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