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 <algorithm>
  27 #include <cstdint>
  28 #include <iterator>
  29 #include <sstream>
  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
  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         if (iconvName_ == "UTF-8" && package_ == none)
 192                 return true;
 193         // platex does not load inputenc: force conversion of supported characters
 194         if (package_ == Encoding::japanese
 195             && ((0xb7 <= c && c <= 0x05ff) // Latin-1 Supplement ... Hebrew
 196                         || (0x1d00 <= c && c <= 0x218f) // Phonetic Extensions ... Number Forms
 197                         || (0x2193 <= c && c <= 0x2aff) // Arrows ... Supplemental Mathematical Operators
 198                         || (0xfb00 <= c && c <= 0xfb4f) // Alphabetic Presentation Forms
 199                         || (0x1d400 <= c && c <= 0x1d7ff))) // Mathematical Alphanumeric Symbols
 200                 return false;
 201         if (c < start_encodable_ && !isForced(c))
 202                 return true;
 203         if (encodable_.find(c) != encodable_.end())
 204                 return true;
 205         return false;
 206 }
 207
 208
 209 pair<docstring, bool> Encoding::latexChar(char_type c) const
 210 {
 211         if (encodable(c))
 212                 return make_pair(docstring(1, c), false);
 213
 214         // c cannot (or should not) be encoded in this encoding
 215         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 216         if (it == unicodesymbols.end())
 217                 throw EncodingException(c);
 218         // at least one of mathcommand and textcommand is nonempty
 219         if (it->second.textcommand().empty())
 220                 return make_pair(
 221                         "\\ensuremath{" + it->second.mathcommand() + '}', false);
 222         return make_pair(it->second.textcommand(), !it->second.textnotermination());
 223 }
 224
 225
 226 pair<docstring, docstring> Encoding::latexString(docstring const & input, bool dryrun) const
 227 {
 228         docstring result;
 229         docstring uncodable;
 230         bool terminate = false;
 231         for (size_t n = 0; n < input.size(); ++n) {
 232                 char_type const c = input[n];
 233                 try {
 234                         pair<docstring, bool> latex_char = latexChar(c);
 235                         docstring const latex = latex_char.first;
 236                         if (terminate && !prefixIs(latex, '\\')
 237                             && !prefixIs(latex, '{')
 238                             && !prefixIs(latex, '}')) {
 239                                         // Prevent eating of a following
 240                                         // space or command corruption by
 241                                         // following characters
 242                                         if (latex == " ")
 243                                                 result += "{}";
 244                                         else
 245                                                 result += " ";
 246                                 }
 247                         result += latex;
 248                         terminate = latex_char.second;
 249                 } catch (EncodingException & /* e */) {
 250                         LYXERR0("Uncodable character <" << docstring(1, c)
 251                                         << "> in latexString!");
 252                         if (dryrun) {
 253                                 result += "<" + _("LyX Warning: ")
 254                                            + _("uncodable character") + " '";
 255                                 result += docstring(1, input[n]);
 256                                 result += "'>";
 257                         } else
 258                                 uncodable += input[n];
 259                 }
 260         }
 261         return make_pair(result, uncodable);
 262 }
 263
 264
 265 vector<char_type> Encoding::symbolsList() const
 266 {
 267         // assure the used encoding is properly initialized
 268         init();
 269
 270         // first all those below start_encodable_
 271         vector<char_type> symbols;
 272         for (char_type c = 0; c < start_encodable_; ++c)
 273                 symbols.push_back(c);
 274         // add all encodable characters
 275         copy(encodable_.begin(), encodable_.end(), back_inserter(symbols));
 276         // now the ones from the unicodesymbols file that are not already there
 277         for (pair<char_type, CharInfo> const & elem : unicodesymbols) {
 278                 if (find(symbols.begin(), symbols.end(), elem.first) == symbols.end())
 279                         symbols.push_back(elem.first);
 280         }
 281         // finally, sort the vector
 282         sort(symbols.begin(), symbols.end());
 283         return symbols;
