src/Encoding.cpp

   1 /**
   2  * \file Encoding.cpp
   3  * This file is part of LyX, the document processor.
   4  * Licence details can be found in the file COPYING.
   5  *
   6  * \author Lars Gullik Bjønnes
   7  * \author Jean-Marc Lasgouttes
   8  * \author Dekel Tsur
   9  *
  10  * Full author contact details are available in file CREDITS.
  11  */
  12
  13 #include <config.h>
  14
  15 #include "Encoding.h"
  16
  17 #include "Lexer.h"
  18
  19 #include "support/debug.h"
  20 #include "support/gettext.h"
  21 #include "support/lstrings.h"
  22 #include "support/mutex.h"
  23 #include "support/textutils.h"
  24 #include "support/unicode.h"
  25
  26 #include <boost/cstdint.hpp>
  27
  28 #include <iterator>
  29 #include <algorithm>
  30 #include <sstream>
  31
  32 using namespace std;
  33 using namespace lyx::support;
  34
  35 namespace lyx {
  36
  37 int const Encoding::any = -1;
  38
  39 Encodings encodings;
  40
  41 Encodings::MathCommandSet Encodings::mathcmd;
  42 Encodings::TextCommandSet Encodings::textcmd;
  43 Encodings::MathSymbolSet  Encodings::mathsym;
  44
  45 namespace {
  46
  47 typedef map<char_type, CharInfo> CharInfoMap;
  48 CharInfoMap unicodesymbols;
  49
  50 typedef set<char_type> CharSet;
  51 typedef map<string, CharSet> CharSetMap;
  52 CharSet forced;
  53 CharSetMap forcedselected;
  54
  55 typedef set<char_type> MathAlphaSet;
  56 MathAlphaSet mathalpha;
  57
  58
  59 /// The highest code point in UCS4 encoding (1<<20 + 1<<16)
  60 char_type const max_ucs4 = 0x110000;
  61
  62 } // namespace
  63
  64
  65 EncodingException::EncodingException(char_type c)
  66         : failed_char(c), par_id(0), pos(0)
  67 {
  68 }
  69
  70
  71 const char * EncodingException::what() const throw()
  72 {
  73         return "Could not find LaTeX command for a character";
  74 }
  75
  76
  77 CharInfo::CharInfo(
  78         docstring const & textcommand, docstring const & mathcommand,
  79         std::string const & textpreamble, std::string const & mathpreamble,
  80         std::string const & tipashortcut, unsigned int flags)
  81         : textcommand_(textcommand), mathcommand_(mathcommand),
  82           textpreamble_(textpreamble), mathpreamble_(mathpreamble),
  83           tipashortcut_(tipashortcut), flags_(flags)
  84 {
  85 }
  86
  87
  88 Encoding::Encoding(string const & n, string const & l, string const & g,
  89                    string const & i, bool f, bool u, Encoding::Package p)
  90         : name_(n), latexName_(l), guiName_(g), iconvName_(i), fixedwidth_(f),
  91           unsafe_(u), forced_(&forcedselected[n]), package_(p)
  92 {
  93         if (n == "ascii") {
  94                 // ASCII can encode 128 code points and nothing else
  95                 start_encodable_ = 128;
  96                 complete_ = true;
  97         } else if (i == "UTF-8") {
  98                 // UTF8 can encode all UCS4 code points
  99                 start_encodable_ = max_ucs4;
 100                 complete_ = true;
 101         } else {
 102                 start_encodable_ = 0;
 103                 complete_ = false;
 104         }
 105 }
 106
 107
 108 void Encoding::init() const
 109 {
 110         // Since the the constructor is the only method which sets complete_
 111         // to false the test for complete_ is thread-safe without mutex.
 112         if (complete_)
 113                 return;
 114
 115         static Mutex mutex;
 116         Mutex::Locker lock(&mutex);
 117
 118         // We need to test again for complete_, since another thread could
 119         // have set it to true while we were waiting for the lock and we must
 120         // not modify an encoding which is already complete.
 121         if (complete_)
 122                 return;
 123
 124         // We do not make any member mutable  so that it can be easily verified
 125         // that all const methods are thread-safe: init() is the only const
 126         // method which changes complete_, encodable_ and start_encodable_, and
 127         // it uses a mutex to ensure thread-safety.
