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 <iterator>
  29 #include <algorithm>
  30 #include <sstream>
  31
  32 using namespace std;
  33 using namespace lyx::support;
  34
  35 namespace lyx {
  36
  37 int const Encoding::any = -1;
  38
  39 Encodings encodings;
  40
  41 Encodings::MathCommandSet Encodings::mathcmd;
  42 Encodings::TextCommandSet Encodings::textcmd;
  43 Encodings::MathSymbolSet  Encodings::mathsym;
  44
  45 namespace {
  46
  47 typedef map<char_type, CharInfo> CharInfoMap;
  48 CharInfoMap unicodesymbols;
  49
  50 typedef set<char_type> CharSet;
  51 typedef map<string, CharSet> CharSetMap;
  52 CharSet forced;
  53 CharSetMap forcedselected;
  54
  55 typedef set<char_type> MathAlphaSet;
  56 MathAlphaSet mathalpha;
  57
  58
  59 /// The highest code point in UCS4 encoding (1<<20 + 1<<16)
  60 char_type const max_ucs4 = 0x110000;
  61
  62 } // namespace
  63
  64
  65 EncodingException::EncodingException(char_type c)
  66         : failed_char(c), par_id(0), pos(0)
  67 {
  68 }
  69
  70
  71 const char * EncodingException::what() const throw()
  72 {
  73         return "Could not find LaTeX command for a character";
  74 }
  75
  76
  77 CharInfo::CharInfo(
  78         docstring const & textcommand, docstring const & mathcommand,
  79         std::string const & textpreamble, std::string const & mathpreamble,
  80         std::string const & tipashortcut, unsigned int flags)
  81         : textcommand_(textcommand), mathcommand_(mathcommand),
  82           textpreamble_(textpreamble), mathpreamble_(mathpreamble),
  83           tipashortcut_(tipashortcut), flags_(flags)
  84 {
  85 }
  86
  87
  88 Encoding::Encoding(string const & n, string const & l, string const & g,
  89                    string const & i, bool f, bool u, Encoding::Package p)
  90         : name_(n), latexName_(l), guiName_(g), iconvName_(i), fixedwidth_(f),
  91           unsafe_(u), forced_(&forcedselected[n]), package_(p)
  92 {
  93         if (n == "ascii") {
  94                 // ASCII can encode 128 code points and nothing else
  95                 start_encodable_ = 128;
  96                 complete_ = true;
  97         } else if (i == "UTF-8") {
  98                 // UTF8 can encode all UCS4 code points
  99                 start_encodable_ = max_ucs4;
 100                 complete_ = true;
 101         } else {
 102                 start_encodable_ = 0;
 103                 complete_ = false;
 104         }
 105 }
 106
 107
 108 void Encoding::init() const
 109 {
 110         // Since the the constructor is the only method which sets complete_
 111         // to false the test for complete_ is thread-safe without mutex.
 112         if (complete_)
 113                 return;
 114
 115         static Mutex mutex;
 116         Mutex::Locker lock(&mutex);
 117
 118         // We need to test again for complete_, since another thread could
 119         // have set it to true while we were waiting for the lock and we must
 120         // not modify an encoding which is already complete.
 121         if (complete_)
 122                 return;
 123
 124         // We do not make any member mutable  so that it can be easily verified
 125         // that all const methods are thread-safe: init() is the only const
 126         // method which changes complete_, encodable_ and start_encodable_, and
 127         // it uses a mutex to ensure thread-safety.
 128         CharSet & encodable = const_cast<Encoding *>(this)->encodable_;
 129         char_type & start_encodable = const_cast<Encoding *>(this)->start_encodable_;
 130
 131         start_encodable = 0;
 132         // temporarily switch off lyxerr, since we will generate iconv errors
 133         lyxerr.disable();
 134         if (fixedwidth_) {
 135                 // We do not need to check all UCS4 code points, it is enough
 136                 // if we check all 256 code points of this encoding.
