]> git.lyx.org Git - lyx.git/blob - src/mathed/MathParser.cpp
3a8e4c2886262da3711196d3b423ee665fe5ba52
[lyx.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                         if (mode == InsetMath::MATH_MODE) {
1221                                 error("bad math environment");
1222                                 break;
1223                         }
1224                         cell->push_back(MathAtom(new InsetMathHull(buf, hullSimple)));
1225                         parse2(cell->back(), FLAG_SIMPLE2, InsetMath::MATH_MODE, false);
1226                 }
1227
1228                 else if (t.cs() == "[") {
1229                         if (mode != InsetMath::UNDECIDED_MODE) {
1230                                 error("bad math environment");
1231                                 break;
1232                         }
1233                         cell->push_back(MathAtom(new InsetMathHull(buf, hullEquation)));
1234                         parse2(cell->back(), FLAG_EQUATION, InsetMath::MATH_MODE, false);
1235                 }
1236
1237                 else if (t.cs() == "protect")
1238                         // ignore \\protect, will hopefully be re-added during output
1239                         ;
1240
1241                 else if (t.cs() == "end") {
1242                         if (flags & FLAG_END) {
1243                                 // eat environment name
1244                                 docstring const name = getArg('{', '}');
1245                                 if (environments_.empty())
1246                                         error("'found \\end{" + name +
1247                                               "}' without matching '\\begin{" +
1248                                               name + "}'");
1249                                 else if (name != environments_.back())
1250                                         error("'\\end{" + name +
1251                                               "}' does not match '\\begin{" +
1252                                               environments_.back() + "}'");
1253                                 else {
1254                                         environments_.pop_back();
1255                                         // Delete empty last row in matrix
1256                                         // like insets.
1257                                         // If you abuse InsetMathGrid for
1258                                         // non-matrix like structures you
1259                                         // probably need to refine this test.
1260                                         // Right now we only have to test for
1261                                         // single line hull insets.
1262                                         if (grid.nrows() > 1 && innerHull(name))
1263                                                 delEmptyLastRow(grid);
1264                                         return success_;
1265                                 }
1266                         } else
1267                                 error("found 'end' unexpectedly");
1268                 }
1269
1270                 else if (t.cs() == ")") {
1271                         if (flags & FLAG_SIMPLE2)
1272                                 return success_;
1273                         error("found '\\)' unexpectedly");
1274                 }
1275
1276                 else if (t.cs() == "]") {
1277                         if (flags & FLAG_EQUATION)
1278                                 return success_;
1279                         error("found '\\]' unexpectedly");
1280                 }
1281
1282                 else if (t.cs() == "\\") {
1283                         if (flags & FLAG_ALIGN)
1284                                 return success_;
1285                         bool added = false;
1286                         if (nextToken().asInput() == "*") {
1287                                 getToken();
1288                                 added = addRow(grid, cellrow, docstring(), false);
1289                         } else if (good())
1290                                 added = addRow(grid, cellrow, getArg('[', ']'));
1291                         else
1292                                 error("missing token after \\\\");
1293                         if (added) {
1294                                 cellcol = 0;
1295                                 if (grid.asHullInset())
1296                                         grid.asHullInset()->numbered(
1297                                                         cellrow, numbered);
1298                                 cell = &grid.cell(grid.index(cellrow,
1299                                                              cellcol));
1300                         }
1301                 }
1302
1303 #if 0
1304                 else if (t.cs() == "multicolumn") {
1305                         // extract column count and insert dummy cells
1306                         MathData count;
1307                         parse(count, FLAG_ITEM, mode);
1308                         int cols = 1;
1309                         if (!extractNumber(count, cols)) {
1310                                 success_ = false;
1311                                 lyxerr << " can't extract number of cells from " << count << endl;
1312                         }
1313                         // resize the table if necessary
1314                         for (int i = 0; i < cols; ++i) {
1315                                 if (addCol(grid, cellcol)) {
