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 <map>
  25 #include <string>
  26
  27 using namespace std;
  28
  29 namespace {
  30
  31 #ifdef WORDS_BIGENDIAN
  32         char const * utf16_codeset = "UTF16-BE";
  33 #else
  34         char const * utf16_codeset = "UTF16-LE";
  35 #endif
  36
  37 }
  38
  39
  40 namespace lyx {
  41
  42 #ifdef WORDS_BIGENDIAN
  43         char const * ucs4_codeset = "UCS-4BE";
  44 #else
  45         char const * ucs4_codeset = "UCS-4LE";
  46 #endif
  47
  48 static const iconv_t invalid_cd = (iconv_t)(-1);
  49
  50
  51 struct IconvProcessor::Impl
  52 {
  53         Impl(string const & to, string const & from)
  54                 : cd(invalid_cd), tocode_(to), fromcode_(from)
  55         {}
  56
  57         ~Impl()
  58         {
  59                 if (cd != invalid_cd && iconv_close(cd) == -1)
  60                                 LYXERR0("Error returned from iconv_close(" << errno << ")");
  61         }
  62
  63         iconv_t cd;
  64         string tocode_;
  65         string fromcode_;
  66 };
  67
  68
  69 IconvProcessor::IconvProcessor(char const * tocode, char const * fromcode)
  70         : pimpl_(new IconvProcessor::Impl(tocode, fromcode))
  71 {
  72 }
  73
  74
  75 IconvProcessor::IconvProcessor(IconvProcessor const & other)
  76         : pimpl_(new IconvProcessor::Impl(other.pimpl_->tocode_, other.pimpl_->fromcode_))
  77 {
  78 }
  79
  80
  81 IconvProcessor::~IconvProcessor()
  82 {
  83         delete pimpl_;
  84 }
  85
  86
  87 void IconvProcessor::operator=(IconvProcessor const & other)
  88 {
  89         if (&other != this)
  90                 pimpl_ = new Impl(other.pimpl_->tocode_, other.pimpl_->fromcode_);
  91 }
  92
  93
  94 bool IconvProcessor::init()
  95 {
  96         if (pimpl_->cd != invalid_cd)
  97                 return true;
  98
  99         pimpl_->cd = iconv_open(pimpl_->tocode_.c_str(), pimpl_->fromcode_.c_str());
 100         if (pimpl_->cd != invalid_cd)
 101                 return true;
 102
 103         lyxerr << "Error returned from iconv_open" << endl;
 104         switch (errno) {
 105                 case EINVAL:
 106                         lyxerr << "EINVAL The conversion from " << pimpl_->fromcode_
 107                                 << " to " << pimpl_->tocode_
 108                                 << " is not supported by the implementation."
 109                                 << endl;
 110                         break;
 111                 default:
 112                         lyxerr << "\tSome other error: " << errno << endl;
 113                         break;
 114         }
 115         return false;
 116 }
 117
 118
 119 int IconvProcessor::convert(char const * buf, size_t buflen,
 120                 char * outbuf, size_t maxoutsize)
 121 {
 122         if (buflen == 0)
 123                 return 0;
 124
 125         if (pimpl_->cd == invalid_cd) {
 126                 if (!init())
 127                         return -1;
 128         }
 129
 130         char ICONV_CONST * inbuf = const_cast<char ICONV_CONST *>(buf);
 131         size_t inbytesleft = buflen;
 132         size_t outbytesleft = maxoutsize;
 133
 134         int res = iconv(pimpl_->cd, &inbuf, &inbytesleft, &outbuf, &outbytesleft);
 135
 136         // flush out remaining data. This is needed because iconv sometimes
 137         // holds back chars in the stream, waiting for a combination character
 138         // (see e.g. http://sources.redhat.com/bugzilla/show_bug.cgi?id=1124)
 139         iconv(pimpl_->cd, NULL, NULL, &outbuf, &outbytesleft);
 140
 141         //lyxerr << dec;
 142         //lyxerr << "Inbytesleft: " << inbytesleft << endl;
 143         //lyxerr << "Outbytesleft: " << outbytesleft << endl;
 144
 145         if (res != -1)
 146                 // Everything went well.
