src/Encoding.cpp

   1 /**
   2  * \file Encoding.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  * \author Jean-Marc Lasgouttes
   8  * \author Dekel Tsur
   9  *
  10  * Full author contact details are available in file CREDITS.
  11  */
  12
  13 #include <config.h>
  14
  15 #include "Encoding.h"
  16
  17 #include "Lexer.h"
  18
  19 #include "support/debug.h"
  20 #include "support/gettext.h"
  21 #include "support/lstrings.h"
  22 #include "support/mutex.h"
  23 #include "support/textutils.h"
  24 #include "support/unicode.h"
  25
  26 #include <boost/cstdint.hpp>
  27
  28 #include <sstream>
  29 #include <algorithm>
  30
  31 using namespace std;
  32 using namespace lyx::support;
  33
  34 namespace lyx {
  35
  36 int const Encoding::any = -1;
  37
  38 Encodings encodings;
  39
  40 Encodings::MathCommandSet Encodings::mathcmd;
  41 Encodings::TextCommandSet Encodings::textcmd;
  42 Encodings::MathSymbolSet  Encodings::mathsym;
  43
  44 namespace {
  45
  46 typedef map<char_type, CharInfo> CharInfoMap;
  47 CharInfoMap unicodesymbols;
  48
  49 typedef set<char_type> CharSet;
  50 typedef map<string, CharSet> CharSetMap;
  51 CharSet forced;
  52 CharSetMap forcedselected;
  53
  54 typedef set<char_type> MathAlphaSet;
  55 MathAlphaSet mathalpha;
  56
  57
  58 /// The highest code point in UCS4 encoding (1<<20 + 1<<16)
  59 char_type const max_ucs4 = 0x110000;
  60
  61 } // namespace anon
  62
  63
  64 EncodingException::EncodingException(char_type c)
  65         : failed_char(c), par_id(0), pos(0)
  66 {
  67 }
  68
  69
  70 const char * EncodingException::what() const throw()
  71 {
  72         return "Could not find LaTeX command for a character";
  73 }
  74
  75
  76 CharInfo::CharInfo(
  77         docstring const & textcommand, docstring const & mathcommand,
  78         std::string const & textpreamble, std::string const & mathpreamble,
  79         std::string const & tipashortcut, unsigned int flags)
  80         : textcommand_(textcommand), mathcommand_(mathcommand),
  81           textpreamble_(textpreamble), mathpreamble_(mathpreamble),
  82           tipashortcut_(tipashortcut), flags_(flags)
  83 {
  84 }
  85
  86
  87 Encoding::Encoding(string const & n, string const & l, string const & g,
  88                    string const & i, bool f, bool u, Encoding::Package p)
  89         : name_(n), latexName_(l), guiName_(g), iconvName_(i), fixedwidth_(f),
  90           unsafe_(u), forced_(&forcedselected[n]), package_(p)
  91 {
  92         if (n == "ascii") {
  93                 // ASCII can encode 128 code points and nothing else
  94                 start_encodable_ = 128;
  95                 complete_ = true;
  96         } else if (i == "UTF-8") {
  97                 // UTF8 can encode all UCS4 code points
  98                 start_encodable_ = max_ucs4;
  99                 complete_ = true;
 100         } else {
 101                 start_encodable_ = 0;
 102                 complete_ = false;
 103         }
 104 }
 105
 106
 107 void Encoding::init() const
 108 {
 109         // Since the the constructor is the only method which sets complete_
 110         // to false the test for complete_ is thread-safe without mutex.
 111         if (complete_)
 112                 return;
 113
 114         static Mutex mutex;
 115         Mutex::Locker lock(&mutex);
 116
 117         // We need to test again for complete_, since another thread could
 118         // have set it to true while we were waiting for the lock and we must
 119         // not modify an encoding which is already complete.
 120         if (complete_)
 121                 return;
 122
 123         // We do not make any member mutable  so that it can be easily verified
 124         // that all const methods are thread-safe: init() is the only const
 125         // method which changes complete_, encodable_ and start_encodable_, and
 126         // it uses a mutex to ensure thread-safety.
