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                 if (it->second.deprecated())
 321                         continue;
 322                 docstring const math = it->second.mathcommand();
 323                 docstring const text = it->second.textcommand();
 324                 if ((cmdtype & MATH_CMD) && math == cmd) {
 325                         combining = it->second.combining();
 326                         needsTermination = !it->second.mathnotermination();
 327                         if (req && it->second.mathfeature() &&
 328                             !it->second.mathpreamble().empty())
 329                                 req->insert(it->second.mathpreamble());
 330                         return it->first;
 331                 }
 332                 if ((cmdtype & TEXT_CMD) && text == cmd) {
 333                         combining = it->second.combining();
 334                         needsTermination = !it->second.textnotermination();
 335                         if (req && it->second.textfeature() &&
 336                             !it->second.textpreamble().empty())
 337                                 req->insert(it->second.textpreamble());
 338                         return it->first;
 339                 }
 340         }
 341         needsTermination = false;
 342         return 0;
 343 }
 344
 345
 346 docstring Encodings::fromLaTeXCommand(docstring const & cmd, int cmdtype,
 347                 bool & needsTermination, docstring & rem, set<string> * req)
 348 {
 349         needsTermination = false;
 350         rem = empty_docstring();
 351         bool const mathmode = cmdtype & MATH_CMD;
 352         bool const textmode = cmdtype & TEXT_CMD;
 353         docstring symbols;
 354         size_t const cmdend = cmd.size();
 355         size_t prefix = 0;
 356         CharInfoMap::const_iterator const uniend = unicodesymbols.end();
 357         for (size_t i = 0, j = 0; j < cmdend; ++j) {
 358                 // Also get the char after a backslash
 359                 if (j + 1 < cmdend && cmd[j] == '\\') {
 360                         ++j;
 361                         prefix = 1;
 362                         // Detect things like \=*{e} as well
 363                         if (j + 3 < cmdend && cmd[j+1] == '*' &&
 364                             cmd[j+2] == '{') {
 365                                 ++j;
 366                                 prefix = 2;
 367                         }
 368                 }
 369                 // position of the last character before a possible macro
 370                 // argument
 371                 size_t m = j;
 372                 // If a macro argument follows, get it, too
 373                 // Do it here only for single character commands. Other
 374                 // combining commands need this too, but they are handled in
 375                 // the loop below for performance reasons.
 376                 if (j + 1 < cmdend && cmd[j + 1] == '{') {
 377                         size_t k = j + 1;
 378                         int count = 1;
 379                         while (k < cmdend && count) {
 380                                 k = cmd.find_first_of(from_ascii("{}"), k + 1);
 381                                 // braces may not be balanced
 382                                 if (k == docstring::npos)
 383                                         break;
 384                                 if (cmd[k] == '{')
 385                                         ++count;
 386                                 else
 387                                         --count;
 388                         }
 389                         if (k != docstring::npos)
 390                                 j = k;
 391                 } else if (m + 1 < cmdend && isAlphaASCII(cmd[m])) {
 392                         while (m + 2 < cmdend && isAlphaASCII(cmd[m+1]))
 393                                 m++;
 394                 }
 395                 // Start with this substring and try augmenting it when it is
 396                 // the prefix of some command in the unicodesymbols file
 397                 docstring subcmd = cmd.substr(i, j - i + 1);
 398
 399                 CharInfoMap::const_iterator it = unicodesymbols.begin();
 400                 // First part of subcmd which might be a combining character
 401                 docstring combcmd = (m == j) ? docstring() : cmd.substr(i, m - i + 1);
 402                 // The combining character of combcmd if it exists
 403                 CharInfoMap::const_iterator combining = uniend;
 404                 size_t unicmd_size = 0;
 405                 char_type c = 0;
 406                 for (; it != uniend; ++it) {
 407                         if (it->second.deprecated())
 408                                 continue;
 409                         docstring const math = mathmode ? it->second.mathcommand()
 410                                                         : docstring();
 411                         docstring const text = textmode ? it->second.textcommand()
 412                                                         : docstring();
 413                         if (!combcmd.empty() && it->second.combining() &&
 414                             (math == combcmd || text == combcmd))
 415                                 combining = it;
 416                         size_t cur_size = max(math.size(), text.size());
 417                         // The current math or text unicode command cannot
 418                         // match, or we already matched a longer one
 419                         if (cur_size < subcmd.size() || cur_size <= unicmd_size)
 420                                 continue;
 421
 422                         docstring tmp = subcmd;
 423                         size_t k = j;
 424                         while (prefixIs(math, tmp) || prefixIs(text, tmp)) {
 425                                 ++k;
 426                                 if (k >= cmdend || cur_size <= tmp.size())
 427                                         break;
 428                                 tmp += cmd[k];
 429                         }
 430                         // No match
 431                         if (k == j)
 432                                 continue;
 433
 434                         // The last added char caused a mismatch, because
 435                         // we didn't exhaust the chars in cmd and didn't
 436                         // exceed the maximum size of the current unicmd
 437                         if (k < cmdend && cur_size > tmp.size())
 438                                 tmp.resize(tmp.size() - 1);
 439
 440                         // If this is an exact match, we found a (longer)
 441                         // matching entry in the unicodesymbols file.
