3 * This file is part of LyX, the document processor.
4 * Licence details can be found in the file COPYING.
6 * \author Lars Gullik Bjønnes
7 * \author Jean-Marc Lasgouttes
10 * Full author contact details are available in file CREDITS.
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"
26 #include <boost/cstdint.hpp>
32 using namespace lyx::support;
36 int const Encoding::any = -1;
40 Encodings::MathCommandSet Encodings::mathcmd;
41 Encodings::TextCommandSet Encodings::textcmd;
42 Encodings::MathSymbolSet Encodings::mathsym;
46 typedef map<char_type, CharInfo> CharInfoMap;
47 CharInfoMap unicodesymbols;
49 typedef set<char_type> CharSet;
50 typedef map<string, CharSet> CharSetMap;
52 CharSetMap forcedselected;
54 typedef set<char_type> MathAlphaSet;
55 MathAlphaSet mathalpha;
58 /// The highest code point in UCS4 encoding (1<<20 + 1<<16)
59 char_type const max_ucs4 = 0x110000;
64 EncodingException::EncodingException(char_type c)
65 : failed_char(c), par_id(0), pos(0)
70 const char * EncodingException::what() const throw()
72 return "Could not find LaTeX command for a character";
77 docstring const & textcommand, docstring const & mathcommand,
78 std::string const & textpreamble, std::string const & mathpreamble,
79 std::string const & tipashortcut, unsigned int flags)
80 : textcommand_(textcommand), mathcommand_(mathcommand),
81 textpreamble_(textpreamble), mathpreamble_(mathpreamble),
82 tipashortcut_(tipashortcut), flags_(flags)
87 Encoding::Encoding(string const & n, string const & l, string const & g,
88 string const & i, bool f, bool u, Encoding::Package p)
89 : name_(n), latexName_(l), guiName_(g), iconvName_(i), fixedwidth_(f),
90 unsafe_(u), forced_(&forcedselected[n]), package_(p)
93 // ASCII can encode 128 code points and nothing else
94 start_encodable_ = 128;
96 } else if (i == "UTF-8") {
97 // UTF8 can encode all UCS4 code points
98 start_encodable_ = max_ucs4;
101 start_encodable_ = 0;
107 void Encoding::init() const
109 // Since the the constructor is the only method which sets complete_
110 // to false the test for complete_ is thread-safe without mutex.
115 Mutex::Locker lock(&mutex);
117 // We need to test again for complete_, since another thread could
118 // have set it to true while we were waiting for the lock and we must
119 // not modify an encoding which is already complete.
123 // We do not make any member mutable so that it can be easily verified
124 // that all const methods are thread-safe: init() is the only const
125 // method which changes complete_, encodable_ and start_encodable_, and
126 // it uses a mutex to ensure thread-safety.
127 CharSet & encodable = const_cast<Encoding *>(this)->encodable_;
128 char_type & start_encodable = const_cast<Encoding *>(this)->start_encodable_;
131 // temporarily switch off lyxerr, since we will generate iconv errors
134 // We do not need to check all UCS4 code points, it is enough
135 // if we check all 256 code points of this encoding.
136 for (unsigned short j = 0; j < 256; ++j) {
137 char const c = char(j);
138 vector<char_type> const ucs4 = eightbit_to_ucs4(&c, 1, iconvName_);
139 if (ucs4.size() != 1)
141 char_type const uc = ucs4[0];
142 CharInfoMap::const_iterator const it = unicodesymbols.find(uc);
143 if (it == unicodesymbols.end())
144 encodable.insert(uc);
145 else if (!it->second.force()) {
146 if (forced_->empty() || forced_->find(uc) == forced_->end())
147 encodable.insert(uc);
151 // We do not know how many code points this encoding has, and
152 // they do not have a direct representation as a single byte,
153 // therefore we need to check all UCS4 code points.
154 // This is expensive!