1316                                         cell = &grid.cell(grid.index(
1317                                                         cellrow, cellcol));
1318                                         // mark this as dummy
1319                                         grid.cellinfo(grid.index(
1320                                                 cellrow, cellcol)).dummy_ = true;
1321                                 }
1322                         }
1323                         // the last cell is the real thing, not a dummy
1324                         grid.cellinfo(grid.index(cellrow, cellcol)).dummy_ = false;
1325
1326                         // read special alignment
1327                         MathData align;
1328                         parse(align, FLAG_ITEM, mode);
1329                         //grid.cellinfo(grid.index(cellrow, cellcol)).align_ = extractString(align);
1330
1331                         // parse the remaining contents into the "real" cell
1332                         parse(*cell, FLAG_ITEM, mode);
1333                 }
1334 #endif
1335
1336                 else if (t.cs() == "limits" || t.cs() == "nolimits") {
1337                         CatCode cat = nextToken().cat();
1338                         if (cat == catSuper || cat == catSub)
1339                                 limits = t.cs() == "limits" ? 1 : -1;
1340                         else {
1341                                 MathAtom at = createInsetMath(t.cs(), buf);
1342                                 cell->push_back(at);
1343                         }
1344                 }
1345
1346                 else if (t.cs() == "nonumber") {
1347                         if (grid.asHullInset())
1348                                 grid.asHullInset()->numbered(cellrow, false);
1349                 }
1350
1351                 else if (t.cs() == "number") {
1352                         if (grid.asHullInset())
1353                                 grid.asHullInset()->numbered(cellrow, true);
1354                 }
1355
1356                 else if (t.cs() == "hline") {
1357                         grid.rowinfo(cellrow).lines_ ++;
1358                 }
1359
1360                 else if (t.cs() == "sqrt") {
1361                         MathData ar;
1362                         parse(ar, FLAG_OPTION, mode);
1363                         if (ar.size()) {
1364                                 cell->push_back(MathAtom(new InsetMathRoot(buf)));
1365                                 cell->back().nucleus()->cell(0) = ar;
1366                                 parse(cell->back().nucleus()->cell(1), FLAG_ITEM, mode);
1367                         } else {
1368                                 cell->push_back(MathAtom(new InsetMathSqrt(buf)));
1369                                 parse(cell->back().nucleus()->cell(0), FLAG_ITEM, mode);
1370                         }
1371                 }
1372
1373                 else if (t.cs() == "unit") {
1374                         // Allowed formats \unit[val]{unit}
1375                         MathData ar;
1376                         parse(ar, FLAG_OPTION, mode);
1377                         if (ar.size()) {
1378                                 cell->push_back(MathAtom(new InsetMathFrac(buf, InsetMathFrac::UNIT)));
1379                                 cell->back().nucleus()->cell(0) = ar;
1380                                 parse(cell->back().nucleus()->cell(1), FLAG_ITEM, mode);
1381                         } else {
1382                                 cell->push_back(MathAtom(new InsetMathFrac(buf, InsetMathFrac::UNIT, 1)));
1383                                 parse(cell->back().nucleus()->cell(0), FLAG_ITEM, mode);
1384                         }
1385                 }
1386
1387                 else if (t.cs() == "unitfrac") {
1388                         // Here allowed formats are \unitfrac[val]{num}{denom}
1389                         MathData ar;
1390                         parse(ar, FLAG_OPTION, mode);
1391                         if (ar.size()) {
1392                                 cell->push_back(MathAtom(new InsetMathFrac(buf, InsetMathFrac::UNITFRAC, 3)));
1393                                 cell->back().nucleus()->cell(2) = ar;
1394                         } else {
1395                                 cell->push_back(MathAtom(new InsetMathFrac(buf, InsetMathFrac::UNITFRAC)));
1396                         }
1397                         parse(cell->back().nucleus()->cell(0), FLAG_ITEM, mode);
1398                         parse(cell->back().nucleus()->cell(1), FLAG_ITEM, mode);
1399                 }
1400
1401                 else if (t.cs() == "cfrac") {
1402                         // allowed formats are \cfrac[pos]{num}{denom}
1403                         docstring const arg = getArg('[', ']');
1404                         //lyxerr << "got so far: '" << arg << "'" << endl;                              
1405                                 if (arg == "l")
1406                                         cell->push_back(MathAtom(new InsetMathFrac(buf, InsetMathFrac::CFRACLEFT)));
1407                                 else if (arg == "r")
1408                                         cell->push_back(MathAtom(new InsetMathFrac(buf, InsetMathFrac::CFRACRIGHT)));
1409                                 else if (arg.empty() || arg == "c")
1410                                         cell->push_back(MathAtom(new InsetMathFrac(buf, InsetMathFrac::CFRAC)));