 442                         if (math != tmp && text != tmp)
 443                                 continue;
 444                         // If we found a combining command, we need to append
 445                         // the macro argument if this has not been done above.
 446                         if (tmp == combcmd && combining != uniend &&
 447                             k < cmdend && cmd[k] == '{') {
 448                                 size_t l = k;
 449                                 int count = 1;
 450                                 while (l < cmdend && count) {
 451                                         l = cmd.find_first_of(from_ascii("{}"), l + 1);
 452                                         // braces may not be balanced
 453                                         if (l == docstring::npos)
 454                                                 break;
 455                                         if (cmd[l] == '{')
 456                                                 ++count;
 457                                         else
 458                                                 --count;
 459                                 }
 460                                 if (l != docstring::npos) {
 461                                         j = l;
 462                                         subcmd = cmd.substr(i, j - i + 1);
 463                                 }
 464                         }
 465                         // If the entry doesn't start with '\', we take note
 466                         // of the match and continue (this is not a ultimate
 467                         // acceptance, as some other entry may match a longer
 468                         // portion of the cmd string). However, if the entry
 469                         // does start with '\', we accept the match only if
 470                         // this is a valid macro, i.e., either it is a single
 471                         // (nonletter) char macro, or nothing else follows,
 472                         // or what follows is a nonletter char, or the last
 473                         // character is a }.
 474                         else if (tmp[0] != '\\'
 475                                    || (tmp.size() == prefix + 1 &&
 476                                        !isAlphaASCII(tmp[1]) &&
 477                                        (prefix == 1 || !isAlphaASCII(tmp[2])))
 478                                    || k == cmdend
 479                                    || !isAlphaASCII(cmd[k])
 480                                    || tmp[tmp.size() - 1] == '}'
 481                                  ) {
 482                                 c = it->first;
 483                                 j = k - 1;
 484                                 i = j + 1;
 485                                 unicmd_size = cur_size;
 486                                 if (math == tmp)
 487                                         needsTermination = !it->second.mathnotermination();
 488                                 else
 489                                         needsTermination = !it->second.textnotermination();
 490                                 if (req) {
 491                                         if (math == tmp && it->second.mathfeature() &&
 492                                             !it->second.mathpreamble().empty())
 493                                                 req->insert(it->second.mathpreamble());
 494                                         if (text == tmp && it->second.textfeature() &&
 495                                             !it->second.textpreamble().empty())
 496                                                 req->insert(it->second.textpreamble());
 497                                 }
 498                         }
 499                 }
 500                 if (unicmd_size)
 501                         symbols += c;
 502                 else if (combining != uniend &&
 503                          prefixIs(subcmd, combcmd + '{')) {
 504                         // We know that subcmd starts with combcmd and
 505                         // contains an argument in braces.
 506                         docstring const arg = subcmd.substr(
 507                                 combcmd.length() + 1,
 508                                 subcmd.length() - combcmd.length() - 2);
 509                         // If arg is a single character we can construct a
 510                         // combining sequence.
 511                         char_type a;
 512                         bool argcomb = false;
 513                         if (arg.size() == 1 && isAlnumASCII(arg[0]))
 514                                 a = arg[0];
 515                         else {
 516                                 // Use the version of fromLaTeXCommand() that
 517                                 // parses only one command, since we cannot
 518                                 // use more than one character.