 137                 for (unsigned short j = 0; j < 256; ++j) {
 138                         char const c = char(j);
 139                         vector<char_type> const ucs4 = eightbit_to_ucs4(&c, 1, iconvName_);
 140                         if (ucs4.size() != 1)
 141                                 continue;
 142                         char_type const uc = ucs4[0];
 143                         CharInfoMap::const_iterator const it = unicodesymbols.find(uc);
 144                         if (it == unicodesymbols.end())
 145                                 encodable.insert(uc);
 146                         else if (!it->second.force()) {
 147                                 if (forced_->empty() || forced_->find(uc) == forced_->end())
 148                                         encodable.insert(uc);
 149                         }
 150                 }
 151         } else {
 152                 // We do not know how many code points this encoding has, and
 153                 // they do not have a direct representation as a single byte,
 154                 // therefore we need to check all UCS4 code points.
 155                 // This is expensive!
 156                 for (char_type c = 0; c < max_ucs4; ++c) {
 157                         vector<char> const eightbit = ucs4_to_eightbit(&c, 1, iconvName_);
 158                         if (!eightbit.empty()) {
 159                                 CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 160                                 if (it == unicodesymbols.end())
 161                                         encodable.insert(c);
 162                                 else if (!it->second.force()) {
 163                                         if (forced_->empty() || forced_->find(c) == forced_->end())
 164                                                 encodable.insert(c);
 165                                 }
 166                         }
 167                 }
 168         }
 169         lyxerr.enable();
 170         CharSet::iterator it = encodable.find(start_encodable);
 171         while (it != encodable.end()) {
 172                 encodable.erase(it);
 173                 ++start_encodable;
 174                 it = encodable.find(start_encodable);
 175         }
 176         const_cast<Encoding *>(this)->complete_ = true;
 177 }
 178
 179
 180 bool Encoding::isForced(char_type c) const
 181 {
 182         if (!forced.empty() && forced.find(c) != forced.end())
 183                 return true;
 184         return !forced_->empty() && forced_->find(c) != forced_->end();
 185 }
 186
 187
 188 bool Encoding::encodable(char_type c) const
 189 {
 190         // assure the used encoding is properly initialized
 191         init();
 192
 193         if (iconvName_ == "UTF-8" && package_ == none)
 194                 return true;
 195         if (c < start_encodable_ && !isForced(c))
 196                 return true;
 197         if (encodable_.find(c) != encodable_.end())
 198                 return true;
 199         return false;
 200 }
 201
 202
 203 pair<docstring, bool> Encoding::latexChar(char_type c) const
 204 {
 205         if (encodable(c))
 206                 return make_pair(docstring(1, c), false);
 207
 208         // c cannot (or should not) be encoded in this encoding
 209         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 210         if (it == unicodesymbols.end())
 211                 throw EncodingException(c);
 212         // at least one of mathcommand and textcommand is nonempty
 213         if (it->second.textcommand().empty())
 214                 return make_pair(
 215                         "\\ensuremath{" + it->second.mathcommand() + '}', false);
 216         return make_pair(it->second.textcommand(), !it->second.textnotermination());
 217 }
 218
 219
 220 pair<docstring, docstring> Encoding::latexString(docstring const & input, bool dryrun) const
 221 {
 222         docstring result;
 223         docstring uncodable;
 224         bool terminate = false;
 225         for (size_t n = 0; n < input.size(); ++n) {
 226                 try {
 227                         char_type const c = input[n];
 228                         pair<docstring, bool> latex_char = latexChar(c);
 229                         docstring const latex = latex_char.first;
 230                         if (terminate && !prefixIs(latex, '\\')
 231                             && !prefixIs(latex, '{')
 232                             && !prefixIs(latex, '}')) {
 233                                         // Prevent eating of a following
 234                                         // space or command corruption by
 235                                         // following characters
 236                                         if (latex == " ")
 237                                                 result += "{}";
 238                                         else
 239                                                 result += " ";
 240                                 }
 241                         result += latex;
 242                         terminate = latex_char.second;
 243                 } catch (EncodingException & /* e */) {
 244                         LYXERR0("Uncodable character in latexString!");
 245                         if (dryrun) {
 246                                 result += "<" + _("LyX Warning: ")
 247                                            + _("uncodable character") + " '";
 248                                 result += docstring(1, input[n]);
 249                                 result += "'>";
 250                         } else
 251                                 uncodable += input[n];
 252                 }
 253         }
 254         return make_pair(result, uncodable);
 255 }
 256
 257
 258 vector<char_type> Encoding::symbolsList() const
