]> git.lyx.org Git - features.git/blob - src/mathed/MathParser.cpp
Fix bug #1337: LyX confused on $ symbol in math equations
[features.git] / src / mathed / MathParser.cpp
1 /**
2  * \file MathParser.cpp
3  * This file is part of LyX, the document processor.
4  * Licence details can be found in the file COPYING.
5  *
6  * \author André Pönitz
7  *
8  * Full author contact details are available in file CREDITS.
9  */
10
11 /*
12
13 If someone desperately needs partial "structures" (such as a few
14 cells of an array inset or similar) (s)he could uses the
15 following hack as starting point to write some macros:
16
17   \newif\ifcomment
18   \commentfalse
19   \ifcomment
20           \def\makeamptab{\catcode`\&=4\relax}
21           \def\makeampletter{\catcode`\&=11\relax}
22     \def\b{\makeampletter\expandafter\makeamptab\bi}
23     \long\def\bi#1\e{}
24   \else
25     \def\b{}\def\e{}
26   \fi
27   ...
28
29   \[\begin{array}{ccc}
30 1
31 &
32
33   \end{array}\]
34
35 */
36
37
38 #include <config.h>
39
40 #include "MathParser.h"
41
42 #include "InsetMathArray.h"
43 #include "InsetMathBig.h"
44 #include "InsetMathBrace.h"
45 #include "InsetMathChar.h"
46 #include "InsetMathColor.h"
47 #include "InsetMathComment.h"
48 #include "InsetMathDelim.h"
49 #include "InsetMathEnsureMath.h"
50 #include "InsetMathEnv.h"
51 #include "InsetMathFrac.h"
52 #include "InsetMathKern.h"
53 #include "MathMacro.h"
54 #include "InsetMathPar.h"
55 #include "InsetMathRef.h"
56 #include "InsetMathRoot.h"
57 #include "InsetMathScript.h"
58 #include "InsetMathSpace.h"
59 #include "InsetMathSplit.h"
60 #include "InsetMathSqrt.h"
61 #include "InsetMathString.h"
62 #include "InsetMathTabular.h"
63 #include "MathMacroTemplate.h"
64 #include "MathFactory.h"
65 #include "MathMacroArgument.h"
66 #include "MathSupport.h"
67
68 #include "Buffer.h"
69 #include "BufferParams.h"
70 #include "Encoding.h"
71 #include "Lexer.h"
72
73 #include "support/debug.h"
74 #include "support/convert.h"
75 #include "support/docstream.h"
76
77 #include <sstream>
78
79 //#define FILEDEBUG
80
81 using namespace std;
82
83 namespace lyx {
84
85 namespace {
86
87 InsetMath::mode_type asMode(InsetMath::mode_type oldmode, docstring const & str)
88 {
89         //lyxerr << "handling mode: '" << str << "'" << endl;
90         if (str == "mathmode")
91                 return InsetMath::MATH_MODE;
92         if (str == "textmode" || str == "forcetext")
93                 return InsetMath::TEXT_MODE;
94         return oldmode;
95 }
96
97
98 bool stared(docstring const & s)
99 {
100         size_t const n = s.size();
101         return n && s[n - 1] == '*';
102 }
103
104
105 docstring const repl(docstring const & oldstr, char_type const c,
106                      docstring const & macro, bool textmode = false)
107 {
108         docstring newstr;
109         size_t i;
110         size_t j;
111
112         for (i = 0, j = 0; i < oldstr.size(); ++i) {
113                 if (c == oldstr[i]) {
114                         newstr.append(oldstr, j, i - j);
115                         newstr.append(macro);
116                         j = i + 1;
117                         if (macro.size() > 2 && j < oldstr.size())
118                                 newstr += (textmode && oldstr[j] == ' ' ? '\\' : ' ');
119                 }
120         }
121
122         // Any substitution?
123         if (j == 0)
124                 return oldstr;
125
126         newstr.append(oldstr, j, i - j);
127         return newstr;
128 }
129
130
131 docstring escapeSpecialChars(docstring const & str, bool textmode)
132 {
133         docstring const backslash = textmode ? from_ascii("\\textbackslash")
134                                              : from_ascii("\\backslash");
135         docstring const caret = textmode ? from_ascii("\\textasciicircum")
136                                          : from_ascii("\\mathcircumflex");
137         docstring const tilde = textmode ? from_ascii("\\textasciitilde")
138                                          : from_ascii("\\sim");
139
140         return repl(repl(repl(repl(repl(repl(repl(repl(repl(repl(str,
141                         '\\', backslash, textmode),
142                         '^', caret, textmode),
143                         '~', tilde, textmode),
144                         '_', from_ascii("\\_")),
145                         '$', from_ascii("\\$")),
146                         '#', from_ascii("\\#")),
147                         '&', from_ascii("\\&")),
148                         '%', from_ascii("\\%")),
149                         '{', from_ascii("\\{")),
150                         '}', from_ascii("\\}"));
151 }
152
153
154 /*!
155  * Add the row \p cellrow to \p grid.
156  * \returns wether the row could be added. Adding a row can fail for
157  * environments like "equation" that have a fixed number of rows.
158  */
159 bool addRow(InsetMathGrid & grid, InsetMathGrid::row_type & cellrow,
160             docstring const & vskip, bool allow_newpage_ = true)
161 {
162         ++cellrow;
163         if (cellrow == grid.nrows()) {
164                 //lyxerr << "adding row " << cellrow << endl;
165                 grid.addRow(cellrow - 1);
166                 if (cellrow == grid.nrows()) {
167                         // We can't add a row to this grid, so let's
168                         // append the content of this cell to the previous
169                         // one.
170                         // This does not happen in well formed .lyx files,
171                         // but LyX versions 1.3.x and older could create
172                         // such files and tex2lyx can still do that.
173                         --cellrow;
174                         lyxerr << "ignoring extra row";
175                         if (!vskip.empty())
176                                 lyxerr << " with extra space " << to_utf8(vskip);
177                         if (!allow_newpage_)
178                                 lyxerr << " with no page break allowed";
179                         lyxerr << '.' << endl;
180                         return false;
181                 }
182         }
183         grid.vcrskip(Length(to_utf8(vskip)), cellrow - 1);
184         grid.rowinfo(cellrow - 1).allow_newpage_ = allow_newpage_;
185         return true;
186 }
187
188
189 /*!
190  * Add the column \p cellcol to \p grid.
191  * \returns wether the column could be added. Adding a column can fail for
192  * environments like "eqnarray" that have a fixed number of columns.
193  */
194 bool addCol(InsetMathGrid & grid, InsetMathGrid::col_type & cellcol)
195 {
196         ++cellcol;
197         if (cellcol == grid.ncols()) {
198                 //lyxerr << "adding column " << cellcol << endl;
199                 grid.addCol(cellcol);
200                 if (cellcol == grid.ncols()) {
201                         // We can't add a column to this grid, so let's
202                         // append the content of this cell to the previous
203                         // one.
204                         // This does not happen in well formed .lyx files,
205                         // but LyX versions 1.3.x and older could create
206                         // such files and tex2lyx can still do that.
207                         --cellcol;
208                         lyxerr << "ignoring extra column." << endl;
209                         return false;
210                 }
211         }
212         return true;
213 }
214
215
216 /*!
217  * Check whether the last row is empty and remove it if yes.
218  * Otherwise the following code
219  * \verbatim
220 \begin{array}{|c|c|}
221 \hline
222 1 & 2 \\ \hline
223 3 & 4 \\ \hline
224 \end{array}
225  * \endverbatim
226  * will result in a grid with 3 rows (+ the dummy row that is always present),
227  * because the last '\\' opens a new row.
228  * Note that this is only needed for inner-hull grid types, such as array
229  * or aligned, but not for outer-hull grid types, such as eqnarray or align.
230  */
231 void delEmptyLastRow(InsetMathGrid & grid)
232 {
233         InsetMathGrid::row_type const row = grid.nrows() - 1;
234         for (InsetMathGrid::col_type col = 0; col < grid.ncols(); ++col) {
235                 if (!grid.cell(grid.index(row, col)).empty())
236                         return;
237         }
238         // Copy the row information of the empty row (which would contain the
239         // last hline in the example above) to the dummy row and delete the
240         // empty row.
241         grid.rowinfo(row + 1) = grid.rowinfo(row);
242         grid.delRow(row);
243 }
244
245
246 /*!
247  * Tell whether the environment name corresponds to an inner-hull grid type.
248  */
249 bool innerHull(docstring const & name)
250 {
251         // For [bB]matrix, [vV]matrix, and pmatrix we can check the suffix only
252         return name == "array" || name == "cases" || name == "aligned"
253                 || name == "alignedat" || name == "gathered" || name == "split"
254                 || name == "subarray" || name == "tabular" || name == "matrix"
255                 || name.substr(1) == "matrix";
256 }
257
258
259 // These are TeX's catcodes
260 enum CatCode {
261         catEscape,     // 0    backslash
262         catBegin,      // 1    {
263         catEnd,        // 2    }
264         catMath,       // 3    $
265         catAlign,      // 4    &
266         catNewline,    // 5    ^^M
267         catParameter,  // 6    #
268         catSuper,      // 7    ^
269         catSub,        // 8    _
270         catIgnore,     // 9
271         catSpace,      // 10   space
272         catLetter,     // 11   a-zA-Z
273         catOther,      // 12   none of the above
274         catActive,     // 13   ~
275         catComment,    // 14   %
276         catInvalid     // 15   <delete>
277 };
278
279 CatCode theCatcode[128];
280
281
282 inline CatCode catcode(char_type c)
283 {
284         /* The only characters that are not catOther lie in the pure ASCII
285          * range. Therefore theCatcode has only 128 entries.