 127         CharSet & encodable = const_cast<Encoding *>(this)->encodable_;
 128         char_type & start_encodable = const_cast<Encoding *>(this)->start_encodable_;
 129
 130         start_encodable = 0;
 131         // temporarily switch off lyxerr, since we will generate iconv errors
 132         lyxerr.disable();
 133         if (fixedwidth_) {
 134                 // We do not need to check all UCS4 code points, it is enough
 135                 // if we check all 256 code points of this encoding.
 136                 for (unsigned short j = 0; j < 256; ++j) {
 137                         char const c = char(j);
 138                         vector<char_type> const ucs4 = eightbit_to_ucs4(&c, 1, iconvName_);
 139                         if (ucs4.size() != 1)
 140                                 continue;
 141                         char_type const uc = ucs4[0];
 142                         CharInfoMap::const_iterator const it = unicodesymbols.find(uc);
 143                         if (it == unicodesymbols.end())
 144                                 encodable.insert(uc);
 145                         else if (!it->second.force()) {
 146                                 if (forced_->empty() || forced_->find(uc) == forced_->end())
 147                                         encodable.insert(uc);
 148                         }
 149                 }
 150         } else {
 151                 // We do not know how many code points this encoding has, and
 152                 // they do not have a direct representation as a single byte,
 153                 // therefore we need to check all UCS4 code points.
 154                 // This is expensive!
 155                 for (char_type c = 0; c < max_ucs4; ++c) {
 156                         vector<char> const eightbit = ucs4_to_eightbit(&c, 1, iconvName_);
 157                         if (!eightbit.empty()) {
 158                                 CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 159                                 if (it == unicodesymbols.end())
 160                                         encodable.insert(c);
 161                                 else if (!it->second.force()) {
 162                                         if (forced_->empty() || forced_->find(c) == forced_->end())
 163                                                 encodable.insert(c);
 164                                 }
 165                         }
 166                 }
 167         }
 168         lyxerr.enable();
 169         CharSet::iterator it = encodable.find(start_encodable);
 170         while (it != encodable.end()) {
 171                 encodable.erase(it);
 172                 ++start_encodable;
 173                 it = encodable.find(start_encodable);
 174         }
 175         const_cast<Encoding *>(this)->complete_ = true;
 176 }
 177
 178
 179 bool Encoding::isForced(char_type c) const
 180 {
 181         if (!forced.empty() && forced.find(c) != forced.end())
 182                 return true;
 183         return !forced_->empty() && forced_->find(c) != forced_->end();
 184 }
 185
 186
 187 bool Encoding::encodable(char_type c) const
 188 {
 189         // assure the used encoding is properly initialized
 190         init();
 191
 192         if (iconvName_ == "UTF-8" && package_ == none)
 193                 return true;
 194         if (c < start_encodable_ && !isForced(c))
 195                 return true;
 196         if (encodable_.find(c) != encodable_.end())
 197                 return true;
 198         return false;
 199 }
 200
 201
 202 pair<docstring, bool> Encoding::latexChar(char_type c) const
 203 {
 204         if (encodable(c))
 205                 return make_pair(docstring(1, c), false);
 206
 207         // c cannot (or should not) be encoded in this encoding
 208         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 209         if (it == unicodesymbols.end())
 210                 throw EncodingException(c);
 211         // at least one of mathcommand and textcommand is nonempty
 212         if (it->second.textcommand().empty())
 213                 return make_pair(
 214                         "\\ensuremath{" + it->second.mathcommand() + '}', false);
 215         return make_pair(it->second.textcommand(), !it->second.textnotermination());
 216 }
 217
 218
 219 pair<docstring, docstring> Encoding::latexString(docstring const & input, bool dryrun) const
 220 {
 221         docstring result;
 222         docstring uncodable;
 223         bool terminate = false;
 224         for (size_t n = 0; n < input.size(); ++n) {
 225                 try {
 226                         char_type const c = input[n];
 227                         pair<docstring, bool> latex_char = latexChar(c);
 228                         docstring const latex = latex_char.first;
 229                         if (terminate && !prefixIs(latex, '\\')
 230                             && !prefixIs(latex, '{')
 231                             && !prefixIs(latex, '}')) {
 232                                         // Prevent eating of a following
 233                                         // space or command corruption by
 234                                         // following characters
 235                                         if (latex == " ")
 236                                                 result += "{}";
 237                                         else
 238                                                 result += " ";
 239                                 }
 240                         result += latex;
 241                         terminate = latex_char.second;
 242                 } catch (EncodingException & /* e */) {
 243                         LYXERR0("Uncodable character in latexString!");
 244                         if (dryrun) {
 245                                 result += "<" + _("LyX Warning: ")
