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/docstring.h"
  21 #include "support/gettext.h"
  22 #include "support/lstrings.h"
  23 #include "support/mutex.h"
  24 #include "support/textutils.h"
  25 #include "support/unicode.h"
  26
  27 #include <algorithm>
  28 #include <cstdint>
  29 #include <iterator>
  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 noexcept
  72 {
  73         return "Could not find LaTeX command for a character";
  74 }
  75
  76
  77 CharInfo::CharInfo(
  78                 docstring const & text_command, docstring const & math_command,
  79                 std::string const & text_preamble, std::string const & math_preamble,
  80                 std::string const & tipa_shortcut, unsigned int flags)
  81         : text_commands_({text_command}), math_commands_({math_command}),
  82           text_preamble_(text_preamble), math_preamble_(math_preamble),
  83           tipa_shortcut_(tipa_shortcut), flags_(flags)
  84 {
  85 }
  86
  87
  88 CharInfo::CharInfo(
  89                 std::vector<docstring> const & text_commands, std::vector<docstring> const & math_commands,
  90                 std::string const & text_preamble, std::string const & math_preamble,
  91                 std::string const & tipa_shortcut, unsigned int flags)
  92         : text_commands_(text_commands), math_commands_(math_commands),
  93           text_preamble_(text_preamble), math_preamble_(math_preamble),
  94           tipa_shortcut_(tipa_shortcut), flags_(flags)
  95 {
  96 }
  97
  98
  99 Encoding::Encoding(string const & n, string const & l, string const & g,
 100                    string const & i, bool f, bool u, Encoding::Package p)
 101         : name_(n), latexName_(l), guiName_(g), iconvName_(i), fixedwidth_(f),
 102           unsafe_(u), forced_(&forcedSelected[n]), package_(p)
 103 {
 104         if (n == "ascii") {
 105                 // ASCII can encode 128 code points and nothing else
 106                 start_encodable_ = 128;
 107                 complete_ = true;
 108         } else if (i == "UTF-8") {
 109                 // UTF8 can encode all UCS4 code points
 110                 start_encodable_ = max_ucs4;
 111                 complete_ = true;
 112         } else {
 113                 start_encodable_ = 0;
 114                 complete_ = false;
 115         }
 116 }
 117
 118
 119 void Encoding::init() const
 120 {
 121         // Since the the constructor is the only method which sets complete_
 122         // to false the test for complete_ is thread-safe without mutex.
 123         if (complete_)
 124                 return;
 125
 126         static Mutex mutex;
 127         Mutex::Locker lock(&mutex);
 128
 129         // We need to test again for complete_, since another thread could
 130         // have set it to true while we were waiting for the lock and we must
 131         // not modify an encoding which is already complete.
 132         if (complete_)
 133                 return;
 134
 135         // We do not make any member mutable  so that it can be easily verified
 136         // that all const methods are thread-safe: init() is the only const
 137         // method which changes complete_, encodable_ and start_encodable_, and
 138         // it uses a mutex to ensure thread-safety.
 139         CharSet & encodable = const_cast<Encoding *>(this)->encodable_;
 140         char_type & start_encodable = const_cast<Encoding *>(this)->start_encodable_;
 141
 142         start_encodable = 0;
 143         // temporarily switch off lyxerr, since we will generate iconv errors
 144         lyxerr.disable();
 145         if (fixedwidth_) {
 146                 // We do not need to check all UCS4 code points, it is enough
 147                 // if we check all 256 code points of this encoding.
 148                 for (unsigned short j = 0; j < 256; ++j) {
 149                         char const c = char(j);
 150                         vector<char_type> const ucs4 = eightbit_to_ucs4(&c, 1, iconvName_);
 151                         if (ucs4.size() != 1)
 152                                 continue;
 153                         char_type const uc = ucs4[0];
 154                         CharInfoMap::const_iterator const it = unicodesymbols.find(uc);
 155                         if (it == unicodesymbols.end())
 156                                 encodable.insert(uc);
 157                         else if (!it->second.force()) {
 158                                 if (forced_->empty() || forced_->find(uc) == forced_->end())
 159                                         encodable.insert(uc);
 160                         }
 161                 }
 162         } else {
 163                 // We do not know how many code points this encoding has, and
 164                 // they do not have a direct representation as a single byte,
 165                 // therefore we need to check all UCS4 code points.
 166                 // This is expensive!