 147                 return maxoutsize - outbytesleft;
 148
 149         // There are some errors in the conversion
 150         lyxerr << "Error returned from iconv" << endl;
 151         switch (errno) {
 152                 case E2BIG:
 153                         lyxerr << "E2BIG  There is not sufficient room at *outbuf." << endl;
 154                         break;
 155                 case EILSEQ:
 156                         lyxerr << "EILSEQ An invalid multibyte sequence"
 157                                 << " has been encountered in the input.\n"
 158                                 << "When converting from " << pimpl_->fromcode_
 159                                 << " to " << pimpl_->tocode_ << ".\n";
 160                         lyxerr << "Input:" << hex;
 161                         for (size_t i = 0; i < buflen; ++i) {
 162                                 // char may be signed, avoid output of
 163                                 // something like 0xffffffc2
 164                                 boost::uint32_t const b =
 165                                         *reinterpret_cast<unsigned char const *>(buf + i);
 166                                 lyxerr << " 0x" << (unsigned int)b;
 167                         }
 168                         lyxerr << dec << endl;
 169                         break;
 170                 case EINVAL:
 171                         lyxerr << "EINVAL An incomplete multibyte sequence"
 172                                 << " has been encountered in the input.\n"
 173                                 << "When converting from " << pimpl_->fromcode_
 174                                 << " to " << pimpl_->tocode_ << ".\n";
 175                         lyxerr << "Input:" << hex;
 176                         for (size_t i = 0; i < buflen; ++i) {
 177                                 // char may be signed, avoid output of
 178                                 // something like 0xffffffc2
 179                                 boost::uint32_t const b =
 180                                         *reinterpret_cast<unsigned char const *>(buf + i);
 181                                 lyxerr << " 0x" << (unsigned int)b;
 182                         }
 183                         lyxerr << dec << endl;
 184                         break;
 185                 default:
 186                         lyxerr << "\tSome other error: " << errno << endl;
 187                         break;
 188         }
 189         // We got an error so we close down the conversion engine
 190         if (iconv_close(pimpl_->cd) == -1) {
 191                 lyxerr << "Error returned from iconv_close("
 192                         << errno << ")" << endl;
 193         }
 194         pimpl_->cd = invalid_cd;
 195         return -1;
 196 }
 197
 198
 199 namespace {
 200
 201
 202 template<typename RetType, typename InType>
 203 vector<RetType>
 204 iconv_convert(IconvProcessor & processor, InType const * buf, size_t buflen)
 205 {
 206         if (buflen == 0)
 207                 return vector<RetType>();
 208
 209         char const * inbuf = reinterpret_cast<char const *>(buf);
 210         size_t inbytesleft = buflen * sizeof(InType);
 211
 212         size_t const outsize = 32768;
 213         static char out[outsize];
 214         char * outbuf = out;
 215
 216         int bytes = processor.convert(inbuf, inbytesleft, outbuf, outsize);
 217         if (bytes <= 0)
 218                 // Conversion failed
 219                 // FIXME Maybe throw an exception and handle that in the caller?