 246                                            + _("uncodable character") + " '";
 247                                 result += docstring(1, input[n]);
 248                                 result += "'>";
 249                         } else
 250                                 uncodable += input[n];
 251                 }
 252         }
 253         return make_pair(result, uncodable);
 254 }
 255
 256
 257 vector<char_type> Encoding::symbolsList() const
 258 {
 259         // assure the used encoding is properly initialized
 260         init();
 261
 262         // first all encodable characters
 263         vector<char_type> symbols(encodable_.begin(), encodable_.end());
 264         // add those below start_encodable_
 265         for (char_type c = 0; c < start_encodable_; ++c)
 266                 symbols.push_back(c);
 267         // now the ones from the unicodesymbols file
 268         CharInfoMap::const_iterator const end = unicodesymbols.end();
 269         CharInfoMap::const_iterator it = unicodesymbols.begin();
 270         for (; it != end; ++it)
 271                 symbols.push_back(it->first);
 272         return symbols;
 273 }
 274
 275
 276 bool Encodings::latexMathChar(char_type c, bool mathmode,
 277                         Encoding const * encoding, docstring & command,
 278                         bool & needsTermination)
 279 {
 280         command = empty_docstring();
 281         if (encoding)
 282                 if (encoding->encodable(c))
 283                         command = docstring(1, c);
 284         needsTermination = false;
 285
 286         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 287         if (it == unicodesymbols.end()) {
 288                 if (!encoding || command.empty())
 289                         throw EncodingException(c);
 290                 if (mathmode)
 291                         addMathSym(c);
 292                 return false;
 293         }
 294         // at least one of mathcommand and textcommand is nonempty
 295         bool use_math = (mathmode && !it->second.mathcommand().empty()) ||
 296                         (!mathmode && it->second.textcommand().empty());
 297         if (use_math) {
 298                 command = it->second.mathcommand();
 299                 needsTermination = !it->second.mathnotermination();
 300                 addMathCmd(c);
 301         } else {
 302                 if (!encoding || command.empty()) {
 303                         command = it->second.textcommand();
 304                         needsTermination = !it->second.textnotermination();
 305                         addTextCmd(c);
 306                 }
 307                 if (mathmode)
 308                         addMathSym(c);
 309         }
 310         return use_math;
 311 }
 312
 313
 314 char_type Encodings::fromLaTeXCommand(docstring const & cmd, int cmdtype,
 315                 bool & combining, bool & needsTermination, set<string> * req)
 316 {
 317         CharInfoMap::const_iterator const end = unicodesymbols.end();
 318         CharInfoMap::const_iterator it = unicodesymbols.begin();
 319         for (combining = false; it != end; ++it) {
 320                 docstring const math = it->second.mathcommand();
 321                 docstring const text = it->second.textcommand();
 322                 if ((cmdtype & MATH_CMD) && math == cmd) {
 323                         combining = it->second.combining();
 324                         needsTermination = !it->second.mathnotermination();
 325                         if (req && it->second.mathfeature() &&
 326                             !it->second.mathpreamble().empty())
 327                                 req->insert(it->second.mathpreamble());
 328                         return it->first;
 329                 }
 330                 if ((cmdtype & TEXT_CMD) && text == cmd) {
 331                         combining = it->second.combining();
 332                         needsTermination = !it->second.textnotermination();
 333                         if (req && it->second.textfeature() &&
 334                             !it->second.textpreamble().empty())
 335                                 req->insert(it->second.textpreamble());
 336                         return it->first;
 337                 }
 338         }
 339         needsTermination = false;
 340         return 0;
 341 }
 342
 343
 344 docstring Encodings::fromLaTeXCommand(docstring const & cmd, int cmdtype,
 345                 bool & needsTermination, docstring & rem, set<string> * req)
 346 {
 347         needsTermination = false;
 348         rem = empty_docstring();
 349         bool const mathmode = cmdtype & MATH_CMD;
 350         bool const textmode = cmdtype & TEXT_CMD;
 351         docstring symbols;
 352         size_t const cmdend = cmd.size();
 353         size_t prefix = 0;
 354         CharInfoMap::const_iterator const uniend = unicodesymbols.end();
 355         for (size_t i = 0, j = 0; j < cmdend; ++j) {
 356                 // Also get the char after a backslash
 357                 if (j + 1 < cmdend && cmd[j] == '\\') {
 358                         ++j;
 359                         prefix = 1;
 360                         // Detect things like \=*{e} as well
 361                         if (j + 3 < cmdend && cmd[j+1] == '*' &&
 362                             cmd[j+2] == '{') {
 363                                 ++j;
 364                                 prefix = 2;
 365                         }
 366                 }
 367                 // position of the last character before a possible macro
 368                 // argument
 369                 size_t m = j;
 370                 // If a macro argument follows, get it, too
 371                 // Do it here only for single character commands. Other
 372                 // combining commands need this too, but they are handled in
 373                 // the loop below for performance reasons.
