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 "support/debug.h"
  18 #include "support/docstring.h"
  19 #include "support/gettext.h"
  20 #include "support/Lexer.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 noexcept
  71 {
  72         return "Could not find LaTeX command for a character";
  73 }
  74
  75
  76 CharInfo::CharInfo(
  77                 docstring const & text_command, docstring const & math_command,
  78                 std::string const & text_preamble, std::string const & math_preamble,
  79                 std::string const & tipa_shortcut, unsigned int flags)
  80         : text_commands_({text_command}), math_commands_({math_command}),
  81           text_preamble_(text_preamble), math_preamble_(math_preamble),
  82           tipa_shortcut_(tipa_shortcut), flags_(flags)
  83 {
  84 }
  85
  86
  87 CharInfo::CharInfo(
  88                 std::vector<docstring> const & text_commands, std::vector<docstring> const & math_commands,
  89                 std::string const & text_preamble, std::string const & math_preamble,
  90                 std::string const & tipa_shortcut, unsigned int flags)
  91         : text_commands_(text_commands), math_commands_(math_commands),
  92           text_preamble_(text_preamble), math_preamble_(math_preamble),
  93           tipa_shortcut_(tipa_shortcut), flags_(flags)
  94 {
  95 }
  96
  97
  98 Encoding::Encoding(string const & n, string const & l, string const & g,
  99                    string const & i, bool f, bool u, Encoding::Package p)
 100         : name_(n), latexName_(l), guiName_(g), iconvName_(i), fixedwidth_(f),
 101           unsafe_(u), forced_(&forcedSelected[n]), package_(p)
 102 {
 103         if (n == "ascii") {
 104                 // ASCII can encode 128 code points and nothing else
 105                 start_encodable_ = 128;
 106                 complete_ = true;
 107         } else if (i == "UTF-8") {
 108                 // UTF8 can encode all UCS4 code points
 109                 start_encodable_ = max_ucs4;
 110                 complete_ = true;
 111         } else {
 112                 start_encodable_ = 0;
 113                 complete_ = false;
 114         }
 115 }
 116
 117
 118 void Encoding::init() const
 119 {
 120         // Since the the constructor is the only method which sets complete_
 121         // to false the test for complete_ is thread-safe without mutex.
 122         if (complete_)
 123                 return;
 124
 125         static Mutex mutex;
 126         Mutex::Locker lock(&mutex);
 127
 128         // We need to test again for complete_, since another thread could
 129         // have set it to true while we were waiting for the lock and we must
 130         // not modify an encoding which is already complete.
 131         if (complete_)
 132                 return;
 133
 134         // We do not make any member mutable  so that it can be easily verified
 135         // that all const methods are thread-safe: init() is the only const
 136         // method which changes complete_, encodable_ and start_encodable_, and
 137         // it uses a mutex to ensure thread-safety.
 138         CharSet & encodable = const_cast<Encoding *>(this)->encodable_;
 139         char_type & start_encodable = const_cast<Encoding *>(this)->start_encodable_;
 140
 141         start_encodable = 0;
 142         // temporarily switch off lyxerr, since we will generate iconv errors
 143         lyxerr.disable();
 144         if (fixedwidth_) {
 145                 // We do not need to check all UCS4 code points, it is enough
 146                 // if we check all 256 code points of this encoding.
 147                 for (unsigned short j = 0; j < 256; ++j) {
 148                         char const c = char(j);
 149                         vector<char_type> const ucs4 = eightbit_to_ucs4(&c, 1, iconvName_);
 150                         if (ucs4.size() != 1)
 151                                 continue;
 152                         char_type const uc = ucs4[0];
 153                         CharInfoMap::const_iterator const it = unicodesymbols.find(uc);
 154                         if (it == unicodesymbols.end())
 155                                 encodable.insert(uc);
 156                         else if (!it->second.force()) {
 157                                 if (forced_->empty() || forced_->find(uc) == forced_->end())
 158                                         encodable.insert(uc);
 159                         }
 160                 }
 161         } else {
 162                 // We do not know how many code points this encoding has, and
 163                 // they do not have a direct representation as a single byte,
 164                 // therefore we need to check all UCS4 code points.
 165                 // This is expensive!