 519                                 bool dummy = false;
 520                                 set<string> r;
 521                                 a = fromLaTeXCommand(arg, cmdtype, argcomb,
 522                                                      dummy, &r);
 523                                 if (a && req && !argcomb)
 524                                         req->insert(r.begin(), r.end());
 525                         }
 526                         if (a && !argcomb) {
 527                                 // In unicode the combining character comes
 528                                 // after its base
 529                                 symbols += a;
 530                                 symbols += combining->first;
 531                                 i = j + 1;
 532                                 unicmd_size = 2;
 533                         }
 534                 }
 535                 if (j + 1 == cmdend && !unicmd_size) {
 536                         // No luck. Return what remains
 537                         rem = cmd.substr(i);
 538                         if (needsTermination && !rem.empty()) {
 539                                 if (rem.substr(0, 2) == "{}") {
 540                                         rem = rem.substr(2);
 541                                         needsTermination = false;
 542                                 } else if (rem[0] == ' ') {
 543                                         needsTermination = false;
 544                                         // LaTeX would swallow all spaces
 545                                         rem = ltrim(rem);
 546                                 }
 547                         }
 548                 }
 549         }
 550         return symbols;
 551 }
 552
 553
 554 CharInfo const & Encodings::unicodeCharInfo(char_type c)
 555 {
 556         static CharInfo empty;
 557         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 558         return it != unicodesymbols.end() ? it->second : empty;
 559 }
 560
 561
 562 bool Encodings::isCombiningChar(char_type c)
 563 {
 564         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 565         if (it != unicodesymbols.end())
 566                 return it->second.combining();
 567         return false;
 568 }
 569
 570
 571 string const Encodings::TIPAShortcut(char_type c)
 572 {
 573         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 574         if (it != unicodesymbols.end())
 575                 return it->second.tipashortcut();
 576         return string();
 577 }
 578
 579
 580 bool Encodings::isKnownScriptChar(char_type const c, string & preamble)
 581 {
 582         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 583
 584         if (it == unicodesymbols.end())
 585                 return false;
 586
 587         if (it->second.textpreamble() != "textgreek" && it->second.textpreamble() != "textcyr")
 588                 return false;
 589
 590         if (preamble.empty()) {
 591                 preamble = it->second.textpreamble();
 592                 return true;
 593         }
 594         return it->second.textpreamble() == preamble;
 595 }
 596
 597
 598 bool Encodings::isMathAlpha(char_type c)
 599 {
 600         return mathalpha.count(c);
 601 }
 602
 603
 604 Encoding const *
 605 Encodings::fromLyXName(string const & name, bool allowUnsafe) const
 606 {
 607         EncodingList::const_iterator const it = encodinglist.find(name);
 608         if (it == encodinglist.end())
 609                 return 0;
 610         if (!allowUnsafe && it->second.unsafe())
 611                 return 0;
 612         return &it->second;
 613 }
 614
 615
 616 Encoding const *
 617 Encodings::fromLaTeXName(string const & n, int const & p, bool allowUnsafe) const
 618 {
 619         string name = n;
 620         // FIXME: if we have to test for too many of these synonyms,
 621         // we should instead extend the format of lib/encodings
 622         if (n == "ansinew")
 623                 name = "cp1252";
 624
 625         // We don't use find_if because it makes copies of the pairs in
 626         // the map.
 627         // This linear search is OK since we don't have many encodings.
 628         // Users could even optimize it by putting the encodings they use
 629         // most at the top of lib/encodings.
 630         EncodingList::const_iterator const end = encodinglist.end();
 631         for (EncodingList::const_iterator it = encodinglist.begin(); it != end; ++it)
 632                 if ((it->second.latexName() == name) && (it->second.package() & p)
 633                                 && (!it->second.unsafe() || allowUnsafe))
 634                         return &it->second;
 635         return 0;
 636 }
 637
 638
 639 Encoding const *
 640 Encodings::fromIconvName(string const & n, int const & p, bool allowUnsafe) const
 641 {
 642         EncodingList::const_iterator const end = encodinglist.end();
 643         for (EncodingList::const_iterator it = encodinglist.begin(); it != end; ++it)
 644                 if ((it->second.iconvName() == n) && (it->second.package() & p)
 645                                 && (!it->second.unsafe() || allowUnsafe))
 646                         return &it->second;
 647         return 0;
 648 }
 649
 650
 651 Encodings::Encodings()
 652 {}
 653
 654
 655 void Encodings::read(FileName const & encfile, FileName const & symbolsfile)
 656 {
 657         // We must read the symbolsfile first, because the Encoding
 658         // constructor depends on it.