 259 {
 260         // assure the used encoding is properly initialized
 261         init();
 262
 263         // first all those below start_encodable_
 264         vector<char_type> symbols;
 265         for (char_type c = 0; c < start_encodable_; ++c)
 266                 symbols.push_back(c);
 267         // add all encodable characters
 268         copy(encodable_.begin(), encodable_.end(), back_inserter(symbols));
 269         // now the ones from the unicodesymbols file that are not already there
 270         for (pair<const char_type, CharInfo> const & elem : unicodesymbols) {
 271                 if (find(symbols.begin(), symbols.end(), elem.first) == symbols.end())
 272                         symbols.push_back(elem.first);
 273         }
 274         // finally, sort the vector
 275         sort(symbols.begin(), symbols.end());
 276         return symbols;
 277 }
 278
 279
 280 bool Encodings::latexMathChar(char_type c, bool mathmode,
 281                         Encoding const * encoding, docstring & command,
 282                         bool & needsTermination)
 283 {
 284         command = empty_docstring();
 285         if (encoding)
 286                 if (encoding->encodable(c))
 287                         command = docstring(1, c);
 288         needsTermination = false;
 289
 290         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 291         if (it == unicodesymbols.end()) {
 292                 if (!encoding || command.empty())
 293                         throw EncodingException(c);
 294                 if (mathmode)
 295                         addMathSym(c);
 296                 return false;
 297         }
 298         // at least one of mathcommand and textcommand is nonempty
 299         bool use_math = (mathmode && !it->second.mathcommand().empty()) ||
 300                         (!mathmode && it->second.textcommand().empty());
 301         if (use_math) {
 302                 command = it->second.mathcommand();
 303                 needsTermination = !it->second.mathnotermination();
 304                 addMathCmd(c);
 305         } else {
 306                 if (!encoding || command.empty()) {
 307                         command = it->second.textcommand();
 308                         needsTermination = !it->second.textnotermination();
 309                 }
 310                 if (mathmode)
 311                         addMathSym(c);
 312                 else
 313                         addTextCmd(c);
 314         }
 315         return use_math;
 316 }
 317
 318
 319 char_type Encodings::fromLaTeXCommand(docstring const & cmd, int cmdtype,
 320                 bool & combining, bool & needsTermination, set<string> * req)
 321 {
 322         CharInfoMap::const_iterator const end = unicodesymbols.end();
 323         CharInfoMap::const_iterator it = unicodesymbols.begin();
 324         for (combining = false; it != end; ++it) {
 325                 if (it->second.deprecated())
 326                         continue;
 327                 docstring const math = it->second.mathcommand();
 328                 docstring const text = it->second.textcommand();
 329                 if ((cmdtype & MATH_CMD) && math == cmd) {
 330                         combining = it->second.combining();
 331                         needsTermination = !it->second.mathnotermination();
 332                         if (req && it->second.mathfeature() &&
 333                             !it->second.mathpreamble().empty())
 334                                 req->insert(it->second.mathpreamble());
 335                         return it->first;
 336                 }
 337                 if ((cmdtype & TEXT_CMD) && text == cmd) {
 338                         combining = it->second.combining();
 339                         needsTermination = !it->second.textnotermination();
 340                         if (req && it->second.textfeature() &&
 341                             !it->second.textpreamble().empty())
 342                                 req->insert(it->second.textpreamble());
 343                         return it->first;
 344                 }
 345         }
 346         needsTermination = false;
 347         return 0;
 348 }
 349
 350
 351 docstring Encodings::fromLaTeXCommand(docstring const & cmd, int cmdtype,
 352                 bool & needsTermination, docstring & rem, set<string> * req)
 353 {
 354         needsTermination = false;
 355         rem = empty_docstring();
 356         bool const mathmode = cmdtype & MATH_CMD;
 357         bool const textmode = cmdtype & TEXT_CMD;
 358         docstring symbols;
 359         size_t const cmdend = cmd.size();
 360         size_t prefix = 0;
 361         CharInfoMap::const_iterator const uniend = unicodesymbols.end();
 362         for (size_t i = 0, j = 0; j < cmdend; ++j) {
 363                 // Also get the char after a backslash
 364                 if (j + 1 < cmdend && cmd[j] == '\\') {
 365                         ++j;
 366                         prefix = 1;
 367                         // Detect things like \=*{e} as well
 368                         if (j + 3 < cmdend && cmd[j+1] == '*' &&
 369                             cmd[j+2] == '{') {
 370                                 ++j;
 371                                 prefix = 2;
 372                         }
 373                 }
 374                 // position of the last character before a possible macro
 375                 // argument
 376                 size_t m = j;
 377                 // If a macro argument follows, get it, too
 378                 // Do it here only for single character commands. Other
 379                 // combining commands need this too, but they are handled in
 380                 // the loop below for performance reasons.
