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 those below start_encodable_
 263         vector<char_type> symbols;
 264         for (char_type c = 0; c < start_encodable_; ++c)
 265                 symbols.push_back(c);
 266         //add all encodable characters
 267         copy(encodable_.begin(), encodable_.end(), back_inserter(symbols));
 268         // now the ones from the unicodesymbols file
 269         for (pair<char_type, CharInfo> const & elem : unicodesymbols)
 270                 symbols.push_back(elem.first);
 271         return symbols;
 272 }
 273
 274
 275 bool Encodings::latexMathChar(char_type c, bool mathmode,
 276                         Encoding const * encoding, docstring & command,
 277                         bool & needsTermination)
 278 {
 279         command = empty_docstring();
 280         if (encoding)
 281                 if (encoding->encodable(c))
 282                         command = docstring(1, c);
 283         needsTermination = false;
 284
 285         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 286         if (it == unicodesymbols.end()) {
 287                 if (!encoding || command.empty())
 288                         throw EncodingException(c);
 289                 if (mathmode)
 290                         addMathSym(c);
 291                 return false;
 292         }
 293         // at least one of mathcommand and textcommand is nonempty
 294         bool use_math = (mathmode && !it->second.mathcommand().empty()) ||
 295                         (!mathmode && it->second.textcommand().empty());
 296         if (use_math) {
 297                 command = it->second.mathcommand();
 298                 needsTermination = !it->second.mathnotermination();
 299                 addMathCmd(c);
 300         } else {
 301                 if (!encoding || command.empty()) {
 302                         command = it->second.textcommand();
 303                         needsTermination = !it->second.textnotermination();
 304                         addTextCmd(c);
 305                 }
 306                 if (mathmode)
 307                         addMathSym(c);
 308         }
 309         return use_math;
 310 }
 311
 312
 313 char_type Encodings::fromLaTeXCommand(docstring const & cmd, int cmdtype,
 314                 bool & combining, bool & needsTermination, set<string> * req)
 315 {
 316         CharInfoMap::const_iterator const end = unicodesymbols.end();
 317         CharInfoMap::const_iterator it = unicodesymbols.begin();
 318         for (combining = false; it != end; ++it) {
 319                 if (it->second.deprecated())
 320                         continue;
 321                 docstring const math = it->second.mathcommand();
 322                 docstring const text = it->second.textcommand();
 323                 if ((cmdtype & MATH_CMD) && math == cmd) {
 324                         combining = it->second.combining();
 325                         needsTermination = !it->second.mathnotermination();
 326                         if (req && it->second.mathfeature() &&
 327                             !it->second.mathpreamble().empty())
 328                                 req->insert(it->second.mathpreamble());
 329                         return it->first;
 330                 }
 331                 if ((cmdtype & TEXT_CMD) && text == cmd) {
 332                         combining = it->second.combining();
 333                         needsTermination = !it->second.textnotermination();
 334                         if (req && it->second.textfeature() &&
 335                             !it->second.textpreamble().empty())
 336                                 req->insert(it->second.textpreamble());
 337                         return it->first;
 338                 }
 339         }
 340         needsTermination = false;
 341         return 0;
 342 }
 343
 344
 345 docstring Encodings::fromLaTeXCommand(docstring const & cmd, int cmdtype,
 346                 bool & needsTermination, docstring & rem, set<string> * req)
 347 {
 348         needsTermination = false;
 349         rem = empty_docstring();
 350         bool const mathmode = cmdtype & MATH_CMD;
 351         bool const textmode = cmdtype & TEXT_CMD;
 352         docstring symbols;
 353         size_t const cmdend = cmd.size();
 354         size_t prefix = 0;
 355         CharInfoMap::const_iterator const uniend = unicodesymbols.end();
 356         for (size_t i = 0, j = 0; j < cmdend; ++j) {
 357                 // Also get the char after a backslash
 358                 if (j + 1 < cmdend && cmd[j] == '\\') {
 359                         ++j;
 360                         prefix = 1;
 361                         // Detect things like \=*{e} as well
 362                         if (j + 3 < cmdend && cmd[j+1] == '*' &&
 363                             cmd[j+2] == '{') {
 364                                 ++j;
 365                                 prefix = 2;
 366                         }
 367                 }
 368                 // position of the last character before a possible macro
 369                 // argument
 370                 size_t m = j;
 371                 // If a macro argument follows, get it, too
 372                 // Do it here only for single character commands. Other
 373                 // combining commands need this too, but they are handled in
 374                 // the loop below for performance reasons.