286          * TeX itself deals with 8bit characters, so if needed this table
287          * could be enlarged to 256 entries.
288          * Any larger value does not make sense, since the fact that we use
289          * unicode internally does not change Knuth's TeX engine.
290          * Apart from that a table for the full 21bit UCS4 range would waste
291          * too much memory. */
292         if (c >= 128)
293                 return catOther;
294
295         return theCatcode[c];
296 }
297
298
299 enum {
300         FLAG_ALIGN      = 1 << 0,  //  next & or \\ ends the parsing process
301         FLAG_BRACE_LAST = 1 << 1,  //  next closing brace ends the parsing
302         FLAG_RIGHT      = 1 << 2,  //  next \\right ends the parsing process
303         FLAG_END        = 1 << 3,  //  next \\end ends the parsing process
304         FLAG_BRACK_LAST = 1 << 4,  //  next closing bracket ends the parsing
305         FLAG_TEXTMODE   = 1 << 5,  //  we are in a box
306         FLAG_ITEM       = 1 << 6,  //  read a (possibly braced) token
307         FLAG_LEAVE      = 1 << 7,  //  leave the loop at the end
308         FLAG_SIMPLE     = 1 << 8,  //  next $ leaves the loop
309         FLAG_EQUATION   = 1 << 9,  //  next \] leaves the loop
310         FLAG_SIMPLE2    = 1 << 10, //  next \) leaves the loop
311         FLAG_OPTION     = 1 << 11, //  read [...] style option
312         FLAG_BRACED     = 1 << 12  //  read {...} style argument
313 };
314
315
316 //
317 // Helper class for parsing
318 //
319
320 class Token {
321 public:
322         ///
323         Token() : cs_(), char_(0), cat_(catIgnore) {}
324         ///
325         Token(char_type c, CatCode cat) : cs_(), char_(c), cat_(cat) {}
326         ///
327         explicit Token(docstring const & cs) : cs_(cs), char_(0), cat_(catIgnore) {}
328
329         ///
330         docstring const & cs() const { return cs_; }
331         ///
332         CatCode cat() const { return cat_; }
333         ///
334         char_type character() const { return char_; }
335         ///
336         docstring asString() const { return cs_.size() ? cs_ : docstring(1, char_); }
337         ///
338         docstring asInput() const { return cs_.size() ? '\\' + cs_ : docstring(1, char_); }
339
340 private:
341         ///
342         docstring cs_;
343         ///
344         char_type char_;
345         ///
346         CatCode cat_;
347 };
348
349
350 ostream & operator<<(ostream & os, Token const & t)
351 {
352         if (t.cs().size()) {
353                 docstring const & cs = t.cs();
354                 // FIXME: For some strange reason, the stream operator instanciate
355                 // a new Token before outputting the contents of t.cs().
356                 // Because of this the line
357                 //     os << '\\' << cs;
358                 // below becomes recursive.
359                 // In order to avoid that we return early:
360                 if (cs == "\\")
361                         return os;
362                 os << '\\' << to_utf8(cs);
363         }
364         else if (t.cat() == catLetter)
365                 os << t.character();
366         else
367                 os << '[' << t.character() << ',' << t.cat() << ']';
368         return os;
369 }
370
371
372 class Parser {
373 public:
374         ///
375         typedef  InsetMath::mode_type mode_type;
376         ///
377         typedef  Parse::flags parse_mode;
378
379         ///
380         Parser(Lexer & lex, parse_mode mode, Buffer * buf);
381         /// Only use this for reading from .lyx file format, for the reason
382         /// see Parser::tokenize(istream &).
383         Parser(istream & is, parse_mode mode, Buffer * buf);
384         ///
385         Parser(docstring const & str, parse_mode mode, Buffer * buf);
386
387         ///
388         bool parse(MathAtom & at);
389         ///
390         bool parse(MathData & array, unsigned flags, mode_type mode);
391         ///
392         bool parse1(InsetMathGrid & grid, unsigned flags, mode_type mode,
393                 bool numbered);
394         ///
395         MathData parse(unsigned flags, mode_type mode);
396         ///
397         int lineno() const { return lineno_; }
398         ///
399         void putback();
400
401 private:
402         ///
403         void parse2(MathAtom & at, unsigned flags, mode_type mode, bool numbered);
404         /// get arg delimited by 'left' and 'right'
405         docstring getArg(char_type left, char_type right);
406         ///
407         char_type getChar();
408         ///
409         void error(string const & msg);
410         void error(docstring const & msg) { error(to_utf8(msg)); }
411         /// dump contents to screen
412         void dump() const;
413         /// Only use this for reading from .lyx file format (see
414         /// implementation for reason)
415         void tokenize(istream & is);
416         ///
417         void tokenize(docstring const & s);
418         ///
419         void skipSpaceTokens(idocstream & is, char_type c);
420         ///
421         void push_back(Token const & t);
422         ///
423         void pop_back();
424         ///
425         Token const & prevToken() const;
426         ///
427         Token const & nextToken() const;
428         ///
429         Token const & getToken();
430         /// skips spaces if any
431         void skipSpaces();
432         ///
433         void lex(docstring const & s);
434         ///
435         bool good() const;
436         ///
437         docstring parse_verbatim_item();
438         ///
439         docstring parse_verbatim_option();
440
441         ///
442         int lineno_;
443         ///
444         vector<Token> tokens_;
445         ///
446         unsigned pos_;
447         /// Stack of active environments
448         vector<docstring> environments_;
449         ///
450         parse_mode mode_;
451         ///
452         bool success_;
453         ///
454         Buffer * buffer_;
455 };
456
457
458 Parser::Parser(Lexer & lexer, parse_mode mode, Buffer * buf)
459         : lineno_(lexer.lineNumber()), pos_(0), mode_(mode), success_(true),
460           buffer_(buf)
461 {
462         tokenize(lexer.getStream());
463         lexer.eatLine();
464 }
465
466
467 Parser::Parser(istream & is, parse_mode mode, Buffer * buf)
468         : lineno_(0), pos_(0), mode_(mode), success_(true), buffer_(buf)
469 {
470         tokenize(is);
471 }
472
473
474 Parser::Parser(docstring const & str, parse_mode mode, Buffer * buf)
475         : lineno_(0), pos_(0), mode_(mode), success_(true), buffer_(buf)
476 {
477         tokenize(str);
478 }
479
480
481 void Parser::push_back(Token const & t)
482 {
483         tokens_.push_back(t);
484 }
485
486
487 void Parser::pop_back()
488 {
489         tokens_.pop_back();
490 }
491
492
493 Token const & Parser::prevToken() const
494 {
495         static const Token dummy;
496         return pos_ > 0 ? tokens_[pos_ - 1] : dummy;
497 }
498
499
500 Token const & Parser::nextToken() const
501 {
502         static const Token dummy;
503         return good() ? tokens_[pos_] : dummy;
504 }
505
506
507 Token const & Parser::getToken()
508 {
509         static const Token dummy;
510         //lyxerr << "looking at token " << tokens_[pos_] << " pos: " << pos_ << endl;
511         return good() ? tokens_[pos_++] : dummy;
512 }
513
514
515 void Parser::skipSpaces()
516 {
517         while (nextToken().cat() == catSpace || nextToken().cat() == catNewline)
518                 getToken();
519 }
520
521
522 void Parser::putback()
523 {
524         --pos_;
525 }
526
527
528 bool Parser::good() const
529 {
530         return pos_ < tokens_.size();
531 }
532
533
534 char_type Parser::getChar()
535 {
536         if (!good()) {
537                 error("The input stream is not well...");
538                 putback();
539                 return 0;
540         }
541         return tokens_[pos_++].character();
542 }
543
544
545 docstring Parser::getArg(char_type left, char_type right)
546 {
547         skipSpaces();
548
549         docstring result;
550         char_type c = getChar();
551
552         if (c != left)
553                 putback();
554         else
555                 while ((c = getChar()) != right && good())
556                         result += c;
557
558         return result;
559 }
560
561
562 void Parser::skipSpaceTokens(idocstream & is, char_type c)
563 {
564         // skip trailing spaces
565         while (catcode(c) == catSpace || catcode(c) == catNewline)
566                 if (!is.get(c))
567                         break;
568         //lyxerr << "putting back: " << c << endl;
569         is.putback(c);
570 }
571
572
573 void Parser::tokenize(istream & is)
574 {
575         // eat everything up to the next \end_inset or end of stream
576         // and store it in s for further tokenization
577         string s;
578         char c;
579         while (is.get(c)) {
580                 s += c;
581                 if (s.size() >= 10 && s.substr(s.size() - 10) == "\\end_inset") {
582                         s = s.substr(0, s.size() - 10);
583                         break;
584                 }
585         }
586         // Remove the space after \end_inset
587         if (is.get(c) && c != ' ')
588                 is.unget();
589
590         // tokenize buffer
591         tokenize(from_utf8(s));
592 }
593
594
595 void Parser::tokenize(docstring const & buffer)
596 {
597         idocstringstream is(mode_ & Parse::VERBATIM
598                         ? escapeSpecialChars(buffer, mode_ & Parse::TEXTMODE)
599                         : buffer, ios::in | ios::binary);
600
601         char_type c;
602         while (is.get(c)) {
603                 //lyxerr << "reading c: " << c << endl;
604
605                 switch (catcode(c)) {
606                         case catNewline: {
607                                 ++lineno_;
608                                 is.get(c);
609                                 if (catcode(c) == catNewline)
610                                         ; //push_back(Token("par"));
611                                 else {
612                                         push_back(Token('\n', catNewline));
613                                         is.putback(c);
614                                 }
615                                 break;
616                         }
617
618 /*
619                         case catComment: {
620                                 while (is.get(c) && catcode(c) != catNewline)
621                                         ;
622                                 ++lineno_;
623                                 break;
624                         }
625 */
626
627                         case catEscape: {
628                                 is.get(c);
629                                 if (!is) {
630                                         error("unexpected end of input");
631                                 } else {
632                                         if (c == '\n')
633                                                 c = ' ';
634                                         docstring s(1, c);
635                                         if (catcode(c) == catLetter) {
636                                                 // collect letters
637                                                 while (is.get(c) && catcode(c) == catLetter)
638                                                         s += c;
639                                                 skipSpaceTokens(is, c);
640                                         }
641                                         push_back(Token(s));
642                                 }
643                                 break;
644                         }
645
646                         case catSuper:
647                         case catSub: {
648                                 push_back(Token(c, catcode(c)));
649                                 is.get(c);
650                                 skipSpaceTokens(is, c);
651                                 break;
652                         }
653
654                         case catIgnore: {
655                                 if (!(mode_ & Parse::QUIET))
656                                         lyxerr << "ignoring a char: " << int(c) << endl;
657                                 break;