 374                 if (j + 1 < cmdend && cmd[j + 1] == '{') {
 375                         size_t k = j + 1;
 376                         int count = 1;
 377                         while (k < cmdend && count) {
 378                                 k = cmd.find_first_of(from_ascii("{}"), k + 1);
 379                                 // braces may not be balanced
 380                                 if (k == docstring::npos)
 381                                         break;
 382                                 if (cmd[k] == '{')
 383                                         ++count;
 384                                 else
 385                                         --count;
 386                         }
 387                         if (k != docstring::npos)
 388                                 j = k;
 389                 } else if (m + 1 < cmdend && isAlphaASCII(cmd[m])) {
 390                         while (m + 2 < cmdend && isAlphaASCII(cmd[m+1]))
 391                                 m++;
 392                 }
 393                 // Start with this substring and try augmenting it when it is
 394                 // the prefix of some command in the unicodesymbols file
 395                 docstring subcmd = cmd.substr(i, j - i + 1);
 396
 397                 CharInfoMap::const_iterator it = unicodesymbols.begin();
 398                 // First part of subcmd which might be a combining character
 399                 docstring combcmd = (m == j) ? docstring() : cmd.substr(i, m - i + 1);
 400                 // The combining character of combcmd if it exists
 401                 CharInfoMap::const_iterator combining = uniend;
 402                 size_t unicmd_size = 0;
 403                 char_type c = 0;
 404                 for (; it != uniend; ++it) {
 405                         docstring const math = mathmode ? it->second.mathcommand()
 406                                                         : docstring();
 407                         docstring const text = textmode ? it->second.textcommand()
 408                                                         : docstring();
 409                         if (!combcmd.empty() && it->second.combining() &&
 410                             (math == combcmd || text == combcmd))
 411                                 combining = it;
 412                         size_t cur_size = max(math.size(), text.size());
 413                         // The current math or text unicode command cannot
 414                         // match, or we already matched a longer one
 415                         if (cur_size < subcmd.size() || cur_size <= unicmd_size)
 416                                 continue;
 417
 418                         docstring tmp = subcmd;
 419                         size_t k = j;
 420                         while (prefixIs(math, tmp) || prefixIs(text, tmp)) {
 421                                 ++k;
 422                                 if (k >= cmdend || cur_size <= tmp.size())
 423                                         break;
 424                                 tmp += cmd[k];
 425                         }
 426                         // No match
 427                         if (k == j)
 428                                 continue;
 429
 430                         // The last added char caused a mismatch, because
 431                         // we didn't exhaust the chars in cmd and didn't
 432                         // exceed the maximum size of the current unicmd
 433                         if (k < cmdend && cur_size > tmp.size())
 434                                 tmp.resize(tmp.size() - 1);
 435
 436                         // If this is an exact match, we found a (longer)
 437                         // matching entry in the unicodesymbols file.
 438                         if (math != tmp && text != tmp)
 439                                 continue;
 440                         // If we found a combining command, we need to append
 441                         // the macro argument if this has not been done above.