 166                 for (char_type c = 0; c < max_ucs4; ++c) {
 167                         vector<char> const eightbit = ucs4_to_eightbit(&c, 1, iconvName_);
 168                         if (!eightbit.empty()) {
 169                                 CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 170                                 if (it == unicodesymbols.end())
 171                                         encodable.insert(c);
 172                                 else if (!it->second.force()) {
 173                                         if (forced_->empty() || forced_->find(c) == forced_->end())
 174                                                 encodable.insert(c);
 175                                 }
 176                         }
 177                 }
 178         }
 179         lyxerr.enable();
 180         CharSet::iterator it = encodable.find(start_encodable);
 181         while (it != encodable.end()) {
 182                 encodable.erase(it);
 183                 ++start_encodable;
 184                 it = encodable.find(start_encodable);
 185         }
 186         const_cast<Encoding *>(this)->complete_ = true;
 187 }
 188
 189
 190 bool Encoding::isForced(char_type c) const
 191 {
 192         if (!forced.empty() && forced.find(c) != forced.end())
 193                 return true;
 194         return !forced_->empty() && forced_->find(c) != forced_->end();
 195 }
 196
 197
 198 bool Encoding::encodable(char_type c) const
 199 {
 200         // assure the used encoding is properly initialized
 201         init();
 202         if (iconvName_ == "UTF-8" && package_ == none)
 203                 return true;
 204         // platex does not load inputenc: force conversion of supported characters
 205         if (package_ == Encoding::japanese
 206             && ((0xb7 <= c && c <= 0x05ff) // Latin-1 Supplement ... Hebrew
 207                         || (0x1d00 <= c && c <= 0x218f) // Phonetic Extensions ... Number Forms
 208                         || (0x2193 <= c && c <= 0x2aff) // Arrows ... Supplemental Mathematical Operators
 209                         || (0xfb00 <= c && c <= 0xfb4f) // Alphabetic Presentation Forms
 210                         || (0x1d400 <= c && c <= 0x1d7ff))) // Mathematical Alphanumeric Symbols
 211                 return false;
 212         if (c < start_encodable_ && !isForced(c))
 213                 return true;
 214         if (encodable_.find(c) != encodable_.end())
 215                 return true;
 216         return false;
 217 }
 218
 219
 220 pair<docstring, bool> Encoding::latexChar(char_type c) const
 221 {
 222         if (encodable(c))
 223                 return make_pair(docstring(1, c), false);
 224
 225         // c cannot (or should not) be encoded in this encoding
 226         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 227         if (it == unicodesymbols.end())
 228                 throw EncodingException(c);
 229         // at least one of mathCommand and textCommand is nonempty
 230         if (it->second.textCommand().empty())
 231                 return make_pair(
 232                                 "\\ensuremath{" + it->second.mathCommand() + '}', false);
 233         return make_pair(it->second.textCommand(), !it->second.textNoTermination());
 234 }
 235
 236
 237 pair<docstring, docstring> Encoding::latexString(docstring const & input, bool dryrun) const
 238 {
 239         docstring result;
 240         docstring uncodable;
 241         bool terminate = false;
 242         for (char_type const c : input) {
 243                 try {
 244                         pair<docstring, bool> latex_char = latexChar(c);
 245                         docstring const latex = latex_char.first;
 246                         if (terminate && !prefixIs(latex, '\\')
 247                             && !prefixIs(latex, '{')
 248                             && !prefixIs(latex, '}')) {
 249                                         // Prevent eating of a following
 250                                         // space or command corruption by
 251                                         // following characters
 252                                         if (latex == " ")
 253                                                 result += "{}";
 254                                         else
 255                                                 result += " ";
 256                                 }
 257                         result += latex;
 258                         terminate = latex_char.second;
 259                 } catch (EncodingException & /* e */) {
 260                         LYXERR0("Uncodable character <" << docstring(1, c)
 261                                         << "> in latexString!");
 262                         if (dryrun) {
 263                                 result += "<" + _("LyX Warning: ")
 264                                            + _("uncodable character") + " '";
 265                                 result += docstring(1, c);
 266                                 result += "'>";
 267                         } else
 268                                 uncodable += c;
 269                 }
 270         }
 271         return make_pair(result, uncodable);
 272 }
 273
 274
 275 vector<char_type> Encoding::symbolsList() const
 276 {
 277         // assure the used encoding is properly initialized
 278         init();
 279
 280         // first all those below start_encodable_
 281         vector<char_type> symbols;
 282         for (char_type c = 0; c < start_encodable_; ++c)
 283                 symbols.push_back(c);
 284         // add all encodable characters
 285         copy(encodable_.begin(), encodable_.end(), back_inserter(symbols));
 286         // now the ones from the unicodesymbols file that are not already there
 287         for (auto const & elem : unicodesymbols) {
 288                 if (find(symbols.begin(), symbols.end(), elem.first) == symbols.end())
 289                         symbols.push_back(elem.first);
 290         }
 291         // finally, sort the vector
 292         sort(symbols.begin(), symbols.end());
 293         return symbols;
 294 }
 295
 296