 167                 for (char_type c = 0; c < max_ucs4; ++c) {
 168                         vector<char> const eightbit = ucs4_to_eightbit(&c, 1, iconvName_);
 169                         if (!eightbit.empty()) {
 170                                 CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 171                                 if (it == unicodesymbols.end())
 172                                         encodable.insert(c);
 173                                 else if (!it->second.force()) {
 174                                         if (forced_->empty() || forced_->find(c) == forced_->end())
 175                                                 encodable.insert(c);
 176                                 }
 177                         }
 178                 }
 179         }
 180         lyxerr.enable();
 181         CharSet::iterator it = encodable.find(start_encodable);
 182         while (it != encodable.end()) {
 183                 encodable.erase(it);
 184                 ++start_encodable;
 185                 it = encodable.find(start_encodable);
 186         }
 187         const_cast<Encoding *>(this)->complete_ = true;
 188 }
 189
 190
 191 bool Encoding::isForced(char_type c) const
 192 {
 193         if (!forced.empty() && forced.find(c) != forced.end())
 194                 return true;
 195         return !forced_->empty() && forced_->find(c) != forced_->end();
 196 }
 197
 198
 199 bool Encoding::encodable(char_type c) const
 200 {
 201         // assure the used encoding is properly initialized
 202         init();
 203         if (iconvName_ == "UTF-8" && package_ == none)
 204                 return true;
 205         // platex does not load inputenc: force conversion of supported characters
 206         if (package_ == Encoding::japanese
 207             && ((0xb7 <= c && c <= 0x05ff) // Latin-1 Supplement ... Hebrew
 208                         || (0x1d00 <= c && c <= 0x218f) // Phonetic Extensions ... Number Forms
 209                         || (0x2193 <= c && c <= 0x2aff) // Arrows ... Supplemental Mathematical Operators
 210                         || (0xfb00 <= c && c <= 0xfb4f) // Alphabetic Presentation Forms
 211                         || (0x1d400 <= c && c <= 0x1d7ff))) // Mathematical Alphanumeric Symbols
 212                 return false;
 213         if (c < start_encodable_ && !isForced(c))
 214                 return true;
 215         if (encodable_.find(c) != encodable_.end())
 216                 return true;
 217         return false;
 218 }
 219
 220
 221 pair<docstring, bool> Encoding::latexChar(char_type c) const
 222 {
 223         if (encodable(c))
 224                 return make_pair(docstring(1, c), false);
 225
 226         // c cannot (or should not) be encoded in this encoding
 227         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 228         if (it == unicodesymbols.end())
 229                 throw EncodingException(c);
 230         // at least one of mathCommand and textCommand is nonempty
 231         if (it->second.textCommand().empty())
 232                 return make_pair(
 233                                 "\\ensuremath{" + it->second.mathCommand() + '}', false);
 234         return make_pair(it->second.textCommand(), !it->second.textNoTermination());
 235 }
 236
 237
 238 pair<docstring, docstring> Encoding::latexString(docstring const & input, bool dryrun) const
 239 {
 240         docstring result;
 241         docstring uncodable;
 242         bool terminate = false;
 243         for (char_type const c : input) {
 244                 try {
 245                         pair<docstring, bool> latex_char = latexChar(c);
 246                         docstring const latex = latex_char.first;
 247                         if (terminate && !prefixIs(latex, '\\')
 248                             && !prefixIs(latex, '{')
 249                             && !prefixIs(latex, '}')) {
 250                                         // Prevent eating of a following
 251                                         // space or command corruption by
 252                                         // following characters
 253                                         if (latex == " ")
 254                                                 result += "{}";
 255                                         else
 256                                                 result += " ";
 257                                 }
 258                         result += latex;
 259                         terminate = latex_char.second;
 260                 } catch (EncodingException & /* e */) {
 261                         LYXERR0("Uncodable character <" << docstring(1, c)
 262                                         << "> in latexString!");
 263                         if (dryrun) {
 264                                 result += "<" + _("LyX Warning: ")
 265                                            + _("uncodable character") + " '";
 266                                 result += docstring(1, c);
 267                                 result += "'>";
 268                         } else
 269                                 uncodable += c;
 270                 }
 271         }
 272         return make_pair(result, uncodable);
 273 }
 274
 275
 276 vector<char_type> Encoding::symbolsList() const
 277 {
 278         // assure the used encoding is properly initialized
 279         init();
 280
 281         // first all those below start_encodable_
 282         vector<char_type> symbols;
 283         for (char_type c = 0; c < start_encodable_; ++c)
 284                 symbols.push_back(c);
 285         // add all encodable characters
 286         copy(encodable_.begin(), encodable_.end(), back_inserter(symbols));
 287         // now the ones from the unicodesymbols file that are not already there