1411                                 else {
1412                                         error("found invalid optional argument");
1413                                         break;
1414                                 }
1415                         parse(cell->back().nucleus()->cell(0), FLAG_ITEM, mode);
1416                         parse(cell->back().nucleus()->cell(1), FLAG_ITEM, mode);
1417                 }
1418
1419                 else if (t.cs() == "xrightarrow" || t.cs() == "xleftarrow") {
1420                         cell->push_back(createInsetMath(t.cs(), buf));
1421                         parse(cell->back().nucleus()->cell(1), FLAG_OPTION, mode);
1422                         parse(cell->back().nucleus()->cell(0), FLAG_ITEM, mode);
1423                 }
1424
1425                 else if (t.cs() == "ref" || t.cs() == "eqref" || t.cs() == "prettyref"
1426                           || t.cs() == "pageref" || t.cs() == "vpageref" || t.cs() == "vref") {
1427                         cell->push_back(MathAtom(new InsetMathRef(buf, t.cs())));
1428                         docstring const opt = parse_verbatim_option();
1429                         docstring const ref = parse_verbatim_item();
1430                         if (!opt.empty()) {
1431                                 cell->back().nucleus()->cell(1).push_back(
1432                                         MathAtom(new InsetMathString(opt)));
1433                         }
1434                         cell->back().nucleus()->cell(0).push_back(
1435                                         MathAtom(new InsetMathString(ref)));
1436                 }
1437
1438                 else if (t.cs() == "left") {
1439                         skipSpaces();
1440                         Token const & tl = getToken();
1441                         // \| and \Vert are equivalent, and InsetMathDelim
1442                         // can't handle \|
1443                         // FIXME: fix this in InsetMathDelim itself!
1444                         docstring const l = tl.cs() == "|" ? from_ascii("Vert") : tl.asString();
1445                         MathData ar;
1446                         parse(ar, FLAG_RIGHT, mode);
1447                         if (!good())
1448                                 break;
1449                         skipSpaces();
1450                         Token const & tr = getToken();
1451                         docstring const r = tr.cs() == "|" ? from_ascii("Vert") : tr.asString();
1452                         cell->push_back(MathAtom(new InsetMathDelim(buf, l, r, ar)));
1453                 }
1454
1455                 else if (t.cs() == "right") {
1456                         if (flags & FLAG_RIGHT)
1457                                 return success_;
1458                         //lyxerr << "got so far: '" << cell << "'" << endl;
1459                         error("Unmatched right delimiter");
1460                         return success_;
1461                 }
1462
1463                 else if (t.cs() == "begin") {
1464                         docstring const name = getArg('{', '}');
1465                         environments_.push_back(name);
1466
1467                         if (name == "array" || name == "subarray") {
1468                                 docstring const valign = parse_verbatim_option() + 'c';
1469                                 docstring const halign = parse_verbatim_item();
1470                                 cell->push_back(MathAtom(new InsetMathArray(buf, name,
1471                                         InsetMathGrid::guessColumns(halign), 1, (char)valign[0], halign)));
1472                                 parse2(cell->back(), FLAG_END, mode, false);
1473                         }
1474
1475                         else if (name == "tabular") {
1476                                 docstring const valign = parse_verbatim_option() + 'c';
1477                                 docstring const halign = parse_verbatim_item();
1478                                 cell->push_back(MathAtom(new InsetMathTabular(buf, name,
1479                                         InsetMathGrid::guessColumns(halign), 1, (char)valign[0], halign)));
1480                                 parse2(cell->back(), FLAG_END, InsetMath::TEXT_MODE, false);
1481                         }
1482
1483                         else if (name == "split" || name == "cases") {
1484                                 cell->push_back(createInsetMath(name, buf));
1485                                 parse2(cell->back(), FLAG_END, mode, false);
1486                         }
1487
1488                         else if (name == "alignedat") {
1489                                 docstring const valign = parse_verbatim_option() + 'c';
1490                                 // ignore this for a while
1491                                 getArg('{', '}');
1492                                 cell->push_back(MathAtom(new InsetMathSplit(buf, name, (char)valign[0])));
1493                                 parse2(cell->back(), FLAG_END, mode, false);
1494                         }
1495
1496                         else if (name == "math") {
1497                                 if (mode == InsetMath::MATH_MODE) {
1498                                         error("bad math environment");
1499                                         break;
1500                                 }
1501                                 cell->push_back(MathAtom(new InsetMathHull(buf, hullSimple)));