 442                         if (tmp == combcmd && combining != uniend &&
 443                             k < cmdend && cmd[k] == '{') {
 444                                 size_t l = k;
 445                                 int count = 1;
 446                                 while (l < cmdend && count) {
 447                                         l = cmd.find_first_of(from_ascii("{}"), l + 1);
 448                                         // braces may not be balanced
 449                                         if (l == docstring::npos)
 450                                                 break;
 451                                         if (cmd[l] == '{')
 452                                                 ++count;
 453                                         else
 454                                                 --count;
 455                                 }
 456                                 if (l != docstring::npos) {
 457                                         j = l;
 458                                         subcmd = cmd.substr(i, j - i + 1);
 459                                 }
 460                         }
 461                         // If the entry doesn't start with '\', we take note
 462                         // of the match and continue (this is not a ultimate
 463                         // acceptance, as some other entry may match a longer
 464                         // portion of the cmd string). However, if the entry
 465                         // does start with '\', we accept the match only if
 466                         // this is a valid macro, i.e., either it is a single
 467                         // (nonletter) char macro, or nothing else follows,
 468                         // or what follows is a nonletter char, or the last
 469                         // character is a }.
 470                         else if (tmp[0] != '\\'
 471                                    || (tmp.size() == prefix + 1 &&
 472                                        !isAlphaASCII(tmp[1]) &&
 473                                        (prefix == 1 || !isAlphaASCII(tmp[2])))
 474                                    || k == cmdend
 475                                    || !isAlphaASCII(cmd[k])
 476                                    || tmp[tmp.size() - 1] == '}'
 477                                  ) {
 478                                 c = it->first;
 479                                 j = k - 1;
 480                                 i = j + 1;
 481                                 unicmd_size = cur_size;
 482                                 if (math == tmp)
 483                                         needsTermination = !it->second.mathnotermination();
 484                                 else
 485                                         needsTermination = !it->second.textnotermination();
 486                                 if (req) {
 487                                         if (math == tmp && it->second.mathfeature() &&
 488                                             !it->second.mathpreamble().empty())
 489                                                 req->insert(it->second.mathpreamble());
 490                                         if (text == tmp && it->second.textfeature() &&
 491                                             !it->second.textpreamble().empty())
 492                                                 req->insert(it->second.textpreamble());
 493                                 }
 494                         }
 495                 }
 496                 if (unicmd_size)
 497                         symbols += c;
 498                 else if (combining != uniend &&
 499                          prefixIs(subcmd, combcmd + '{')) {
 500                         // We know that subcmd starts with combcmd and
 501                         // contains an argument in braces.
 502                         docstring const arg = subcmd.substr(
 503                                 combcmd.length() + 1,
 504                                 subcmd.length() - combcmd.length() - 2);
 505                         // If arg is a single character we can construct a
 506                         // combining sequence.
 507                         char_type a;
 508                         bool argcomb = false;
 509                         if (arg.size() == 1 && isAlnumASCII(arg[0]))
 510                                 a = arg[0];
 511                         else {
 512                                 // Use the version of fromLaTeXCommand() that
 513                                 // parses only one command, since we cannot
 514                                 // use more than one character.
