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         docstring symbols;
 385         size_t const cmdend = cmd.size();
 386         size_t prefix = 0;
 387         CharInfoMap::const_iterator const uniend = unicodesymbols.end();
 388         for (size_t i = 0, j = 0; j < cmdend; ++j) {
 389                 // Also get the char after a backslash
 390                 if (j + 1 < cmdend && cmd[j] == '\\') {
 391                         ++j;
 392                         prefix = 1;
 393                         // Detect things like \=*{e} as well
 394                         if (j + 3 < cmdend && cmd[j+1] == '*' &&
 395                             cmd[j+2] == '{') {
 396                                 ++j;
 397                                 prefix = 2;
 398                         }
 399                 }
 400                 // position of the last character before a possible macro
 401                 // argument
 402                 size_t m = j;
 403                 // If a macro argument follows, get it, too
 404                 // Do it here only for single character commands. Other
 405                 // combining commands need this too, but they are handled in
 406                 // the loop below for performance reasons.
 407                 if (j + 1 < cmdend && cmd[j + 1] == '{') {
 408                         size_t k = j + 1;
 409                         int count = 1;
 410                         while (k < cmdend && count) {
 411                                 k = cmd.find_first_of(from_ascii("{}"), k + 1);
 412                                 // braces may not be balanced
 413                                 if (k == docstring::npos)
 414                                         break;
 415                                 if (cmd[k] == '{')
 416                                         ++count;
 417                                 else
 418                                         --count;
 419                         }
 420                         if (k != docstring::npos)
 421                                 j = k;
 422                 } else if (m + 1 < cmdend && isAlphaASCII(cmd[m])) {
 423                         while (m + 2 < cmdend && isAlphaASCII(cmd[m+1]))
 424                                 m++;
 425                 }
 426                 // Start with this substring and try augmenting it when it is
 427                 // the prefix of some command in the unicodesymbols file
 428                 docstring subcmd = cmd.substr(i, j - i + 1);
 429
 430                 CharInfoMap::const_iterator it = unicodesymbols.begin();
 431                 // First part of subcmd which might be a combining character
 432                 docstring combcmd = (m == j) ? docstring() : cmd.substr(i, m - i + 1);
 433                 // The combining character of combcmd if it exists
 434                 CharInfoMap::const_iterator combining = uniend;
 435                 size_t unicmd_size = 0;
 436                 char_type c = 0;
 437                 for (; it != uniend; ++it) {
 438                         if (it->second.deprecated())
 439                                 continue;
 440                         docstring const math = mathmode ? it->second.mathCommand()
 441                                                         : docstring();
 442                         docstring const text = textmode ? it->second.textCommand()
 443                                                         : docstring();
 444                         if (!combcmd.empty() && it->second.combining() &&
 445                             (math == combcmd || text == combcmd))
 446                                 combining = it;
 447                         size_t cur_size = max(math.size(), text.size());
 448                         // The current math or text unicode command cannot
 449                         // match, or we already matched a longer one
 450                         if (cur_size < subcmd.size() || cur_size <= unicmd_size)
 451                                 continue;
 452
 453                         docstring tmp = subcmd;
 454                         size_t k = j;
 455                         while (prefixIs(math, tmp) || prefixIs(text, tmp)) {
 456                                 ++k;
 457                                 if (k >= cmdend || cur_size <= tmp.size())
 458                                         break;
 459                                 tmp += cmd[k];
 460                         }
 461                         // No match
 462                         if (k == j)
 463                                 continue;
 464
 465                         // The last added char caused a mismatch, because
 466                         // we didn't exhaust the chars in cmd and didn't
 467                         // exceed the maximum size of the current unicmd
 468                         if (k < cmdend && cur_size > tmp.size())
 469                                 tmp.resize(tmp.size() - 1);
 470
 471                         // If this is an exact match, we found a (longer)
 472                         // matching entry in the unicodesymbols file.
 473                         if (math != tmp && text != tmp)
 474                                 continue;
 475                         // If we found a combining command, we need to append
 476                         // the macro argument if this has not been done above.