 297 bool Encodings::latexMathChar(char_type c, bool mathmode,
 298                         Encoding const * encoding, docstring & command,
 299                         bool & needsTermination)
 300 {
 301         command = empty_docstring();
 302         if (encoding)
 303                 if (encoding->encodable(c))
 304                         command = docstring(1, c);
 305         needsTermination = false;
 306
 307         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 308         if (it == unicodesymbols.end()) {
 309                 if (!encoding || command.empty())
 310                         throw EncodingException(c);
 311                 if (mathmode)
 312                         addMathSym(c);
 313                 return false;
 314         }
 315         // at least one of mathCommand and textCommand is nonempty
 316         bool use_math = (mathmode && !it->second.mathCommand().empty()) ||
 317                         (!mathmode && it->second.textCommand().empty());
 318         if (use_math) {
 319                 command = it->second.mathCommand();
 320                 needsTermination = !it->second.mathNoTermination();
 321                 addMathCmd(c);
 322         } else {
 323                 if (!encoding || command.empty()) {
 324                         command = it->second.textCommand();
 325                         needsTermination = !it->second.textNoTermination();
 326                 }
 327                 if (mathmode)
 328                         addMathSym(c);
 329                 else
 330                         addTextCmd(c);
 331         }
 332         return use_math;
 333 }
 334
 335
 336 char_type Encodings::fromLaTeXCommand(docstring const & cmd, int cmdtype,
 337                 bool & combining, bool & needsTermination, set<string> * req)
 338 {
 339         CharInfoMap::const_iterator const end = unicodesymbols.end();
 340         CharInfoMap::const_iterator it = unicodesymbols.begin();
 341         for (combining = false; it != end; ++it) {
 342                 if (it->second.deprecated())
 343                         continue;
 344
 345                 if (cmdtype & MATH_CMD) {
 346                         for (const docstring& math : it->second.mathCommands()) {
 347                                 if ((cmdtype & MATH_CMD) && math == cmd) {
 348                                         combining = it->second.combining();
 349                                         needsTermination = !it->second.mathNoTermination();
 350                                         if (req && it->second.mathFeature() &&
 351                                             !it->second.mathPreamble().empty())
 352                                                 req->insert(it->second.mathPreamble());
 353                                         return it->first;
 354                                 }
 355                         }
 356                 }
 357
 358                 if (cmdtype & TEXT_CMD) {
 359                         for (const docstring& text : it->second.textCommands()) {
 360                                 if (text == cmd) {
 361                                         combining = it->second.combining();
 362                                         needsTermination = !it->second.textNoTermination();
 363                                         if (req && it->second.textFeature() &&
 364                                             !it->second.textPreamble().empty())
 365                                                 req->insert(it->second.textPreamble());
 366                                         return it->first;
 367                                 }
 368                         }
 369                 }
 370         }
 371         needsTermination = false;
 372         return 0;
 373 }
 374
 375
 376 docstring Encodings::fromLaTeXCommand(docstring const & cmd, int cmdtype,
 377                 bool & needsTermination, docstring & rem, set<string> * req)
 378 {
 379         needsTermination = false;
 380         rem = empty_docstring();
 381         bool const mathmode = cmdtype & MATH_CMD;
 382         bool const textmode = cmdtype & TEXT_CMD;
 383
 384         // Easy case: the command is a complete entry of unicodesymbols.
 385         for (const auto & unicodeSymbol : unicodesymbols) {
 386                 if (mathmode) {
 387                         for (const auto & command : unicodeSymbol.second.mathCommands()) {
 388                                 if (command == cmd) {
 389                                         docstring value;
 390                                         value += unicodeSymbol.first;
 391                                         needsTermination = !unicodeSymbol.second.mathNoTermination();
 392                                         if (req && unicodeSymbol.second.mathFeature()
 393                                                 && !unicodeSymbol.second.mathPreamble().empty())
 394                                                         req->insert(unicodeSymbol.second.mathPreamble());
 395                                         return value;
 396                                 }
 397                         }
 398                 }
 399                 if (textmode) {
 400                         for (const auto & command : unicodeSymbol.second.textCommands()) {
 401                                 if (command == cmd) {
 402                                         docstring value;
 403                                         value += unicodeSymbol.first;
 404                                         needsTermination = !unicodeSymbol.second.textNoTermination();
 405                                         if (req && unicodeSymbol.second.textFeature()
 406                                                 && !unicodeSymbol.second.textPreamble().empty())
 407                                                 req->insert(unicodeSymbol.second.textPreamble());
 408                                         return value;
 409                                 }
 410                         }
 411                 }
 412         }
 413
 414         // Otherwise, try to map as many commands as possible, matching prefixes of the command.