 515                                 bool dummy = false;
 516                                 set<string> r;
 517                                 a = fromLaTeXCommand(arg, cmdtype, argcomb,
 518                                                      dummy, &r);
 519                                 if (a && req && !argcomb)
 520                                         req->insert(r.begin(), r.end());
 521                         }
 522                         if (a && !argcomb) {
 523                                 // In unicode the combining character comes
 524                                 // after its base
 525                                 symbols += a;
 526                                 symbols += combining->first;
 527                                 i = j + 1;
 528                                 unicmd_size = 2;
 529                         }
 530                 }
 531                 if (j + 1 == cmdend && !unicmd_size) {
 532                         // No luck. Return what remains
 533                         rem = cmd.substr(i);
 534                         if (needsTermination && !rem.empty()) {
 535                                 if (rem.substr(0, 2) == "{}") {
 536                                         rem = rem.substr(2);
 537                                         needsTermination = false;
 538                                 } else if (rem[0] == ' ') {
 539                                         needsTermination = false;
 540                                         // LaTeX would swallow all spaces
 541                                         rem = ltrim(rem);
 542                                 }
 543                         }
 544                 }
 545         }
 546         return symbols;
 547 }
 548
 549
 550 CharInfo const & Encodings::unicodeCharInfo(char_type c)
 551 {
 552         static CharInfo empty;
 553         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 554         return it != unicodesymbols.end() ? it->second : empty;
 555 }
 556
 557
 558 bool Encodings::isCombiningChar(char_type c)
 559 {
 560         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 561         if (it != unicodesymbols.end())
 562                 return it->second.combining();
 563         return false;
 564 }
 565
 566
 567 string const Encodings::TIPAShortcut(char_type c)
 568 {
 569         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 570         if (it != unicodesymbols.end())
 571                 return it->second.tipashortcut();
 572         return string();
 573 }
 574
 575
 576 bool Encodings::isKnownScriptChar(char_type const c, string & preamble)
 577 {
 578         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 579
 580         if (it == unicodesymbols.end())
 581                 return false;
 582
 583         if (it->second.textpreamble() != "textgreek" && it->second.textpreamble() != "textcyr")
 584                 return false;
 585
 586         if (preamble.empty()) {
 587                 preamble = it->second.textpreamble();
 588                 return true;
 589         }
 590         return it->second.textpreamble() == preamble;
 591 }
 592
 593
 594 bool Encodings::isMathAlpha(char_type c)
 595 {
 596         return mathalpha.count(c);
 597 }
 598
 599
 600 Encoding const *
 601 Encodings::fromLyXName(string const & name, bool allowUnsafe) const
 602 {
 603         EncodingList::const_iterator const it = encodinglist.find(name);
 604         if (it == encodinglist.end())
 605                 return 0;
 606         if (!allowUnsafe && it->second.unsafe())
 607                 return 0;
 608         return &it->second;
 609 }
 610
 611
 612 Encoding const *
 613 Encodings::fromLaTeXName(string const & n, int const & p, bool allowUnsafe) const
 614 {
 615         string name = n;
 616         // FIXME: if we have to test for too many of these synonyms,
 617         // we should instead extend the format of lib/encodings
 618         if (n == "ansinew")
 619                 name = "cp1252";
 620
 621         // We don't use find_if because it makes copies of the pairs in
 622         // the map.
 623         // This linear search is OK since we don't have many encodings.
 624         // Users could even optimize it by putting the encodings they use
 625         // most at the top of lib/encodings.
 626         EncodingList::const_iterator const end = encodinglist.end();
 627         for (EncodingList::const_iterator it = encodinglist.begin(); it != end; ++it)
 628                 if ((it->second.latexName() == name) && (it->second.package() & p)
 629                                 && (!it->second.unsafe() || allowUnsafe))
 630                         return &it->second;
 631         return 0;
 632 }
 633
 634
 635 Encoding const *
 636 Encodings::fromIconvName(string const & n, int const & p, bool allowUnsafe) const
 637 {
 638         EncodingList::const_iterator const end = encodinglist.end();
 639         for (EncodingList::const_iterator it = encodinglist.begin(); it != end; ++it)
 640                 if ((it->second.iconvName() == n) && (it->second.package() & p)
 641                                 && (!it->second.unsafe() || allowUnsafe))
 642                         return &it->second;
 643         return 0;
 644 }
 645
 646
 647 Encodings::Encodings()
 648 {}
 649
 650
 651 void Encodings::read(FileName const & encfile, FileName const & symbolsfile)
 652 {
 653         // We must read the symbolsfile first, because the Encoding
 654         // constructor depends on it.
 655         CharSetMap forcednotselected;
 656         Lexer symbolslex;
 657         symbolslex.setFile(symbolsfile);
 658         bool getNextToken = true;
 659         while (symbolslex.isOK()) {
 660                 char_type symbol;
 661
 662                 if (getNextToken) {
 663                         if (!symbolslex.next(true))
 664                                 break;
 665                 } else
 666                         getNextToken = true;
 667
 668                 istringstream is(symbolslex.getString());
 669                 // reading symbol directly does not work if
 670                 // char_type == wchar_t.