 128         CharSet & encodable = const_cast<Encoding *>(this)->encodable_;
 129         char_type & start_encodable = const_cast<Encoding *>(this)->start_encodable_;
 130
 131         start_encodable = 0;
 132         // temporarily switch off lyxerr, since we will generate iconv errors
 133         lyxerr.disable();
 134         if (fixedwidth_) {
 135                 // We do not need to check all UCS4 code points, it is enough
 136                 // if we check all 256 code points of this encoding.
 137                 for (unsigned short j = 0; j < 256; ++j) {
 138                         char const c = char(j);
 139                         vector<char_type> const ucs4 = eightbit_to_ucs4(&c, 1, iconvName_);
 140                         if (ucs4.size() != 1)
 141                                 continue;
 142                         char_type const uc = ucs4[0];
 143                         CharInfoMap::const_iterator const it = unicodesymbols.find(uc);
 144                         if (it == unicodesymbols.end())
 145                                 encodable.insert(uc);
 146                         else if (!it->second.force()) {
 147                                 if (forced_->empty() || forced_->find(uc) == forced_->end())
 148                                         encodable.insert(uc);
 149                         }
 150                 }
 151         } else {
 152                 // We do not know how many code points this encoding has, and
 153                 // they do not have a direct representation as a single byte,
 154                 // therefore we need to check all UCS4 code points.
 155                 // This is expensive!
 156                 for (char_type c = 0; c < max_ucs4; ++c) {
 157                         vector<char> const eightbit = ucs4_to_eightbit(&c, 1, iconvName_);
 158                         if (!eightbit.empty()) {
 159                                 CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 160                                 if (it == unicodesymbols.end())
 161                                         encodable.insert(c);
 162                                 else if (!it->second.force()) {
 163                                         if (forced_->empty() || forced_->find(c) == forced_->end())
 164                                                 encodable.insert(c);
 165                                 }
 166                         }
 167                 }
 168         }
 169         lyxerr.enable();
 170         CharSet::iterator it = encodable.find(start_encodable);
 171         while (it != encodable.end()) {
 172                 encodable.erase(it);
 173                 ++start_encodable;
 174                 it = encodable.find(start_encodable);
 175         }
 176         const_cast<Encoding *>(this)->complete_ = true;
 177 }
 178
 179
 180 bool Encoding::isForced(char_type c) const
 181 {
 182         if (!forced.empty() && forced.find(c) != forced.end())
 183                 return true;
 184         return !forced_->empty() && forced_->find(c) != forced_->end();
 185 }
 186
 187
 188 bool Encoding::encodable(char_type c) const
 189 {
 190         // assure the used encoding is properly initialized
 191         init();
 192         if (iconvName_ == "UTF-8" && package_ == none)
 193                 return true;
 194         // platex does not load inputenc: force conversion of supported characters
 195         if (package_ == Encoding::japanese
 196             && ((0xb7 <= c && c <= 0x05ff) // Latin-1 Supplement ... Hebrew
 197                         || (0x1d00 <= c && c <= 0x218f) // Phonetic Extensions ... Number Forms
 198                         || (0x2193 <= c && c <= 0x2aff) // Arrows ... Supplemental Mathematical Operators
 199                         || (0xfb00 <= c && c <= 0xfb4f) // Alphabetic Presentation Forms
 200                         || (0x1d400 <= c && c <= 0x1d7ff))) // Mathematical Alphanumeric Symbols
 201                 return false;
 202         if (c < start_encodable_ && !isForced(c))
 203                 return true;
 204         if (encodable_.find(c) != encodable_.end())
 205                 return true;
 206         return false;
 207 }
 208
 209
 210 pair<docstring, bool> Encoding::latexChar(char_type c) const
 211 {
 212         if (encodable(c))
 213                 return make_pair(docstring(1, c), false);
 214
 215         // c cannot (or should not) be encoded in this encoding
 216         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 217         if (it == unicodesymbols.end())
 218                 throw EncodingException(c);
 219         // at least one of mathcommand and textcommand is nonempty
 220         if (it->second.textcommand().empty())
 221                 return make_pair(
 222                         "\\ensuremath{" + it->second.mathcommand() + '}', false);
 223         return make_pair(it->second.textcommand(), !it->second.textnotermination());
 224 }
 225
 226
 227 pair<docstring, docstring> Encoding::latexString(docstring const & input, bool dryrun) const
 228 {
 229         docstring result;
 230         docstring uncodable;
 231         bool terminate = false;
 232         for (size_t n = 0; n < input.size(); ++n) {
 233                 char_type const c = input[n];
 234                 try {
 235                         pair<docstring, bool> latex_char = latexChar(c);
 236                         docstring const latex = latex_char.first;
 237                         if (terminate && !prefixIs(latex, '\\')
 238                             && !prefixIs(latex, '{')
