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 <ostream>
  26 #include <map>
  27
  28 using std::endl;
  29 using std::map;
  30 using std::make_pair;
  31 using std::string;
  32 using std::vector;
  33
  34 namespace {
  35
  36 #ifdef WORDS_BIGENDIAN
  37         char const * utf16_codeset = "UTF16-BE";
  38 #else
  39         char const * utf16_codeset = "UTF16-LE";
  40 #endif
  41
  42 }
  43
  44
  45 namespace lyx {
  46
  47 #ifdef WORDS_BIGENDIAN
  48         char const * ucs4_codeset = "UCS-4BE";
  49 #else
  50         char const * ucs4_codeset = "UCS-4LE";
  51 #endif
  52
  53 static const iconv_t invalid_cd = (iconv_t)(-1);
  54
  55
  56 struct IconvProcessor::Private {
  57         Private(): cd(invalid_cd) {}
  58         ~Private()
  59         {
  60                 if (cd != invalid_cd) {
  61                         if (iconv_close(cd) == -1) {
  62                                 lyxerr << "Error returned from iconv_close("
  63                                        << errno << ")" << endl;
  64                         }
  65                 }
  66         }
  67         iconv_t cd;
  68 };
  69
  70
  71 IconvProcessor::IconvProcessor(char const * tocode, char const * fromcode)
  72         : tocode_(tocode), fromcode_(fromcode),
  73                 pimpl_(new IconvProcessor::Private)
  74 {
  75 }
  76
  77
  78 IconvProcessor::IconvProcessor(IconvProcessor const & other)
  79         : tocode_(other.tocode_), fromcode_(other.fromcode_),
  80           pimpl_(new IconvProcessor::Private)
  81 {
  82 }
  83
  84
  85 IconvProcessor & IconvProcessor::operator=(IconvProcessor const & other)
  86 {
  87         if (&other == this)
  88                 return *this;
  89         tocode_ = other.tocode_;
  90         fromcode_ = other.fromcode_;
  91         pimpl_.reset(new Private);
  92         return *this;
  93 }
  94
  95
  96 IconvProcessor::~IconvProcessor() {}
  97
  98
  99 bool IconvProcessor::init()
 100 {
 101         if (pimpl_->cd != invalid_cd)
 102                 return true;
 103
 104         pimpl_->cd = iconv_open(tocode_.c_str(), fromcode_.c_str());
 105         if (pimpl_->cd != invalid_cd)
 106                 return true;
 107
 108         lyxerr << "Error returned from iconv_open" << endl;
 109         switch (errno) {
 110                 case EINVAL:
 111                         lyxerr << "EINVAL The conversion from " << fromcode_
 112                                 << " to " << tocode_
 113                                 << " is not supported by the implementation."
 114                                 << endl;
 115                         break;
 116                 default:
 117                         lyxerr << "\tSome other error: " << errno << endl;
 118                         break;
 119         }
 120         return false;
 121 }
 122
 123
 124 int IconvProcessor::convert(char const * buf, size_t buflen,
 125                 char * outbuf, size_t maxoutsize)
 126 {
 127         if (buflen == 0)
 128                 return 0;
 129
 130         if (pimpl_->cd == invalid_cd) {
 131                 if (!init())
 132                         return -1;
 133         }
 134
 135         char ICONV_CONST * inbuf = const_cast<char ICONV_CONST *>(buf);
 136         size_t inbytesleft = buflen;
 137         size_t outbytesleft = maxoutsize;
 138
 139         int res = iconv(pimpl_->cd, &inbuf, &inbytesleft, &outbuf, &outbytesleft);
 140
 141         //lyxerr << std::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 " << fromcode_
 159                                 << " to " << tocode_ << ".\n";
 160                         lyxerr << "Input:" << std::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 << std::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 " << fromcode_
 174                                 << " to " << tocode_ << ".\n";
 175                         lyxerr << "Input:" << std::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 << std::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,
 205               InType const * buf,
 206               size_t buflen)
 207 {
 208         if (buflen == 0)
 209                 return vector<RetType>();
 210
 211         char const * inbuf = reinterpret_cast<char const *>(buf);
 212         size_t inbytesleft = buflen * sizeof(InType);
 213
 214         size_t const outsize = 32768;
 215         static char out[outsize];
 216         char * outbuf = out;
 217
 218         int bytes = processor.convert(inbuf, inbytesleft, outbuf, outsize);
 219         if (bytes <= 0)
 220                 // Conversion failed
 221                 // FIXME Maybe throw an exception and handle that in the caller?