155 for (char_type c = 0; c < max_ucs4; ++c) {
156 vector<char> const eightbit = ucs4_to_eightbit(&c, 1, iconvName_);
157 if (!eightbit.empty()) {
158 CharInfoMap::const_iterator const it = unicodesymbols.find(c);
159 if (it == unicodesymbols.end())
161 else if (!it->second.force()) {
162 if (forced_->empty() || forced_->find(c) == forced_->end())
169 CharSet::iterator it = encodable.find(start_encodable);
170 while (it != encodable.end()) {
173 it = encodable.find(start_encodable);
175 const_cast<Encoding *>(this)->complete_ = true;
179 bool Encoding::isForced(char_type c) const
181 if (!forced.empty() && forced.find(c) != forced.end())
183 return !forced_->empty() && forced_->find(c) != forced_->end();
187 bool Encoding::encodable(char_type c) const
189 // assure the used encoding is properly initialized
192 if (iconvName_ == "UTF-8" && package_ == none)
194 if (c < start_encodable_ && !isForced(c))
196 if (encodable_.find(c) != encodable_.end())
202 pair<docstring, bool> Encoding::latexChar(char_type c) const
205 return make_pair(docstring(1, c), false);
207 // c cannot (or should not) be encoded in this encoding
208 CharInfoMap::const_iterator const it = unicodesymbols.find(c);
209 if (it == unicodesymbols.end())
210 throw EncodingException(c);
211 // at least one of mathcommand and textcommand is nonempty
212 if (it->second.textcommand().empty())
214 "\\ensuremath{" + it->second.mathcommand() + '}', false);
215 return make_pair(it->second.textcommand(), !it->second.textnotermination());
219 pair<docstring, docstring> Encoding::latexString(docstring const & input, bool dryrun) const
223 bool terminate = false;
224 for (size_t n = 0; n < input.size(); ++n) {
226 char_type const c = input[n];
227 pair<docstring, bool> latex_char = latexChar(c);
228 docstring const latex = latex_char.first;
229 if (terminate && !prefixIs(latex, '\\')
230 && !prefixIs(latex, '{')
231 && !prefixIs(latex, '}')) {
232 // Prevent eating of a following
233 // space or command corruption by
234 // following characters
241 terminate = latex_char.second;
242 } catch (EncodingException & /* e */) {
243 LYXERR0("Uncodable character in latexString!");
245 result += "<" + _("LyX Warning: ")
246 + _("uncodable character") + " '";
247 result += docstring(1, input[n]);
250 uncodable += input[n];
253 return make_pair(result, uncodable);
257 vector<char_type> Encoding::symbolsList() const
259 // assure the used encoding is properly initialized
262 // first all those below start_encodable_
263 vector<char_type> symbols;
264 for (char_type c = 0; c < start_encodable_; ++c)
265 symbols.push_back(c);
266 //add all encodable characters
267 copy(encodable_.begin(), encodable_.end(), back_inserter(symbols));
268 // now the ones from the unicodesymbols file
269 for (pair<char_type, CharInfo> const & elem : unicodesymbols)
270 symbols.push_back(elem.first);
275 bool Encodings::latexMathChar(char_type c, bool mathmode,
276 Encoding const * encoding, docstring & command,
277 bool & needsTermination)
279 command = empty_docstring();
281 if (encoding->encodable(c))
282 command = docstring(1, c);
283 needsTermination = false;
285 CharInfoMap::const_iterator const it = unicodesymbols.find(c);
286 if (it == unicodesymbols.end()) {
287 if (!encoding || command.empty())
288 throw EncodingException(c);
293 // at least one of mathcommand and textcommand is nonempty
294 bool use_math = (mathmode && !it->second.mathcommand().empty()) ||
295 (!mathmode && it->second.textcommand().empty());
297 command = it->second.mathcommand();
298 needsTermination = !it->second.mathnotermination();
301 if (!encoding || command.empty()) {
302 command = it->second.textcommand();
303 needsTermination = !it->second.textnotermination();
