug.h"
  17
  18 #include <iconv.h>
  19
  20 #include <boost/cstdint.hpp>
  21
  22 #include <cerrno>
  23 #include <iomanip>
  24 #include <ostream>
  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::Impl
  56 {
  57         Impl(std::string const & to, std::string const & from)
  58                 : cd(invalid_cd), tocode_(to), fromcode_(from)
  59         {}
  60
  61         ~Impl()
  62         {
  63                 if (cd != invalid_cd && iconv_close(cd) == -1)
  64                                 LYXERR0("Error returned from iconv_close(" << errno << ")");
  65         }
  66
  67         iconv_t cd;
  68         std::string tocode_;
  69         std::string fromcode_;
  70 };
  71
  72
  73 IconvProcessor::IconvProcessor(char const * tocode, char const * fromcode)
  74         : pimpl_(new IconvProcessor::Impl(tocode, fromcode))
  75 {
  76 }
  77
  78
  79 IconvProcessor::IconvProcessor(IconvProcessor const & other)
  80         : pimpl_(new IconvProcessor::Impl(other.pimpl_->tocode_, other.pimpl_->fromcode_))
  81 {
  82 }
  83
  84
  85 IconvProcessor::~IconvProcessor()
  86 {
  87         delete pimpl_;
  88 }
  89
  90
  91 void IconvProcessor::operator=(IconvProcessor const & other)
  92 {
  93         if (&other != this)
  94                 pimpl_ = new Impl(other.pimpl_->tocode_, other.pimpl_->fromcode_);
  95 }
  96
  97
  98 bool IconvProcessor::init()
  99 {
 100         if (pimpl_->cd != invalid_cd)
 101                 return true;
 102
 103         pimpl_->cd = iconv_open(pimpl_->tocode_.c_str(), pimpl_->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 " << pimpl_->fromcode_
 111                                 << " to " << pimpl_->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 " << pimpl_->fromcode_
 158                                 << " to " << pimpl_->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" << (unsigned int)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 " << pimpl_->fromcode_
 173                                 << " to " << pimpl_->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" << (unsigned int)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, InType const * buf, 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