658                         }
659
660                         default:
661                                 push_back(Token(c, catcode(c)));
662                 }
663         }
664
665 #ifdef FILEDEBUG
666         dump();
667 #endif
668 }
669
670
671 void Parser::dump() const
672 {
673         lyxerr << "\nTokens: ";
674         for (unsigned i = 0; i < tokens_.size(); ++i) {
675                 if (i == pos_)
676                         lyxerr << " <#> ";
677                 lyxerr << tokens_[i];
678         }
679         lyxerr << " pos: " << pos_ << endl;
680 }
681
682
683 void Parser::error(string const & msg)
684 {
685         success_ = false;
686         if (!(mode_ & Parse::QUIET)) {
687                 lyxerr << "Line ~" << lineno_ << ": Math parse error: "
688                        << msg << endl;
689                 dump();
690         }
691 }
692
693
694 bool Parser::parse(MathAtom & at)
695 {
696         skipSpaces();
697         MathData ar(buffer_);
698         parse(ar, false, InsetMath::UNDECIDED_MODE);
699         if (ar.size() != 1 || ar.front()->getType() == hullNone) {
700                 if (!(mode_ & Parse::QUIET))
701                         lyxerr << "unusual contents found: " << ar << endl;
702                 at = MathAtom(new InsetMathPar(buffer_, ar));
703                 //if (at->nargs() > 0)
704                 //      at.nucleus()->cell(0) = ar;
705                 //else
706                 //      lyxerr << "unusual contents found: " << ar << endl;
707                 success_ = false;
708         } else
709                 at = ar[0];
710         return success_;
711 }
712
713
714 docstring Parser::parse_verbatim_option()
715 {
716         skipSpaces();
717         docstring res;
718         if (nextToken().character() == '[') {
719                 Token t = getToken();
720                 for (Token t = getToken(); t.character() != ']' && good(); t = getToken()) {
721                         if (t.cat() == catBegin) {
722                                 putback();
723                                 res += '{' + parse_verbatim_item() + '}';
724                         } else
725                                 res += t.asInput();
726                 }
727         }
728         return res;
729 }
730
731
732 docstring Parser::parse_verbatim_item()
733 {
734         skipSpaces();
735         docstring res;
736         if (nextToken().cat() == catBegin) {
737                 Token t = getToken();
738                 for (Token t = getToken(); t.cat() != catEnd && good(); t = getToken()) {
739                         if (t.cat() == catBegin) {
740                                 putback();
741                                 res += '{' + parse_verbatim_item() + '}';
742                         }
743                         else
744                                 res += t.asInput();
745                 }
746         }
747         return res;
748 }
749
750
751 MathData Parser::parse(unsigned flags, mode_type mode)
752 {
753         MathData ar(buffer_);
754         parse(ar, flags, mode);
755         return ar;
756 }
757
758
759 bool Parser::parse(MathData & array, unsigned flags, mode_type mode)
760 {
761         InsetMathGrid grid(buffer_, 1, 1);
762         parse1(grid, flags, mode, false);
763         array = grid.cell(0);
764         return success_;
765 }
766
767
768 void Parser::parse2(MathAtom & at, const unsigned flags, const mode_type mode,
769         const bool numbered)
770 {
771         parse1(*(at.nucleus()->asGridInset()), flags, mode, numbered);
772 }
773
774
775 bool Parser::parse1(InsetMathGrid & grid, unsigned flags,
776         const mode_type mode, const bool numbered)
777 {
778         int limits = 0;
779         InsetMathGrid::row_type cellrow = 0;
780         InsetMathGrid::col_type cellcol = 0;
781         MathData * cell = &grid.cell(grid.index(cellrow, cellcol));
782         Buffer * buf = buffer_;
783
784         if (grid.asHullInset())
785                 grid.asHullInset()->numbered(cellrow, numbered);
786
787         //dump();
788         //lyxerr << " flags: " << flags << endl;
789         //lyxerr << " mode: " << mode  << endl;
790         //lyxerr << "grid: " << grid << endl;
791
792         while (good()) {
793                 Token const & t = getToken();
794
795 #ifdef FILEDEBUG
796                 lyxerr << "t: " << t << " flags: " << flags << endl;
797                 lyxerr << "mode: " << mode  << endl;
798                 cell->dump();
799                 lyxerr << endl;
800 #endif
801
802                 if (flags & FLAG_ITEM) {
803
804                         if (t.cat() == catBegin) {
805                                 // skip the brace and collect everything to the next matching
806                                 // closing brace
807                                 parse1(grid, FLAG_BRACE_LAST, mode, numbered);
808                                 return success_;
809                         }
810
811                         // handle only this single token, leave the loop if done
812                         flags = FLAG_LEAVE;
813                 }
814
815
816                 if (flags & FLAG_BRACED) {
817                         if (t.cat() == catSpace)
818                                 continue;
819
820                         if (t.cat() != catBegin) {
821                                 error("opening brace expected");
822                                 return success_;
823                         }
824
825                         // skip the brace and collect everything to the next matching
826                         // closing brace
827                         flags = FLAG_BRACE_LAST;
828                 }
829
830
831                 if (flags & FLAG_OPTION) {
832                         if (t.cat() == catOther && t.character() == '[') {
833                                 MathData ar;
834                                 parse(ar, FLAG_BRACK_LAST, mode);
835                                 cell->append(ar);
836                         } else {
837                                 // no option found, put back token and we are done
838                                 putback();
839                         }
840                         return success_;
841                 }
842
843                 //
844                 // cat codes
845                 //
846                 if (t.cat() == catMath) {
847                         if (mode != InsetMath::MATH_MODE) {
848                                 // we are inside some text mode thingy, so opening new math is allowed
849                                 Token const & n = getToken();
850                                 if (n.cat() == catMath) {
851                                         // TeX's $$...$$ syntax for displayed math
852                                         cell->push_back(MathAtom(new InsetMathHull(buf, hullEquation)));
853                                         parse2(cell->back(), FLAG_SIMPLE, InsetMath::MATH_MODE, false);
854                                         getToken(); // skip the second '$' token
855                                 } else {
856                                         // simple $...$  stuff
857                                         putback();
858                                         if (mode == InsetMath::UNDECIDED_MODE) {
859                                                 cell->push_back(MathAtom(new InsetMathHull(buf, hullSimple)));
860                                                 parse2(cell->back(), FLAG_SIMPLE, InsetMath::MATH_MODE, false);
861                                         } else {
862                                                 // Don't create nested math hulls (bug #5392)
863                                                 cell->push_back(MathAtom(new InsetMathEnsureMath(buf)));
864                                                 parse(cell->back().nucleus()->cell(0), FLAG_SIMPLE, InsetMath::MATH_MODE);
865                                         }
866                                 }
867                         }
868
869                         else if (flags & FLAG_SIMPLE) {
870                                 // this is the end of the formula
871                                 return success_;
872                         }
873
874                         else {
875                                 error("something strange in the parser");
876                                 break;
877                         }
878                 }
879
880                 else if (t.cat() == catLetter)
881                         cell->push_back(MathAtom(new InsetMathChar(t.character())));
882
883                 else if (t.cat() == catSpace && mode != InsetMath::MATH_MODE) {
884                         if (cell->empty() || cell->back()->getChar() != ' ')
885                                 cell->push_back(MathAtom(new InsetMathChar(t.character())));
886                 }
887
888                 else if (t.cat() == catNewline && mode != InsetMath::MATH_MODE) {
889                         if (cell->empty() || cell->back()->getChar() != ' ')
890                                 cell->push_back(MathAtom(new InsetMathChar(' ')));
891                 }
892
893                 else if (t.cat() == catParameter) {
894                         Token const & n = getToken();
895                         cell->push_back(MathAtom(new MathMacroArgument(n.character()-'0')));
896                 }
897
898                 else if (t.cat() == catActive)
899                         cell->push_back(MathAtom(new InsetMathChar(t.character())));
900
901                 else if (t.cat() == catBegin) {
902                         MathData ar;
903                         parse(ar, FLAG_BRACE_LAST, mode);
904                         // do not create a BraceInset if they were written by LyX
905                         // this helps to keep the annoyance of  "a choose b"  to a minimum
906                         if (ar.size() == 1 && ar[0]->extraBraces())
907                                 cell->append(ar);
908                         else
909                                 cell->push_back(MathAtom(new InsetMathBrace(ar)));
910                 }
911
912                 else if (t.cat() == catEnd) {
913                         if (flags & FLAG_BRACE_LAST)
914                                 return success_;
915                         error("found '}' unexpectedly");
916                         //LASSERT(false, /**/);
917                         //add(cell, '}', LM_TC_TEX);
918                 }
919
920                 else if (t.cat() == catAlign) {
921                         //lyxerr << " column now " << (cellcol + 1)
922                         //       << " max: " << grid.ncols() << endl;
923                         if (flags & FLAG_ALIGN)
924                                 return success_;
925                         if (addCol(grid, cellcol))
926                                 cell = &grid.cell(grid.index(cellrow, cellcol));
927                 }
928
929                 else if (t.cat() == catSuper || t.cat() == catSub) {
930                         bool up = (t.cat() == catSuper);
931                         // we need no new script inset if the last thing was a scriptinset,
932                         // which has that script already not the same script already
933                         if (!cell->size())
934                                 cell->push_back(MathAtom(new InsetMathScript(buf, up)));
935                         else if (cell->back()->asScriptInset() &&
936                                         !cell->back()->asScriptInset()->has(up))
937                                 cell->back().nucleus()->asScriptInset()->ensure(up);
938                         else if (cell->back()->asScriptInset())
939                                 cell->push_back(MathAtom(new InsetMathScript(buf, up)));
940                         else
941                                 cell->back() = MathAtom(new InsetMathScript(buf, cell->back(), up));
942                         InsetMathScript * p = cell->back().nucleus()->asScriptInset();
943                         // special handling of {}-bases
944                         // Here we could remove the brace inset for things
945                         // like {a'}^2 and add the braces back in
946                         // InsetMathScript::write().