 239                             && !prefixIs(latex, '}')) {
 240                                         // Prevent eating of a following
 241                                         // space or command corruption by
 242                                         // following characters
 243                                         if (latex == " ")
 244                                                 result += "{}";
 245                                         else
 246                                                 result += " ";
 247                                 }
 248                         result += latex;
 249                         terminate = latex_char.second;
 250                 } catch (EncodingException & /* e */) {
 251                         LYXERR0("Uncodable character <" << docstring(1, c)
 252                                         << "> in latexString!");
 253                         if (dryrun) {
 254                                 result += "<" + _("LyX Warning: ")
 255                                            + _("uncodable character") + " '";
 256                                 result += docstring(1, input[n]);
 257                                 result += "'>";
 258                         } else
 259                                 uncodable += input[n];
 260                 }
 261         }
 262         return make_pair(result, uncodable);
 263 }
 264
 265
 266 vector<char_type> Encoding::symbolsList() const
 267 {
 268         // assure the used encoding is properly initialized
 269         init();
 270
 271         // first all those below start_encodable_
 272         vector<char_type> symbols;
 273         for (char_type c = 0; c < start_encodable_; ++c)
 274                 symbols.push_back(c);
 275         // add all encodable characters
 276         copy(encodable_.begin(), encodable_.end(), back_inserter(symbols));
 277         // now the ones from the unicodesymbols file that are not already there
 278         for (pair<char_type, CharInfo> const & elem : unicodesymbols) {
 279                 if (find(symbols.begin(), symbols.end(), elem.first) == symbols.end())
 280                         symbols.push_back(elem.first);
 281         }
 282         // finally, sort the vector
 283         sort(symbols.begin(), symbols.end());
 284         return symbols;
 285 }
 286
 287
 288 bool Encodings::latexMathChar(char_type c, bool mathmode,
 289                         Encoding const * encoding, docstring & command,
 290                         bool & needsTermination)
 291 {
 292         command = empty_docstring();
 293         if (encoding)
 294                 if (encoding->encodable(c))
 295                         command = docstring(1, c);
 296         needsTermination = false;
 297
 298         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 299         if (it == unicodesymbols.end()) {
 300                 if (!encoding || command.empty())
 301                         throw EncodingException(c);
 302                 if (mathmode)
 303                         addMathSym(c);
 304                 return false;
 305         }
 306         // at least one of mathcommand and textcommand is nonempty
 307         bool use_math = (mathmode && !it->second.mathcommand().empty()) ||
 308                         (!mathmode && it->second.textcommand().empty());
 309         if (use_math) {
 310                 command = it->second.mathcommand();
 311                 needsTermination = !it->second.mathnotermination();
 312                 addMathCmd(c);
 313         } else {
 314                 if (!encoding || command.empty()) {
 315                         command = it->second.textcommand();
 316                         needsTermination = !it->second.textnotermination();
 317                 }
 318                 if (mathmode)
 319                         addMathSym(c);
 320                 else
 321                         addTextCmd(c);
 322         }
 323         return use_math;
 324 }
 325
 326
 327 char_type Encodings::fromLaTeXCommand(docstring const & cmd, int cmdtype,
 328                 bool & combining, bool & needsTermination, set<string> * req)
 329 {
 330         CharInfoMap::const_iterator const end = unicodesymbols.end();
 331         CharInfoMap::const_iterator it = unicodesymbols.begin();
 332         for (combining = false; it != end; ++it) {
 333                 if (it->second.deprecated())
 334                         continue;
 335                 docstring const math = it->second.mathcommand();
 336                 docstring const text = it->second.textcommand();
 337                 if ((cmdtype & MATH_CMD) && math == cmd) {
 338                         combining = it->second.combining();
 339                         needsTermination = !it->second.mathnotermination();
 340                         if (req && it->second.mathfeature() &&
 341                             !it->second.mathpreamble().empty())
 342                                 req->insert(it->second.mathpreamble());
 343                         return it->first;
 344                 }
 345                 if ((cmdtype & TEXT_CMD) && text == cmd) {
 346                         combining = it->second.combining();
 347                         needsTermination = !it->second.textnotermination();
 348                         if (req && it->second.textfeature() &&
 349                             !it->second.textpreamble().empty())
 350                                 req->insert(it->second.textpreamble());
 351                         return it->first;