 477                         if (tmp == combcmd && combining != uniend &&
 478                             k < cmdend && cmd[k] == '{') {
 479                                 size_t l = k;
 480                                 int count = 1;
 481                                 while (l < cmdend && count) {
 482                                         l = cmd.find_first_of(from_ascii("{}"), l + 1);
 483                                         // braces may not be balanced
 484                                         if (l == docstring::npos)
 485                                                 break;
 486                                         if (cmd[l] == '{')
 487                                                 ++count;
 488                                         else
 489                                                 --count;
 490                                 }
 491                                 if (l != docstring::npos) {
 492                                         j = l;
 493                                         subcmd = cmd.substr(i, j - i + 1);
 494                                 }
 495                         }
 496                         // If the entry doesn't start with '\', we take note
 497                         // of the match and continue (this is not a ultimate
 498                         // acceptance, as some other entry may match a longer
 499                         // portion of the cmd string). However, if the entry
 500                         // does start with '\', we accept the match only if
 501                         // this is a valid macro, i.e., either it is a single
 502                         // (nonletter) char macro, or nothing else follows,
 503                         // or what follows is a nonletter char, or the last
 504                         // character is a }.
 505                         else if (tmp[0] != '\\'
 506                                    || (tmp.size() == prefix + 1 &&
 507                                        !isAlphaASCII(tmp[1]) &&
 508                                        (prefix == 1 || !isAlphaASCII(tmp[2])))
 509                                    || k == cmdend
 510                                    || !isAlphaASCII(cmd[k])
 511                                    || tmp[tmp.size() - 1] == '}'
 512                                  ) {
 513                                 c = it->first;
 514                                 j = k - 1;
 515                                 i = j + 1;
 516                                 unicmd_size = cur_size;
 517                                 if (math == tmp)
 518                                         needsTermination = !it->second.mathNoTermination();
 519                                 else
 520                                         needsTermination = !it->second.textNoTermination();
 521                                 if (req) {
 522                                         if (math == tmp && it->second.mathFeature() &&
 523                                             !it->second.mathPreamble().empty())
 524                                                 req->insert(it->second.mathPreamble());
 525                                         if (text == tmp && it->second.textFeature() &&
 526                                             !it->second.textPreamble().empty())
 527                                                 req->insert(it->second.textPreamble());
 528                                 }
 529                         }
 530                 }
 531                 if (unicmd_size)
 532                         symbols += c;
 533                 else if (combining != uniend &&
 534                          prefixIs(subcmd, combcmd + '{')) {
 535                         // We know that subcmd starts with combcmd and
 536                         // contains an argument in braces.
 537                         docstring const arg = subcmd.substr(
 538                                 combcmd.length() + 1,
 539                                 subcmd.length() - combcmd.length() - 2);
 540                         // If arg is a single character we can construct a
 541                         // combining sequence.
 542                         char_type a;
 543                         bool argcomb = false;
 544                         if (arg.size() == 1 && isAlnumASCII(arg[0]))
 545                                 a = arg[0];
 546                         else {
 547                                 // Use the version of fromLaTeXCommand() that
 548                                 // parses only one command, since we cannot
 549                                 // use more than one character.
 550                                 bool dummy = false;
 551                                 set<string> r;
 552                                 a = fromLaTeXCommand(arg, cmdtype, argcomb,
 553                                                      dummy, &r);
 554                                 if (a && req && !argcomb)
 555                                         req->insert(r.begin(), r.end());
 556                         }
 557                         if (a && !argcomb) {
 558                                 // In unicode the combining character comes
 559                                 // after its base
 560                                 symbols += a;
 561                                 symbols += combining->first;
 562                                 i = j + 1;
 563                                 unicmd_size = 2;
 564                         }
 565                 }
 566                 if (j + 1 == cmdend && !unicmd_size) {
 567                         // No luck. Return what remains
 568                         rem = cmd.substr(i);
 569                         if (needsTermination && !rem.empty()) {