 284 }
 285
 286
 287 bool Encodings::latexMathChar(char_type c, bool mathmode,
 288                         Encoding const * encoding, docstring & command,
 289                         bool & needsTermination)
 290 {
 291         command = empty_docstring();
 292         if (encoding)
 293                 if (encoding->encodable(c))
 294                         command = docstring(1, c);
 295         needsTermination = false;
 296
 297         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 298         if (it == unicodesymbols.end()) {
 299                 if (!encoding || command.empty())
 300                         throw EncodingException(c);
 301                 if (mathmode)
 302                         addMathSym(c);
 303                 return false;
 304         }
 305         // at least one of mathcommand and textcommand is nonempty
 306         bool use_math = (mathmode && !it->second.mathcommand().empty()) ||
 307                         (!mathmode && it->second.textcommand().empty());
 308         if (use_math) {
 309                 command = it->second.mathcommand();
 310                 needsTermination = !it->second.mathnotermination();
 311                 addMathCmd(c);
 312         } else {
 313                 if (!encoding || command.empty()) {
 314                         command = it->second.textcommand();
 315                         needsTermination = !it->second.textnotermination();
 316                 }
 317                 if (mathmode)
 318                         addMathSym(c);
 319                 else
 320                         addTextCmd(c);
 321         }
 322         return use_math;
 323 }
 324
 325
 326 char_type Encodings::fromLaTeXCommand(docstring const & cmd, int cmdtype,
 327                 bool & combining, bool & needsTermination, set<string> * req)
 328 {
 329         CharInfoMap::const_iterator const end = unicodesymbols.end();
 330         CharInfoMap::const_iterator it = unicodesymbols.begin();
 331         for (combining = false; it != end; ++it) {
 332                 if (it->second.deprecated())
 333                         continue;
 334                 docstring const math = it->second.mathcommand();
 335                 docstring const text = it->second.textcommand();
 336                 if ((cmdtype & MATH_CMD) && math == cmd) {
 337                         combining = it->second.combining();
 338                         needsTermination = !it->second.mathnotermination();
 339                         if (req && it->second.mathfeature() &&
 340                             !it->second.mathpreamble().empty())
 341                                 req->insert(it->second.mathpreamble());
 342                         return it->first;
 343                 }
 344                 if ((cmdtype & TEXT_CMD) && text == cmd) {
 345                         combining = it->second.combining();
 346                         needsTermination = !it->second.textnotermination();
 347                         if (req && it->second.textfeature() &&
 348                             !it->second.textpreamble().empty())
 349                                 req->insert(it->second.textpreamble());
 350                         return it->first;
 351                 }
 352         }
 353         needsTermination = false;
 354         return 0;
 355 }
 356
 357
 358 docstring Encodings::fromLaTeXCommand(docstring const & cmd, int cmdtype,
 359                 bool & needsTermination, docstring & rem, set<string> * req)
 360 {
 361         needsTermination = false;
 362         rem = empty_docstring();
 363         bool const mathmode = cmdtype & MATH_CMD;
 364         bool const textmode = cmdtype & TEXT_CMD;
 365         docstring symbols;
 366         size_t const cmdend = cmd.size();
 367         size_t prefix = 0;
 368         CharInfoMap::const_iterator const uniend = unicodesymbols.end();
 369         for (size_t i = 0, j = 0; j < cmdend; ++j) {
 370                 // Also get the char after a backslash
 371                 if (j + 1 < cmdend && cmd[j] == '\\') {
 372                         ++j;
 373                         prefix = 1;
 374                         // Detect things like \=*{e} as well
 375                         if (j + 3 < cmdend && cmd[j+1] == '*' &&
 376                             cmd[j+2] == '{') {
 377                                 ++j;
 378                                 prefix = 2;
 379                         }
 380                 }
 381                 // position of the last character before a possible macro
 382                 // argument
 383                 size_t m = j;
 384                 // If a macro argument follows, get it, too
 385                 // Do it here only for single character commands. Other
 386                 // combining commands need this too, but they are handled in
 387                 // the loop below for performance reasons.