 375                 if (j + 1 < cmdend && cmd[j + 1] == '{') {
 376                         size_t k = j + 1;
 377                         int count = 1;
 378                         while (k < cmdend && count) {
 379                                 k = cmd.find_first_of(from_ascii("{}"), k + 1);
 380                                 // braces may not be balanced
 381                                 if (k == docstring::npos)
 382                                         break;
 383                                 if (cmd[k] == '{')
 384                                         ++count;
 385                                 else
 386                                         --count;
 387                         }
 388                         if (k != docstring::npos)
 389                                 j = k;
 390                 } else if (m + 1 < cmdend && isAlphaASCII(cmd[m])) {
 391                         while (m + 2 < cmdend && isAlphaASCII(cmd[m+1]))
 392                                 m++;
 393                 }
 394                 // Start with this substring and try augmenting it when it is
 395                 // the prefix of some command in the unicodesymbols file
 396                 docstring subcmd = cmd.substr(i, j - i + 1);
 397
 398                 CharInfoMap::const_iterator it = unicodesymbols.begin();
 399                 // First part of subcmd which might be a combining character
 400                 docstring combcmd = (m == j) ? docstring() : cmd.substr(i, m - i + 1);
 401                 // The combining character of combcmd if it exists
 402                 CharInfoMap::const_iterator combining = uniend;
 403                 size_t unicmd_size = 0;
 404                 char_type c = 0;
 405                 for (; it != uniend; ++it) {
 406                         if (it->second.deprecated())
 407                                 continue;
 408                         docstring const math = mathmode ? it->second.mathcommand()
 409                                                         : docstring();
 410                         docstring const text = textmode ? it->second.textcommand()
 411                                                         : docstring();
 412                         if (!combcmd.empty() && it->second.combining() &&
 413                             (math == combcmd || text == combcmd))
 414                                 combining = it;
 415                         size_t cur_size = max(math.size(), text.size());
 416                         // The current math or text unicode command cannot
 417                         // match, or we already matched a longer one
 418                         if (cur_size < subcmd.size() || cur_size <= unicmd_size)
 419                                 continue;
 420
 421                         docstring tmp = subcmd;
 422                         size_t k = j;
 423                         while (prefixIs(math, tmp) || prefixIs(text, tmp)) {
 424                                 ++k;
 425                                 if (k >= cmdend || cur_size <= tmp.size())
 426                                         break;
 427                                 tmp += cmd[k];
 428                         }
 429                         // No match
 430                         if (k == j)
 431                                 continue;
 432
 433                         // The last added char caused a mismatch, because
 434                         // we didn't exhaust the chars in cmd and didn't
 435                         // exceed the maximum size of the current unicmd
 436                         if (k < cmdend && cur_size > tmp.size())
 437                                 tmp.resize(tmp.size() - 1);
 438
 439                         // If this is an exact match, we found a (longer)
 440                         // matching entry in the unicodesymbols file.
 441                         if (math != tmp && text != tmp)
 442                                 continue;
 443                         // If we found a combining command, we need to append
 444                         // the macro argument if this has not been done above.
 445                         if (tmp == combcmd && combining != uniend &&
 446                             k < cmdend && cmd[k] == '{') {
 447                                 size_t l = k;
 448                                 int count = 1;
 449                                 while (l < cmdend && count) {
 450                                         l = cmd.find_first_of(from_ascii("{}"), l + 1);
 451                                         // braces may not be balanced
 452                                         if (l == docstring::npos)
 453                                                 break;
 454                                         if (cmd[l] == '{')
 455                                                 ++count;
 456                                         else
 457                                                 --count;
 458                                 }
 459                                 if (l != docstring::npos) {
 460                                         j = l;
 461                                         subcmd = cmd.substr(i, j - i + 1);
 462                                 }
 463                         }
 464                         // If the entry doesn't start with '\', we take note
 465                         // of the match and continue (this is not a ultimate
 466                         // acceptance, as some other entry may match a longer
 467                         // portion of the cmd string). However, if the entry
 468                         // does start with '\', we accept the match only if
 469                         // this is a valid macro, i.e., either it is a single
 470                         // (nonletter) char macro, or nothing else follows,
 471                         // or what follows is a nonletter char, or the last
 472                         // character is a }.