 288         for (auto const & elem : unicodesymbols) {
 289                 if (find(symbols.begin(), symbols.end(), elem.first) == symbols.end())
 290                         symbols.push_back(elem.first);
 291         }
 292         // finally, sort the vector
 293         sort(symbols.begin(), symbols.end());
 294         return symbols;
 295 }
 296
 297
 298 bool Encodings::latexMathChar(char_type c, bool mathmode,
 299                         Encoding const * encoding, docstring & command,
 300                         bool & needsTermination)
 301 {
 302         command = empty_docstring();
 303         if (encoding)
 304                 if (encoding->encodable(c))
 305                         command = docstring(1, c);
 306         needsTermination = false;
 307
 308         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 309         if (it == unicodesymbols.end()) {
 310                 if (!encoding || command.empty())
 311                         throw EncodingException(c);
 312                 if (mathmode)
 313                         addMathSym(c);
 314                 return false;
 315         }
 316         // at least one of mathCommand and textCommand is nonempty
 317         bool use_math = (mathmode && !it->second.mathCommand().empty()) ||
 318                         (!mathmode && it->second.textCommand().empty());
 319         if (use_math) {
 320                 command = it->second.mathCommand();
 321                 needsTermination = !it->second.mathNoTermination();
 322                 addMathCmd(c);
 323         } else {
 324                 if (!encoding || command.empty()) {
 325                         command = it->second.textCommand();
 326                         needsTermination = !it->second.textNoTermination();
 327                 }
 328                 if (mathmode)
 329                         addMathSym(c);
 330                 else
 331                         addTextCmd(c);
 332         }
 333         return use_math;
 334 }
 335
 336
 337 char_type Encodings::fromLaTeXCommand(docstring const & cmd, int cmdtype,
 338                 bool & combining, bool & needsTermination, set<string> * req)
 339 {
 340         CharInfoMap::const_iterator const end = unicodesymbols.end();
 341         CharInfoMap::const_iterator it = unicodesymbols.begin();
 342         for (combining = false; it != end; ++it) {
 343                 if (it->second.deprecated())
 344                         continue;
 345
 346                 if (cmdtype & MATH_CMD) {
 347                         for (const docstring& math : it->second.mathCommands()) {
 348                                 if ((cmdtype & MATH_CMD) && math == cmd) {
 349                                         combining = it->second.combining();
 350                                         needsTermination = !it->second.mathNoTermination();
 351                                         if (req && it->second.mathFeature() &&
 352                                             !it->second.mathPreamble().empty())
 353                                                 req->insert(it->second.mathPreamble());
 354                                         return it->first;
 355                                 }
 356                         }
 357                 }
 358
 359                 if (cmdtype & TEXT_CMD) {
 360                         for (const docstring& text : it->second.textCommands()) {
 361                                 if (text == cmd) {
 362                                         combining = it->second.combining();
 363                                         needsTermination = !it->second.textNoTermination();
 364                                         if (req && it->second.textFeature() &&
 365                                             !it->second.textPreamble().empty())
 366                                                 req->insert(it->second.textPreamble());
 367                                         return it->first;
 368                                 }
 369                         }
 370                 }
 371         }
 372         needsTermination = false;
 373         return 0;
 374 }
 375
 376
 377 docstring Encodings::fromLaTeXCommand(docstring const & cmd, int cmdtype,
 378                 bool & needsTermination, docstring & rem, set<string> * req)
 379 {
 380         needsTermination = false;
 381         rem = empty_docstring();
 382         bool const mathmode = cmdtype & MATH_CMD;
 383         bool const textmode = cmdtype & TEXT_CMD;
 384
 385         // Easy case: the command is a complete entry of unicodesymbols.
 386         for (const auto & unicodeSymbol : unicodesymbols) {
 387                 if (mathmode) {
 388                         for (const auto & command : unicodeSymbol.second.mathCommands()) {
 389                                 if (command == cmd) {
 390                                         docstring value;
 391                                         value += unicodeSymbol.first;
 392                                         return value;
 393                                 }
 394                         }
 395                 }
 396                 if (textmode) {
 397                         for (const auto & command : unicodeSymbol.second.textCommands()) {
 398                                 if (command == cmd) {
 399                                         docstring value;
 400                                         value += unicodeSymbol.first;
 401                                         return value;
 402                                 }
 403                         }
 404                 }
 405         }
 406
 407         // Otherwise, try to map as many commands as possible, matching prefixes of the command.