947                         // We do not do it, since it is not possible to detect
948                         // reliably whether the braces are needed because the
949                         // nucleus contains more than one symbol, or whether
950                         // they are needed for unknown commands like \xx{a}_0
951                         // or \yy{a}{b}_0. This was done in revision 14819
952                         // in an unreliable way. See this thread
953                         // http://www.mail-archive.com/lyx-devel%40lists.lyx.org/msg104917.html
954                         // for more details.
955                         // However, we remove empty braces because they look
956                         // ugly on screen and we are sure that they were added
957                         // by the write() method (and will be re-added on save).
958                         if (p->nuc().size() == 1 &&
959                             p->nuc().back()->asBraceInset() &&
960                             p->nuc().back()->asBraceInset()->cell(0).empty())
961                                 p->nuc().erase(0);
962
963                         parse(p->cell(p->idxOfScript(up)), FLAG_ITEM, mode);
964                         if (limits) {
965                                 p->limits(limits);
966                                 limits = 0;
967                         }
968                 }
969
970                 else if (t.character() == ']' && (flags & FLAG_BRACK_LAST)) {
971                         //lyxerr << "finished reading option" << endl;
972                         return success_;
973                 }
974
975                 else if (t.cat() == catOther) {
976                         char_type c = t.character();
977                         if (isAsciiOrMathAlpha(c)
978                             || mode_ & Parse::VERBATIM
979                             || !(mode_ & Parse::USETEXT)
980                             || mode == InsetMath::TEXT_MODE) {
981                                 cell->push_back(MathAtom(new InsetMathChar(c)));
982                         } else {
983                                 MathAtom at = createInsetMath("text", buf);
984                                 at.nucleus()->cell(0).push_back(MathAtom(new InsetMathChar(c)));
985                                 while (nextToken().cat() == catOther
986                                        && !isAsciiOrMathAlpha(nextToken().character())) {
987                                         c = getToken().character();
988                                         at.nucleus()->cell(0).push_back(MathAtom(new InsetMathChar(c)));
989                                 }
990                                 cell->push_back(at);
991                         }
992                 }
993
994                 else if (t.cat() == catComment) {
995                         docstring s;
996                         while (good()) {
997                                 Token const & t = getToken();
998                                 if (t.cat() == catNewline)
999                                         break;
1000                                 s += t.asString();
1001                         }
1002                         cell->push_back(MathAtom(new InsetMathComment(buf, s)));
1003                         skipSpaces();
1004                 }
1005
1006                 //
1007                 // control sequences
1008                 //
1009
1010                 else if (t.cs() == "lyxlock") {
1011                         if (cell->size())
1012                                 cell->back().nucleus()->lock(true);
1013                 }
1014
1015                 else if ((t.cs() == "global" && nextToken().cs() == "def") ||
1016                          t.cs() == "def") {
1017                         if (t.cs() == "global")
1018                                 getToken();
1019                         
1020                         // get name
1021                         docstring name = getToken().cs();
1022                         
1023                         // read parameters
1024                         int nargs = 0;
1025                         docstring pars;
1026                         while (good() && nextToken().cat() != catBegin) {
1027                                 pars += getToken().cs();
1028                                 ++nargs;
1029                         }
1030                         nargs /= 2;
1031                         
1032                         // read definition
1033                         MathData def;
1034                         parse(def, FLAG_ITEM, InsetMath::UNDECIDED_MODE);
1035                         
1036                         // is a version for display attached?
1037                         skipSpaces();
1038                         MathData display;
1039                         if (nextToken().cat() == catBegin)
1040                                 parse(display, FLAG_ITEM, InsetMath::MATH_MODE);
1041                         
1042                         cell->push_back(MathAtom(new MathMacroTemplate(buf,
1043                                 name, nargs, 0, MacroTypeDef,
1044                                 vector<MathData>(), def, display)));
1045
1046                         if (buf && (mode_ & Parse::TRACKMACRO))
1047                                 buf->usermacros.insert(name);
1048                 }
1049                 
1050                 else if (t.cs() == "newcommand" ||
1051                          t.cs() == "renewcommand" ||
1052                          t.cs() == "newlyxcommand") {
1053                         // get name
1054                         if (getToken().cat() != catBegin) {
1055                                 error("'{' in \\newcommand expected (1) ");
1056                                 return success_;
1057                         }
1058                         docstring name = getToken().cs();
1059                         if (getToken().cat() != catEnd) {
1060                                 error("'}' in \\newcommand expected");
1061                                 return success_;
1062                         }
1063                                 
1064                         // get arity
1065                         docstring const arg = getArg('[', ']');
1066                         int nargs = 0;
1067                         if (!arg.empty())
1068                                 nargs = convert<int>(arg);
1069                                 
1070                         // optional argument given?
1071                         skipSpaces();
1072                         int optionals = 0;
1073                         vector<MathData> optionalValues;
1074                         while (nextToken().character() == '[') {
1075                                 getToken();
1076                                 optionalValues.push_back(MathData());
1077                                 parse(optionalValues[optionals], FLAG_BRACK_LAST, mode);
1078                                 ++optionals;
1079                         }
1080                         
1081                         MathData def;
1082                         parse(def, FLAG_ITEM, InsetMath::UNDECIDED_MODE);
1083                         
1084                         // is a version for display attached?
1085                         skipSpaces();
1086                         MathData display;
1087                         if (nextToken().cat() == catBegin)
1088                                 parse(display, FLAG_ITEM, InsetMath::MATH_MODE);
1089                         
1090                         cell->push_back(MathAtom(new MathMacroTemplate(buf,
1091                                 name, nargs, optionals, MacroTypeNewcommand,
1092                                 optionalValues, def, display)));
1093
1094                         if (buf && (mode_ & Parse::TRACKMACRO))
1095                                 buf->usermacros.insert(name);
1096                 }
1097                 
1098                 else if (t.cs() == "newcommandx" ||
1099                          t.cs() == "renewcommandx") {
1100                         // \newcommandx{\foo}[2][usedefault, addprefix=\global,1=default]{#1,#2}
1101                         // get name
1102                         docstring name;
1103                         if (nextToken().cat() == catBegin) {
1104                                 getToken();
1105                                 name = getToken().cs();
1106                                 if (getToken().cat() != catEnd) {
1107                                         error("'}' in \\newcommandx expected");
1108                                         return success_;
1109                                 }
1110                         } else
1111                                 name = getToken().cs();
1112                                 
1113                         // get arity
1114                         docstring const arg = getArg('[', ']');
1115                         if (arg.empty()) {
1116                                 error("[num] in \\newcommandx expected");
1117                                 return success_;
1118                         }
1119                         int nargs = convert<int>(arg);
1120                         
1121                         // get options
1122                         int optionals = 0;
1123                         vector<MathData> optionalValues;
1124                         if (nextToken().character() == '[') {
1125                                 // skip '['
1126                                 getToken();
1127                                         
1128                                 // handle 'opt=value' options, separated by ','.