 671                 boost::uint32_t tmp;
 672                 if(!(is >> hex >> tmp))
 673                         break;
 674                 symbol = tmp;
 675
 676                 if (!symbolslex.next(true))
 677                         break;
 678                 docstring textcommand = symbolslex.getDocString();
 679                 if (!symbolslex.next(true))
 680                         break;
 681                 string textpreamble = symbolslex.getString();
 682                 if (!symbolslex.next(true))
 683                         break;
 684                 string sflags = symbolslex.getString();
 685
 686                 string tipashortcut;
 687                 int flags = 0;
 688
 689                 if (suffixIs(textcommand, '}'))
 690                         flags |= CharInfoTextNoTermination;
 691                 while (!sflags.empty()) {
 692                         string flag;
 693                         sflags = split(sflags, flag, ',');
 694                         if (flag == "combining") {
 695                                 flags |= CharInfoCombining;
 696                         } else if (flag == "force") {
 697                                 flags |= CharInfoForce;
 698                                 forced.insert(symbol);
 699                         } else if (prefixIs(flag, "force=")) {
 700                                 vector<string> encodings =
 701                                         getVectorFromString(flag.substr(6), ";");
 702                                 for (size_t i = 0; i < encodings.size(); ++i)
 703                                         forcedselected[encodings[i]].insert(symbol);
 704                                 flags |= CharInfoForceSelected;
 705                         } else if (prefixIs(flag, "force!=")) {
 706                                 vector<string> encodings =
 707                                         getVectorFromString(flag.substr(7), ";");
 708                                 for (size_t i = 0; i < encodings.size(); ++i)
 709                                         forcednotselected[encodings[i]].insert(symbol);
 710                                 flags |= CharInfoForceSelected;
 711                         } else if (flag == "mathalpha") {
 712                                 mathalpha.insert(symbol);
 713                         } else if (flag == "notermination=text") {
 714                                 flags |= CharInfoTextNoTermination;
 715                         } else if (flag == "notermination=math") {
 716                                 flags |= CharInfoMathNoTermination;
 717                         } else if (flag == "notermination=both") {
 718                                 flags |= CharInfoTextNoTermination;
 719                                 flags |= CharInfoMathNoTermination;
 720                         } else if (flag == "notermination=none") {
 721                                 flags &= ~CharInfoTextNoTermination;
 722                                 flags &= ~CharInfoMathNoTermination;
 723                         } else if (contains(flag, "tipashortcut=")) {
 724                                 tipashortcut = split(flag, '=');
 725                         } else {
 726                                 lyxerr << "Ignoring unknown flag `" << flag
 727                                        << "' for symbol `0x"
 728                                        << hex << symbol << dec
 729                                        << "'." << endl;
 730                         }
 731                 }
 732                 // mathcommand and mathpreamble have been added for 1.6.0.
 733                 // make them optional so that old files still work.
 734                 int const lineno = symbolslex.lineNumber();
 735                 bool breakout = false;
 736                 docstring mathcommand;
 737                 string mathpreamble;
 738                 if (symbolslex.next(true)) {
 739                         if (symbolslex.lineNumber() != lineno) {
 740                                 // line in old format without mathcommand and mathpreamble
 741                                 getNextToken = false;
 742                         } else {
 743                                 mathcommand = symbolslex.getDocString();
 744                                 if (suffixIs(mathcommand, '}'))
 745                                         flags |= CharInfoMathNoTermination;
 746                                 if (symbolslex.next(true)) {
 747                                         if (symbolslex.lineNumber() != lineno) {
 748                                                 // line in new format with mathcommand only
 749                                                 getNextToken = false;
 750                                         } else {
 751                                                 // line in new format with mathcommand and mathpreamble
 752                                                 mathpreamble = symbolslex.getString();
 753                                         }
 754                                 } else
 755                                         breakout = true;
 756                         }
 757                 } else {
 758                         breakout = true;
 759                 }
 760
 761                 // backward compatibility
 762                 if (mathpreamble == "esintoramsmath")