313 char_type Encodings::fromLaTeXCommand(docstring const & cmd, int cmdtype,
314 bool & combining, bool & needsTermination, set<string> * req)
316 CharInfoMap::const_iterator const end = unicodesymbols.end();
317 CharInfoMap::const_iterator it = unicodesymbols.begin();
318 for (combining = false; it != end; ++it) {
319 if (it->second.deprecated())
321 docstring const math = it->second.mathcommand();
322 docstring const text = it->second.textcommand();
323 if ((cmdtype & MATH_CMD) && math == cmd) {
324 combining = it->second.combining();
325 needsTermination = !it->second.mathnotermination();
326 if (req && it->second.mathfeature() &&
327 !it->second.mathpreamble().empty())
328 req->insert(it->second.mathpreamble());
331 if ((cmdtype & TEXT_CMD) && text == cmd) {
332 combining = it->second.combining();
333 needsTermination = !it->second.textnotermination();
334 if (req && it->second.textfeature() &&
335 !it->second.textpreamble().empty())
336 req->insert(it->second.textpreamble());
340 needsTermination = false;
345 docstring Encodings::fromLaTeXCommand(docstring const & cmd, int cmdtype,
346 bool & needsTermination, docstring & rem, set<string> * req)
348 needsTermination = false;
349 rem = empty_docstring();
350 bool const mathmode = cmdtype & MATH_CMD;
351 bool const textmode = cmdtype & TEXT_CMD;
353 size_t const cmdend = cmd.size();
355 CharInfoMap::const_iterator const uniend = unicodesymbols.end();
356 for (size_t i = 0, j = 0; j < cmdend; ++j) {
357 // Also get the char after a backslash
358 if (j + 1 < cmdend && cmd[j] == '\\') {
361 // Detect things like \=*{e} as well
362 if (j + 3 < cmdend && cmd[j+1] == '*' &&
368 // position of the last character before a possible macro
371 // If a macro argument follows, get it, too
372 // Do it here only for single character commands. Other
373 // combining commands need this too, but they are handled in
374 // the loop below for performance reasons.
375 if (j + 1 < cmdend && cmd[j + 1] == '{') {
378 while (k < cmdend && count) {
379 k = cmd.find_first_of(from_ascii("{}"), k + 1);
380 // braces may not be balanced
381 if (k == docstring::npos)
388 if (k != docstring::npos)
390 } else if (m + 1 < cmdend && isAlphaASCII(cmd[m])) {
391 while (m + 2 < cmdend && isAlphaASCII(cmd[m+1]))
394 // Start with this substring and try augmenting it when it is
395 // the prefix of some command in the unicodesymbols file
396 docstring subcmd = cmd.substr(i, j - i + 1);
398 CharInfoMap::const_iterator it = unicodesymbols.begin();
399 // First part of subcmd which might be a combining character
400 docstring combcmd = (m == j) ? docstring() : cmd.substr(i, m - i + 1);
401 // The combining character of combcmd if it exists
402 CharInfoMap::const_iterator combining = uniend;
403 size_t unicmd_size = 0;
405 for (; it != uniend; ++it) {
406 if (it->second.deprecated())
408 docstring const math = mathmode ? it->second.mathcommand()
410 docstring const text = textmode ? it->second.textcommand()
412 if (!combcmd.empty() && it->second.combining() &&
413 (math == combcmd || text == combcmd))
415 size_t cur_size = max(math.size(), text.size());
416 // The current math or text unicode command cannot
417 // match, or we already matched a longer one
418 if (cur_size < subcmd.size() || cur_size <= unicmd_size)
421 docstring tmp = subcmd;
423 while (prefixIs(math, tmp) || prefixIs(text, tmp)) {
425 if (k >= cmdend || cur_size <= tmp.size())
433 // The last added char caused a mismatch, because
434 // we didn't exhaust the chars in cmd and didn't
435 // exceed the maximum size of the current unicmd
436 if (k < cmdend && cur_size > tmp.size())
437 tmp.resize(tmp.size() - 1);
439 // If this is an exact match, we found a (longer)
440 // matching entry in the unicodesymbols file.