 659         CharSetMap forcednotselected;
 660         Lexer symbolslex;
 661         symbolslex.setFile(symbolsfile);
 662         bool getNextToken = true;
 663         while (symbolslex.isOK()) {
 664                 char_type symbol;
 665
 666                 if (getNextToken) {
 667                         if (!symbolslex.next(true))
 668                                 break;
 669                 } else
 670                         getNextToken = true;
 671
 672                 istringstream is(symbolslex.getString());
 673                 // reading symbol directly does not work if
 674                 // char_type == wchar_t.
 675                 boost::uint32_t tmp;
 676                 if(!(is >> hex >> tmp))
 677                         break;
 678                 symbol = tmp;
 679
 680                 if (!symbolslex.next(true))
 681                         break;
 682                 docstring textcommand = symbolslex.getDocString();
 683                 if (!symbolslex.next(true))
 684                         break;
 685                 string textpreamble = symbolslex.getString();
 686                 if (!symbolslex.next(true))
 687                         break;
 688                 string sflags = symbolslex.getString();
 689
 690                 string tipashortcut;
 691                 int flags = 0;
 692
 693                 if (suffixIs(textcommand, '}'))
 694                         flags |= CharInfoTextNoTermination;
 695                 while (!sflags.empty()) {
 696                         string flag;
 697                         sflags = split(sflags, flag, ',');
 698                         if (flag == "combining") {
 699                                 flags |= CharInfoCombining;
 700                         } else if (flag == "force") {
 701                                 flags |= CharInfoForce;
 702                                 forced.insert(symbol);
 703                         } else if (prefixIs(flag, "force=")) {
 704                                 vector<string> encodings =
 705                                         getVectorFromString(flag.substr(6), ";");
 706                                 for (size_t i = 0; i < encodings.size(); ++i)
 707                                         forcedselected[encodings[i]].insert(symbol);
 708                                 flags |= CharInfoForceSelected;
 709                         } else if (prefixIs(flag, "force!=")) {
 710                                 vector<string> encodings =
 711                                         getVectorFromString(flag.substr(7), ";");
 712                                 for (size_t i = 0; i < encodings.size(); ++i)
 713                                         forcednotselected[encodings[i]].insert(symbol);
 714                                 flags |= CharInfoForceSelected;
 715                         } else if (flag == "mathalpha") {
 716                                 mathalpha.insert(symbol);
 717                         } else if (flag == "notermination=text") {
 718                                 flags |= CharInfoTextNoTermination;
 719                         } else if (flag == "notermination=math") {
 720                                 flags |= CharInfoMathNoTermination;
 721                         } else if (flag == "notermination=both") {
 722                                 flags |= CharInfoTextNoTermination;
 723                                 flags |= CharInfoMathNoTermination;
 724                         } else if (flag == "notermination=none") {
 725                                 flags &= ~CharInfoTextNoTermination;
 726                                 flags &= ~CharInfoMathNoTermination;
 727                         } else if (contains(flag, "tipashortcut=")) {
 728                                 tipashortcut = split(flag, '=');
 729                         } else if (flag == "deprecated") {
 730                                 flags |= CharInfoDeprecated;
 731                         } else {
 732                                 lyxerr << "Ignoring unknown flag `" << flag
 733                                        << "' for symbol `0x"
 734                                        << hex << symbol << dec
 735                                        << "'." << endl;
 736                         }
 737                 }
 738                 // mathcommand and mathpreamble have been added for 1.6.0.
 739                 // make them optional so that old files still work.