 352                 }
 353         }
 354         needsTermination = false;
 355         return 0;
 356 }
 357
 358
 359 docstring Encodings::fromLaTeXCommand(docstring const & cmd, int cmdtype,
 360                 bool & needsTermination, docstring & rem, set<string> * req)
 361 {
 362         needsTermination = false;
 363         rem = empty_docstring();
 364         bool const mathmode = cmdtype & MATH_CMD;
 365         bool const textmode = cmdtype & TEXT_CMD;
 366         docstring symbols;
 367         size_t const cmdend = cmd.size();
 368         size_t prefix = 0;
 369         CharInfoMap::const_iterator const uniend = unicodesymbols.end();
 370         for (size_t i = 0, j = 0; j < cmdend; ++j) {
 371                 // Also get the char after a backslash
 372                 if (j + 1 < cmdend && cmd[j] == '\\') {
 373                         ++j;
 374                         prefix = 1;
 375                         // Detect things like \=*{e} as well
 376                         if (j + 3 < cmdend && cmd[j+1] == '*' &&
 377                             cmd[j+2] == '{') {
 378                                 ++j;
 379                                 prefix = 2;
 380                         }
 381                 }
 382                 // position of the last character before a possible macro
 383                 // argument
 384                 size_t m = j;
 385                 // If a macro argument follows, get it, too
 386                 // Do it here only for single character commands. Other
 387                 // combining commands need this too, but they are handled in
 388                 // the loop below for performance reasons.
 389                 if (j + 1 < cmdend && cmd[j + 1] == '{') {
 390                         size_t k = j + 1;
 391                         int count = 1;
 392                         while (k < cmdend && count) {
 393                                 k = cmd.find_first_of(from_ascii("{}"), k + 1);
 394                                 // braces may not be balanced
 395                                 if (k == docstring::npos)
 396                                         break;
 397                                 if (cmd[k] == '{')
 398                                         ++count;
 399                                 else
 400                                         --count;
 401                         }
 402                         if (k != docstring::npos)
 403                                 j = k;
 404                 } else if (m + 1 < cmdend && isAlphaASCII(cmd[m])) {
 405                         while (m + 2 < cmdend && isAlphaASCII(cmd[m+1]))
 406                                 m++;
 407                 }
 408                 // Start with this substring and try augmenting it when it is
 409                 // the prefix of some command in the unicodesymbols file
 410                 docstring subcmd = cmd.substr(i, j - i + 1);
 411
 412                 CharInfoMap::const_iterator it = unicodesymbols.begin();
 413                 // First part of subcmd which might be a combining character
 414                 docstring combcmd = (m == j) ? docstring() : cmd.substr(i, m - i + 1);
 415                 // The combining character of combcmd if it exists
 416                 CharInfoMap::const_iterator combining = uniend;
 417                 size_t unicmd_size = 0;
 418                 char_type c = 0;
 419                 for (; it != uniend; ++it) {
 420                         if (it->second.deprecated())
 421                                 continue;
 422                         docstring const math = mathmode ? it->second.mathcommand()
 423                                                         : docstring();
 424                         docstring const text = textmode ? it->second.textcommand()
 425                                                         : docstring();
 426                         if (!combcmd.empty() && it->second.combining() &&
 427                             (math == combcmd || text == combcmd))
 428                                 combining = it;
 429                         size_t cur_size = max(math.size(), text.size());
 430                         // The current math or text unicode command cannot
 431                         // match, or we already matched a longer one
 432                         if (cur_size < subcmd.size() || cur_size <= unicmd_size)
 433                                 continue;
 434
 435                         docstring tmp = subcmd;
 436                         size_t k = j;
 437                         while (prefixIs(math, tmp) || prefixIs(text, tmp)) {
 438                                 ++k;
 439                                 if (k >= cmdend || cur_size <= tmp.size())
 440                                         break;
 441                                 tmp += cmd[k];
 442                         }
 443                         // No match
 444                         if (k == j)
 445                                 continue;
 446
 447                         // The last added char caused a mismatch, because
 448                         // we didn't exhaust the chars in cmd and didn't
 449                         // exceed the maximum size of the current unicmd
 450                         if (k < cmdend && cur_size > tmp.size())
 451                                 tmp.resize(tmp.size() - 1);
 452
 453                         // If this is an exact match, we found a (longer)
 454                         // matching entry in the unicodesymbols file.