 388                 if (j + 1 < cmdend && cmd[j + 1] == '{') {
 389                         size_t k = j + 1;
 390                         int count = 1;
 391                         while (k < cmdend && count) {
 392                                 k = cmd.find_first_of(from_ascii("{}"), k + 1);
 393                                 // braces may not be balanced
 394                                 if (k == docstring::npos)
 395                                         break;
 396                                 if (cmd[k] == '{')
 397                                         ++count;
 398                                 else
 399                                         --count;
 400                         }
 401                         if (k != docstring::npos)
 402                                 j = k;
 403                 } else if (m + 1 < cmdend && isAlphaASCII(cmd[m])) {
 404                         while (m + 2 < cmdend && isAlphaASCII(cmd[m+1]))
 405                                 m++;
 406                 }
 407                 // Start with this substring and try augmenting it when it is
 408                 // the prefix of some command in the unicodesymbols file
 409                 docstring subcmd = cmd.substr(i, j - i + 1);
 410
 411                 CharInfoMap::const_iterator it = unicodesymbols.begin();
 412                 // First part of subcmd which might be a combining character
 413                 docstring combcmd = (m == j) ? docstring() : cmd.substr(i, m - i + 1);
 414                 // The combining character of combcmd if it exists
 415                 CharInfoMap::const_iterator combining = uniend;
 416                 size_t unicmd_size = 0;
 417                 char_type c = 0;
 418                 for (; it != uniend; ++it) {
 419                         if (it->second.deprecated())
 420                                 continue;
 421                         docstring const math = mathmode ? it->second.mathcommand()
 422                                                         : docstring();
 423                         docstring const text = textmode ? it->second.textcommand()
 424                                                         : docstring();
 425                         if (!combcmd.empty() && it->second.combining() &&
 426                             (math == combcmd || text == combcmd))
 427                                 combining = it;
 428                         size_t cur_size = max(math.size(), text.size());
 429                         // The current math or text unicode command cannot
 430                         // match, or we already matched a longer one
 431                         if (cur_size < subcmd.size() || cur_size <= unicmd_size)
 432                                 continue;
 433
 434                         docstring tmp = subcmd;
 435                         size_t k = j;
 436                         while (prefixIs(math, tmp) || prefixIs(text, tmp)) {
 437                                 ++k;
 438                                 if (k >= cmdend || cur_size <= tmp.size())
 439                                         break;
 440                                 tmp += cmd[k];
 441                         }
 442                         // No match
 443                         if (k == j)
 444                                 continue;
 445
 446                         // The last added char caused a mismatch, because
 447                         // we didn't exhaust the chars in cmd and didn't
 448                         // exceed the maximum size of the current unicmd
 449                         if (k < cmdend && cur_size > tmp.size())
 450                                 tmp.resize(tmp.size() - 1);
 451
 452                         // If this is an exact match, we found a (longer)
 453                         // matching entry in the unicodesymbols file.
 454                         if (math != tmp && text != tmp)
 455                                 continue;
 456                         // If we found a combining command, we need to append
 457                         // the macro argument if this has not been done above.
 458                         if (tmp == combcmd && combining != uniend &&
 459                             k < cmdend && cmd[k] == '{') {
 460                                 size_t l = k;
 461                                 int count = 1;
 462                                 while (l < cmdend && count) {
 463                                         l = cmd.find_first_of(from_ascii("{}"), l + 1);
 464                                         // braces may not be balanced
 465                                         if (l == docstring::npos)
 466                                                 break;
 467                                         if (cmd[l] == '{')
 468                                                 ++count;
 469                                         else
 470                                                 --count;
 471                                 }
 472                                 if (l != docstring::npos) {
 473                                         j = l;
 474                                         subcmd = cmd.substr(i, j - i + 1);
 475                                 }
 476                         }
 477                         // If the entry doesn't start with '\', we take note
 478                         // of the match and continue (this is not a ultimate
 479                         // acceptance, as some other entry may match a longer
 480                         // portion of the cmd string). However, if the entry
 481                         // does start with '\', we accept the match only if
 482                         // this is a valid macro, i.e., either it is a single
 483                         // (nonletter) char macro, or nothing else follows,
 484                         // or what follows is a nonletter char, or the last
 485                         // character is a }.