1129                                 skipSpaces();
1130                                 while (nextToken().character() != ']' && good()) {
1131                                         if (nextToken().character() >= '1'
1132                                             && nextToken().character() <= '9') {
1133                                                 // optional value -> get parameter number
1134                                                 int n = getChar() - '0';
1135                                                 if (n > nargs) {
1136                                                         error("Arity of \\newcommandx too low "
1137                                                               "for given optional parameter.");
1138                                                         return success_;
1139                                                 }
1140                                                 
1141                                                 // skip '='
1142                                                 if (getToken().character() != '=') {
1143                                                         error("'=' and optional parameter value "
1144                                                               "expected for \\newcommandx");
1145                                                         return success_;
1146                                                 }
1147                                                 
1148                                                 // get value
1149                                                 int optNum = max(size_t(n), optionalValues.size());
1150                                                 optionalValues.resize(optNum);
1151                                                 optionalValues[n - 1].clear();
1152                                                 while (nextToken().character() != ']'
1153                                                        && nextToken().character() != ',') {
1154                                                         MathData data;
1155                                                         parse(data, FLAG_ITEM, InsetMath::UNDECIDED_MODE);
1156                                                         optionalValues[n - 1].append(data);
1157                                                 }
1158                                                 optionals = max(n, optionals);
1159                                         } else if (nextToken().cat() == catLetter) {
1160                                                 // we in fact ignore every non-optional
1161                                                 // parameter
1162                                                 
1163                                                 // get option name
1164                                                 docstring opt;
1165                                                 while (nextToken().cat() == catLetter)
1166                                                         opt += getChar();
1167                                         
1168                                                 // value?
1169                                                 skipSpaces();
1170                                                 MathData value;
1171                                                 if (nextToken().character() == '=') {
1172                                                         getToken();
1173                                                         while (nextToken().character() != ']'
1174                                                                 && nextToken().character() != ',')
1175                                                                 parse(value, FLAG_ITEM, 
1176                                                                       InsetMath::UNDECIDED_MODE);
1177                                                 }
1178                                         } else {
1179                                                 error("option for \\newcommandx expected");
1180                                                 return success_;
1181                                         }
1182                                         
1183                                         // skip komma
1184                                         skipSpaces();
1185                                         if (nextToken().character() == ',') {
1186                                                 getChar();
1187                                                 skipSpaces();
1188                                         } else if (nextToken().character() != ']') {
1189                                                 error("Expecting ',' or ']' in options "
1190                                                       "of \\newcommandx");
1191                                                 return success_;
1192                                         }
1193                                 }
1194                                 
1195                                 // skip ']'
1196                                 if (!good())
1197                                         return success_;
1198                                 getToken();
1199                         }
1200
1201                         // get definition
1202                         MathData def;
1203                         parse(def, FLAG_ITEM, InsetMath::UNDECIDED_MODE);
1204
1205                         // is a version for display attached?
1206                         skipSpaces();
1207                         MathData display;
1208                         if (nextToken().cat() == catBegin)
1209                                 parse(display, FLAG_ITEM, InsetMath::MATH_MODE);
1210
1211                         cell->push_back(MathAtom(new MathMacroTemplate(buf,
1212                                 name, nargs, optionals, MacroTypeNewcommandx,
1213                                 optionalValues, def, display)));
1214
1215                         if (buf && (mode_ & Parse::TRACKMACRO))
1216                                 buf->usermacros.insert(name);
1217                 }
1218
1219                 else if (t.cs() == "(") {
1220                         cell->push_back(MathAtom(new InsetMathEnsureMath(buf)));
1221                         parse(cell->back().nucleus()->cell(0), FLAG_SIMPLE2, InsetMath::MATH_MODE);
1222                 }
1223
1224                 else if (t.cs() == "[") {
1225                         if (mode != InsetMath::UNDECIDED_MODE) {
1226                                 error("bad math environment");
1227                                 break;
1228                         }
1229                         cell->push_back(MathAtom(new InsetMathHull(buf, hullEquation)));
1230                         parse2(cell->back(), FLAG_EQUATION, InsetMath::MATH_MODE, false);
1231                 }
1232
1233                 else if (t.cs() == "protect")
1234                         // ignore \\protect, will hopefully be re-added during output
1235                         ;
1236
1237                 else if (t.cs() == "end") {
1238                         if (flags & FLAG_END) {
1239                                 // eat environment name
1240                                 docstring const name = getArg('{', '}');
1241                                 if (environments_.empty())
1242                                         error("'found \\end{" + name +
1243                                               "}' without matching '\\begin{" +
1244                                               name + "}'");
1245                                 else if (name != environments_.back())
1246                                         error("'\\end{" + name +
1247                                               "}' does not match '\\begin{" +
1248                                               environments_.back() + "}'");
1249                                 else {
1250                                         environments_.pop_back();
1251                                         // Delete empty last row in matrix
1252                                         // like insets.
1253                                         // If you abuse InsetMathGrid for
1254                                         // non-matrix like structures you
1255                                         // probably need to refine this test.
1256                                         // Right now we only have to test for
1257                                         // single line hull insets.
1258                                         if (grid.nrows() > 1 && innerHull(name))
1259                                                 delEmptyLastRow(grid);
1260                                         return success_;
1261                                 }
1262                         } else
1263                                 error("found 'end' unexpectedly");
1264                 }
1265
1266                 else if (t.cs() == ")") {
1267                         if (flags & FLAG_SIMPLE2)
1268                                 return success_;
1269                         error("found '\\)' unexpectedly");
1270                 }
1271
1272                 else if (t.cs() == "]") {
1273                         if (flags & FLAG_EQUATION)
1274                                 return success_;
1275                         error("found '\\]' unexpectedly");
1276                 }
1277
1278                 else if (t.cs() == "\\") {
1279                         if (flags & FLAG_ALIGN)
1280                                 return success_;
1281                         bool added = false;
1282                         if (nextToken().asInput() == "*") {
1283                                 getToken();
1284                                 added = addRow(grid, cellrow, docstring(), false);
1285                         } else if (good())
1286                                 added = addRow(grid, cellrow, getArg('[', ']'));
1287                         else
1288                                 error("missing token after \\\\");
1289                         if (added) {
1290                                 cellcol = 0;
1291                                 if (grid.asHullInset())
1292                                         grid.asHullInset()->numbered(
1293                                                         cellrow, numbered);
1294                                 cell = &grid.cell(grid.index(cellrow,
1295                                                              cellcol));
1296                         }
1297                 }
1298
1299 #if 0
1300                 else if (t.cs() == "multicolumn") {
1301                         // extract column count and insert dummy cells
1302                         MathData count;
1303                         parse(count, FLAG_ITEM, mode);
1304                         int cols = 1;
1305                         if (!extractNumber(count, cols)) {
1306                                 success_ = false;
1307                                 lyxerr << " can't extract number of cells from " << count << endl;
1308                         }
1309                         // resize the table if necessary
1310                         for (int i = 0; i < cols; ++i) {
1311                                 if (addCol(grid, cellcol)) {
1312                                         cell = &grid.cell(grid.index(
1313                                                         cellrow, cellcol));
1314                                         // mark this as dummy
1315                                         grid.cellinfo(grid.index(
1316                                                 cellrow, cellcol)).dummy_ = true;
1317                                 }
1318                         }
1319                         // the last cell is the real thing, not a dummy
1320                         grid.cellinfo(grid.index(cellrow, cellcol)).dummy_ = false;
1321
1322                         // read special alignment
1323                         MathData align;
1324                         parse(align, FLAG_ITEM, mode);
1325                         //grid.cellinfo(grid.index(cellrow, cellcol)).align_ = extractString(align);
1326
1327                         // parse the remaining contents into the "real" cell
1328                         parse(*cell, FLAG_ITEM, mode);
1329                 }
1330 #endif
1331
1332                 else if (t.cs() == "limits" || t.cs() == "nolimits") {
1333                         CatCode cat = nextToken().cat();
1334                         if (cat == catSuper || cat == catSub)
1335                                 limits = t.cs() == "limits" ? 1 : -1;
1336                         else {
1337                                 MathAtom at = createInsetMath(t.cs(), buf);
1338                                 cell->push_back(at);
1339                         }
1340                 }
1341
1342                 else if (t.cs() == "nonumber") {
1343                         if (grid.asHullInset())
1344                                 grid.asHullInset()->numbered(cellrow, false);
1345                 }
1346
1347                 else if (t.cs() == "number") {
1348                         if (grid.asHullInset())
1349                                 grid.asHullInset()->numbered(cellrow, true);
1350                 }
1351
1352                 else if (t.cs() == "hline") {
1353                         grid.rowinfo(cellrow).lines_ ++;
1354                 }
1355
1356                 else if (t.cs() == "sqrt") {
1357                         MathData ar;
1358                         parse(ar, FLAG_OPTION, mode);
1359                         if (ar.size()) {
1360                                 cell->push_back(MathAtom(new InsetMathRoot(buf)));
1361                                 cell->back().nucleus()->cell(0) = ar;
1362                                 parse(cell->back().nucleus()->cell(1), FLAG_ITEM, mode);
1363                         } else {
1364                                 cell->push_back(MathAtom(new InsetMathSqrt(buf)));
1365                                 parse(cell->back().nucleus()->cell(0), FLAG_ITEM, mode);
1366                         }
1367                 }
1368
1369                 else if (t.cs() == "unit") {
1370                         // Allowed formats \unit[val]{unit}
1371                         MathData ar;
1372                         parse(ar, FLAG_OPTION, mode);
1373                         if (ar.size()) {
1374                                 cell->push_back(MathAtom(new InsetMathFrac(buf, InsetMathFrac::UNIT)));
1375                                 cell->back().nucleus()->cell(0) = ar;
1376                                 parse(cell->back().nucleus()->cell(1), FLAG_ITEM, mode);
1377                         } else {
1378                                 cell->push_back(MathAtom(new InsetMathFrac(buf, InsetMathFrac::UNIT, 1)));
1379                                 parse(cell->back().nucleus()->cell(0), FLAG_ITEM, mode);
1380                         }