 415         docstring symbols;
 416         size_t const cmdend = cmd.size();
 417         size_t prefix = 0;
 418         CharInfoMap::const_iterator const uniend = unicodesymbols.end();
 419         for (size_t i = 0, j = 0; j < cmdend; ++j) {
 420                 // Also get the char after a backslash
 421                 if (j + 1 < cmdend && cmd[j] == '\\') {
 422                         ++j;
 423                         prefix = 1;
 424                         // Detect things like \=*{e} as well
 425                         if (j + 3 < cmdend && cmd[j+1] == '*' &&
 426                             cmd[j+2] == '{') {
 427                                 ++j;
 428                                 prefix = 2;
 429                         }
 430                 }
 431                 // position of the last character before a possible macro
 432                 // argument
 433                 size_t m = j;
 434                 // If a macro argument follows, get it, too
 435                 // Do it here only for single character commands. Other
 436                 // combining commands need this too, but they are handled in
 437                 // the loop below for performance reasons.
 438                 if (j + 1 < cmdend && cmd[j + 1] == '{') {
 439                         size_t k = j + 1;
 440                         int count = 1;
 441                         while (k < cmdend && count) {
 442                                 k = cmd.find_first_of(from_ascii("{}"), k + 1);
 443                                 // braces may not be balanced
 444                                 if (k == docstring::npos)
 445                                         break;
 446                                 if (cmd[k] == '{')
 447                                         ++count;
 448                                 else
 449                                         --count;
 450                         }
 451                         if (k != docstring::npos)
 452                                 j = k;
 453                 } else if (m + 1 < cmdend && isAlphaASCII(cmd[m])) {
 454                         while (m + 2 < cmdend && isAlphaASCII(cmd[m+1]))
 455                                 m++;
 456                 }
 457                 // Start with this substring and try augmenting it when it is
 458                 // the prefix of some command in the unicodesymbols file
 459                 docstring subcmd = cmd.substr(i, j - i + 1);
 460
 461                 // First part of subcmd which might be a combining character
 462                 docstring combcmd = (m == j) ? docstring() : cmd.substr(i, m - i + 1);
 463                 // The combining character of combcmd if it exists
 464                 size_t unicmd_size = 0;
 465                 char_type c = 0;
 466                 CharInfoMap::const_iterator it = unicodesymbols.begin();
 467                 CharInfoMap::const_iterator combining = uniend;
 468                 for (; it != uniend; ++it) {
 469                         if (it->second.deprecated())
 470                                 continue;
 471                         docstring const math = mathmode ? it->second.mathCommand()
 472                                                         : docstring();
 473                         docstring const text = textmode ? it->second.textCommand()
 474                                                         : docstring();
 475                         if (!combcmd.empty() && it->second.combining() &&
 476                             (math == combcmd || text == combcmd))
 477                                 combining = it;
 478                         size_t cur_size = max(math.size(), text.size());
 479                         // The current math or text unicode command cannot
 480                         // match, or we already matched a longer one
 481                         if (cur_size < subcmd.size() || cur_size <= unicmd_size)
 482                                 continue;
 483
 484                         docstring tmp = subcmd;
 485                         size_t k = j;
 486                         while (prefixIs(math, tmp) || prefixIs(text, tmp)) {
 487                                 ++k;
 488                                 if (k >= cmdend || cur_size <= tmp.size())
 489                                         break;
 490                                 tmp += cmd[k];
 491                         }
 492                         // No match
 493                         if (k == j)
 494                                 continue;
 495
 496                         // The last added char caused a mismatch, because
 497                         // we didn't exhaust the chars in cmd and didn't
 498                         // exceed the maximum size of the current unicmd
 499                         if (k < cmdend && cur_size > tmp.size())
 500                                 tmp.resize(tmp.size() - 1);
 501
 502                         // If this is an exact match, we found a (longer)
 503                         // matching entry in the unicodesymbols file.
 504                         if (math != tmp && text != tmp)
 505                                 continue;
 506                         // If we found a combining command, we need to append
 507                         // the macro argument if this has not been done above.