441 if (math != tmp && text != tmp)
443 // If we found a combining command, we need to append
444 // the macro argument if this has not been done above.
445 if (tmp == combcmd && combining != uniend &&
446 k < cmdend && cmd[k] == '{') {
449 while (l < cmdend && count) {
450 l = cmd.find_first_of(from_ascii("{}"), l + 1);
451 // braces may not be balanced
452 if (l == docstring::npos)
459 if (l != docstring::npos) {
461 subcmd = cmd.substr(i, j - i + 1);
464 // If the entry doesn't start with '\', we take note
465 // of the match and continue (this is not a ultimate
466 // acceptance, as some other entry may match a longer
467 // portion of the cmd string). However, if the entry
468 // does start with '\', we accept the match only if
469 // this is a valid macro, i.e., either it is a single
470 // (nonletter) char macro, or nothing else follows,
471 // or what follows is a nonletter char, or the last
473 else if (tmp[0] != '\\'
474 || (tmp.size() == prefix + 1 &&
475 !isAlphaASCII(tmp[1]) &&
476 (prefix == 1 || !isAlphaASCII(tmp[2])))
478 || !isAlphaASCII(cmd[k])
479 || tmp[tmp.size() - 1] == '}'
484 unicmd_size = cur_size;
486 needsTermination = !it->second.mathnotermination();
488 needsTermination = !it->second.textnotermination();
490 if (math == tmp && it->second.mathfeature() &&
491 !it->second.mathpreamble().empty())
492 req->insert(it->second.mathpreamble());
493 if (text == tmp && it->second.textfeature() &&
494 !it->second.textpreamble().empty())
495 req->insert(it->second.textpreamble());
501 else if (combining != uniend &&
502 prefixIs(subcmd, combcmd + '{')) {
503 // We know that subcmd starts with combcmd and
504 // contains an argument in braces.
505 docstring const arg = subcmd.substr(
506 combcmd.length() + 1,
507 subcmd.length() - combcmd.length() - 2);
508 // If arg is a single character we can construct a
509 // combining sequence.
511 bool argcomb = false;
512 if (arg.size() == 1 && isAlnumASCII(arg[0]))
515 // Use the version of fromLaTeXCommand() that
516 // parses only one command, since we cannot
517 // use more than one character.
520 a = fromLaTeXCommand(arg, cmdtype, argcomb,
522 if (a && req && !argcomb)
523 req->insert(r.begin(), r.end());
526 // In unicode the combining character comes
529 symbols += combining->first;
534 if (j + 1 == cmdend && !unicmd_size) {
535 // No luck. Return what remains
537 if (needsTermination && !rem.empty()) {
538 if (rem.substr(0, 2) == "{}") {
540 needsTermination = false;
541 } else if (rem[0] == ' ') {
542 needsTermination = false;
543 // LaTeX would swallow all spaces
553 CharInfo const & Encodings::unicodeCharInfo(char_type c)
555 static CharInfo empty;
556 CharInfoMap::const_iterator const it = unicodesymbols.find(c);
557 return it != unicodesymbols.end() ? it->second : empty;
561 bool Encodings::isCombiningChar(char_type c)
563 CharInfoMap::const_iterator const it = unicodesymbols.find(c);
564 if (it != unicodesymbols.end())
565 return it->second.combining();
570 string const Encodings::TIPAShortcut(char_type c)
572 CharInfoMap::const_iterator const it = unicodesymbols.find(c);
573 if (it != unicodesymbols.end())
574 return it->second.tipashortcut();
579 bool Encodings::isKnownScriptChar(char_type const c, string & preamble)
581 CharInfoMap::const_iterator const it = unicodesymbols.find(c);
583 if (it == unicodesymbols.end())
586 if (it->second.textpreamble() != "textgreek" && it->second.textpreamble() != "textcyr")
589 if (preamble.empty()) {
590 preamble = it->second.textpreamble();
593 return it->second.textpreamble() == preamble;
597 bool Encodings::isMathAlpha(char_type c)
599 return mathalpha.count(c);
603 bool Encodings::isUnicodeTextOnly(char_type c)
605 if (isASCII(c) || isMathAlpha(c))
608 CharInfoMap::const_iterator const it = unicodesymbols.find(c);
609 return it == unicodesymbols.end() || it->second.mathcommand().empty();
614 Encodings::fromLyXName(string const & name, bool allowUnsafe) const
616 EncodingList::const_iterator const it = encodinglist.find(name);
617 if (it == encodinglist.end())
619 if (!allowUnsafe && it->second.unsafe())
626 Encodings::fromLaTeXName(string const & n, int const & p, bool allowUnsafe) const
629 // FIXME: if we have to test for too many of these synonyms,
630 // we should instead extend the format of lib/encodings
634 // We don't use find_if because it makes copies of the pairs in
636 // This linear search is OK since we don't have many encodings.