1381                 }
1382
1383                 else if (t.cs() == "unitfrac") {
1384                         // Here allowed formats are \unitfrac[val]{num}{denom}
1385                         MathData ar;
1386                         parse(ar, FLAG_OPTION, mode);
1387                         if (ar.size()) {
1388                                 cell->push_back(MathAtom(new InsetMathFrac(buf, InsetMathFrac::UNITFRAC, 3)));
1389                                 cell->back().nucleus()->cell(2) = ar;
1390                         } else {
1391                                 cell->push_back(MathAtom(new InsetMathFrac(buf, InsetMathFrac::UNITFRAC)));
1392                         }
1393                         parse(cell->back().nucleus()->cell(0), FLAG_ITEM, mode);
1394                         parse(cell->back().nucleus()->cell(1), FLAG_ITEM, mode);
1395                 }
1396
1397                 else if (t.cs() == "cfrac") {
1398                         // allowed formats are \cfrac[pos]{num}{denom}
1399                         docstring const arg = getArg('[', ']');
1400                         //lyxerr << "got so far: '" << arg << "'" << endl;                              
1401                                 if (arg == "l")
1402                                         cell->push_back(MathAtom(new InsetMathFrac(buf, InsetMathFrac::CFRACLEFT)));
1403                                 else if (arg == "r")
1404                                         cell->push_back(MathAtom(new InsetMathFrac(buf, InsetMathFrac::CFRACRIGHT)));
1405                                 else if (arg.empty() || arg == "c")
1406                                         cell->push_back(MathAtom(new InsetMathFrac(buf, InsetMathFrac::CFRAC)));
1407                                 else {
1408                                         error("found invalid optional argument");
1409                                         break;
1410                                 }
1411                         parse(cell->back().nucleus()->cell(0), FLAG_ITEM, mode);
1412                         parse(cell->back().nucleus()->cell(1), FLAG_ITEM, mode);
1413                 }
1414
1415                 else if (t.cs() == "xrightarrow" || t.cs() == "xleftarrow") {
1416                         cell->push_back(createInsetMath(t.cs(), buf));
1417                         parse(cell->back().nucleus()->cell(1), FLAG_OPTION, mode);
1418                         parse(cell->back().nucleus()->cell(0), FLAG_ITEM, mode);
1419                 }
1420
1421                 else if (t.cs() == "ref" || t.cs() == "eqref" || t.cs() == "prettyref"
1422                           || t.cs() == "pageref" || t.cs() == "vpageref" || t.cs() == "vref") {
1423                         cell->push_back(MathAtom(new InsetMathRef(buf, t.cs())));
1424                         docstring const opt = parse_verbatim_option();
1425                         docstring const ref = parse_verbatim_item();
1426                         if (!opt.empty()) {
1427                                 cell->back().nucleus()->cell(1).push_back(
1428                                         MathAtom(new InsetMathString(opt)));
1429                         }
1430                         cell->back().nucleus()->cell(0).push_back(
1431                                         MathAtom(new InsetMathString(ref)));
1432                 }
1433
1434                 else if (t.cs() == "left") {
1435                         skipSpaces();
1436                         Token const & tl = getToken();
1437                         // \| and \Vert are equivalent, and InsetMathDelim
1438                         // can't handle \|
1439                         // FIXME: fix this in InsetMathDelim itself!
1440                         docstring const l = tl.cs() == "|" ? from_ascii("Vert") : tl.asString();
1441                         MathData ar;
1442                         parse(ar, FLAG_RIGHT, mode);
1443                         if (!good())
1444                                 break;
1445                         skipSpaces();
1446                         Token const & tr = getToken();
1447                         docstring const r = tr.cs() == "|" ? from_ascii("Vert") : tr.asString();
1448                         cell->push_back(MathAtom(new InsetMathDelim(buf, l, r, ar)));
1449                 }
1450
1451                 else if (t.cs() == "right") {
1452                         if (flags & FLAG_RIGHT)
1453                                 return success_;
1454                         //lyxerr << "got so far: '" << cell << "'" << endl;
1455                         error("Unmatched right delimiter");
1456                         return success_;
1457                 }
1458
1459                 else if (t.cs() == "begin") {
1460                         docstring const name = getArg('{', '}');
1461                         environments_.push_back(name);
1462
1463                         if (name == "array" || name == "subarray") {
1464                                 docstring const valign = parse_verbatim_option() + 'c';
1465                                 docstring const halign = parse_verbatim_item();
1466                                 cell->push_back(MathAtom(new InsetMathArray(buf, name,
1467                                         InsetMathGrid::guessColumns(halign), 1, (char)valign[0], halign)));
1468                                 parse2(cell->back(), FLAG_END, mode, false);
1469                         }
1470
1471                         else if (name == "tabular") {
1472                                 docstring const valign = parse_verbatim_option() + 'c';
1473                                 docstring const halign = parse_verbatim_item();
1474                                 cell->push_back(MathAtom(new InsetMathTabular(buf, name,
1475                                         InsetMathGrid::guessColumns(halign), 1, (char)valign[0], halign)));
1476                                 parse2(cell->back(), FLAG_END, InsetMath::TEXT_MODE, false);
1477                         }
1478
1479                         else if (name == "split" || name == "cases") {
1480                                 cell->push_back(createInsetMath(name, buf));
1481                                 parse2(cell->back(), FLAG_END, mode, false);
1482                         }
1483
1484                         else if (name == "alignedat") {
1485                                 docstring const valign = parse_verbatim_option() + 'c';
1486                                 // ignore this for a while
1487                                 getArg('{', '}');
1488                                 cell->push_back(MathAtom(new InsetMathSplit(buf, name, (char)valign[0])));
1489                                 parse2(cell->back(), FLAG_END, mode, false);
1490                         }
1491
1492                         else if (name == "math") {
1493                                 cell->push_back(MathAtom(new InsetMathEnsureMath(buf)));
1494                                 parse(cell->back().nucleus()->cell(0), FLAG_END, InsetMath::MATH_MODE);
1495                         }
1496
1497                         else if (name == "equation" || name == "equation*"
1498                                         || name == "displaymath") {
1499                                 if (mode != InsetMath::UNDECIDED_MODE) {
1500                                         error("bad math environment");
1501                                         break;
1502                                 }
1503                                 cell->push_back(MathAtom(new InsetMathHull(buf, hullEquation)));
1504                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, (name == "equation"));
1505                         }
1506
1507                         else if (name == "eqnarray" || name == "eqnarray*") {
1508                                 if (mode != InsetMath::UNDECIDED_MODE) {
1509                                         error("bad math environment");
1510                                         break;
1511                                 }
1512                                 cell->push_back(MathAtom(new InsetMathHull(buf, hullEqnArray)));
1513                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1514                         }
1515
1516                         else if (name == "align" || name == "align*") {
1517                                 if (mode != InsetMath::UNDECIDED_MODE) {
1518                                         error("bad math environment");
1519                                         break;
1520                                 }
1521                                 cell->push_back(MathAtom(new InsetMathHull(buf, hullAlign)));
1522                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1523                         }
1524
1525                         else if (name == "flalign" || name == "flalign*") {
1526                                 if (mode != InsetMath::UNDECIDED_MODE) {
1527                                         error("bad math environment");
1528                                         break;
1529                                 }
1530                                 cell->push_back(MathAtom(new InsetMathHull(buf, hullFlAlign)));
1531                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1532                         }
1533
1534                         else if (name == "alignat" || name == "alignat*") {
1535                                 if (mode != InsetMath::UNDECIDED_MODE) {
1536                                         error("bad math environment");
1537                                         break;
1538                                 }
1539                                 // ignore this for a while
1540                                 getArg('{', '}');
1541                                 cell->push_back(MathAtom(new InsetMathHull(buf, hullAlignAt)));
1542                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1543                         }
1544
1545                         else if (name == "xalignat" || name == "xalignat*") {
1546                                 if (mode != InsetMath::UNDECIDED_MODE) {
1547                                         error("bad math environment");
1548                                         break;
1549                                 }
1550                                 // ignore this for a while
1551                                 getArg('{', '}');
1552                                 cell->push_back(MathAtom(new InsetMathHull(buf, hullXAlignAt)));
1553                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1554                         }
1555
1556                         else if (name == "xxalignat") {
1557                                 if (mode != InsetMath::UNDECIDED_MODE) {
1558                                         error("bad math environment");
1559                                         break;
1560                                 }
1561                                 // ignore this for a while
1562                                 getArg('{', '}');
1563                                 cell->push_back(MathAtom(new InsetMathHull(buf, hullXXAlignAt)));
1564                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1565                         }
1566
1567                         else if (name == "multline" || name == "multline*") {
1568                                 if (mode != InsetMath::UNDECIDED_MODE) {
1569                                         error("bad math environment");
1570                                         break;
1571                                 }
1572                                 cell->push_back(MathAtom(new InsetMathHull(buf, hullMultline)));
1573                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1574                         }
1575
1576                         else if (name == "gather" || name == "gather*") {
1577                                 if (mode != InsetMath::UNDECIDED_MODE) {
1578                                         error("bad math environment");
1579                                         break;
1580                                 }
1581                                 cell->push_back(MathAtom(new InsetMathHull(buf, hullGather)));
1582                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1583                         }
1584
1585                         else if (latexkeys const * l = in_word_set(name)) {
1586                                 if (l->inset == "matrix") {
1587                                         cell->push_back(createInsetMath(name, buf));
1588                                         parse2(cell->back(), FLAG_END, mode, false);
1589                                 } else if (l->inset == "split") {
1590                                         docstring const valign = parse_verbatim_option() + 'c';
1591                                         cell->push_back(MathAtom(
1592                                                 new InsetMathSplit(buf, name, (char)valign[0])));
1593                                         parse2(cell->back(), FLAG_END, mode, false);
1594                                 } else {
1595                                         success_ = false;
1596                                         if (!(mode_ & Parse::QUIET)) {
1597                                                 dump();
1598                                                 lyxerr << "found math environment `"
1599                                                        << to_utf8(name)
1600                                                        << "' in symbols file with unsupported inset `"
1601                                                        << to_utf8(l->inset)
1602                                                        << "'." << endl;
1603                                         }
1604                                         // create generic environment inset
1605                                         cell->push_back(MathAtom(new InsetMathEnv(buf, name)));
1606                                         parse(cell->back().nucleus()->cell(0), FLAG_END, mode);
1607                                 }
1608                         }
1609
1610                         else {
1611                                 success_ = false;
1612                                 if (!(mode_ & Parse::QUIET)) {
1613                                         dump();
1614                                         lyxerr << "found unknown math environment '"
1615                                                << to_utf8(name) << "'" << endl;
1616                                 }
1617                                 // create generic environment inset
1618                                 cell->push_back(MathAtom(new InsetMathEnv(buf, name)));
1619                                 parse(cell->back().nucleus()->cell(0), FLAG_END, mode);
1620                         }
1621                 }
1622
1623                 else if (t.cs() == "kern") {
1624                         // FIXME: A hack...