 220                 return vector<RetType>();
 221
 222         RetType const * tmp = reinterpret_cast<RetType const *>(out);
 223         return vector<RetType>(tmp, tmp + bytes / sizeof(RetType));
 224 }
 225
 226 } // anon namespace
 227
 228
 229 vector<char_type> utf8_to_ucs4(vector<char> const & utf8str)
 230 {
 231         if (utf8str.empty())
 232                 return vector<char_type>();
 233
 234         return utf8_to_ucs4(&utf8str[0], utf8str.size());
 235 }
 236
 237
 238 vector<char_type>
 239 utf8_to_ucs4(char const * utf8str, size_t ls)
 240 {
 241         static IconvProcessor processor(ucs4_codeset, "UTF-8");
 242         return iconv_convert<char_type>(processor, utf8str, ls);
 243 }
 244
 245
 246 vector<char_type>
 247 utf16_to_ucs4(unsigned short const * s, size_t ls)
 248 {
 249         static IconvProcessor processor(ucs4_codeset, utf16_codeset);
 250         return iconv_convert<char_type>(processor, s, ls);
 251 }
 252
 253
 254 vector<unsigned short>
 255 ucs4_to_utf16(char_type const * s, size_t ls)
 256 {
 257         static IconvProcessor processor(utf16_codeset, ucs4_codeset);
 258         return iconv_convert<unsigned short>(processor, s, ls);
 259 }
 260
 261
 262 vector<char>
 263 ucs4_to_utf8(char_type c)
 264 {
 265         static IconvProcessor processor("UTF-8", ucs4_codeset);
 266         return iconv_convert<char>(processor, &c, 1);
 267 }
 268
 269
 270 vector<char>
 271 ucs4_to_utf8(vector<char_type> const & ucs4str)
 272 {
 273         if (ucs4str.empty())
 274                 return vector<char>();
 275
 276         return ucs4_to_utf8(&ucs4str[0], ucs4str.size());
 277 }
 278
 279
 280 vector<char>
 281 ucs4_to_utf8(char_type const * ucs4str, size_t ls)
 282 {
 283         static IconvProcessor processor("UTF-8", ucs4_codeset);
 284         return iconv_convert<char>(processor, ucs4str, ls);
 285 }
 286
 287
 288 vector<char_type>
 289 eightbit_to_ucs4(char const * s, size_t ls, string const & encoding)
 290 {
 291         static map<string, IconvProcessor> processors;
 292         if (processors.find(encoding) == processors.end()) {
 293                 IconvProcessor processor(ucs4_codeset, encoding.c_str());
 294                 processors.insert(make_pair(encoding, processor));
 295         }
 296         return iconv_convert<char_type>(processors[encoding], s, ls);
 297 }
 298
 299
 300 vector<char>
 301 ucs4_to_eightbit(char_type const * ucs4str, size_t ls, string const & encoding)
 302 {
 303         static map<string, IconvProcessor> processors;
 304         if (processors.find(encoding) == processors.end()) {
 305                 IconvProcessor processor(encoding.c_str(), ucs4_codeset);
 306                 processors.insert(make_pair(encoding, processor));
 307         }
 308         return iconv_convert<char>(processors[encoding], ucs4str, ls);
 309 }
 310
 311
 312 char ucs4_to_eightbit(char_type ucs4, string const & encoding)
 313 {
 314         static map<string, IconvProcessor> processors;
 315         map<string, IconvProcessor>::iterator it = processors.find(encoding);
 316         if (it == processors.end()) {
 317                 IconvProcessor processor(encoding.c_str(), ucs4_codeset);
 318                 it = processors.insert(make_pair(encoding, processor)).first;
 319         }
 320
 321         char out;
 322         int const bytes = it->second.convert((char *)(&ucs4), 4, &out, 1);
 323         if (bytes > 0)
 324                 return out;
 325         return 0;
 326 }
 327
 328
 329 void ucs4_to_multibytes(char_type ucs4, vector<char> & out,
 330         string const & encoding)
 331 {
 332         static map<string, IconvProcessor> processors;
 333         map<string, IconvProcessor>::iterator it = processors.find(encoding);
 334         if (it == processors.end()) {
 335                 IconvProcessor processor(encoding.c_str(), ucs4_codeset);
 336                 it = processors.insert(make_pair(encoding, processor)).first;
 337         }
 338
 339         out.resize(4);
 340         int bytes = it->second.convert((char *)(&ucs4), 4, &out[0], 4);
 341         if (bytes > 0)
 342                 out.resize(bytes);
 343         else
 344                 out.clear();
 345 }
 346
 347 } // namespace lyx