 473                         else if (tmp[0] != '\\'
 474                                    || (tmp.size() == prefix + 1 &&
 475                                        !isAlphaASCII(tmp[1]) &&
 476                                        (prefix == 1 || !isAlphaASCII(tmp[2])))
 477                                    || k == cmdend
 478                                    || !isAlphaASCII(cmd[k])
 479                                    || tmp[tmp.size() - 1] == '}'
 480                                  ) {
 481                                 c = it->first;
 482                                 j = k - 1;
 483                                 i = j + 1;
 484                                 unicmd_size = cur_size;
 485                                 if (math == tmp)
 486                                         needsTermination = !it->second.mathnotermination();
 487                                 else
 488                                         needsTermination = !it->second.textnotermination();
 489                                 if (req) {
 490                                         if (math == tmp && it->second.mathfeature() &&
 491                                             !it->second.mathpreamble().empty())
 492                                                 req->insert(it->second.mathpreamble());
 493                                         if (text == tmp && it->second.textfeature() &&
 494                                             !it->second.textpreamble().empty())
 495                                                 req->insert(it->second.textpreamble());
 496                                 }
 497                         }
 498                 }
 499                 if (unicmd_size)
 500                         symbols += c;
 501                 else if (combining != uniend &&
 502                          prefixIs(subcmd, combcmd + '{')) {
 503                         // We know that subcmd starts with combcmd and
 504                         // contains an argument in braces.
 505                         docstring const arg = subcmd.substr(
 506                                 combcmd.length() + 1,
 507                                 subcmd.length() - combcmd.length() - 2);
 508                         // If arg is a single character we can construct a
 509                         // combining sequence.
 510                         char_type a;
 511                         bool argcomb = false;
 512                         if (arg.size() == 1 && isAlnumASCII(arg[0]))
 513                                 a = arg[0];
 514                         else {
 515                                 // Use the version of fromLaTeXCommand() that
 516                                 // parses only one command, since we cannot
 517                                 // use more than one character.
 518                                 bool dummy = false;
 519                                 set<string> r;
 520                                 a = fromLaTeXCommand(arg, cmdtype, argcomb,
 521                                                      dummy, &r);
 522                                 if (a && req && !argcomb)
 523                                         req->insert(r.begin(), r.end());
 524                         }
 525                         if (a && !argcomb) {
 526                                 // In unicode the combining character comes
 527                                 // after its base
 528                                 symbols += a;
 529                                 symbols += combining->first;
 530                                 i = j + 1;
 531                                 unicmd_size = 2;
 532                         }
 533                 }
 534                 if (j + 1 == cmdend && !unicmd_size) {
 535                         // No luck. Return what remains
 536                         rem = cmd.substr(i);
 537                         if (needsTermination && !rem.empty()) {
 538                                 if (rem.substr(0, 2) == "{}") {
 539                                         rem = rem.substr(2);
 540                                         needsTermination = false;
 541                                 } else if (rem[0] == ' ') {
 542                                         needsTermination = false;
 543                                         // LaTeX would swallow all spaces
 544                                         rem = ltrim(rem);
 545                                 }
 546                         }
 547                 }
 548         }
 549         return symbols;
 550 }
 551
 552
 553 CharInfo const & Encodings::unicodeCharInfo(char_type c)