1625                         docstring s;
1626                         int num_tokens = 0;
1627                         while (true) {
1628                                 Token const & t = getToken();
1629                                 ++num_tokens;
1630                                 if (!good()) {
1631                                         s.clear();
1632                                         while (num_tokens--)
1633                                                 putback();
1634                                         break;
1635                                 }
1636                                 s += t.character();
1637                                 if (isValidLength(to_utf8(s)))
1638                                         break;
1639                         }
1640                         if (s.empty())
1641                                 cell->push_back(MathAtom(new InsetMathKern));
1642                         else
1643                                 cell->push_back(MathAtom(new InsetMathKern(s)));
1644                 }
1645
1646                 else if (t.cs() == "label") {
1647                         // FIXME: This is swallowed in inline formulas
1648                         docstring label = parse_verbatim_item();
1649                         MathData ar;
1650                         asArray(label, ar);
1651                         if (grid.asHullInset()) {
1652                                 grid.asHullInset()->label(cellrow, label);
1653                         } else {
1654                                 cell->push_back(createInsetMath(t.cs(), buf));
1655                                 cell->push_back(MathAtom(new InsetMathBrace(ar)));
1656                         }
1657                 }
1658
1659                 else if (t.cs() == "choose" || t.cs() == "over"
1660                                 || t.cs() == "atop" || t.cs() == "brace"
1661                                 || t.cs() == "brack") {
1662                         MathAtom at = createInsetMath(t.cs(), buf);
1663                         at.nucleus()->cell(0) = *cell;
1664                         cell->clear();
1665                         parse(at.nucleus()->cell(1), flags, mode);
1666                         cell->push_back(at);
1667                         return success_;
1668                 }
1669
1670                 else if (t.cs() == "color") {
1671                         docstring const color = parse_verbatim_item();
1672                         cell->push_back(MathAtom(new InsetMathColor(buf, true, color)));
1673                         parse(cell->back().nucleus()->cell(0), flags, mode);
1674                         return success_;
1675                 }
1676
1677                 else if (t.cs() == "textcolor") {
1678                         docstring const color = parse_verbatim_item();
1679                         cell->push_back(MathAtom(new InsetMathColor(buf, false, color)));
1680                         parse(cell->back().nucleus()->cell(0), FLAG_ITEM, InsetMath::TEXT_MODE);
1681                 }
1682
1683                 else if (t.cs() == "normalcolor") {
1684                         cell->push_back(createInsetMath(t.cs(), buf));
1685                         parse(cell->back().nucleus()->cell(0), flags, mode);
1686                         return success_;
1687                 }
1688
1689                 else if (t.cs() == "substack") {
1690                         cell->push_back(createInsetMath(t.cs(), buf));
1691                         parse2(cell->back(), FLAG_ITEM, mode, false);
1692                         // Delete empty last row if present
1693                         InsetMathGrid & subgrid =
1694                                 *(cell->back().nucleus()->asGridInset());
1695                         if (subgrid.nrows() > 1)
1696                                 delEmptyLastRow(subgrid);
1697                 }
1698
1699                 else if (t.cs() == "xymatrix") {
1700                         odocstringstream os;
1701                         while (good() && nextToken().cat() != catBegin)
1702                                 os << getToken().asInput();
1703                         cell->push_back(createInsetMath(t.cs() + os.str(), buf));
1704                         parse2(cell->back(), FLAG_ITEM, mode, false);
1705                         // Delete empty last row if present
1706                         InsetMathGrid & subgrid =
1707                                 *(cell->back().nucleus()->asGridInset());
1708                         if (subgrid.nrows() > 1)
1709                                 delEmptyLastRow(subgrid);
1710                 }
1711
1712                 else if (t.cs() == "framebox" || t.cs() == "makebox") {
1713                         cell->push_back(createInsetMath(t.cs(), buf));
1714                         parse(cell->back().nucleus()->cell(0), FLAG_OPTION, InsetMath::TEXT_MODE);
1715                         parse(cell->back().nucleus()->cell(1), FLAG_OPTION, InsetMath::TEXT_MODE);
1716                         parse(cell->back().nucleus()->cell(2), FLAG_ITEM, InsetMath::TEXT_MODE);
1717                 }
1718
1719                 else if (t.cs() == "tag") {
1720                         if (nextToken().character() == '*') {
1721                                 getToken();
1722                                 cell->push_back(createInsetMath(t.cs() + '*', buf));
1723                         } else
1724                                 cell->push_back(createInsetMath(t.cs(), buf));
1725                         parse(cell->back().nucleus()->cell(0), FLAG_ITEM, InsetMath::TEXT_MODE);
1726                 }
1727
1728                 else if (t.cs() == "hspace" && nextToken().character() != '*') {
1729                         docstring const name = t.cs();
1730                         docstring const arg = parse_verbatim_item();
1731                         Length length;
1732                         if (isValidLength(to_utf8(arg), &length))
1733                                 cell->push_back(MathAtom(new InsetMathSpace(length)));
1734                         else {
1735                                 // Since the Length class cannot use length variables
1736                                 // we must not create an InsetMathSpace.