 486                         else if (tmp[0] != '\\'
 487                                    || (tmp.size() == prefix + 1 &&
 488                                        !isAlphaASCII(tmp[1]) &&
 489                                        (prefix == 1 || !isAlphaASCII(tmp[2])))
 490                                    || k == cmdend
 491                                    || !isAlphaASCII(cmd[k])
 492                                    || tmp[tmp.size() - 1] == '}'
 493                                  ) {
 494                                 c = it->first;
 495                                 j = k - 1;
 496                                 i = j + 1;
 497                                 unicmd_size = cur_size;
 498                                 if (math == tmp)
 499                                         needsTermination = !it->second.mathnotermination();
 500                                 else
 501                                         needsTermination = !it->second.textnotermination();
 502                                 if (req) {
 503                                         if (math == tmp && it->second.mathfeature() &&
 504                                             !it->second.mathpreamble().empty())
 505                                                 req->insert(it->second.mathpreamble());
 506                                         if (text == tmp && it->second.textfeature() &&
 507                                             !it->second.textpreamble().empty())
 508                                                 req->insert(it->second.textpreamble());
 509                                 }
 510                         }
 511                 }
 512                 if (unicmd_size)
 513                         symbols += c;
 514                 else if (combining != uniend &&
 515                          prefixIs(subcmd, combcmd + '{')) {
 516                         // We know that subcmd starts with combcmd and
 517                         // contains an argument in braces.
 518                         docstring const arg = subcmd.substr(
 519                                 combcmd.length() + 1,
 520                                 subcmd.length() - combcmd.length() - 2);
 521                         // If arg is a single character we can construct a
 522                         // combining sequence.
 523                         char_type a;
 524                         bool argcomb = false;
 525                         if (arg.size() == 1 && isAlnumASCII(arg[0]))
 526                                 a = arg[0];
 527                         else {
 528                                 // Use the version of fromLaTeXCommand() that
 529                                 // parses only one command, since we cannot
 530                                 // use more than one character.
 531                                 bool dummy = false;
 532                                 set<string> r;
 533                                 a = fromLaTeXCommand(arg, cmdtype, argcomb,
 534                                                      dummy, &r);
 535                                 if (a && req && !argcomb)
 536                                         req->insert(r.begin(), r.end());
 537                         }
 538                         if (a && !argcomb) {
 539                                 // In unicode the combining character comes
 540                                 // after its base
 541                                 symbols += a;
 542                                 symbols += combining->first;
 543                                 i = j + 1;
 544                                 unicmd_size = 2;
 545                         }
 546                 }
 547                 if (j + 1 == cmdend && !unicmd_size) {
 548                         // No luck. Return what remains
 549                         rem = cmd.substr(i);
 550                         if (needsTermination && !rem.empty()) {
 551                                 if (rem.substr(0, 2) == "{}") {
 552                                         rem = rem.substr(2);
 553                                         needsTermination = false;
 554                                 } else if (rem[0] == ' ') {
 555                                         needsTermination = false;
 556                                         // LaTeX would swallow all spaces
 557                                         rem = ltrim(rem);
 558                                 }
 559                         }
 560                 }
 561         }
 562         return symbols;
 563 }
 564
 565
 566 CharInfo const & Encodings::unicodeCharInfo(char_type c)
 567 {
 568         static CharInfo empty;
 569         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 570         return it != unicodesymbols.end() ? it->second : empty;
 571 }
 572
 573
 574 bool Encodings::isCombiningChar(char_type c)