 381                 if (j + 1 < cmdend && cmd[j + 1] == '{') {
 382                         size_t k = j + 1;
 383                         int count = 1;
 384                         while (k < cmdend && count) {
 385                                 k = cmd.find_first_of(from_ascii("{}"), k + 1);
 386                                 // braces may not be balanced
 387                                 if (k == docstring::npos)
 388                                         break;
 389                                 if (cmd[k] == '{')
 390                                         ++count;
 391                                 else
 392                                         --count;
 393                         }
 394                         if (k != docstring::npos)
 395                                 j = k;
 396                 } else if (m + 1 < cmdend && isAlphaASCII(cmd[m])) {
 397                         while (m + 2 < cmdend && isAlphaASCII(cmd[m+1]))
 398                                 m++;
 399                 }
 400                 // Start with this substring and try augmenting it when it is
 401                 // the prefix of some command in the unicodesymbols file
 402                 docstring subcmd = cmd.substr(i, j - i + 1);
 403
 404                 CharInfoMap::const_iterator it = unicodesymbols.begin();
 405                 // First part of subcmd which might be a combining character
 406                 docstring combcmd = (m == j) ? docstring() : cmd.substr(i, m - i + 1);
 407                 // The combining character of combcmd if it exists
 408                 CharInfoMap::const_iterator combining = uniend;
 409                 size_t unicmd_size = 0;
 410                 char_type c = 0;
 411                 for (; it != uniend; ++it) {
 412                         if (it->second.deprecated())
 413                                 continue;
 414                         docstring const math = mathmode ? it->second.mathcommand()
 415                                                         : docstring();
 416                         docstring const text = textmode ? it->second.textcommand()
 417                                                         : docstring();
 418                         if (!combcmd.empty() && it->second.combining() &&
 419                             (math == combcmd || text == combcmd))
 420                                 combining = it;
 421                         size_t cur_size = max(math.size(), text.size());
 422                         // The current math or text unicode command cannot
 423                         // match, or we already matched a longer one
 424                         if (cur_size < subcmd.size() || cur_size <= unicmd_size)
 425                                 continue;
 426
 427                         docstring tmp = subcmd;
 428                         size_t k = j;
 429                         while (prefixIs(math, tmp) || prefixIs(text, tmp)) {
 430                                 ++k;
 431                                 if (k >= cmdend || cur_size <= tmp.size())
 432                                         break;
 433                                 tmp += cmd[k];
 434                         }
 435                         // No match
 436                         if (k == j)
 437                                 continue;
 438
 439                         // The last added char caused a mismatch, because
 440                         // we didn't exhaust the chars in cmd and didn't
 441                         // exceed the maximum size of the current unicmd
 442                         if (k < cmdend && cur_size > tmp.size())
 443                                 tmp.resize(tmp.size() - 1);
 444
 445                         // If this is an exact match, we found a (longer)
 446                         // matching entry in the unicodesymbols file.
 447                         if (math != tmp && text != tmp)
 448                                 continue;
 449                         // If we found a combining command, we need to append
 450                         // the macro argument if this has not been done above.