 740                 int const lineno = symbolslex.lineNumber();
 741                 bool breakout = false;
 742                 docstring mathcommand;
 743                 string mathpreamble;
 744                 if (symbolslex.next(true)) {
 745                         if (symbolslex.lineNumber() != lineno) {
 746                                 // line in old format without mathcommand and mathpreamble
 747                                 getNextToken = false;
 748                         } else {
 749                                 mathcommand = symbolslex.getDocString();
 750                                 if (suffixIs(mathcommand, '}'))
 751                                         flags |= CharInfoMathNoTermination;
 752                                 if (symbolslex.next(true)) {
 753                                         if (symbolslex.lineNumber() != lineno) {
 754                                                 // line in new format with mathcommand only
 755                                                 getNextToken = false;
 756                                         } else {
 757                                                 // line in new format with mathcommand and mathpreamble
 758                                                 mathpreamble = symbolslex.getString();
 759                                         }
 760                                 } else
 761                                         breakout = true;
 762                         }
 763                 } else {
 764                         breakout = true;
 765                 }
 766
 767                 // backward compatibility
 768                 if (mathpreamble == "esintoramsmath")
 769                         mathpreamble = "esint|amsmath";
 770
 771                 if (!textpreamble.empty())
 772                         if (textpreamble[0] != '\\')
 773                                 flags |= CharInfoTextFeature;
 774                 if (!mathpreamble.empty())
 775                         if (mathpreamble[0] != '\\')
 776                                 flags |= CharInfoMathFeature;
 777
 778                 CharInfo info = CharInfo(
 779                         textcommand, mathcommand,
 780                         textpreamble, mathpreamble,
 781                         tipashortcut, flags);
 782                 LYXERR(Debug::INFO, "Read unicode symbol " << symbol << " '"
 783                            << to_utf8(info.textcommand()) << "' '" << info.textpreamble()
 784                            << " '" << info.textfeature() << ' ' << info.textnotermination()
 785                            << ' ' << to_utf8(info.mathcommand()) << "' '" << info.mathpreamble()
 786                            << "' " << info.mathfeature() << ' ' << info.mathnotermination()
 787                            << ' ' << info.combining() << ' ' << info.force()
 788                            << ' ' << info.forceselected());
 789
 790                 // we assume that at least one command is nonempty when using unicodesymbols
 791                 if (info.isUnicodeSymbol()) {
 792                         unicodesymbols[symbol] = info;
 793                 }
 794
 795                 if (breakout)
 796                         break;
 797         }
 798
 799         // Now read the encodings
 800         enum {
 801                 et_encoding = 1,
 802                 et_end
 803         };
 804
 805         LexerKeyword encodingtags[] = {
 806                 { "encoding", et_encoding },
 807                 { "end", et_end }
 808         };
 809
 810         Lexer lex(encodingtags);
 811         lex.setFile(encfile);
 812         lex.setContext("Encodings::read");
 813         while (lex.isOK()) {
 814                 switch (lex.lex()) {
 815                 case et_encoding:
 816                 {
 817                         lex.next();
 818                         string const name = lex.getString();
 819                         lex.next();
 820                         string const latexname = lex.getString();
 821                         lex.next();
 822                         string const guiname = lex.getString();
 823                         lex.next();
 824                         string const iconvname = lex.getString();
 825                         lex.next();
 826                         string const width = lex.getString();
 827                         bool fixedwidth = false;
 828                         bool unsafe = false;
 829                         if (width == "fixed")
 830                                 fixedwidth = true;
 831                         else if (width == "variable")
 832                                 fixedwidth = false;
 833                         else if (width == "variableunsafe") {
 834                                 fixedwidth = false;
 835                                 unsafe = true;
 836                         }
 837                         else
 838                                 lex.printError("Unknown width");
 839
 840                         lex.next();
 841                         string const p = lex.getString();
 842                         Encoding::Package package = Encoding::none;
 843                         if (p == "none")
 844                                 package = Encoding::none;
 845                         else if (p == "inputenc")
 846                                 package = Encoding::inputenc;
 847                         else if (p == "CJK")
 848                                 package = Encoding::CJK;
 849                         else if (p == "japanese")
 850                                 package = Encoding::japanese;
 851                         else
 852                                 lex.printError("Unknown package");
 853
 854                         LYXERR(Debug::INFO, "Reading encoding " << name);
 855                         encodinglist[name] = Encoding(name, latexname,
 856                                 guiname, iconvname, fixedwidth, unsafe,
 857                                 package);
 858
 859                         if (lex.lex() != et_end)
 860                                 lex.printError("Missing end");
 861                         break;
 862                 }
 863                 case et_end:
 864                         lex.printError("Misplaced end");
 865                         break;
 866                 case Lexer::LEX_FEOF:
 867                         break;
 868                 default:
 869                         lex.printError("Unknown tag");
 870                         break;
 871                 }
 872         }
 873
 874         // Move all information from forcednotselected to forcedselected
 875         for (CharSetMap::const_iterator it1 = forcednotselected.begin(); it1 != forcednotselected.end(); ++it1) {
 876                 for (CharSetMap::iterator it2 = forcedselected.begin(); it2 != forcedselected.end(); ++it2) {
 877                         if (it2->first != it1->first)
 878                                 it2->second.insert(it1->second.begin(), it1->second.end());
 879                 }
 880         }
 881
 882 }
 883
 884
 885 } // namespace lyx