 763                         mathpreamble = "esint|amsmath";
 764
 765                 if (!textpreamble.empty())
 766                         if (textpreamble[0] != '\\')
 767                                 flags |= CharInfoTextFeature;
 768                 if (!mathpreamble.empty())
 769                         if (mathpreamble[0] != '\\')
 770                                 flags |= CharInfoMathFeature;
 771
 772                 CharInfo info = CharInfo(
 773                         textcommand, mathcommand,
 774                         textpreamble, mathpreamble,
 775                         tipashortcut, flags);
 776                 LYXERR(Debug::INFO, "Read unicode symbol " << symbol << " '"
 777                            << to_utf8(info.textcommand()) << "' '" << info.textpreamble()
 778                            << " '" << info.textfeature() << ' ' << info.textnotermination()
 779                            << ' ' << to_utf8(info.mathcommand()) << "' '" << info.mathpreamble()
 780                            << "' " << info.mathfeature() << ' ' << info.mathnotermination()
 781                            << ' ' << info.combining() << ' ' << info.force()
 782                            << ' ' << info.forceselected());
 783
 784                 // we assume that at least one command is nonempty when using unicodesymbols
 785                 if (info.isUnicodeSymbol()) {
 786                         unicodesymbols[symbol] = info;
 787                 }
 788
 789                 if (breakout)
 790                         break;
 791         }
 792
 793         // Now read the encodings
 794         enum {
 795                 et_encoding = 1,
 796                 et_end
 797         };
 798
 799         LexerKeyword encodingtags[] = {
 800                 { "encoding", et_encoding },
 801                 { "end", et_end }
 802         };
 803
 804         Lexer lex(encodingtags);
 805         lex.setFile(encfile);
 806         lex.setContext("Encodings::read");
 807         while (lex.isOK()) {
 808                 switch (lex.lex()) {
 809                 case et_encoding:
 810                 {
 811                         lex.next();
 812                         string const name = lex.getString();
 813                         lex.next();
 814                         string const latexname = lex.getString();
 815                         lex.next();
 816                         string const guiname = lex.getString();
 817                         lex.next();
 818                         string const iconvname = lex.getString();
 819                         lex.next();
 820                         string const width = lex.getString();
 821                         bool fixedwidth = false;
 822                         bool unsafe = false;
 823                         if (width == "fixed")
 824                                 fixedwidth = true;
 825                         else if (width == "variable")
 826                                 fixedwidth = false;
 827                         else if (width == "variableunsafe") {
 828                                 fixedwidth = false;
 829                                 unsafe = true;
 830                         }
 831                         else
 832                                 lex.printError("Unknown width");
 833
 834                         lex.next();
 835                         string const p = lex.getString();
 836                         Encoding::Package package = Encoding::none;
 837                         if (p == "none")
 838                                 package = Encoding::none;
 839                         else if (p == "inputenc")
 840                                 package = Encoding::inputenc;
 841                         else if (p == "CJK")
 842                                 package = Encoding::CJK;
 843                         else if (p == "japanese")
 844                                 package = Encoding::japanese;
 845                         else
 846                                 lex.printError("Unknown package");
 847
 848                         LYXERR(Debug::INFO, "Reading encoding " << name);
 849                         encodinglist[name] = Encoding(name, latexname,
 850                                 guiname, iconvname, fixedwidth, unsafe,
 851                                 package);
 852
 853                         if (lex.lex() != et_end)
 854                                 lex.printError("Missing end");
 855                         break;
 856                 }
 857                 case et_end:
 858                         lex.printError("Misplaced end");
 859                         break;
 860                 case Lexer::LEX_FEOF:
 861                         break;
 862                 default:
 863                         lex.printError("Unknown tag");
 864                         break;
 865                 }
 866         }
 867
 868         // Move all information from forcednotselected to forcedselected
 869         for (CharSetMap::const_iterator it1 = forcednotselected.begin(); it1 != forcednotselected.end(); ++it1) {
 870                 for (CharSetMap::iterator it2 = forcedselected.begin(); it2 != forcedselected.end(); ++it2) {
 871                         if (it2->first != it1->first)
 872                                 it2->second.insert(it1->second.begin(), it1->second.end());
 873                 }
 874         }
 875
 876 }
 877
 878
 879 } // namespace lyx