 222                 return vector<RetType>();
 223
 224         RetType const * tmp = reinterpret_cast<RetType const *>(out);
 225         return vector<RetType>(tmp, tmp + bytes / sizeof(RetType));
 226 }
 227
 228 } // anon namespace
 229
 230
 231 vector<char_type> utf8_to_ucs4(vector<char> const & utf8str)
 232 {
 233         if (utf8str.empty())
 234                 return vector<char_type>();
 235
 236         return utf8_to_ucs4(&utf8str[0], utf8str.size());
 237 }
 238
 239
 240 vector<char_type>
 241 utf8_to_ucs4(char const * utf8str, size_t ls)
 242 {
 243         static IconvProcessor processor(ucs4_codeset, "UTF-8");
 244         return iconv_convert<char_type>(processor, utf8str, ls);
 245 }
 246
 247
 248 vector<char_type>
 249 utf16_to_ucs4(unsigned short const * s, size_t ls)
 250 {
 251         static IconvProcessor processor(ucs4_codeset, utf16_codeset);
 252         return iconv_convert<char_type>(processor, s, ls);
 253 }
 254
 255
 256 vector<unsigned short>
 257 ucs4_to_utf16(char_type const * s, size_t ls)
 258 {
 259         static IconvProcessor processor(utf16_codeset, ucs4_codeset);
 260         return iconv_convert<unsigned short>(processor, s, ls);
 261 }
 262
 263
 264 vector<char>
 265 ucs4_to_utf8(char_type c)
 266 {
 267         static IconvProcessor processor("UTF-8", ucs4_codeset);
 268         return iconv_convert<char>(processor, &c, 1);
 269 }
 270
 271
 272 vector<char>
 273 ucs4_to_utf8(vector<char_type> const & ucs4str)
 274 {
 275         if (ucs4str.empty())
 276                 return vector<char>();
 277
 278         return ucs4_to_utf8(&ucs4str[0], ucs4str.size());
 279 }
 280
 281
 282 vector<char>
 283 ucs4_to_utf8(char_type const * ucs4str, size_t ls)
 284 {
 285         static IconvProcessor processor("UTF-8", ucs4_codeset);
 286         return iconv_convert<char>(processor, ucs4str, ls);
 287 }
 288
 289
 290 vector<char_type>
 291 eightbit_to_ucs4(char const * s, size_t ls, string const & encoding)
 292 {
 293         static map<string, IconvProcessor> processors;
 294         if (processors.find(encoding) == processors.end()) {
 295                 IconvProcessor processor(ucs4_codeset, encoding.c_str());
 296                 processors.insert(make_pair(encoding, processor));
 297         }
 298         return iconv_convert<char_type>(processors[encoding], s, ls);
 299 }
 300
 301
 302 vector<char>
 303 ucs4_to_eightbit(char_type const * ucs4str, size_t ls, string const & encoding)
 304 {
 305         static map<string, IconvProcessor> processors;
 306         if (processors.find(encoding) == processors.end()) {
 307                 IconvProcessor processor(encoding.c_str(), ucs4_codeset);
 308                 processors.insert(make_pair(encoding, processor));
 309         }
 310         return iconv_convert<char>(processors[encoding], ucs4str, ls);
 311 }
 312
 313
 314 char ucs4_to_eightbit(char_type ucs4, string const & encoding)
 315 {
 316         static map<string, IconvProcessor> processors;
 317         map<string, IconvProcessor>::iterator it = processors.find(encoding);
 318         if (it == processors.end()) {
 319                 IconvProcessor processor(encoding.c_str(), ucs4_codeset);
 320                 it = processors.insert(make_pair(encoding, processor)).first;
 321         }
 322
 323         char out;
 324         int const bytes = it->second.convert((char *)(&ucs4), 4, &out, 1);
 325         if (bytes > 0)
 326                 return out;
 327         return 0;
 328 }
 329
 330
 331 void ucs4_to_multibytes(char_type ucs4, vector<char> & out,
 332         string const & encoding)
 333 {
 334         static map<string, IconvProcessor> processors;
 335         map<string, IconvProcessor>::iterator it = processors.find(encoding);
 336         if (it == processors.end()) {
 337                 IconvProcessor processor(encoding.c_str(), ucs4_codeset);
 338                 it = processors.insert(make_pair(encoding, processor)).first;
 339         }
 340
 341         out.resize(4);
 342         int bytes = it->second.convert((char *)(&ucs4), 4, &out[0], 4);
 343         if (bytes > 0)
 344                 out.resize(bytes);
 345         else
 346                 out.clear();
 347 }
 348
 349 } // namespace lyx