 570                                 if (rem.substr(0, 2) == "{}") {
 571                                         rem = rem.substr(2);
 572                                         needsTermination = false;
 573                                 } else if (rem[0] == ' ') {
 574                                         needsTermination = false;
 575                                         // LaTeX would swallow all spaces
 576                                         rem = ltrim(rem);
 577                                 }
 578                         }
 579                 }
 580         }
 581         return symbols;
 582 }
 583
 584
 585 CharInfo const & Encodings::unicodeCharInfo(char_type c)
 586 {
 587         static CharInfo empty;
 588         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 589         return it != unicodesymbols.end() ? it->second : empty;
 590 }
 591
 592
 593 bool Encodings::isCombiningChar(char_type c)
 594 {
 595         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 596         if (it != unicodesymbols.end())
 597                 return it->second.combining();
 598         return false;
 599 }
 600
 601
 602 string const Encodings::TIPAShortcut(char_type c)
 603 {
 604         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 605         if (it != unicodesymbols.end())
 606                 return it->second.tipaShortcut();
 607         return string();
 608 }
 609
 610
 611 string const Encodings::isKnownScriptChar(char_type const c)
 612 {
 613         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 614
 615         if (it == unicodesymbols.end())
 616                 return string();
 617         // FIXME: parse complex textPreamble (may be list or alternatives,
 618         //                e.g., "subscript,textgreek" or "textcomp|textgreek")
 619         if (it->second.textPreamble() == "textgreek"
 620                 || it->second.textPreamble() == "textcyrillic")
 621                 return it->second.textPreamble();
 622         return string();
 623 }
 624
 625
 626 bool Encodings::fontencSupportsScript(string const & fontenc, string const & script)
 627 {
 628         if (script == "textgreek")
 629                 return (fontenc == "LGR" || fontenc == "TU");
 630         if (script == "textcyrillic")
 631                 return (fontenc == "T2A" || fontenc == "T2B" || fontenc == "T2C"
 632                                 || fontenc == "X2" || fontenc == "TU");
 633         return false;
 634 }
 635
 636
 637 bool Encodings::isMathAlpha(char_type c)
 638 {
 639         return mathalpha.count(c);
 640 }
 641
 642
 643 bool Encodings::isUnicodeTextOnly(char_type c)
 644 {
 645         if (isASCII(c) || isMathAlpha(c))
 646                 return false;
 647
 648         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 649         return it == unicodesymbols.end() || it->second.mathCommand().empty();
 650 }
 651
 652
 653 Encoding const *
 654 Encodings::fromLyXName(string const & name, bool allowUnsafe) const
 655 {
 656         EncodingList::const_iterator const it = encodinglist.find(name);
 657         if (it == encodinglist.end())
 658                 return nullptr;
 659         if (!allowUnsafe && it->second.unsafe())
 660                 return nullptr;
 661         return &it->second;
 662 }
 663
 664
 665 Encoding const *
 666 Encodings::fromLaTeXName(string const & n, int p, bool allowUnsafe) const
 667 {
 668         string name = n;
 669         // FIXME: if we have to test for too many of these synonyms,
 670         // we should instead extend the format of lib/encodings
 671         if (n == "ansinew")
 672                 name = "cp1252";
 673
 674         // We don't use find_if because it makes copies of the pairs in
 675         // the map.
 676         // This linear search is OK since we don't have many encodings.
 677         // Users could even optimize it by putting the encodings they use
 678         // most at the top of lib/encodings.
 679         EncodingList::const_iterator const end = encodinglist.end();
 680         for (EncodingList::const_iterator it = encodinglist.begin(); it != end; ++it)
 681                 if ((it->second.latexName() == name) && (it->second.package() & p)
 682                                 && (!it->second.unsafe() || allowUnsafe))
 683                         return &it->second;
 684         return nullptr;
 685 }
 686
 687
 688 Encoding const *
 689 Encodings::fromIconvName(string const & n, int p, bool allowUnsafe) const
 690 {
 691         EncodingList::const_iterator const end = encodinglist.end();
 692         for (EncodingList::const_iterator it = encodinglist.begin(); it != end; ++it)
 693                 if ((it->second.iconvName() == n) && (it->second.package() & p)
 694                                 && (!it->second.unsafe() || allowUnsafe))
 695                         return &it->second;
 696         return nullptr;
 697 }
 698
 699
 700 Encodings::Encodings()
 701 {}
 702
 703
 704 void Encodings::read(FileName const & encfile, FileName const & symbolsfile)
 705 {
 706         // We must read the symbolsfile first, because the Encoding
 707         // constructor depends on it.
 708         CharSetMap forcednotselected;
 709         Lexer symbolslex;
 710         symbolslex.setFile(symbolsfile);
 711         bool getNextToken = true;
 712         while (symbolslex.isOK()) {
 713                 char_type symbol;
 714
 715                 if (getNextToken) {
 716                         if (!symbolslex.next(true))
 717                                 break;
 718                 } else
 719                         getNextToken = true;
 720
 721                 istringstream is(symbolslex.getString());
 722                 // reading symbol directly does not work if
 723                 // char_type == wchar_t.