 508                         if (tmp == combcmd && combining != uniend &&
 509                             k < cmdend && cmd[k] == '{') {
 510                                 size_t l = k;
 511                                 int count = 1;
 512                                 while (l < cmdend && count) {
 513                                         l = cmd.find_first_of(from_ascii("{}"), l + 1);
 514                                         // braces may not be balanced
 515                                         if (l == docstring::npos)
 516                                                 break;
 517                                         if (cmd[l] == '{')
 518                                                 ++count;
 519                                         else
 520                                                 --count;
 521                                 }
 522                                 if (l != docstring::npos) {
 523                                         j = l;
 524                                         subcmd = cmd.substr(i, j - i + 1);
 525                                 }
 526                         }
 527                         // If the entry doesn't start with '\', we take note
 528                         // of the match and continue (this is not a ultimate
 529                         // acceptance, as some other entry may match a longer
 530                         // portion of the cmd string). However, if the entry
 531                         // does start with '\', we accept the match only if
 532                         // this is a valid macro, i.e., either it is a single
 533                         // (nonletter) char macro, or nothing else follows,
 534                         // or what follows is a nonletter char, or the last
 535                         // character is a }.
 536                         else if (tmp[0] != '\\'
 537                                    || (tmp.size() == prefix + 1 &&
 538                                        !isAlphaASCII(tmp[1]) &&
 539                                        (prefix == 1 || !isAlphaASCII(tmp[2])))
 540                                    || k == cmdend
 541                                    || !isAlphaASCII(cmd[k])
 542                                    || tmp[tmp.size() - 1] == '}'
 543                                  ) {
 544                                 c = it->first;
 545                                 j = k - 1;
 546                                 i = j + 1;
 547                                 unicmd_size = cur_size;
 548                                 if (math == tmp)
 549                                         needsTermination = !it->second.mathNoTermination();
 550                                 else
 551                                         needsTermination = !it->second.textNoTermination();
 552                                 if (req) {
 553                                         if (math == tmp && it->second.mathFeature() &&
 554                                             !it->second.mathPreamble().empty())
 555                                                 req->insert(it->second.mathPreamble());
 556                                         if (text == tmp && it->second.textFeature() &&
 557                                             !it->second.textPreamble().empty())
 558                                                 req->insert(it->second.textPreamble());
 559                                 }
 560                         }
 561                 }
 562                 if (unicmd_size)
 563                         symbols += c;
 564                 else if (combining != uniend &&
 565                          prefixIs(subcmd, combcmd + '{')) {
 566                         // We know that subcmd starts with combcmd and
 567                         // contains an argument in braces.
 568                         docstring const arg = subcmd.substr(
 569                                 combcmd.length() + 1,
 570                                 subcmd.length() - combcmd.length() - 2);
 571                         // If arg is a single character we can construct a
 572                         // combining sequence.
 573                         char_type a;
 574                         bool argcomb = false;
 575                         if (arg.size() == 1 && isAlnumASCII(arg[0]))
 576                                 a = arg[0];
 577                         else {
 578                                 // Use the version of fromLaTeXCommand() that
 579                                 // parses only one command, since we cannot
 580                                 // use more than one character.
 581                                 bool dummy = false;
 582                                 set<string> r;
 583                                 a = fromLaTeXCommand(arg, cmdtype, argcomb,
 584                                                      dummy, &r);
 585                                 if (a && req && !argcomb)
 586                                         req->insert(r.begin(), r.end());
 587                         }
 588                         if (a && !argcomb) {
 589                                 // In unicode the combining character comes
 590                                 // after its base
 591                                 symbols += a;
 592                                 symbols += combining->first;
 593                                 i = j + 1;
 594                                 unicmd_size = 2;
 595                         }
 596                 }
 597                 if (j + 1 == cmdend && !unicmd_size) {
 598                         // No luck. Return what remains
 599                         rem = cmd.substr(i);
 600                         if (needsTermination && !rem.empty()) {
 601                                 if (rem.substr(0, 2) == "{}") {
 602                                         rem = rem.substr(2);