 554 {
 555         static CharInfo empty;
 556         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 557         return it != unicodesymbols.end() ? it->second : empty;
 558 }
 559
 560
 561 bool Encodings::isCombiningChar(char_type c)
 562 {
 563         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 564         if (it != unicodesymbols.end())
 565                 return it->second.combining();
 566         return false;
 567 }
 568
 569
 570 string const Encodings::TIPAShortcut(char_type c)
 571 {
 572         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 573         if (it != unicodesymbols.end())
 574                 return it->second.tipashortcut();
 575         return string();
 576 }
 577
 578
 579 bool Encodings::isKnownScriptChar(char_type const c, string & preamble)
 580 {
 581         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 582
 583         if (it == unicodesymbols.end())
 584                 return false;
 585
 586         if (it->second.textpreamble() != "textgreek" && it->second.textpreamble() != "textcyr")
 587                 return false;
 588
 589         if (preamble.empty()) {
 590                 preamble = it->second.textpreamble();
 591                 return true;
 592         }
 593         return it->second.textpreamble() == preamble;
 594 }
 595
 596
 597 bool Encodings::isMathAlpha(char_type c)
 598 {
 599         return mathalpha.count(c);
 600 }
 601
 602
 603 bool Encodings::isUnicodeTextOnly(char_type c)
 604 {
 605         if (isASCII(c) || isMathAlpha(c))
 606                 return false;
 607
 608         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 609         return it == unicodesymbols.end() || it->second.mathcommand().empty();
 610 }
 611
 612
 613 Encoding const *
 614 Encodings::fromLyXName(string const & name, bool allowUnsafe) const
 615 {
 616         EncodingList::const_iterator const it = encodinglist.find(name);
 617         if (it == encodinglist.end())
 618                 return 0;
 619         if (!allowUnsafe && it->second.unsafe())
 620                 return 0;
 621         return &it->second;
 622 }
 623
 624
 625 Encoding const *
 626 Encodings::fromLaTeXName(string const & n, int const & p, bool allowUnsafe) const
 627 {
 628         string name = n;
 629         // FIXME: if we have to test for too many of these synonyms,
 630         // we should instead extend the format of lib/encodings
 631         if (n == "ansinew")
 632                 name = "cp1252";
 633
 634         // We don't use find_if because it makes copies of the pairs in
 635         // the map.
 636         // This linear search is OK since we don't have many encodings.
 637         // Users could even optimize it by putting the encodings they use
 638         // most at the top of lib/encodings.
 639         EncodingList::const_iterator const end = encodinglist.end();
 640         for (EncodingList::const_iterator it = encodinglist.begin(); it != end; ++it)
 641                 if ((it->second.latexName() == name) && (it->second.package() & p)
 642                                 && (!it->second.unsafe() || allowUnsafe))
 643                         return &it->second;
 644         return 0;
 645 }
 646
 647
 648 Encoding const *
 649 Encodings::fromIconvName(string const & n, int const & p, bool allowUnsafe) const
 650 {
 651         EncodingList::const_iterator const end = encodinglist.end();
 652         for (EncodingList::const_iterator it = encodinglist.begin(); it != end; ++it)
 653                 if ((it->second.iconvName() == n) && (it->second.package() & p)
 654                                 && (!it->second.unsafe() || allowUnsafe))
 655                         return &it->second;
 656         return 0;
 657 }
 658
 659
 660 Encodings::Encodings()
 661 {}
 662
 663
 664 void Encodings::read(FileName const & encfile, FileName const & symbolsfile)
 665 {
 666         // We must read the symbolsfile first, because the Encoding
 667         // constructor depends on it.
 668         CharSetMap forcednotselected;
 669         Lexer symbolslex;
 670         symbolslex.setFile(symbolsfile);
 671         bool getNextToken = true;
 672         while (symbolslex.isOK()) {
 673                 char_type symbol;
 674
 675                 if (getNextToken) {
 676                         if (!symbolslex.next(true))
 677                                 break;
 678                 } else
 679                         getNextToken = true;
 680
 681                 istringstream is(symbolslex.getString());
 682                 // reading symbol directly does not work if
 683                 // char_type == wchar_t.