 724                 uint32_t tmp;
 725                 if(!(is >> hex >> tmp))
 726                         break;
 727                 symbol = tmp;
 728
 729                 if (!symbolslex.next(true))
 730                         break;
 731                 docstring textcommand = symbolslex.getDocString();
 732                 if (!symbolslex.next(true))
 733                         break;
 734                 string textpreamble = symbolslex.getString();
 735                 if (!symbolslex.next(true))
 736                         break;
 737                 string sflags = symbolslex.getString();
 738
 739                 string tipashortcut;
 740                 int flags = 0;
 741
 742                 if (suffixIs(textcommand, '}'))
 743                         flags |= CharInfoTextNoTermination;
 744                 while (!sflags.empty()) {
 745                         string flag;
 746                         sflags = split(sflags, flag, ',');
 747                         if (flag == "combining") {
 748                                 flags |= CharInfoCombining;
 749                         } else if (flag == "force") {
 750                                 flags |= CharInfoForce;
 751                                 forced.insert(symbol);
 752                         } else if (prefixIs(flag, "force=")) {
 753                                 vector<string> encs =
 754                                         getVectorFromString(flag.substr(6), ";");
 755                                 for (auto const & enc : encs)
 756                                         forcedselected[enc].insert(symbol);
 757                                 flags |= CharInfoForceSelected;
 758                         } else if (prefixIs(flag, "force!=")) {
 759                                 vector<string> encs =
 760                                         getVectorFromString(flag.substr(7), ";");
 761                                 for (auto const & enc : encs)
 762                                         forcednotselected[enc].insert(symbol);
 763                                 flags |= CharInfoForceSelected;
 764                         } else if (flag == "mathalpha") {
 765                                 mathalpha.insert(symbol);
 766                         } else if (flag == "notermination=text") {
 767                                 flags |= CharInfoTextNoTermination;
 768                         } else if (flag == "notermination=math") {
 769                                 flags |= CharInfoMathNoTermination;
 770                         } else if (flag == "notermination=both") {
 771                                 flags |= CharInfoTextNoTermination;
 772                                 flags |= CharInfoMathNoTermination;
 773                         } else if (flag == "notermination=none") {
 774                                 flags &= ~CharInfoTextNoTermination;
 775                                 flags &= ~CharInfoMathNoTermination;
 776                         } else if (contains(flag, "tipaShortcut=")) {
 777                                 tipashortcut = split(flag, '=');
 778                         } else if (flag == "deprecated") {
 779                                 flags |= CharInfoDeprecated;
 780                         } else {
 781                                 lyxerr << "Ignoring unknown flag `" << flag
 782                                        << "' for symbol `0x"
 783                                        << hex << symbol << dec
 784                                        << "'." << endl;
 785                         }
 786                 }
 787                 // mathCommand and mathPreamble have been added for 1.6.0.
 788                 // make them optional so that old files still work.