 451                         if (tmp == combcmd && combining != uniend &&
 452                             k < cmdend && cmd[k] == '{') {
 453                                 size_t l = k;
 454                                 int count = 1;
 455                                 while (l < cmdend && count) {
 456                                         l = cmd.find_first_of(from_ascii("{}"), l + 1);
 457                                         // braces may not be balanced
 458                                         if (l == docstring::npos)
 459                                                 break;
 460                                         if (cmd[l] == '{')
 461                                                 ++count;
 462                                         else
 463                                                 --count;
 464                                 }
 465                                 if (l != docstring::npos) {
 466                                         j = l;
 467                                         subcmd = cmd.substr(i, j - i + 1);
 468                                 }
 469                         }
 470                         // If the entry doesn't start with '\', we take note
 471                         // of the match and continue (this is not a ultimate
 472                         // acceptance, as some other entry may match a longer
 473                         // portion of the cmd string). However, if the entry
 474                         // does start with '\', we accept the match only if
 475                         // this is a valid macro, i.e., either it is a single
 476                         // (nonletter) char macro, or nothing else follows,
 477                         // or what follows is a nonletter char, or the last
 478                         // character is a }.
 479                         else if (tmp[0] != '\\'
 480                                    || (tmp.size() == prefix + 1 &&
 481                                        !isAlphaASCII(tmp[1]) &&
 482                                        (prefix == 1 || !isAlphaASCII(tmp[2])))
 483                                    || k == cmdend
 484                                    || !isAlphaASCII(cmd[k])
 485                                    || tmp[tmp.size() - 1] == '}'
 486                                  ) {
 487                                 c = it->first;
 488                                 j = k - 1;
 489                                 i = j + 1;
 490                                 unicmd_size = cur_size;
 491                                 if (math == tmp)
 492                                         needsTermination = !it->second.mathnotermination();
 493                                 else
 494                                         needsTermination = !it->second.textnotermination();
 495                                 if (req) {
 496                                         if (math == tmp && it->second.mathfeature() &&
 497                                             !it->second.mathpreamble().empty())
 498                                                 req->insert(it->second.mathpreamble());
 499                                         if (text == tmp && it->second.textfeature() &&
 500                                             !it->second.textpreamble().empty())
 501                                                 req->insert(it->second.textpreamble());
 502                                 }
 503                         }
 504                 }
 505                 if (unicmd_size)
 506                         symbols += c;
 507                 else if (combining != uniend &&
 508                          prefixIs(subcmd, combcmd + '{')) {
 509                         // We know that subcmd starts with combcmd and
 510                         // contains an argument in braces.
 511                         docstring const arg = subcmd.substr(
 512                                 combcmd.length() + 1,
 513                                 subcmd.length() - combcmd.length() - 2);
 514                         // If arg is a single character we can construct a
 515                         // combining sequence.
 516                         char_type a;
 517                         bool argcomb = false;
 518                         if (arg.size() == 1 && isAlnumASCII(arg[0]))
 519                                 a = arg[0];
 520                         else {
 521                                 // Use the version of fromLaTeXCommand() that
 522                                 // parses only one command, since we cannot
 523                                 // use more than one character.
 524                                 bool dummy = false;
 525                                 set<string> r;
 526                                 a = fromLaTeXCommand(arg, cmdtype, argcomb,
 527                                                      dummy, &r);
 528                                 if (a && req && !argcomb)
 529                                         req->insert(r.begin(), r.end());
 530                         }
 531                         if (a && !argcomb) {
 532                                 // In unicode the combining character comes
 533                                 // after its base
 534                                 symbols += a;
 535                                 symbols += combining->first;
 536                                 i = j + 1;
 537                                 unicmd_size = 2;
 538                         }
 539                 }