637 // Users could even optimize it by putting the encodings they use
638 // most at the top of lib/encodings.
639 EncodingList::const_iterator const end = encodinglist.end();
640 for (EncodingList::const_iterator it = encodinglist.begin(); it != end; ++it)
641 if ((it->second.latexName() == name) && (it->second.package() & p)
642 && (!it->second.unsafe() || allowUnsafe))
649 Encodings::fromIconvName(string const & n, int const & p, bool allowUnsafe) const
651 EncodingList::const_iterator const end = encodinglist.end();
652 for (EncodingList::const_iterator it = encodinglist.begin(); it != end; ++it)
653 if ((it->second.iconvName() == n) && (it->second.package() & p)
654 && (!it->second.unsafe() || allowUnsafe))
660 Encodings::Encodings()
664 void Encodings::read(FileName const & encfile, FileName const & symbolsfile)
666 // We must read the symbolsfile first, because the Encoding
667 // constructor depends on it.
668 CharSetMap forcednotselected;
670 symbolslex.setFile(symbolsfile);
671 bool getNextToken = true;
672 while (symbolslex.isOK()) {
676 if (!symbolslex.next(true))
681 istringstream is(symbolslex.getString());
682 // reading symbol directly does not work if
683 // char_type == wchar_t.
685 if(!(is >> hex >> tmp))
689 if (!symbolslex.next(true))
691 docstring textcommand = symbolslex.getDocString();
692 if (!symbolslex.next(true))
694 string textpreamble = symbolslex.getString();
695 if (!symbolslex.next(true))
697 string sflags = symbolslex.getString();
702 if (suffixIs(textcommand, '}'))
703 flags |= CharInfoTextNoTermination;
704 while (!sflags.empty()) {
706 sflags = split(sflags, flag, ',');
707 if (flag == "combining") {
708 flags |= CharInfoCombining;
709 } else if (flag == "force") {
710 flags |= CharInfoForce;
711 forced.insert(symbol);
712 } else if (prefixIs(flag, "force=")) {
713 vector<string> encodings =
714 getVectorFromString(flag.substr(6), ";");
715 for (size_t i = 0; i < encodings.size(); ++i)
716 forcedselected[encodings[i]].insert(symbol);
717 flags |= CharInfoForceSelected;
718 } else if (prefixIs(flag, "force!=")) {
719 vector<string> encodings =
720 getVectorFromString(flag.substr(7), ";");
721 for (size_t i = 0; i < encodings.size(); ++i)
722 forcednotselected[encodings[i]].insert(symbol);
723 flags |= CharInfoForceSelected;
724 } else if (flag == "mathalpha") {
725 mathalpha.insert(symbol);
726 } else if (flag == "notermination=text") {
727 flags |= CharInfoTextNoTermination;
728 } else if (flag == "notermination=math") {
729 flags |= CharInfoMathNoTermination;
730 } else if (flag == "notermination=both") {
731 flags |= CharInfoTextNoTermination;
732 flags |= CharInfoMathNoTermination;
733 } else if (flag == "notermination=none") {
734 flags &= ~CharInfoTextNoTermination;
735 flags &= ~CharInfoMathNoTermination;
736 } else if (contains(flag, "tipashortcut=")) {
737 tipashortcut = split(flag, '=');
738 } else if (flag == "deprecated") {
739 flags |= CharInfoDeprecated;
741 lyxerr << "Ignoring unknown flag `" << flag
742 << "' for symbol `0x"
743 << hex << symbol << dec
747 // mathcommand and mathpreamble have been added for 1.6.0.