 408         docstring symbols;
 409         size_t const cmdend = cmd.size();
 410         size_t prefix = 0;
 411         CharInfoMap::const_iterator const uniend = unicodesymbols.end();
 412         for (size_t i = 0, j = 0; j < cmdend; ++j) {
 413                 // Also get the char after a backslash
 414                 if (j + 1 < cmdend && cmd[j] == '\\') {
 415                         ++j;
 416                         prefix = 1;
 417                         // Detect things like \=*{e} as well
 418                         if (j + 3 < cmdend && cmd[j+1] == '*' &&
 419                             cmd[j+2] == '{') {
 420                                 ++j;
 421                                 prefix = 2;
 422                         }
 423                 }
 424                 // position of the last character before a possible macro
 425                 // argument
 426                 size_t m = j;
 427                 // If a macro argument follows, get it, too
 428                 // Do it here only for single character commands. Other
 429                 // combining commands need this too, but they are handled in
 430                 // the loop below for performance reasons.
 431                 if (j + 1 < cmdend && cmd[j + 1] == '{') {
 432                         size_t k = j + 1;
 433                         int count = 1;
 434                         while (k < cmdend && count) {
 435                                 k = cmd.find_first_of(from_ascii("{}"), k + 1);
 436                                 // braces may not be balanced
 437                                 if (k == docstring::npos)
 438                                         break;
 439                                 if (cmd[k] == '{')
 440                                         ++count;
 441                                 else
 442                                         --count;
 443                         }
 444                         if (k != docstring::npos)
 445                                 j = k;
 446                 } else if (m + 1 < cmdend && isAlphaASCII(cmd[m])) {
 447                         while (m + 2 < cmdend && isAlphaASCII(cmd[m+1]))
 448                                 m++;
 449                 }
 450                 // Start with this substring and try augmenting it when it is
 451                 // the prefix of some command in the unicodesymbols file
 452                 docstring subcmd = cmd.substr(i, j - i + 1);
 453
 454                 CharInfoMap::const_iterator it = unicodesymbols.begin();
 455                 // First part of subcmd which might be a combining character
 456                 docstring combcmd = (m == j) ? docstring() : cmd.substr(i, m - i + 1);
 457                 // The combining character of combcmd if it exists
 458                 CharInfoMap::const_iterator combining = uniend;
 459                 size_t unicmd_size = 0;
 460                 char_type c = 0;
 461                 for (; it != uniend; ++it) {
 462                         if (it->second.deprecated())
 463                                 continue;
 464                         docstring const math = mathmode ? it->second.mathCommand()
 465                                                         : docstring();
 466                         docstring const text = textmode ? it->second.textCommand()
 467                                                         : docstring();
 468                         if (!combcmd.empty() && it->second.combining() &&
 469                             (math == combcmd || text == combcmd))
 470                                 combining = it;
 471                         size_t cur_size = max(math.size(), text.size());
 472                         // The current math or text unicode command cannot
 473                         // match, or we already matched a longer one
 474                         if (cur_size < subcmd.size() || cur_size <= unicmd_size)
 475                                 continue;
 476
 477                         docstring tmp = subcmd;
 478                         size_t k = j;
 479                         while (prefixIs(math, tmp) || prefixIs(text, tmp)) {
 480                                 ++k;
 481                                 if (k >= cmdend || cur_size <= tmp.size())
 482                                         break;
 483                                 tmp += cmd[k];
 484                         }
 485                         // No match
 486                         if (k == j)
 487                                 continue;
 488
 489                         // The last added char caused a mismatch, because
 490                         // we didn't exhaust the chars in cmd and didn't
 491                         // exceed the maximum size of the current unicmd
 492                         if (k < cmdend && cur_size > tmp.size())
 493                                 tmp.resize(tmp.size() - 1);
 494
 495                         // If this is an exact match, we found a (longer)
 496                         // matching entry in the unicodesymbols file.
 497                         if (math != tmp && text != tmp)
 498                                 continue;
 499                         // If we found a combining command, we need to append
 500                         // the macro argument if this has not been done above.