 540                 if (j + 1 == cmdend && !unicmd_size) {
 541                         // No luck. Return what remains
 542                         rem = cmd.substr(i);
 543                         if (needsTermination && !rem.empty()) {
 544                                 if (rem.substr(0, 2) == "{}") {
 545                                         rem = rem.substr(2);
 546                                         needsTermination = false;
 547                                 } else if (rem[0] == ' ') {
 548                                         needsTermination = false;
 549                                         // LaTeX would swallow all spaces
 550                                         rem = ltrim(rem);
 551                                 }
 552                         }
 553                 }
 554         }
 555         return symbols;
 556 }
 557
 558
 559 CharInfo const & Encodings::unicodeCharInfo(char_type c)
 560 {
 561         static CharInfo empty;
 562         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 563         return it != unicodesymbols.end() ? it->second : empty;
 564 }
 565
 566
 567 bool Encodings::isCombiningChar(char_type c)
 568 {
 569         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 570         if (it != unicodesymbols.end())
 571                 return it->second.combining();
 572         return false;
 573 }
 574
 575
 576 string const Encodings::TIPAShortcut(char_type c)
 577 {
 578         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 579         if (it != unicodesymbols.end())
 580                 return it->second.tipashortcut();
 581         return string();
 582 }
 583
 584
 585 bool Encodings::isKnownScriptChar(char_type const c, string & preamble)
 586 {
 587         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 588
 589         if (it == unicodesymbols.end())
 590                 return false;
 591
 592         if (it->second.textpreamble() != "textgreek" && it->second.textpreamble() != "textcyr")
 593                 return false;
 594
 595         if (preamble.empty()) {
 596                 preamble = it->second.textpreamble();
 597                 return true;
 598         }
 599         return it->second.textpreamble() == preamble;
 600 }
 601
 602
 603 bool Encodings::isMathAlpha(char_type c)
 604 {
 605         return mathalpha.count(c);
 606 }
 607
 608
 609 bool Encodings::isUnicodeTextOnly(char_type c)
 610 {
 611         if (isASCII(c) || isMathAlpha(c))
 612                 return false;
 613
 614         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 615         return it == unicodesymbols.end() || it->second.mathcommand().empty();
 616 }
 617
 618
 619 Encoding const *
 620 Encodings::fromLyXName(string const & name, bool allowUnsafe) const
 621 {
 622         EncodingList::const_iterator const it = encodinglist.find(name);
 623         if (it == encodinglist.end())
 624                 return 0;
 625         if (!allowUnsafe && it->second.unsafe())
 626                 return 0;
 627         return &it->second;
 628 }
 629
 630
 631 Encoding const *
 632 Encodings::fromLaTeXName(string const & n, int const & p, bool allowUnsafe) const
 633 {
 634         string name = n;
 635         // FIXME: if we have to test for too many of these synonyms,
 636         // we should instead extend the format of lib/encodings
 637         if (n == "ansinew")
 638                 name = "cp1252";
 639
 640         // We don't use find_if because it makes copies of the pairs in
 641         // the map.
 642         // This linear search is OK since we don't have many encodings.
 643         // Users could even optimize it by putting the encodings they use
 644         // most at the top of lib/encodings.
 645         EncodingList::const_iterator const end = encodinglist.end();
 646         for (EncodingList::const_iterator it = encodinglist.begin(); it != end; ++it)
 647                 if ((it->second.latexName() == name) && (it->second.package() & p)
 648                                 && (!it->second.unsafe() || allowUnsafe))
 649                         return &it->second;
 650         return 0;
 651 }
 652
 653
 654 Encoding const *
 655 Encodings::fromIconvName(string const & n, int const & p, bool allowUnsafe) const
 656 {
 657         EncodingList::const_iterator const end = encodinglist.end();
 658         for (EncodingList::const_iterator it = encodinglist.begin(); it != end; ++it)
 659                 if ((it->second.iconvName() == n) && (it->second.package() & p)
 660                                 && (!it->second.unsafe() || allowUnsafe))
 661                         return &it->second;
 662         return 0;
 663 }
 664
 665
 666 Encodings::Encodings()
 667 {}
 668
 669
 670 void Encodings::read(FileName const & encfile, FileName const & symbolsfile)
 671 {
 672         // We must read the symbolsfile first, because the Encoding
 673         // constructor depends on it.
 674         CharSetMap forcednotselected;
 675         Lexer symbolslex;
 676         symbolslex.setFile(symbolsfile);
 677         bool getNextToken = true;
 678         while (symbolslex.isOK()) {
 679                 char_type symbol;
 680
 681                 if (getNextToken) {
 682                         if (!symbolslex.next(true))
 683                                 break;
 684                 } else
 685                         getNextToken = true;
 686
 687                 istringstream is(symbolslex.getString());
 688                 // reading symbol directly does not work if
 689                 // char_type == wchar_t.