1737                                 cell->push_back(MathAtom(new MathMacro(buf, name)));
1738                                 MathData ar;
1739                                 mathed_parse_cell(ar, '{' + arg + '}', mode_);
1740                                 cell->append(ar);
1741                         }
1742                 }
1743
1744 #if 0
1745                 else if (t.cs() == "infer") {
1746                         MathData ar;
1747                         parse(ar, FLAG_OPTION, mode);
1748                         cell->push_back(createInsetMath(t.cs(), buf));
1749                         parse2(cell->back(), FLAG_ITEM, mode, false);
1750                 }
1751
1752                 // Disabled
1753                 else if (1 && t.cs() == "ar") {
1754                         auto_ptr<InsetMathXYArrow> p(new InsetMathXYArrow);
1755                         // try to read target
1756                         parse(p->cell(0), FLAG_OTPTION, mode);
1757                         // try to read label
1758                         if (nextToken().cat() == catSuper || nextToken().cat() == catSub) {
1759                                 p->up_ = nextToken().cat() == catSuper;
1760                                 getToken();
1761                                 parse(p->cell(1), FLAG_ITEM, mode);
1762                                 //lyxerr << "read label: " << p->cell(1) << endl;
1763                         }
1764
1765                         cell->push_back(MathAtom(p.release()));
1766                         //lyxerr << "read cell: " << cell << endl;
1767                 }
1768 #endif
1769
1770                 else if (t.cs() == "lyxmathsym") {
1771                         skipSpaces();
1772                         if (getToken().cat() != catBegin) {
1773                                 error("'{' expected in \\" + t.cs());
1774                                 return success_;
1775                         }
1776                         int count = 0;
1777                         docstring cmd;
1778                         CatCode cat = nextToken().cat();
1779                         while (good() && (count || cat != catEnd)) {
1780                                 if (cat == catBegin)
1781                                         ++count;
1782                                 else if (cat == catEnd)
1783                                         --count;
1784                                 cmd += getToken().asInput();
1785                                 cat = nextToken().cat();
1786                         }
1787                         if (getToken().cat() != catEnd) {
1788                                 error("'}' expected in \\" + t.cs());
1789                                 return success_;
1790                         }
1791                         docstring rem;
1792                         do {
1793                                 cmd = Encodings::fromLaTeXCommand(cmd, rem);
1794                                 for (size_t i = 0; i < cmd.size(); ++i)
1795                                         cell->push_back(MathAtom(new InsetMathChar(cmd[i])));
1796                                 if (rem.size()) {
1797                                         char_type c = rem[0];
1798                                         cell->push_back(MathAtom(new InsetMathChar(c)));
1799                                         cmd = rem.substr(1);
1800                                         rem.clear();
1801                                 } else
1802                                         cmd.clear();
1803                         } while (cmd.size());
1804                 }
1805
1806                 else if (t.cs().size()) {
1807                         bool const no_mhchem =
1808                                 (t.cs() == "ce" || t.cs() == "cf") && buf
1809                                 && buf->params().use_mhchem == BufferParams::package_off;
1810                         bool const is_user_macro = no_mhchem ||
1811                                 (buf && (mode_ & Parse::TRACKMACRO
1812                                         ? buf->usermacros.count(t.cs()) != 0
1813                                         : buf->getMacro(t.cs(), false) != 0));
1814                         latexkeys const * l = in_word_set(t.cs());
1815                         if (l && !is_user_macro) {
1816                                 if (l->inset == "big") {
1817                                         skipSpaces();
1818                                         docstring const delim = getToken().asInput();
1819                                         if (InsetMathBig::isBigInsetDelim(delim))
1820                                                 cell->push_back(MathAtom(
1821                                                         new InsetMathBig(t.cs(), delim)));
1822                                         else {
1823                                                 cell->push_back(createInsetMath(t.cs(), buf));
1824                                                 putback();
1825                                         }
1826                                 }
1827
1828                                 else if (l->inset == "font") {
1829                                         cell->push_back(createInsetMath(t.cs(), buf));
1830                                         parse(cell->back().nucleus()->cell(0),
1831                                                 FLAG_ITEM, asMode(mode, l->extra));
1832                                 }
1833
1834                                 else if (l->inset == "oldfont") {
1835                                         cell->push_back(createInsetMath(t.cs(), buf));
1836                                         parse(cell->back().nucleus()->cell(0),
1837                                                 flags | FLAG_ALIGN, asMode(mode, l->extra));
1838                                         if (prevToken().cat() != catAlign &&
1839                                             prevToken().cs() != "\\")
1840                                                 return success_;
1841                                         putback();
1842                                 }
1843
1844                                 else if (l->inset == "style") {
1845                                         cell->push_back(createInsetMath(t.cs(), buf));
1846                                         parse(cell->back().nucleus()->cell(0),
1847                                                 flags | FLAG_ALIGN, mode);
1848                                         if (prevToken().cat() != catAlign &&
1849                                             prevToken().cs() != "\\")
1850                                                 return success_;
1851                                         putback();
1852                                 }
1853
1854                                 else {
1855                                         MathAtom at = createInsetMath(t.cs(), buf);
1856                                         for (InsetMath::idx_type i = 0; i < at->nargs(); ++i)
1857                                                 parse(at.nucleus()->cell(i),
1858                                                         FLAG_ITEM, asMode(mode, l->extra));
1859                                         cell->push_back(at);
1860                                 }
1861                         }
1862
1863                         else {
1864                                 bool is_unicode_symbol = false;
1865                                 if (mode == InsetMath::TEXT_MODE && !is_user_macro) {
1866                                         int num_tokens = 0;
1867                                         docstring cmd = prevToken().asInput();
1868                                         CatCode cat = nextToken().cat();
1869                                         if (cat == catBegin) {
1870                                                 int count = 0;
1871                                                 while (good() && (count || cat != catEnd)) {
1872                                                         cat = nextToken().cat();
1873                                                         cmd += getToken().asInput();
1874                                                         ++num_tokens;
1875                                                         if (cat == catBegin)
1876                                                                 ++count;
1877                                                         else if (cat == catEnd)
1878                                                                 --count;
1879                                                 }
1880                                         }
1881                                         bool is_combining;
1882                                         char_type c =
1883                                                 Encodings::fromLaTeXCommand(cmd, is_combining);
1884                                         if (is_combining) {
1885                                                 if (cat == catLetter)
1886                                                         cmd += '{';
1887                                                 cmd += getToken().asInput();
1888                                                 ++num_tokens;
1889                                                 if (cat == catLetter)
1890                                                         cmd += '}';
1891                                                 c = Encodings::fromLaTeXCommand(cmd, is_combining);
1892                                         }
1893                                         if (c) {
1894                                                 is_unicode_symbol = true;
1895                                                 cell->push_back(MathAtom(new InsetMathChar(c)));
1896                                         } else {
1897                                                 while (num_tokens--)
1898                                                         putback();
1899                                         }
1900                                 }
1901                                 if (!is_unicode_symbol) {
1902                                         MathAtom at = is_user_macro ?
1903                                                 MathAtom(new MathMacro(buf, t.cs()))
1904                                                 : createInsetMath(t.cs(), buf);
1905                                         InsetMath::mode_type m = mode;
1906                                         //if (m == InsetMath::UNDECIDED_MODE)
1907                                         //lyxerr << "default creation: m1: " << m << endl;
1908                                         if (at->currentMode() != InsetMath::UNDECIDED_MODE)
1909                                                 m = at->currentMode();
1910                                         //lyxerr << "default creation: m2: " << m << endl;
1911                                         InsetMath::idx_type start = 0;
1912                                         // this fails on \bigg[...\bigg]
1913                                         //MathData opt;
1914                                         //parse(opt, FLAG_OPTION, InsetMath::VERBATIM_MODE);
1915                                         //if (opt.size()) {
1916                                         //      start = 1;
1917                                         //      at.nucleus()->cell(0) = opt;
1918                                         //}
1919                                         for (InsetMath::idx_type i = start; i < at->nargs(); ++i) {
1920                                                 parse(at.nucleus()->cell(i), FLAG_ITEM, m);
1921                                                 skipSpaces();
1922                                         }
1923                                         cell->push_back(at);
1924                                 }
1925                         }
1926                 }
1927
1928
1929                 if (flags & FLAG_LEAVE) {
1930                         flags &= ~FLAG_LEAVE;
1931                         break;
1932                 }
1933         }
1934         return success_;
1935 }
1936
1937
1938
1939 } // anonymous namespace
1940
1941
1942 bool mathed_parse_cell(MathData & ar, docstring const & str, Parse::flags f)
1943 {
1944         return Parser(str, f, ar.buffer()).parse(ar, 0, f & Parse::TEXTMODE ?
1945                                 InsetMath::TEXT_MODE : InsetMath::MATH_MODE);
1946 }
1947
1948
1949 bool mathed_parse_cell(MathData & ar, istream & is, Parse::flags f)
1950 {
1951         return Parser(is, f, ar.buffer()).parse(ar, 0, f & Parse::TEXTMODE ?
1952                                 InsetMath::TEXT_MODE : InsetMath::MATH_MODE);
1953 }
1954
1955
1956 bool mathed_parse_normal(Buffer * buf, MathAtom & t, docstring const & str,
1957                          Parse::flags f)
1958 {
1959         return Parser(str, f, buf).parse(t);
1960 }
1961
1962
1963 bool mathed_parse_normal(Buffer * buf, MathAtom & t, Lexer & lex,
1964                          Parse::flags f)
1965 {
1966         return Parser(lex, f, buf).parse(t);
1967 }
1968
1969
1970 bool mathed_parse_normal(InsetMathGrid & grid, docstring const & str,
1971                          Parse::flags f)
1972 {
1973         return Parser(str, f, &grid.buffer()).parse1(grid, 0, f & Parse::TEXTMODE ?
1974                         InsetMath::TEXT_MODE : InsetMath::MATH_MODE, false);
1975 }
1976
1977
1978 void initParser()
1979 {
1980         fill(theCatcode, theCatcode + 128, catOther);
1981         fill(theCatcode + 'a', theCatcode + 'z' + 1, catLetter);
1982         fill(theCatcode + 'A', theCatcode + 'Z' + 1, catLetter);
1983
1984         theCatcode[int('\\')] = catEscape;
1985         theCatcode[int('{')]  = catBegin;
1986         theCatcode[int('}')]  = catEnd;
1987         theCatcode[int('$')]  = catMath;
1988         theCatcode[int('&')]  = catAlign;
1989         theCatcode[int('\n')] = catNewline;
1990         theCatcode[int('#')]  = catParameter;
1991         theCatcode[int('^')]  = catSuper;
1992         theCatcode[int('_')]  = catSub;
1993         theCatcode[int(0x7f)] = catIgnore;
1994         theCatcode[int(' ')]  = catSpace;
1995         theCatcode[int('\t')] = catSpace;
1996         theCatcode[int('\r')] = catNewline;
1997         theCatcode[int('~')]  = catActive;
1998         theCatcode[int('%')]  = catComment;
1999 }
2000
2001
2002 } // namespace lyx