 501                         if (tmp == combcmd && combining != uniend &&
 502                             k < cmdend && cmd[k] == '{') {
 503                                 size_t l = k;
 504                                 int count = 1;
 505                                 while (l < cmdend && count) {
 506                                         l = cmd.find_first_of(from_ascii("{}"), l + 1);
 507                                         // braces may not be balanced
 508                                         if (l == docstring::npos)
 509                                                 break;
 510                                         if (cmd[l] == '{')
 511                                                 ++count;
 512                                         else
 513                                                 --count;
 514                                 }
 515                                 if (l != docstring::npos) {
 516                                         j = l;
 517                                         subcmd = cmd.substr(i, j - i + 1);
 518                                 }
 519                         }
 520                         // If the entry doesn't start with '\', we take note
 521                         // of the match and continue (this is not a ultimate
 522                         // acceptance, as some other entry may match a longer
 523                         // portion of the cmd string). However, if the entry
 524                         // does start with '\', we accept the match only if
 525                         // this is a valid macro, i.e., either it is a single
 526                         // (nonletter) char macro, or nothing else follows,
 527                         // or what follows is a nonletter char, or the last
 528                         // character is a }.
 529                         else if (tmp[0] != '\\'
 530                                    || (tmp.size() == prefix + 1 &&
 531                                        !isAlphaASCII(tmp[1]) &&
 532                                        (prefix == 1 || !isAlphaASCII(tmp[2])))
 533                                    || k == cmdend
 534                                    || !isAlphaASCII(cmd[k])
 535                                    || tmp[tmp.size() - 1] == '}'
 536                                  ) {
 537                                 c = it->first;
 538                                 j = k - 1;
 539                                 i = j + 1;
 540                                 unicmd_size = cur_size;
 541                                 if (math == tmp)
 542                                         needsTermination = !it->second.mathNoTermination();
 543                                 else
 544                                         needsTermination = !it->second.textNoTermination();
 545                                 if (req) {
 546                                         if (math == tmp && it->second.mathFeature() &&
 547                                             !it->second.mathPreamble().empty())
 548                                                 req->insert(it->second.mathPreamble());
 549                                         if (text == tmp && it->second.textFeature() &&
 550                                             !it->second.textPreamble().empty())
 551                                                 req->insert(it->second.textPreamble());
 552                                 }
 553                         }
 554                 }
 555                 if (unicmd_size)
 556                         symbols += c;
 557                 else if (combining != uniend &&
 558                          prefixIs(subcmd, combcmd + '{')) {
 559                         // We know that subcmd starts with combcmd and
 560                         // contains an argument in braces.
 561                         docstring const arg = subcmd.substr(
 562                                 combcmd.length() + 1,
 563                                 subcmd.length() - combcmd.length() - 2);
 564                         // If arg is a single character we can construct a
 565                         // combining sequence.
 566                         char_type a;
 567                         bool argcomb = false;
 568                         if (arg.size() == 1 && isAlnumASCII(arg[0]))
 569                                 a = arg[0];
 570                         else {
 571                                 // Use the version of fromLaTeXCommand() that
 572                                 // parses only one command, since we cannot
 573                                 // use more than one character.
 574                                 bool dummy = false;
 575                                 set<string> r;
 576                                 a = fromLaTeXCommand(arg, cmdtype, argcomb,
 577                                                      dummy, &r);
 578                                 if (a && req && !argcomb)
 579                                         req->insert(r.begin(), r.end());
 580                         }
 581                         if (a && !argcomb) {
 582                                 // In unicode the combining character comes
 583                                 // after its base
 584                                 symbols += a;
 585                                 symbols += combining->first;
 586                                 i = j + 1;
 587                                 unicmd_size = 2;
 588                         }
 589                 }
 590                 if (j + 1 == cmdend && !unicmd_size) {
 591                         // No luck. Return what remains
 592                         rem = cmd.substr(i);
 593                         if (needsTermination && !rem.empty()) {
 594                                 if (rem.substr(0, 2) == "{}") {
 595                                         rem = rem.substr(2);