 690                 boost::uint32_t tmp;
 691                 if(!(is >> hex >> tmp))
 692                         break;
 693                 symbol = tmp;
 694
 695                 if (!symbolslex.next(true))
 696                         break;
 697                 docstring textcommand = symbolslex.getDocString();
 698                 if (!symbolslex.next(true))
 699                         break;
 700                 string textpreamble = symbolslex.getString();
 701                 if (!symbolslex.next(true))
 702                         break;
 703                 string sflags = symbolslex.getString();
 704
 705                 string tipashortcut;
 706                 int flags = 0;
 707
 708                 if (suffixIs(textcommand, '}'))
 709                         flags |= CharInfoTextNoTermination;
 710                 while (!sflags.empty()) {
 711                         string flag;
 712                         sflags = split(sflags, flag, ',');
 713                         if (flag == "combining") {
 714                                 flags |= CharInfoCombining;
 715                         } else if (flag == "force") {
 716                                 flags |= CharInfoForce;
 717                                 forced.insert(symbol);
 718                         } else if (prefixIs(flag, "force=")) {
 719                                 vector<string> encodings =
 720                                         getVectorFromString(flag.substr(6), ";");
 721                                 for (size_t i = 0; i < encodings.size(); ++i)
 722                                         forcedselected[encodings[i]].insert(symbol);
 723                                 flags |= CharInfoForceSelected;
 724                         } else if (prefixIs(flag, "force!=")) {
 725                                 vector<string> encodings =
 726                                         getVectorFromString(flag.substr(7), ";");
 727                                 for (size_t i = 0; i < encodings.size(); ++i)
 728                                         forcednotselected[encodings[i]].insert(symbol);
 729                                 flags |= CharInfoForceSelected;
 730                         } else if (flag == "mathalpha") {
 731                                 mathalpha.insert(symbol);
 732                         } else if (flag == "notermination=text") {
 733                                 flags |= CharInfoTextNoTermination;
 734                         } else if (flag == "notermination=math") {
 735                                 flags |= CharInfoMathNoTermination;
 736                         } else if (flag == "notermination=both") {
 737                                 flags |= CharInfoTextNoTermination;
 738                                 flags |= CharInfoMathNoTermination;
 739                         } else if (flag == "notermination=none") {
 740                                 flags &= ~CharInfoTextNoTermination;
 741                                 flags &= ~CharInfoMathNoTermination;
 742                         } else if (contains(flag, "tipashortcut=")) {
 743                                 tipashortcut = split(flag, '=');
 744                         } else if (flag == "deprecated") {
 745                                 flags |= CharInfoDeprecated;
 746                         } else {
 747                                 lyxerr << "Ignoring unknown flag `" << flag
 748                                        << "' for symbol `0x"
 749                                        << hex << symbol << dec
 750                                        << "'." << endl;
 751                         }
 752                 }
 753                 // mathcommand and mathpreamble have been added for 1.6.0.
 754                 // make them optional so that old files still work.