748 // make them optional so that old files still work.
749 int const lineno = symbolslex.lineNumber();
750 bool breakout = false;
751 docstring mathcommand;
753 if (symbolslex.next(true)) {
754 if (symbolslex.lineNumber() != lineno) {
755 // line in old format without mathcommand and mathpreamble
756 getNextToken = false;
758 mathcommand = symbolslex.getDocString();
759 if (suffixIs(mathcommand, '}'))
760 flags |= CharInfoMathNoTermination;
761 if (symbolslex.next(true)) {
762 if (symbolslex.lineNumber() != lineno) {
763 // line in new format with mathcommand only
764 getNextToken = false;
766 // line in new format with mathcommand and mathpreamble
767 mathpreamble = symbolslex.getString();
776 // backward compatibility
777 if (mathpreamble == "esintoramsmath")
778 mathpreamble = "esint|amsmath";
780 if (!textpreamble.empty())
781 if (textpreamble[0] != '\\')
782 flags |= CharInfoTextFeature;
783 if (!mathpreamble.empty())
784 if (mathpreamble[0] != '\\')
785 flags |= CharInfoMathFeature;
787 CharInfo info = CharInfo(
788 textcommand, mathcommand,
789 textpreamble, mathpreamble,
790 tipashortcut, flags);
791 LYXERR(Debug::INFO, "Read unicode symbol " << symbol << " '"
792 << to_utf8(info.textcommand()) << "' '" << info.textpreamble()
793 << " '" << info.textfeature() << ' ' << info.textnotermination()
794 << ' ' << to_utf8(info.mathcommand()) << "' '" << info.mathpreamble()
795 << "' " << info.mathfeature() << ' ' << info.mathnotermination()
796 << ' ' << info.combining() << ' ' << info.force()
797 << ' ' << info.forceselected());
799 // we assume that at least one command is nonempty when using unicodesymbols
800 if (info.isUnicodeSymbol()) {
801 unicodesymbols[symbol] = info;
808 // Now read the encodings
814 LexerKeyword encodingtags[] = {
815 { "encoding", et_encoding },
819 Lexer lex(encodingtags);
820 lex.setFile(encfile);
821 lex.setContext("Encodings::read");
827 string const name = lex.getString();
829 string const latexname = lex.getString();
831 string const guiname = lex.getString();
833 string const iconvname = lex.getString();
835 string const width = lex.getString();
836 bool fixedwidth = false;
838 if (width == "fixed")
840 else if (width == "variable")
842 else if (width == "variableunsafe") {
847 lex.printError("Unknown width");
850 string const p = lex.getString();
851 Encoding::Package package = Encoding::none;
853 package = Encoding::none;
854 else if (p == "inputenc")
855 package = Encoding::inputenc;
857 package = Encoding::CJK;
858 else if (p == "japanese")
859 package = Encoding::japanese;
861 lex.printError("Unknown package");
863 LYXERR(Debug::INFO, "Reading encoding " << name);
864 encodinglist[name] = Encoding(name, latexname,
865 guiname, iconvname, fixedwidth, unsafe,
868 if (lex.lex() != et_end)
869 lex.printError("Missing end");
873 lex.printError("Misplaced end");
875 case Lexer::LEX_FEOF:
878 lex.printError("Unknown tag");
883 // Move all information from forcednotselected to forcedselected
884 for (CharSetMap::const_iterator it1 = forcednotselected.begin(); it1 != forcednotselected.end(); ++it1) {
885 for (CharSetMap::iterator it2 = forcedselected.begin(); it2 != forcedselected.end(); ++it2) {
886 if (it2->first != it1->first)
887 it2->second.insert(it1->second.begin(), it1->second.end());