 455                         if (math != tmp && text != tmp)
 456                                 continue;
 457                         // If we found a combining command, we need to append
 458                         // the macro argument if this has not been done above.
 459                         if (tmp == combcmd && combining != uniend &&
 460                             k < cmdend && cmd[k] == '{') {
 461                                 size_t l = k;
 462                                 int count = 1;
 463                                 while (l < cmdend && count) {
 464                                         l = cmd.find_first_of(from_ascii("{}"), l + 1);
 465                                         // braces may not be balanced
 466                                         if (l == docstring::npos)
 467                                                 break;
 468                                         if (cmd[l] == '{')
 469                                                 ++count;
 470                                         else
 471                                                 --count;
 472                                 }
 473                                 if (l != docstring::npos) {
 474                                         j = l;
 475                                         subcmd = cmd.substr(i, j - i + 1);
 476                                 }
 477                         }
 478                         // If the entry doesn't start with '\', we take note
 479                         // of the match and continue (this is not a ultimate
 480                         // acceptance, as some other entry may match a longer
 481                         // portion of the cmd string). However, if the entry
 482                         // does start with '\', we accept the match only if
 483                         // this is a valid macro, i.e., either it is a single
 484                         // (nonletter) char macro, or nothing else follows,
 485                         // or what follows is a nonletter char, or the last
 486                         // character is a }.
 487                         else if (tmp[0] != '\\'
 488                                    || (tmp.size() == prefix + 1 &&
 489                                        !isAlphaASCII(tmp[1]) &&
 490                                        (prefix == 1 || !isAlphaASCII(tmp[2])))
 491                                    || k == cmdend
 492                                    || !isAlphaASCII(cmd[k])
 493                                    || tmp[tmp.size() - 1] == '}'
 494                                  ) {
 495                                 c = it->first;
 496                                 j = k - 1;
 497                                 i = j + 1;
 498                                 unicmd_size = cur_size;
 499                                 if (math == tmp)
 500                                         needsTermination = !it->second.mathnotermination();
 501                                 else
 502                                         needsTermination = !it->second.textnotermination();
 503                                 if (req) {
 504                                         if (math == tmp && it->second.mathfeature() &&
 505                                             !it->second.mathpreamble().empty())
 506                                                 req->insert(it->second.mathpreamble());
 507                                         if (text == tmp && it->second.textfeature() &&
 508                                             !it->second.textpreamble().empty())
 509                                                 req->insert(it->second.textpreamble());
 510                                 }
 511                         }
 512                 }
 513                 if (unicmd_size)
 514                         symbols += c;
 515                 else if (combining != uniend &&
 516                          prefixIs(subcmd, combcmd + '{')) {
 517                         // We know that subcmd starts with combcmd and
 518                         // contains an argument in braces.
 519                         docstring const arg = subcmd.substr(
 520                                 combcmd.length() + 1,
 521                                 subcmd.length() - combcmd.length() - 2);
 522                         // If arg is a single character we can construct a
 523                         // combining sequence.
 524                         char_type a;
 525                         bool argcomb = false;
 526                         if (arg.size() == 1 && isAlnumASCII(arg[0]))
 527                                 a = arg[0];
 528                         else {
 529                                 // Use the version of fromLaTeXCommand() that
 530                                 // parses only one command, since we cannot
 531                                 // use more than one character.