 596                                         needsTermination = false;
 597                                 } else if (rem[0] == ' ') {
 598                                         needsTermination = false;
 599                                         // LaTeX would swallow all spaces
 600                                         rem = ltrim(rem);
 601                                 }
 602                         }
 603                 }
 604         }
 605         return symbols;
 606 }
 607
 608
 609 CharInfo const & Encodings::unicodeCharInfo(char_type c)
 610 {
 611         static CharInfo empty;
 612         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 613         return it != unicodesymbols.end() ? it->second : empty;
 614 }
 615
 616
 617 bool Encodings::isCombiningChar(char_type c)
 618 {
 619         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 620         if (it != unicodesymbols.end())
 621                 return it->second.combining();
 622         return false;
 623 }
 624
 625
 626 string const Encodings::TIPAShortcut(char_type c)
 627 {
 628         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 629         if (it != unicodesymbols.end())
 630                 return it->second.tipaShortcut();
 631         return string();
 632 }
 633
 634
 635 string const Encodings::isKnownScriptChar(char_type const c)
 636 {
 637         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 638
 639         if (it == unicodesymbols.end())
 640                 return string();
 641         // FIXME: parse complex textPreamble (may be list or alternatives,
 642         //                e.g., "subscript,textgreek" or "textcomp|textgreek")
 643         if (it->second.textPreamble() == "textgreek"
 644                 || it->second.textPreamble() == "textcyrillic")
 645                 return it->second.textPreamble();
 646         return string();
 647 }
 648
 649
 650 bool Encodings::fontencSupportsScript(string const & fontenc, string const & script)
 651 {
 652         if (script == "textgreek")
 653                 return (fontenc == "LGR" || fontenc == "TU");
 654         if (script == "textcyrillic")
 655                 return (fontenc == "T2A" || fontenc == "T2B" || fontenc == "T2C"
 656                                 || fontenc == "X2" || fontenc == "TU");
 657         return false;
 658 }
 659
 660
 661 bool Encodings::isMathAlpha(char_type c)
 662 {
 663         return mathalpha.count(c);
 664 }
 665
 666
 667 bool Encodings::isUnicodeTextOnly(char_type c)
 668 {
 669         if (isASCII(c) || isMathAlpha(c))
 670                 return false;
 671
 672         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 673         return it == unicodesymbols.end() || it->second.mathCommand().empty();
 674 }
 675
 676
 677 Encoding const *
 678 Encodings::fromLyXName(string const & name, bool allowUnsafe) const
 679 {
 680         EncodingList::const_iterator const it = encodinglist.find(name);
 681         if (it == encodinglist.end())
 682                 return nullptr;
 683         if (!allowUnsafe && it->second.unsafe())
 684                 return nullptr;
 685         return &it->second;
 686 }
 687
 688
 689 Encoding const *
 690 Encodings::fromLaTeXName(string const & n, int p, bool allowUnsafe) const
 691 {
 692         string name = n;
 693         // FIXME: if we have to test for too many of these synonyms,
 694         // we should instead extend the format of lib/encodings
 695         if (n == "ansinew")
 696                 name = "cp1252";
 697
 698         // We don't use find_if because it makes copies of the pairs in
 699         // the map.
 700         // This linear search is OK since we don't have many encodings.
 701         // Users could even optimize it by putting the encodings they use
 702         // most at the top of lib/encodings.
 703         EncodingList::const_iterator const end = encodinglist.end();
 704         for (EncodingList::const_iterator it = encodinglist.begin(); it != end; ++it)
 705                 if ((it->second.latexName() == name) && (it->second.package() & p)
 706                                 && (!it->second.unsafe() || allowUnsafe))
 707                         return &it->second;
 708         return nullptr;
 709 }
 710
 711
 712 Encoding const *
 713 Encodings::fromIconvName(string const & n, int p, bool allowUnsafe) const
 714 {
 715         EncodingList::const_iterator const end = encodinglist.end();
 716         for (EncodingList::const_iterator it = encodinglist.begin(); it != end; ++it)
 717                 if ((it->second.iconvName() == n) && (it->second.package() & p)
 718                                 && (!it->second.unsafe() || allowUnsafe))
 719                         return &it->second;
 720         return nullptr;
 721 }
 722
 723
 724 Encodings::Encodings()
 725 {}
 726
 727
 728 void Encodings::read(FileName const & encfile, FileName const & symbolsfile)
 729 {
 730         // We must read the symbolsfile first, because the Encoding
 731         // constructor depends on it.
 732         CharSetMap forcedNotSelected;
 733         Lexer symbolsLex;
 734         symbolsLex.setFile(symbolsfile);
 735         bool getNextToken = true;
 736         while (symbolsLex.isOK()) {
 737                 char_type symbol;
 738
 739                 if (getNextToken) {
 740                         if (!symbolsLex.next(true))
 741                                 break;
 742                 } else
 743                         getNextToken = true;
 744
 745                 istringstream is(symbolsLex.getString());
 746                 // reading symbol directly does not work if
 747                 // char_type == wchar_t.
 748                 uint32_t tmp;
 749                 if(!(is >> hex >> tmp))
 750                         break;
 751                 symbol = tmp;
 752
 753                 // Special case: more than one entry for one character (to add other LaTeX commands).