 603                                         needsTermination = false;
 604                                 } else if (rem[0] == ' ') {
 605                                         needsTermination = false;
 606                                         // LaTeX would swallow all spaces
 607                                         rem = ltrim(rem);
 608                                 }
 609                         }
 610                 }
 611         }
 612         return symbols;
 613 }
 614
 615
 616 CharInfo const & Encodings::unicodeCharInfo(char_type c)
 617 {
 618         static CharInfo empty;
 619         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 620         return it != unicodesymbols.end() ? it->second : empty;
 621 }
 622
 623
 624 bool Encodings::isCombiningChar(char_type c)
 625 {
 626         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 627         if (it != unicodesymbols.end())
 628                 return it->second.combining();
 629         return false;
 630 }
 631
 632
 633 string const Encodings::TIPAShortcut(char_type c)
 634 {
 635         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 636         if (it != unicodesymbols.end())
 637                 return it->second.tipaShortcut();
 638         return string();
 639 }
 640
 641
 642 string const Encodings::isKnownScriptChar(char_type const c)
 643 {
 644         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 645
 646         if (it == unicodesymbols.end())
 647                 return string();
 648         // FIXME: parse complex textPreamble (may be list or alternatives,
 649         //                e.g., "subscript,textgreek" or "textcomp|textgreek")
 650         if (it->second.textPreamble() == "textgreek"
 651                 || it->second.textPreamble() == "textcyrillic")
 652                 return it->second.textPreamble();
 653         return string();
 654 }
 655
 656
 657 bool Encodings::fontencSupportsScript(string const & fontenc, string const & script)
 658 {
 659         if (script == "textgreek")
 660                 return (fontenc == "LGR" || fontenc == "TU");
 661         if (script == "textcyrillic")
 662                 return (fontenc == "T2A" || fontenc == "T2B" || fontenc == "T2C"
 663                                 || fontenc == "X2" || fontenc == "TU");
 664         return false;
 665 }
 666
 667
 668 bool Encodings::isMathAlpha(char_type c)
 669 {
 670         return mathalpha.count(c);
 671 }
 672
 673
 674 bool Encodings::isUnicodeTextOnly(char_type c)
 675 {
 676         if (isASCII(c) || isMathAlpha(c))
 677                 return false;
 678
 679         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 680         return it == unicodesymbols.end() || it->second.mathCommand().empty();
 681 }
 682
 683
 684 Encoding const *
 685 Encodings::fromLyXName(string const & name, bool allowUnsafe) const
 686 {
 687         EncodingList::const_iterator const it = encodinglist.find(name);
 688         if (it == encodinglist.end())
 689                 return nullptr;
 690         if (!allowUnsafe && it->second.unsafe())
 691                 return nullptr;
 692         return &it->second;
 693 }
 694
 695
 696 Encoding const *
 697 Encodings::fromLaTeXName(string const & n, int p, bool allowUnsafe) const
 698 {
 699         string name = n;
 700         // FIXME: if we have to test for too many of these synonyms,
 701         // we should instead extend the format of lib/encodings
 702         if (n == "ansinew")
 703                 name = "cp1252";
 704
 705         // We don't use find_if because it makes copies of the pairs in
 706         // the map.
 707         // This linear search is OK since we don't have many encodings.
 708         // Users could even optimize it by putting the encodings they use
 709         // most at the top of lib/encodings.
 710         EncodingList::const_iterator const end = encodinglist.end();
 711         for (EncodingList::const_iterator it = encodinglist.begin(); it != end; ++it)
 712                 if ((it->second.latexName() == name) && (it->second.package() & p)
 713                                 && (!it->second.unsafe() || allowUnsafe))
 714                         return &it->second;
 715         return nullptr;
 716 }
 717
 718
 719 Encoding const *
 720 Encodings::fromIconvName(string const & n, int p, bool allowUnsafe) const
 721 {
 722         EncodingList::const_iterator const end = encodinglist.end();
 723         for (EncodingList::const_iterator it = encodinglist.begin(); it != end; ++it)
 724                 if ((it->second.iconvName() == n) && (it->second.package() & p)
 725                                 && (!it->second.unsafe() || allowUnsafe))
 726                         return &it->second;
 727         return nullptr;
 728 }
 729
 730
 731 Encodings::Encodings()
 732 {}
 733
 734
 735 void Encodings::read(FileName const & encfile, FileName const & symbolsfile)
 736 {
 737         // We must read the symbolsfile first, because the Encoding
 738         // constructor depends on it.
 739         CharSetMap forcedNotSelected;
 740         Lexer symbolsLex;
 741         symbolsLex.setFile(symbolsfile);
 742         bool getNextToken = true;
 743         while (symbolsLex.isOK()) {
 744                 char_type symbol;
 745
 746                 if (getNextToken) {
 747                         if (!symbolsLex.next(true))
 748                                 break;
 749                 } else
 750                         getNextToken = true;
 751
 752                 istringstream is(symbolsLex.getString());
 753                 // reading symbol directly does not work if
 754                 // char_type == wchar_t.
 755                 uint32_t tmp;
 756                 if(!(is >> hex >> tmp))
 757                         break;
 758                 symbol = tmp;
 759
 760                 // Special case: more than one entry for one character (to add other LaTeX commands).