 532                                 bool dummy = false;
 533                                 set<string> r;
 534                                 a = fromLaTeXCommand(arg, cmdtype, argcomb,
 535                                                      dummy, &r);
 536                                 if (a && req && !argcomb)
 537                                         req->insert(r.begin(), r.end());
 538                         }
 539                         if (a && !argcomb) {
 540                                 // In unicode the combining character comes
 541                                 // after its base
 542                                 symbols += a;
 543                                 symbols += combining->first;
 544                                 i = j + 1;
 545                                 unicmd_size = 2;
 546                         }
 547                 }
 548                 if (j + 1 == cmdend && !unicmd_size) {
 549                         // No luck. Return what remains
 550                         rem = cmd.substr(i);
 551                         if (needsTermination && !rem.empty()) {
 552                                 if (rem.substr(0, 2) == "{}") {
 553                                         rem = rem.substr(2);
 554                                         needsTermination = false;
 555                                 } else if (rem[0] == ' ') {
 556                                         needsTermination = false;
 557                                         // LaTeX would swallow all spaces
 558                                         rem = ltrim(rem);
 559                                 }
 560                         }
 561                 }
 562         }
 563         return symbols;
 564 }
 565
 566
 567 CharInfo const & Encodings::unicodeCharInfo(char_type c)
 568 {
 569         static CharInfo empty;
 570         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 571         return it != unicodesymbols.end() ? it->second : empty;
 572 }
 573
 574
 575 bool Encodings::isCombiningChar(char_type c)
 576 {
 577         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 578         if (it != unicodesymbols.end())
 579                 return it->second.combining();
 580         return false;
 581 }
 582
 583
 584 string const Encodings::TIPAShortcut(char_type c)
 585 {
 586         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 587         if (it != unicodesymbols.end())
 588                 return it->second.tipashortcut();
 589         return string();
 590 }
 591
 592
 593 string const Encodings::isKnownScriptChar(char_type const c)
 594 {
 595         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 596
 597         if (it == unicodesymbols.end())
 598                 return string();
 599         // FIXME: parse complex textpreamble (may be list or alternatives,
 600         //                e.g., "subscript,textgreek" or "textcomp|textgreek")
 601         if (it->second.textpreamble() == "textgreek"
 602                 || it->second.textpreamble() == "textcyrillic")
 603                 return it->second.textpreamble();
 604         return string();
 605 }
 606
 607
 608 bool Encodings::fontencSupportsScript(string const & fontenc, string const & script)
 609 {
 610         if (script == "textgreek")
 611                 return (fontenc == "LGR" || fontenc == "TU");
 612         if (script == "textcyrillic")
 613                 return (fontenc == "T2A" || fontenc == "T2B" || fontenc == "T2C"
 614                                 || fontenc == "X2" || fontenc == "TU");
 615         return false;
 616 }
 617
 618
 619 bool Encodings::isMathAlpha(char_type c)
 620 {
 621         return mathalpha.count(c);
 622 }
 623
 624
 625 bool Encodings::isUnicodeTextOnly(char_type c)
 626 {
 627         if (isASCII(c) || isMathAlpha(c))
 628                 return false;
 629
 630         CharInfoMap::const_iterator const it = unicodesymbols.find(c);
 631         return it == unicodesymbols.end() || it->second.mathcommand().empty();
 632 }
 633
 634
 635 Encoding const *
 636 Encodings::fromLyXName(string const & name, bool allowUnsafe) const
 637 {
 638         EncodingList::const_iterator const it = encodinglist.find(name);
 639         if (it == encodinglist.end())
 640                 return 0;
 641         if (!allowUnsafe && it->second.unsafe())
 642                 return 0;
 643         return &it->second;
 644 }
 645
 646
 647 Encoding const *
 648 Encodings::fromLaTeXName(string const & n, int const & p, bool allowUnsafe) const
 649 {
 650         string name = n;
 651         // FIXME: if we have to test for too many of these synonyms,
 652         // we should instead extend the format of lib/encodings
 653         if (n == "ansinew")
 654                 name = "cp1252";
 655
 656         // We don't use find_if because it makes copies of the pairs in
 657         // the map.
 658         // This linear search is OK since we don't have many encodings.
 659         // Users could even optimize it by putting the encodings they use
 660         // most at the top of lib/encodings.
 661         EncodingList::const_iterator const end = encodinglist.end();
 662         for (EncodingList::const_iterator it = encodinglist.begin(); it != end; ++it)
 663                 if ((it->second.latexName() == name) && (it->second.package() & p)
 664                                 && (!it->second.unsafe() || allowUnsafe))
 665                         return &it->second;
 666         return 0;
 667 }
 668
 669
 670 Encoding const *
 671 Encodings::fromIconvName(string const & n, int const & p, bool allowUnsafe) const
 672 {
 673         EncodingList::const_iterator const end = encodinglist.end();
 674         for (EncodingList::const_iterator it = encodinglist.begin(); it != end; ++it)
 675                 if ((it->second.iconvName() == n) && (it->second.package() & p)
 676                                 && (!it->second.unsafe() || allowUnsafe))
 677                         return &it->second;
 678         return 0;
 679 }
 680
 681
 682 Encodings::Encodings()
 683 {}
 684
 685
 686 void Encodings::read(FileName const & encfile, FileName const & symbolsfile)
 687 {
 688         // We must read the symbolsfile first, because the Encoding
 689         // constructor depends on it.