 575 {
 576         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 577         if (it != unicodesymbols.end())
 578                 return it->second.combining();
 579         return false;
 580 }
 581
 582
 583 string const Encodings::TIPAShortcut(char_type c)
 584 {
 585         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 586         if (it != unicodesymbols.end())
 587                 return it->second.tipashortcut();
 588         return string();
 589 }
 590
 591
 592 string const Encodings::isKnownScriptChar(char_type const c)
 593 {
 594         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 595
 596         if (it == unicodesymbols.end())
 597                 return string();
 598         // FIXME: parse complex textpreamble (may be list or alternatives,
 599         //                e.g., "subscript,textgreek" or "textcomp|textgreek")
 600         if (it->second.textpreamble() == "textgreek"
 601                 || it->second.textpreamble() == "textcyrillic")
 602                 return it->second.textpreamble();
 603         return string();
 604 }
 605
 606
 607 bool Encodings::fontencSupportsScript(string const & fontenc, string const & script)
 608 {
 609         if (script == "textgreek")
 610                 return (fontenc == "LGR" || fontenc == "TU");
 611         if (script == "textcyrillic")
 612                 return (fontenc == "T2A" || fontenc == "T2B" || fontenc == "T2C"
 613                                 || fontenc == "X2" || fontenc == "TU");
 614         return false;
 615 }
 616
 617
 618 bool Encodings::isMathAlpha(char_type c)
 619 {
 620         return mathalpha.count(c);
 621 }
 622
 623
 624 bool Encodings::isUnicodeTextOnly(char_type c)
 625 {
 626         if (isASCII(c) || isMathAlpha(c))
 627                 return false;
 628
 629         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 630         return it == unicodesymbols.end() || it->second.mathcommand().empty();
 631 }
 632
 633
 634 Encoding const *
 635 Encodings::fromLyXName(string const & name, bool allowUnsafe) const
 636 {
 637         EncodingList::const_iterator const it = encodinglist.find(name);
 638         if (it == encodinglist.end())
 639                 return 0;
 640         if (!allowUnsafe && it->second.unsafe())
 641                 return 0;
 642         return &it->second;
 643 }
 644
 645
 646 Encoding const *
 647 Encodings::fromLaTeXName(string const & n, int const & p, bool allowUnsafe) const
 648 {
 649         string name = n;
 650         // FIXME: if we have to test for too many of these synonyms,
 651         // we should instead extend the format of lib/encodings
 652         if (n == "ansinew")
 653                 name = "cp1252";
 654
 655         // We don't use find_if because it makes copies of the pairs in
 656         // the map.
 657         // This linear search is OK since we don't have many encodings.
 658         // Users could even optimize it by putting the encodings they use
 659         // most at the top of lib/encodings.
 660         EncodingList::const_iterator const end = encodinglist.end();
 661         for (EncodingList::const_iterator it = encodinglist.begin(); it != end; ++it)
 662                 if ((it->second.latexName() == name) && (it->second.package() & p)
 663                                 && (!it->second.unsafe() || allowUnsafe))
 664                         return &it->second;
 665         return 0;
 666 }
 667
 668
 669 Encoding const *
 670 Encodings::fromIconvName(string const & n, int const & p, bool allowUnsafe) const
 671 {
 672         EncodingList::const_iterator const end = encodinglist.end();
 673         for (EncodingList::const_iterator it = encodinglist.begin(); it != end; ++it)
 674                 if ((it->second.iconvName() == n) && (it->second.package() & p)
 675                                 && (!it->second.unsafe() || allowUnsafe))
 676                         return &it->second;
 677         return 0;
 678 }
 679
 680
 681 Encodings::Encodings()
 682 {}
 683
 684
 685 void Encodings::read(FileName const & encfile, FileName const & symbolsfile)
 686 {
 687         // We must read the symbolsfile first, because the Encoding
 688         // constructor depends on it.
 689         CharSetMap forcednotselected;
 690         Lexer symbolslex;
 691         symbolslex.setFile(symbolsfile);
 692         bool getNextToken = true;
 693         while (symbolslex.isOK()) {
 694                 char_type symbol;
 695
 696                 if (getNextToken) {
 697                         if (!symbolslex.next(true))
 698                                 break;
 699                 } else
 700                         getNextToken = true;
 701
 702                 istringstream is(symbolslex.getString());
 703                 // reading symbol directly does not work if
 704                 // char_type == wchar_t.