 754                 if (unicodesymbols.find(symbol) != unicodesymbols.end()) {
 755                         if (!symbolsLex.next(true))
 756                                 break;
 757                         docstring textCommand = symbolsLex.getDocString();
 758                         if (!symbolsLex.next(true))
 759                                 break;
 760                         string mathCommand = symbolsLex.getString();
 761
 762                         if (!textCommand.empty())
 763                                 unicodesymbols.at(symbol).addTextCommand(textCommand);
 764                         if (!mathCommand.empty())
 765                                 unicodesymbols.at(symbol).addMathCommand(textCommand);
 766
 767                         continue;
 768                 }
 769
 770                 // If the symbol is not the same as the previous entry, consider it is a totally new symbol.
 771                 if (!symbolsLex.next(true))
 772                         break;
 773                 docstring textCommand = symbolsLex.getDocString();
 774                 if (!symbolsLex.next(true))
 775                         break;
 776                 string textPreamble = symbolsLex.getString();
 777                 if (!symbolsLex.next(true))
 778                         break;
 779                 string sflags = symbolsLex.getString();
 780
 781                 string tipaShortcut;
 782                 int flags = 0;
 783
 784                 if (suffixIs(textCommand, '}'))
 785                         flags |= CharInfoTextNoTermination;
 786                 while (!sflags.empty()) {
 787                         string flag;
 788                         sflags = split(sflags, flag, ',');
 789                         if (flag == "combining") {
 790                                 flags |= CharInfoCombining;
 791                         } else if (flag == "force") {
 792                                 flags |= CharInfoForce;
 793                                 forced.insert(symbol);
 794                         } else if (prefixIs(flag, "force=")) {
 795                                 vector<string> encs =
 796                                         getVectorFromString(flag.substr(6), ";");
 797                                 for (auto const & enc : encs)
 798                                         forcedSelected[enc].insert(symbol);
 799                                 flags |= CharInfoForceSelected;
 800                         } else if (prefixIs(flag, "force!=")) {
 801                                 vector<string> encs =
 802                                         getVectorFromString(flag.substr(7), ";");
 803                                 for (auto const & enc : encs)
 804                                         forcedNotSelected[enc].insert(symbol);
 805                                 flags |= CharInfoForceSelected;
 806                         } else if (flag == "mathalpha") {
 807                                 mathalpha.insert(symbol);
 808                         } else if (flag == "notermination=text") {
 809                                 flags |= CharInfoTextNoTermination;
 810                         } else if (flag == "notermination=math") {
 811                                 flags |= CharInfoMathNoTermination;
 812                         } else if (flag == "notermination=both") {
 813                                 flags |= CharInfoTextNoTermination;
 814                                 flags |= CharInfoMathNoTermination;
 815                         } else if (flag == "notermination=none") {
 816                                 flags &= ~CharInfoTextNoTermination;
 817                                 flags &= ~CharInfoMathNoTermination;
 818                         } else if (contains(flag, "tipashortcut=")) {
 819                                 tipaShortcut = split(flag, '=');
 820                         } else if (flag == "deprecated") {
 821                                 flags |= CharInfoDeprecated;
 822                         } else {
 823                                 lyxerr << "Ignoring unknown flag `" << flag
 824                                        << "' for symbol `0x"
 825                                        << hex << symbol << dec
 826                                        << "'." << endl;
 827                         }
 828                 }
 829                 // mathCommand and mathPreamble have been added for 1.6.0.
 830                 // make them optional so that old files still work.