 690         CharSetMap forcednotselected;
 691         Lexer symbolslex;
 692         symbolslex.setFile(symbolsfile);
 693         bool getNextToken = true;
 694         while (symbolslex.isOK()) {
 695                 char_type symbol;
 696
 697                 if (getNextToken) {
 698                         if (!symbolslex.next(true))
 699                                 break;
 700                 } else
 701                         getNextToken = true;
 702
 703                 istringstream is(symbolslex.getString());
 704                 // reading symbol directly does not work if
 705                 // char_type == wchar_t.
 706                 boost::uint32_t tmp;
 707                 if(!(is >> hex >> tmp))
 708                         break;
 709                 symbol = tmp;
 710
 711                 if (!symbolslex.next(true))
 712                         break;
 713                 docstring textcommand = symbolslex.getDocString();
 714                 if (!symbolslex.next(true))
 715                         break;
 716                 string textpreamble = symbolslex.getString();
 717                 if (!symbolslex.next(true))
 718                         break;
 719                 string sflags = symbolslex.getString();
 720
 721                 string tipashortcut;
 722                 int flags = 0;
 723
 724                 if (suffixIs(textcommand, '}'))
 725                         flags |= CharInfoTextNoTermination;
 726                 while (!sflags.empty()) {
 727                         string flag;
 728                         sflags = split(sflags, flag, ',');
 729                         if (flag == "combining") {
 730                                 flags |= CharInfoCombining;
 731                         } else if (flag == "force") {
 732                                 flags |= CharInfoForce;
 733                                 forced.insert(symbol);
 734                         } else if (prefixIs(flag, "force=")) {
 735                                 vector<string> encs =
 736                                         getVectorFromString(flag.substr(6), ";");
 737                                 for (size_t i = 0; i < encs.size(); ++i)
 738                                         forcedselected[encs[i]].insert(symbol);
 739                                 flags |= CharInfoForceSelected;
 740                         } else if (prefixIs(flag, "force!=")) {
 741                                 vector<string> encs =
 742                                         getVectorFromString(flag.substr(7), ";");
 743                                 for (size_t i = 0; i < encs.size(); ++i)
 744                                         forcednotselected[encs[i]].insert(symbol);
 745                                 flags |= CharInfoForceSelected;
 746                         } else if (flag == "mathalpha") {
 747                                 mathalpha.insert(symbol);
 748                         } else if (flag == "notermination=text") {
 749                                 flags |= CharInfoTextNoTermination;
 750                         } else if (flag == "notermination=math") {
 751                                 flags |= CharInfoMathNoTermination;
 752                         } else if (flag == "notermination=both") {
 753                                 flags |= CharInfoTextNoTermination;
 754                                 flags |= CharInfoMathNoTermination;
 755                         } else if (flag == "notermination=none") {
 756                                 flags &= ~CharInfoTextNoTermination;
 757                                 flags &= ~CharInfoMathNoTermination;
 758                         } else if (contains(flag, "tipashortcut=")) {
 759                                 tipashortcut = split(flag, '=');
 760                         } else if (flag == "deprecated") {
 761                                 flags |= CharInfoDeprecated;
 762                         } else {
 763                                 lyxerr << "Ignoring unknown flag `" << flag
 764                                        << "' for symbol `0x"
 765                                        << hex << symbol << dec
 766                                        << "'." << endl;
 767                         }
 768                 }
 769                 // mathcommand and mathpreamble have been added for 1.6.0.
 770                 // make them optional so that old files still work.