 789                 int const lineno = symbolslex.lineNumber();
 790                 bool breakout = false;
 791                 docstring mathcommand;
 792                 string mathpreamble;
 793                 if (symbolslex.next(true)) {
 794                         if (symbolslex.lineNumber() != lineno) {
 795                                 // line in old format without mathCommand and mathPreamble
 796                                 getNextToken = false;
 797                         } else {
 798                                 mathcommand = symbolslex.getDocString();
 799                                 if (suffixIs(mathcommand, '}'))
 800                                         flags |= CharInfoMathNoTermination;
 801                                 if (symbolslex.next(true)) {
 802                                         if (symbolslex.lineNumber() != lineno) {
 803                                                 // line in new format with mathCommand only
 804                                                 getNextToken = false;
 805                                         } else {
 806                                                 // line in new format with mathCommand and mathPreamble
 807                                                 mathpreamble = symbolslex.getString();
 808                                         }
 809                                 } else
 810                                         breakout = true;
 811                         }
 812                 } else {
 813                         breakout = true;
 814                 }
 815
 816                 // backward compatibility
 817                 if (mathpreamble == "esintoramsmath")
 818                         mathpreamble = "esint|amsmath";
 819
 820                 if (!textpreamble.empty())
 821                         if (textpreamble[0] != '\\')
 822                                 flags |= CharInfoTextFeature;
 823                 if (!mathpreamble.empty())
 824                         if (mathpreamble[0] != '\\')
 825                                 flags |= CharInfoMathFeature;
 826
 827                 CharInfo info = CharInfo(
 828                         textcommand, mathcommand,
 829                         textpreamble, mathpreamble,
 830                         tipashortcut, flags);
 831                 LYXERR(Debug::INFO, "Read unicode symbol " << symbol << " '"
 832                                                            << to_utf8(info.textCommand()) << "' '" << info.textPreamble()
 833                                                            << " '" << info.textFeature() << ' ' << info.textNoTermination()
 834                                                            << ' ' << to_utf8(info.mathCommand()) << "' '" << info.mathPreamble()
 835                                                            << "' " << info.mathFeature() << ' ' << info.mathNoTermination()
 836                            << ' ' << info.combining() << ' ' << info.force()
 837                            << ' ' << info.forceSelected());
 838
 839                 // we assume that at least one command is nonempty when using unicodesymbols
 840                 if (info.isUnicodeSymbol()) {
 841                         unicodesymbols[symbol] = info;
 842                 }
 843
 844                 if (breakout)
 845                         break;
 846         }
 847
 848         // Now read the encodings
 849         enum {
 850                 et_encoding = 1,
 851                 et_end
 852         };
 853
 854         LexerKeyword encodingtags[] = {
 855                 { "encoding", et_encoding },
 856                 { "end", et_end }
 857         };
 858
 859         Lexer lex(encodingtags);
 860         lex.setFile(encfile);
 861         lex.setContext("Encodings::read");
 862         while (lex.isOK()) {
 863                 switch (lex.lex()) {
 864                 case et_encoding:
 865                 {
 866                         lex.next();
 867                         string const name = lex.getString();
 868                         lex.next();
 869                         string const latexname = lex.getString();
 870                         lex.next();
 871                         string const guiname = lex.getString();
 872                         lex.next();
 873                         string const iconvname = lex.getString();
 874                         lex.next();
 875                         string const width = lex.getString();
 876                         bool fixedwidth = false;
 877                         bool unsafe = false;
 878                         if (width == "fixed")
 879                                 fixedwidth = true;
 880                         else if (width == "variable")
 881                                 fixedwidth = false;
 882                         else if (width == "variableunsafe") {
 883                                 fixedwidth = false;
 884                                 unsafe = true;
 885                         }
 886                         else
 887                                 lex.printError("Unknown width");
 888
 889                         lex.next();
 890                         string const p = lex.getString();
 891                         Encoding::Package package = Encoding::none;
 892                         if (p == "none")
 893                                 package = Encoding::none;
 894                         else if (p == "inputenc")
 895                                 package = Encoding::inputenc;
 896                         else if (p == "CJK")
 897                                 package = Encoding::CJK;
 898                         else if (p == "japanese")
 899                                 package = Encoding::japanese;
 900                         else
 901                                 lex.printError("Unknown package");
 902
 903                         LYXERR(Debug::INFO, "Reading encoding " << name);
 904                         encodinglist[name] = Encoding(name, latexname,
 905                                 guiname, iconvname, fixedwidth, unsafe,
 906                                 package);
 907
 908                         if (lex.lex() != et_end)
 909                                 lex.printError("Missing end");
 910                         break;
 911                 }
 912                 case et_end:
 913                         lex.printError("Misplaced end");
 914                         break;
 915                 case Lexer::LEX_FEOF:
 916                         break;
 917                 default:
 918                         lex.printError("Unknown tag");
 919                         break;
 920                 }
 921         }
 922
 923         // Move all information from forcednotselected to forcedselected
 924         for (CharSetMap::const_iterator it1 = forcednotselected.begin(); it1 != forcednotselected.end(); ++it1) {
 925                 for (CharSetMap::iterator it2 = forcedselected.begin(); it2 != forcedselected.end(); ++it2) {
 926                         if (it2->first != it1->first)
 927                                 it2->second.insert(it1->second.begin(), it1->second.end());
 928                 }
 929         }
 930
 931 }
 932
 933
 934 } // namespace lyx