 705                 uint32_t tmp;
 706                 if(!(is >> hex >> tmp))
 707                         break;
 708                 symbol = tmp;
 709
 710                 if (!symbolslex.next(true))
 711                         break;
 712                 docstring textcommand = symbolslex.getDocString();
 713                 if (!symbolslex.next(true))
 714                         break;
 715                 string textpreamble = symbolslex.getString();
 716                 if (!symbolslex.next(true))
 717                         break;
 718                 string sflags = symbolslex.getString();
 719
 720                 string tipashortcut;
 721                 int flags = 0;
 722
 723                 if (suffixIs(textcommand, '}'))
 724                         flags |= CharInfoTextNoTermination;
 725                 while (!sflags.empty()) {
 726                         string flag;
 727                         sflags = split(sflags, flag, ',');
 728                         if (flag == "combining") {
 729                                 flags |= CharInfoCombining;
 730                         } else if (flag == "force") {
 731                                 flags |= CharInfoForce;
 732                                 forced.insert(symbol);
 733                         } else if (prefixIs(flag, "force=")) {
 734                                 vector<string> encs =
 735                                         getVectorFromString(flag.substr(6), ";");
 736                                 for (size_t i = 0; i < encs.size(); ++i)
 737                                         forcedselected[encs[i]].insert(symbol);
 738                                 flags |= CharInfoForceSelected;
 739                         } else if (prefixIs(flag, "force!=")) {
 740                                 vector<string> encs =
 741                                         getVectorFromString(flag.substr(7), ";");
 742                                 for (size_t i = 0; i < encs.size(); ++i)
 743                                         forcednotselected[encs[i]].insert(symbol);
 744                                 flags |= CharInfoForceSelected;
 745                         } else if (flag == "mathalpha") {
 746                                 mathalpha.insert(symbol);
 747                         } else if (flag == "notermination=text") {
 748                                 flags |= CharInfoTextNoTermination;
 749                         } else if (flag == "notermination=math") {
 750                                 flags |= CharInfoMathNoTermination;
 751                         } else if (flag == "notermination=both") {
 752                                 flags |= CharInfoTextNoTermination;
 753                                 flags |= CharInfoMathNoTermination;
 754                         } else if (flag == "notermination=none") {
 755                                 flags &= ~CharInfoTextNoTermination;
 756                                 flags &= ~CharInfoMathNoTermination;
 757                         } else if (contains(flag, "tipashortcut=")) {
 758                                 tipashortcut = split(flag, '=');
 759                         } else if (flag == "deprecated") {
 760                                 flags |= CharInfoDeprecated;
 761                         } else {
 762                                 lyxerr << "Ignoring unknown flag `" << flag
 763                                        << "' for symbol `0x"
 764                                        << hex << symbol << dec
 765                                        << "'." << endl;
 766                         }
 767                 }
 768                 // mathcommand and mathpreamble have been added for 1.6.0.
 769                 // make them optional so that old files still work.
 770                 int const lineno = symbolslex.lineNumber();
 771                 bool breakout = false;
 772                 docstring mathcommand;
 773                 string mathpreamble;
 774                 if (symbolslex.next(true)) {
 775                         if (symbolslex.lineNumber() != lineno) {
 776                                 // line in old format without mathcommand and mathpreamble
 777                                 getNextToken = false;
 778                         } else {
 779                                 mathcommand = symbolslex.getDocString();
 780                                 if (suffixIs(mathcommand, '}'))
 781                                         flags |= CharInfoMathNoTermination;
 782                                 if (symbolslex.next(true)) {
 783                                         if (symbolslex.lineNumber() != lineno) {
 784                                                 // line in new format with mathcommand only
 785                                                 getNextToken = false;
 786                                         } else {
 787                                                 // line in new format with mathcommand and mathpreamble
 788                                                 mathpreamble = symbolslex.getString();
 789                                         }
 790                                 } else
 791                                         breakout = true;
 792                         }
 793                 } else {
 794                         breakout = true;
 795                 }
 796
 797                 // backward compatibility
 798                 if (mathpreamble == "esintoramsmath")