 771                 int const lineno = symbolslex.lineNumber();
 772                 bool breakout = false;
 773                 docstring mathcommand;
 774                 string mathpreamble;
 775                 if (symbolslex.next(true)) {
 776                         if (symbolslex.lineNumber() != lineno) {
 777                                 // line in old format without mathcommand and mathpreamble
 778                                 getNextToken = false;
 779                         } else {
 780                                 mathcommand = symbolslex.getDocString();
 781                                 if (suffixIs(mathcommand, '}'))
 782                                         flags |= CharInfoMathNoTermination;
 783                                 if (symbolslex.next(true)) {
 784                                         if (symbolslex.lineNumber() != lineno) {
 785                                                 // line in new format with mathcommand only
 786                                                 getNextToken = false;
 787                                         } else {
 788                                                 // line in new format with mathcommand and mathpreamble
 789                                                 mathpreamble = symbolslex.getString();
 790                                         }
 791                                 } else
 792                                         breakout = true;
 793                         }
 794                 } else {
 795                         breakout = true;
 796                 }
 797
 798                 // backward compatibility
 799                 if (mathpreamble == "esintoramsmath")
 800                         mathpreamble = "esint|amsmath";
 801
 802                 if (!textpreamble.empty())
 803                         if (textpreamble[0] != '\\')
 804                                 flags |= CharInfoTextFeature;
 805                 if (!mathpreamble.empty())
 806                         if (mathpreamble[0] != '\\')
 807                                 flags |= CharInfoMathFeature;
 808
 809                 CharInfo info = CharInfo(
 810                         textcommand, mathcommand,
 811                         textpreamble, mathpreamble,
 812                         tipashortcut, flags);
 813                 LYXERR(Debug::INFO, "Read unicode symbol " << symbol << " '"
 814                            << to_utf8(info.textcommand()) << "' '" << info.textpreamble()
 815                            << " '" << info.textfeature() << ' ' << info.textnotermination()
 816                            << ' ' << to_utf8(info.mathcommand()) << "' '" << info.mathpreamble()
 817                            << "' " << info.mathfeature() << ' ' << info.mathnotermination()
 818                            << ' ' << info.combining() << ' ' << info.force()
 819                            << ' ' << info.forceselected());
 820
 821                 // we assume that at least one command is nonempty when using unicodesymbols
 822                 if (info.isUnicodeSymbol()) {
 823                         unicodesymbols[symbol] = info;
 824                 }
 825
 826                 if (breakout)
 827                         break;
 828         }
 829
 830         // Now read the encodings
 831         enum {
 832                 et_encoding = 1,
 833                 et_end
 834         };
 835
 836         LexerKeyword encodingtags[] = {
 837                 { "encoding", et_encoding },
 838                 { "end", et_end }
 839         };
 840
 841         Lexer lex(encodingtags);
 842         lex.setFile(encfile);
 843         lex.setContext("Encodings::read");
 844         while (lex.isOK()) {
 845                 switch (lex.lex()) {
 846                 case et_encoding:
 847                 {
 848                         lex.next();
 849                         string const name = lex.getString();
 850                         lex.next();
 851                         string const latexname = lex.getString();
 852                         lex.next();
 853                         string const guiname = lex.getString();
 854                         lex.next();
 855                         string const iconvname = lex.getString();
 856                         lex.next();
 857                         string const width = lex.getString();
 858                         bool fixedwidth = false;
 859                         bool unsafe = false;
 860                         if (width == "fixed")
 861                                 fixedwidth = true;
 862                         else if (width == "variable")
 863                                 fixedwidth = false;
 864                         else if (width == "variableunsafe") {
 865                                 fixedwidth = false;
 866                                 unsafe = true;
 867                         }
 868                         else
 869                                 lex.printError("Unknown width");
 870
 871                         lex.next();
 872                         string const p = lex.getString();
 873                         Encoding::Package package = Encoding::none;
 874                         if (p == "none")
 875                                 package = Encoding::none;
 876                         else if (p == "inputenc")
 877                                 package = Encoding::inputenc;
 878                         else if (p == "CJK")
 879                                 package = Encoding::CJK;
 880                         else if (p == "japanese")
 881                                 package = Encoding::japanese;
 882                         else
 883                                 lex.printError("Unknown package");
 884
 885                         LYXERR(Debug::INFO, "Reading encoding " << name);
 886                         encodinglist[name] = Encoding(name, latexname,
 887                                 guiname, iconvname, fixedwidth, unsafe,
 888                                 package);
 889
 890                         if (lex.lex() != et_end)
 891                                 lex.printError("Missing end");
 892                         break;
 893                 }
 894                 case et_end:
 895                         lex.printError("Misplaced end");
 896                         break;
 897                 case Lexer::LEX_FEOF:
 898                         break;
 899                 default:
 900                         lex.printError("Unknown tag");
 901                         break;
 902                 }
 903         }
 904
 905         // Move all information from forcednotselected to forcedselected
 906         for (CharSetMap::const_iterator it1 = forcednotselected.begin(); it1 != forcednotselected.end(); ++it1) {
 907                 for (CharSetMap::iterator it2 = forcedselected.begin(); it2 != forcedselected.end(); ++it2) {
 908                         if (it2->first != it1->first)
 909                                 it2->second.insert(it1->second.begin(), it1->second.end());
 910                 }
 911         }
 912
 913 }
 914
 915
 916 } // namespace lyx