 684                 boost::uint32_t tmp;
 685                 if(!(is >> hex >> tmp))
 686                         break;
 687                 symbol = tmp;
 688
 689                 if (!symbolslex.next(true))
 690                         break;
 691                 docstring textcommand = symbolslex.getDocString();
 692                 if (!symbolslex.next(true))
 693                         break;
 694                 string textpreamble = symbolslex.getString();
 695                 if (!symbolslex.next(true))
 696                         break;
 697                 string sflags = symbolslex.getString();
 698
 699                 string tipashortcut;
 700                 int flags = 0;
 701
 702                 if (suffixIs(textcommand, '}'))
 703                         flags |= CharInfoTextNoTermination;
 704                 while (!sflags.empty()) {
 705                         string flag;
 706                         sflags = split(sflags, flag, ',');
 707                         if (flag == "combining") {
 708                                 flags |= CharInfoCombining;
 709                         } else if (flag == "force") {
 710                                 flags |= CharInfoForce;
 711                                 forced.insert(symbol);
 712                         } else if (prefixIs(flag, "force=")) {
 713                                 vector<string> encodings =
 714                                         getVectorFromString(flag.substr(6), ";");
 715                                 for (size_t i = 0; i < encodings.size(); ++i)
 716                                         forcedselected[encodings[i]].insert(symbol);
 717                                 flags |= CharInfoForceSelected;
 718                         } else if (prefixIs(flag, "force!=")) {
 719                                 vector<string> encodings =
 720                                         getVectorFromString(flag.substr(7), ";");
 721                                 for (size_t i = 0; i < encodings.size(); ++i)
 722                                         forcednotselected[encodings[i]].insert(symbol);
 723                                 flags |= CharInfoForceSelected;
 724                         } else if (flag == "mathalpha") {
 725                                 mathalpha.insert(symbol);
 726                         } else if (flag == "notermination=text") {
 727                                 flags |= CharInfoTextNoTermination;
 728                         } else if (flag == "notermination=math") {
 729                                 flags |= CharInfoMathNoTermination;
 730                         } else if (flag == "notermination=both") {
 731                                 flags |= CharInfoTextNoTermination;
 732                                 flags |= CharInfoMathNoTermination;
 733                         } else if (flag == "notermination=none") {
 734                                 flags &= ~CharInfoTextNoTermination;
 735                                 flags &= ~CharInfoMathNoTermination;
 736                         } else if (contains(flag, "tipashortcut=")) {
 737                                 tipashortcut = split(flag, '=');
 738                         } else if (flag == "deprecated") {
 739                                 flags |= CharInfoDeprecated;
 740                         } else {
 741                                 lyxerr << "Ignoring unknown flag `" << flag
 742                                        << "' for symbol `0x"
 743                                        << hex << symbol << dec
 744                                        << "'." << endl;
 745                         }
 746                 }
 747                 // mathcommand and mathpreamble have been added for 1.6.0.
 748                 // make them optional so that old files still work.
 749                 int const lineno = symbolslex.lineNumber();
 750                 bool breakout = false;
 751                 docstring mathcommand;
 752                 string mathpreamble;
 753                 if (symbolslex.next(true)) {
 754                         if (symbolslex.lineNumber() != lineno) {
 755                                 // line in old format without mathcommand and mathpreamble
 756                                 getNextToken = false;
 757                         } else {
 758                                 mathcommand = symbolslex.getDocString();
 759                                 if (suffixIs(mathcommand, '}'))
 760                                         flags |= CharInfoMathNoTermination;
 761                                 if (symbolslex.next(true)) {
 762                                         if (symbolslex.lineNumber() != lineno) {
 763                                                 // line in new format with mathcommand only
 764                                                 getNextToken = false;
 765                                         } else {
 766                                                 // line in new format with mathcommand and mathpreamble
 767                                                 mathpreamble = symbolslex.getString();
 768                                         }
 769                                 } else
 770                                         breakout = true;
 771                         }
 772                 } else {
 773                         breakout = true;
 774                 }
 775
 776                 // backward compatibility