 831                 int const lineNo = symbolsLex.lineNumber();
 832                 bool breakout = false;
 833                 docstring mathCommand;
 834                 string mathPreamble;
 835                 if (symbolsLex.next(true)) {
 836                         if (symbolsLex.lineNumber() != lineNo) {
 837                                 // line in old format without mathCommand and mathPreamble
 838                                 getNextToken = false;
 839                         } else {
 840                                 mathCommand = symbolsLex.getDocString();
 841                                 if (suffixIs(mathCommand, '}'))
 842                                         flags |= CharInfoMathNoTermination;
 843                                 if (symbolsLex.next(true)) {
 844                                         if (symbolsLex.lineNumber() != lineNo) {
 845                                                 // line in new format with mathCommand only
 846                                                 getNextToken = false;
 847                                         } else {
 848                                                 // line in new format with mathCommand and mathPreamble
 849                                                 mathPreamble = symbolsLex.getString();
 850                                         }
 851                                 } else
 852                                         breakout = true;
 853                         }
 854                 } else {
 855                         breakout = true;
 856                 }
 857
 858                 // backward compatibility
 859                 if (mathPreamble == "esintoramsmath")
 860                         mathPreamble = "esint|amsmath";
 861
 862                 if (!textPreamble.empty())
 863                         if (textPreamble[0] != '\\')
 864                                 flags |= CharInfoTextFeature;
 865                 if (!mathPreamble.empty())
 866                         if (mathPreamble[0] != '\\')
 867                                 flags |= CharInfoMathFeature;
 868
 869                 CharInfo info = CharInfo(
 870                                 textCommand, mathCommand,
 871                                 textPreamble, mathPreamble,
 872                                 tipaShortcut, flags);
 873                 LYXERR(Debug::INFO, "Read unicode symbol " << symbol << " '"
 874                                                            << to_utf8(info.textCommand()) << "' '" << info.textPreamble()
 875                                                            << " '" << info.textFeature() << ' ' << info.textNoTermination()
 876                                                            << ' ' << to_utf8(info.mathCommand()) << "' '" << info.mathPreamble()
 877                                                            << "' " << info.mathFeature() << ' ' << info.mathNoTermination()
 878                            << ' ' << info.combining() << ' ' << info.force()
 879                            << ' ' << info.forceSelected());
 880
 881                 // we assume that at least one command is nonempty when using unicodesymbols
 882                 if (info.isUnicodeSymbol()) {
 883                         unicodesymbols[symbol] = info;
 884                 }
 885
 886                 if (breakout)
 887                         break;
 888         }
 889
 890         // Now read the encodings
 891         enum {
 892                 et_encoding = 1,
 893                 et_end
 894         };
 895
 896         LexerKeyword encodingTags[] = {
 897                 { "encoding", et_encoding },
 898                 { "end", et_end }
 899         };
 900
 901         Lexer lex(encodingTags);
 902         lex.setFile(encfile);
 903         lex.setContext("Encodings::read");
 904         while (lex.isOK()) {
 905                 switch (lex.lex()) {
 906                 case et_encoding:
 907                 {
 908                         lex.next();
 909                         string const name = lex.getString();
 910                         lex.next();
 911                         string const latexName = lex.getString();
 912                         lex.next();
 913                         string const guiName = lex.getString();
 914                         lex.next();
 915                         string const iconvName = lex.getString();
 916                         lex.next();
 917                         string const width = lex.getString();
 918                         bool fixedWidth = false;
 919                         bool unsafe = false;
 920                         if (width == "fixed")
 921                                 fixedWidth = true;
 922                         else if (width == "variable")
 923                                 fixedWidth = false;
 924                         else if (width == "variableunsafe") {
 925                                 fixedWidth = false;
 926                                 unsafe = true;
 927                         }
 928                         else
 929                                 lex.printError("Unknown width");
 930
 931                         lex.next();
 932                         string const p = lex.getString();
 933                         Encoding::Package package = Encoding::none;
 934                         if (p == "none")
 935                                 package = Encoding::none;
 936                         else if (p == "inputenc")
 937                                 package = Encoding::inputenc;
 938                         else if (p == "CJK")
 939                                 package = Encoding::CJK;
 940                         else if (p == "japanese")
 941                                 package = Encoding::japanese;
 942                         else
 943                                 lex.printError("Unknown package");
 944
 945                         LYXERR(Debug::INFO, "Reading encoding " << name);
 946                         encodinglist[name] = Encoding(name, latexName,
 947                                                       guiName, iconvName, fixedWidth, unsafe,
 948                                                       package);
 949
 950                         if (lex.lex() != et_end)
 951                                 lex.printError("Missing end");
 952                         break;
 953                 }
 954                 case et_end:
 955                         lex.printError("Misplaced end");
 956                         break;
 957                 case Lexer::LEX_FEOF:
 958                         break;
 959                 default:
 960                         lex.printError("Unknown tag");
 961                         break;
 962                 }
 963         }
 964
 965         // Move all information from forcedNotSelected to forcedSelected
 966         for (CharSetMap::const_iterator it1 = forcedNotSelected.begin(); it1 != forcedNotSelected.end(); ++it1) {
 967                 for (CharSetMap::iterator it2 = forcedSelected.begin(); it2 != forcedSelected.end(); ++it2) {
 968                         if (it2->first != it1->first)
 969                                 it2->second.insert(it1->second.begin(), it1->second.end());
 970                 }
 971         }
 972
 973 }
 974
 975
 976 } // namespace lyx