 755                 int const lineno = symbolslex.lineNumber();
 756                 bool breakout = false;
 757                 docstring mathcommand;
 758                 string mathpreamble;
 759                 if (symbolslex.next(true)) {
 760                         if (symbolslex.lineNumber() != lineno) {
 761                                 // line in old format without mathcommand and mathpreamble
 762                                 getNextToken = false;
 763                         } else {
 764                                 mathcommand = symbolslex.getDocString();
 765                                 if (suffixIs(mathcommand, '}'))
 766                                         flags |= CharInfoMathNoTermination;
 767                                 if (symbolslex.next(true)) {
 768                                         if (symbolslex.lineNumber() != lineno) {
 769                                                 // line in new format with mathcommand only
 770                                                 getNextToken = false;
 771                                         } else {
 772                                                 // line in new format with mathcommand and mathpreamble
 773                                                 mathpreamble = symbolslex.getString();
 774                                         }
 775                                 } else
 776                                         breakout = true;
 777                         }
 778                 } else {
 779                         breakout = true;
 780                 }
 781
 782                 // backward compatibility
 783                 if (mathpreamble == "esintoramsmath")
 784                         mathpreamble = "esint|amsmath";
 785
 786                 if (!textpreamble.empty())
 787                         if (textpreamble[0] != '\\')
 788                                 flags |= CharInfoTextFeature;
 789                 if (!mathpreamble.empty())
 790                         if (mathpreamble[0] != '\\')
 791                                 flags |= CharInfoMathFeature;
 792
 793                 CharInfo info = CharInfo(
 794                         textcommand, mathcommand,
 795                         textpreamble, mathpreamble,
 796                         tipashortcut, flags);
 797                 LYXERR(Debug::INFO, "Read unicode symbol " << symbol << " '"
 798                            << to_utf8(info.textcommand()) << "' '" << info.textpreamble()
 799                            << " '" << info.textfeature() << ' ' << info.textnotermination()
 800                            << ' ' << to_utf8(info.mathcommand()) << "' '" << info.mathpreamble()
 801                            << "' " << info.mathfeature() << ' ' << info.mathnotermination()
 802                            << ' ' << info.combining() << ' ' << info.force()
 803                            << ' ' << info.forceselected());
 804
 805                 // we assume that at least one command is nonempty when using unicodesymbols
 806                 if (info.isUnicodeSymbol()) {
 807                         unicodesymbols[symbol] = info;
 808                 }
 809
 810                 if (breakout)
 811                         break;
 812         }
 813
 814         // Now read the encodings
 815         enum {
 816                 et_encoding = 1,
 817                 et_end
 818         };
 819
 820         LexerKeyword encodingtags[] = {
 821                 { "encoding", et_encoding },
 822                 { "end", et_end }
 823         };
 824
 825         Lexer lex(encodingtags);
 826         lex.setFile(encfile);
 827         lex.setContext("Encodings::read");
 828         while (lex.isOK()) {
 829                 switch (lex.lex()) {
 830                 case et_encoding:
 831                 {
 832                         lex.next();
 833                         string const name = lex.getString();
 834                         lex.next();
 835                         string const latexname = lex.getString();
 836                         lex.next();
 837                         string const guiname = lex.getString();
 838                         lex.next();
 839                         string const iconvname = lex.getString();
 840                         lex.next();
 841                         string const width = lex.getString();
 842                         bool fixedwidth = false;
 843                         bool unsafe = false;
 844                         if (width == "fixed")
 845                                 fixedwidth = true;
 846                         else if (width == "variable")
 847                                 fixedwidth = false;
 848                         else if (width == "variableunsafe") {
 849                                 fixedwidth = false;
 850                                 unsafe = true;
 851                         }
 852                         else
 853                                 lex.printError("Unknown width");
 854
 855                         lex.next();
 856                         string const p = lex.getString();
 857                         Encoding::Package package = Encoding::none;
 858                         if (p == "none")
 859                                 package = Encoding::none;
 860                         else if (p == "inputenc")
 861                                 package = Encoding::inputenc;
 862                         else if (p == "CJK")
 863                                 package = Encoding::CJK;
 864                         else if (p == "japanese")
 865                                 package = Encoding::japanese;
 866                         else
 867                                 lex.printError("Unknown package");
 868
 869                         LYXERR(Debug::INFO, "Reading encoding " << name);
 870                         encodinglist[name] = Encoding(name, latexname,
 871                                 guiname, iconvname, fixedwidth, unsafe,
 872                                 package);
 873
 874                         if (lex.lex() != et_end)
 875                                 lex.printError("Missing end");
 876                         break;
 877                 }
 878                 case et_end:
 879                         lex.printError("Misplaced end");
 880                         break;
 881                 case Lexer::LEX_FEOF:
 882                         break;
 883                 default:
 884                         lex.printError("Unknown tag");
 885                         break;
 886                 }
 887         }
 888
 889         // Move all information from forcednotselected to forcedselected
 890         for (CharSetMap::const_iterator it1 = forcednotselected.begin(); it1 != forcednotselected.end(); ++it1) {
 891                 for (CharSetMap::iterator it2 = forcedselected.begin(); it2 != forcedselected.end(); ++it2) {
 892                         if (it2->first != it1->first)
 893                                 it2->second.insert(it1->second.begin(), it1->second.end());
 894                 }
 895         }
 896
 897 }
 898
 899
 900 } // namespace lyx