 777                 if (mathpreamble == "esintoramsmath")
 778                         mathpreamble = "esint|amsmath";
 779
 780                 if (!textpreamble.empty())
 781                         if (textpreamble[0] != '\\')
 782                                 flags |= CharInfoTextFeature;
 783                 if (!mathpreamble.empty())
 784                         if (mathpreamble[0] != '\\')
 785                                 flags |= CharInfoMathFeature;
 786
 787                 CharInfo info = CharInfo(
 788                         textcommand, mathcommand,
 789                         textpreamble, mathpreamble,
 790                         tipashortcut, flags);
 791                 LYXERR(Debug::INFO, "Read unicode symbol " << symbol << " '"
 792                            << to_utf8(info.textcommand()) << "' '" << info.textpreamble()
 793                            << " '" << info.textfeature() << ' ' << info.textnotermination()
 794                            << ' ' << to_utf8(info.mathcommand()) << "' '" << info.mathpreamble()
 795                            << "' " << info.mathfeature() << ' ' << info.mathnotermination()
 796                            << ' ' << info.combining() << ' ' << info.force()
 797                            << ' ' << info.forceselected());
 798
 799                 // we assume that at least one command is nonempty when using unicodesymbols
 800                 if (info.isUnicodeSymbol()) {
 801                         unicodesymbols[symbol] = info;
 802                 }
 803
 804                 if (breakout)
 805                         break;
 806         }
 807
 808         // Now read the encodings
 809         enum {
 810                 et_encoding = 1,
 811                 et_end
 812         };
 813
 814         LexerKeyword encodingtags[] = {
 815                 { "encoding", et_encoding },
 816                 { "end", et_end }
 817         };
 818
 819         Lexer lex(encodingtags);
 820         lex.setFile(encfile);
 821         lex.setContext("Encodings::read");
 822         while (lex.isOK()) {
 823                 switch (lex.lex()) {
 824                 case et_encoding:
 825                 {
 826                         lex.next();
 827                         string const name = lex.getString();
 828                         lex.next();
 829                         string const latexname = lex.getString();
 830                         lex.next();
 831                         string const guiname = lex.getString();
 832                         lex.next();
 833                         string const iconvname = lex.getString();
 834                         lex.next();
 835                         string const width = lex.getString();
 836                         bool fixedwidth = false;
 837                         bool unsafe = false;
 838                         if (width == "fixed")
 839                                 fixedwidth = true;
 840                         else if (width == "variable")
 841                                 fixedwidth = false;
 842                         else if (width == "variableunsafe") {
 843                                 fixedwidth = false;
 844                                 unsafe = true;
 845                         }
 846                         else
 847                                 lex.printError("Unknown width");
 848
 849                         lex.next();
 850                         string const p = lex.getString();
 851                         Encoding::Package package = Encoding::none;
 852                         if (p == "none")
 853                                 package = Encoding::none;
 854                         else if (p == "inputenc")
 855                                 package = Encoding::inputenc;
 856                         else if (p == "CJK")
 857                                 package = Encoding::CJK;
 858                         else if (p == "japanese")
 859                                 package = Encoding::japanese;
 860                         else
 861                                 lex.printError("Unknown package");
 862
 863                         LYXERR(Debug::INFO, "Reading encoding " << name);
 864                         encodinglist[name] = Encoding(name, latexname,
 865                                 guiname, iconvname, fixedwidth, unsafe,
 866                                 package);
 867
 868                         if (lex.lex() != et_end)
 869                                 lex.printError("Missing end");
 870                         break;
 871                 }
 872                 case et_end:
 873                         lex.printError("Misplaced end");
 874                         break;
 875                 case Lexer::LEX_FEOF:
 876                         break;
 877                 default:
 878                         lex.printError("Unknown tag");
 879                         break;
 880                 }
 881         }
 882
 883         // Move all information from forcednotselected to forcedselected
 884         for (CharSetMap::const_iterator it1 = forcednotselected.begin(); it1 != forcednotselected.end(); ++it1) {
 885                 for (CharSetMap::iterator it2 = forcedselected.begin(); it2 != forcedselected.end(); ++it2) {
 886                         if (it2->first != it1->first)
 887                                 it2->second.insert(it1->second.begin(), it1->second.end());
 888                 }
 889         }
 890
 891 }
 892
 893
 894 } // namespace lyx