 799                         mathpreamble = "esint|amsmath";
 800
 801                 if (!textpreamble.empty())
 802                         if (textpreamble[0] != '\\')
 803                                 flags |= CharInfoTextFeature;
 804                 if (!mathpreamble.empty())
 805                         if (mathpreamble[0] != '\\')
 806                                 flags |= CharInfoMathFeature;
 807
 808                 CharInfo info = CharInfo(
 809                         textcommand, mathcommand,
 810                         textpreamble, mathpreamble,
 811                         tipashortcut, flags);
 812                 LYXERR(Debug::INFO, "Read unicode symbol " << symbol << " '"
 813                            << to_utf8(info.textcommand()) << "' '" << info.textpreamble()
 814                            << " '" << info.textfeature() << ' ' << info.textnotermination()
 815                            << ' ' << to_utf8(info.mathcommand()) << "' '" << info.mathpreamble()
 816                            << "' " << info.mathfeature() << ' ' << info.mathnotermination()
 817                            << ' ' << info.combining() << ' ' << info.force()
 818                            << ' ' << info.forceselected());
 819
 820                 // we assume that at least one command is nonempty when using unicodesymbols
 821                 if (info.isUnicodeSymbol()) {
 822                         unicodesymbols[symbol] = info;
 823                 }
 824
 825                 if (breakout)
 826                         break;
 827         }
 828
 829         // Now read the encodings
 830         enum {
 831                 et_encoding = 1,
 832                 et_end
 833         };
 834
 835         LexerKeyword encodingtags[] = {
 836                 { "encoding", et_encoding },
 837                 { "end", et_end }
 838         };
 839
 840         Lexer lex(encodingtags);
 841         lex.setFile(encfile);
 842         lex.setContext("Encodings::read");
 843         while (lex.isOK()) {
 844                 switch (lex.lex()) {
 845                 case et_encoding:
 846                 {
 847                         lex.next();
 848                         string const name = lex.getString();
 849                         lex.next();
 850                         string const latexname = lex.getString();
 851                         lex.next();
 852                         string const guiname = lex.getString();
 853                         lex.next();
 854                         string const iconvname = lex.getString();
 855                         lex.next();
 856                         string const width = lex.getString();
 857                         bool fixedwidth = false;
 858                         bool unsafe = false;
 859                         if (width == "fixed")
 860                                 fixedwidth = true;
 861                         else if (width == "variable")
 862                                 fixedwidth = false;
 863                         else if (width == "variableunsafe") {
 864                                 fixedwidth = false;
 865                                 unsafe = true;
 866                         }
 867                         else
 868                                 lex.printError("Unknown width");
 869
 870                         lex.next();
 871                         string const p = lex.getString();
 872                         Encoding::Package package = Encoding::none;
 873                         if (p == "none")
 874                                 package = Encoding::none;
 875                         else if (p == "inputenc")
 876                                 package = Encoding::inputenc;
 877                         else if (p == "CJK")
 878                                 package = Encoding::CJK;
 879                         else if (p == "japanese")
 880                                 package = Encoding::japanese;
 881                         else
 882                                 lex.printError("Unknown package");
 883
 884                         LYXERR(Debug::INFO, "Reading encoding " << name);
 885                         encodinglist[name] = Encoding(name, latexname,
 886                                 guiname, iconvname, fixedwidth, unsafe,
 887                                 package);
 888
 889                         if (lex.lex() != et_end)
 890                                 lex.printError("Missing end");
 891                         break;
 892                 }
 893                 case et_end:
 894                         lex.printError("Misplaced end");
 895                         break;
 896                 case Lexer::LEX_FEOF:
 897                         break;
 898                 default:
 899                         lex.printError("Unknown tag");
 900                         break;
 901                 }
 902         }
 903
 904         // Move all information from forcednotselected to forcedselected
 905         for (CharSetMap::const_iterator it1 = forcednotselected.begin(); it1 != forcednotselected.end(); ++it1) {
 906                 for (CharSetMap::iterator it2 = forcedselected.begin(); it2 != forcedselected.end(); ++it2) {
 907                         if (it2->first != it1->first)
 908                                 it2->second.insert(it1->second.begin(), it1->second.end());
 909                 }
 910         }
 911
 912 }
 913
 914
 915 } // namespace lyx