 761                 if (unicodesymbols.find(symbol) != unicodesymbols.end()) {
 762                         if (!symbolsLex.next(true))
 763                                 break;
 764                         docstring textCommand = symbolsLex.getDocString();
 765                         if (!symbolsLex.next(true))
 766                                 break;
 767                         string mathCommand = symbolsLex.getString();
 768
 769                         if (!textCommand.empty())
 770                                 unicodesymbols.at(symbol).addTextCommand(textCommand);
 771                         if (!mathCommand.empty())
 772                                 unicodesymbols.at(symbol).addMathCommand(textCommand);
 773
 774                         continue;
 775                 }
 776
 777                 // If the symbol is not the same as the previous entry, consider it is a totally new symbol.
 778                 if (!symbolsLex.next(true))
 779                         break;
 780                 docstring textCommand = symbolsLex.getDocString();
 781                 if (!symbolsLex.next(true))
 782                         break;
 783                 string textPreamble = symbolsLex.getString();
 784                 if (!symbolsLex.next(true))
 785                         break;
 786                 string sflags = symbolsLex.getString();
 787
 788                 string tipaShortcut;
 789                 int flags = 0;
 790
 791                 if (suffixIs(textCommand, '}'))
 792                         flags |= CharInfoTextNoTermination;
 793                 while (!sflags.empty()) {
 794                         string flag;
 795                         sflags = split(sflags, flag, ',');
 796                         if (flag == "combining") {
 797                                 flags |= CharInfoCombining;
 798                         } else if (flag == "force") {
 799                                 flags |= CharInfoForce;
 800                                 forced.insert(symbol);
 801                         } else if (prefixIs(flag, "force=")) {
 802                                 vector<string> encs =
 803                                         getVectorFromString(flag.substr(6), ";");
 804                                 for (auto const & enc : encs)
 805                                         forcedSelected[enc].insert(symbol);
 806                                 flags |= CharInfoForceSelected;
 807                         } else if (prefixIs(flag, "force!=")) {
 808                                 vector<string> encs =
 809                                         getVectorFromString(flag.substr(7), ";");
 810                                 for (auto const & enc : encs)
 811                                         forcedNotSelected[enc].insert(symbol);
 812                                 flags |= CharInfoForceSelected;
 813                         } else if (flag == "mathalpha") {
 814                                 mathalpha.insert(symbol);
 815                         } else if (flag == "notermination=text") {
 816                                 flags |= CharInfoTextNoTermination;
 817                         } else if (flag == "notermination=math") {
 818                                 flags |= CharInfoMathNoTermination;
 819                         } else if (flag == "notermination=both") {
 820                                 flags |= CharInfoTextNoTermination;
 821                                 flags |= CharInfoMathNoTermination;
 822                         } else if (flag == "notermination=none") {
 823                                 flags &= ~CharInfoTextNoTermination;
 824                                 flags &= ~CharInfoMathNoTermination;
 825                         } else if (contains(flag, "tipashortcut=")) {
 826                                 tipaShortcut = split(flag, '=');
 827                         } else if (flag == "deprecated") {
 828                                 flags |= CharInfoDeprecated;
 829                         } else {
 830                                 lyxerr << "Ignoring unknown flag `" << flag
 831                                        << "' for symbol `0x"
 832                                        << hex << symbol << dec
 833                                        << "'." << endl;
 834                         }
 835                 }
 836                 // mathCommand and mathPreamble have been added for 1.6.0.
 837                 // make them optional so that old files still work.
 838                 int const lineNo = symbolsLex.lineNumber();
 839                 bool breakout = false;
 840                 docstring mathCommand;
 841                 string mathPreamble;
 842                 if (symbolsLex.next(true)) {
 843                         if (symbolsLex.lineNumber() != lineNo) {
 844                                 // line in old format without mathCommand and mathPreamble
 845                                 getNextToken = false;
 846                         } else {
 847                                 mathCommand = symbolsLex.getDocString();
 848                                 if (suffixIs(mathCommand, '}'))
 849                                         flags |= CharInfoMathNoTermination;
 850                                 if (symbolsLex.next(true)) {
 851                                         if (symbolsLex.lineNumber() != lineNo) {
 852                                                 // line in new format with mathCommand only
 853                                                 getNextToken = false;
 854                                         } else {
 855                                                 // line in new format with mathCommand and mathPreamble
 856                                                 mathPreamble = symbolsLex.getString();
 857                                         }
 858                                 } else
 859                                         breakout = true;
 860                         }
 861                 } else {
 862                         breakout = true;
 863                 }
 864
 865                 // backward compatibility
 866                 if (mathPreamble == "esintoramsmath")
 867                         mathPreamble = "esint|amsmath";
 868
 869                 if (!textPreamble.empty())
 870                         if (textPreamble[0] != '\\')
 871                                 flags |= CharInfoTextFeature;
 872                 if (!mathPreamble.empty())
 873                         if (mathPreamble[0] != '\\')
 874                                 flags |= CharInfoMathFeature;
 875
 876                 CharInfo info = CharInfo(
 877                                 textCommand, mathCommand,
 878                                 textPreamble, mathPreamble,
 879                                 tipaShortcut, flags);
 880                 LYXERR(Debug::INFO, "Read unicode symbol " << symbol << " '"
 881                                                            << to_utf8(info.textCommand()) << "' '" << info.textPreamble()
 882                                                            << " '" << info.textFeature() << ' ' << info.textNoTermination()
 883                                                            << ' ' << to_utf8(info.mathCommand()) << "' '" << info.mathPreamble()
 884                                                            << "' " << info.mathFeature() << ' ' << info.mathNoTermination()
 885                            << ' ' << info.combining() << ' ' << info.force()
 886                            << ' ' << info.forceSelected());
 887
 888                 // we assume that at least one command is nonempty when using unicodesymbols
 889                 if (info.isUnicodeSymbol()) {
 890                         unicodesymbols[symbol] = info;
 891                 }
 892
 893                 if (breakout)
 894                         break;
 895         }
 896
 897         // Now read the encodings
 898         enum {
 899                 et_encoding = 1,
 900                 et_end
 901         };
 902
 903         LexerKeyword encodingTags[] = {
 904                 { "encoding", et_encoding },
 905                 { "end", et_end }
 906         };
 907
 908         Lexer lex(encodingTags);
 909         lex.setFile(encfile);
 910         lex.setContext("Encodings::read");
 911         while (lex.isOK()) {
 912                 switch (lex.lex()) {
 913                 case et_encoding:
 914                 {
 915                         lex.next();
 916                         string const name = lex.getString();
 917                         lex.next();
 918                         string const latexName = lex.getString();
 919                         lex.next();
 920                         string const guiName = lex.getString();
 921                         lex.next();
 922                         string const iconvName = lex.getString();
 923                         lex.next();
 924                         string const width = lex.getString();
 925                         bool fixedWidth = false;
 926                         bool unsafe = false;
 927                         if (width == "fixed")
 928                                 fixedWidth = true;
 929                         else if (width == "variable")
 930                                 fixedWidth = false;
 931                         else if (width == "variableunsafe") {
 932                                 fixedWidth = false;
 933                                 unsafe = true;
 934                         }
 935                         else
 936                                 lex.printError("Unknown width");
 937
 938                         lex.next();
 939                         string const p = lex.getString();
 940                         Encoding::Package package = Encoding::none;
 941                         if (p == "none")
 942                                 package = Encoding::none;
 943                         else if (p == "inputenc")
 944                                 package = Encoding::inputenc;
 945                         else if (p == "CJK")
 946                                 package = Encoding::CJK;
 947                         else if (p == "japanese")
 948                                 package = Encoding::japanese;
 949                         else
 950                                 lex.printError("Unknown package");
 951
 952                         LYXERR(Debug::INFO, "Reading encoding " << name);
 953                         encodinglist[name] = Encoding(name, latexName,
 954                                                       guiName, iconvName, fixedWidth, unsafe,
 955                                                       package);
 956
 957                         if (lex.lex() != et_end)
 958                                 lex.printError("Missing end");
 959                         break;
 960                 }
 961                 case et_end:
 962                         lex.printError("Misplaced end");
 963                         break;
 964                 case Lexer::LEX_FEOF:
 965                         break;
 966                 default:
 967                         lex.printError("Unknown tag");
 968                         break;
 969                 }
 970         }
 971
 972         // Move all information from forcedNotSelected to forcedSelected
 973         for (CharSetMap::const_iterator it1 = forcedNotSelected.begin(); it1 != forcedNotSelected.end(); ++it1) {
 974                 for (CharSetMap::iterator it2 = forcedSelected.begin(); it2 != forcedSelected.end(); ++it2) {
 975                         if (it2->first != it1->first)
 976                                 it2->second.insert(it1->second.begin(), it1->second.end());
 977                 }
 978         }
 979
 980 }
 981
 982
 983 } // namespace lyx