1502                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, true);
1503                         }
1504
1505                         else if (name == "equation" || name == "equation*"
1506                                         || name == "displaymath") {
1507                                 if (mode != InsetMath::UNDECIDED_MODE) {
1508                                         error("bad math environment");
1509                                         break;
1510                                 }
1511                                 cell->push_back(MathAtom(new InsetMathHull(buf, hullEquation)));
1512                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, (name == "equation"));
1513                         }
1514
1515                         else if (name == "eqnarray" || name == "eqnarray*") {
1516                                 if (mode != InsetMath::UNDECIDED_MODE) {
1517                                         error("bad math environment");
1518                                         break;
1519                                 }
1520                                 cell->push_back(MathAtom(new InsetMathHull(buf, hullEqnArray)));
1521                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1522                         }
1523
1524                         else if (name == "align" || name == "align*") {
1525                                 if (mode != InsetMath::UNDECIDED_MODE) {
1526                                         error("bad math environment");
1527                                         break;
1528                                 }
1529                                 cell->push_back(MathAtom(new InsetMathHull(buf, hullAlign)));
1530                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1531                         }
1532
1533                         else if (name == "flalign" || name == "flalign*") {
1534                                 if (mode != InsetMath::UNDECIDED_MODE) {
1535                                         error("bad math environment");
1536                                         break;
1537                                 }
1538                                 cell->push_back(MathAtom(new InsetMathHull(buf, hullFlAlign)));
1539                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1540                         }
1541
1542                         else if (name == "alignat" || name == "alignat*") {
1543                                 if (mode != InsetMath::UNDECIDED_MODE) {
1544                                         error("bad math environment");
1545                                         break;
1546                                 }
1547                                 // ignore this for a while
1548                                 getArg('{', '}');
1549                                 cell->push_back(MathAtom(new InsetMathHull(buf, hullAlignAt)));
1550                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1551                         }
1552
1553                         else if (name == "xalignat" || name == "xalignat*") {
1554                                 if (mode != InsetMath::UNDECIDED_MODE) {
1555                                         error("bad math environment");
1556                                         break;
1557                                 }
1558                                 // ignore this for a while
1559                                 getArg('{', '}');
1560                                 cell->push_back(MathAtom(new InsetMathHull(buf, hullXAlignAt)));
1561                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1562                         }
1563
1564                         else if (name == "xxalignat") {
1565                                 if (mode != InsetMath::UNDECIDED_MODE) {
1566                                         error("bad math environment");
1567                                         break;
1568                                 }
1569                                 // ignore this for a while
1570                                 getArg('{', '}');
1571                                 cell->push_back(MathAtom(new InsetMathHull(buf, hullXXAlignAt)));
1572                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1573                         }
1574
1575                         else if (name == "multline" || name == "multline*") {
1576                                 if (mode != InsetMath::UNDECIDED_MODE) {
1577                                         error("bad math environment");
1578                                         break;
1579                                 }
1580                                 cell->push_back(MathAtom(new InsetMathHull(buf, hullMultline)));
1581                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1582                         }
1583
1584                         else if (name == "gather" || name == "gather*") {
1585                                 if (mode != InsetMath::UNDECIDED_MODE) {
1586                                         error("bad math environment");
1587                                         break;
1588                                 }
1589                                 cell->push_back(MathAtom(new InsetMathHull(buf, hullGather)));
1590                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1591                         }
1592
1593                         else if (latexkeys const * l = in_word_set(name)) {
1594                                 if (l->inset == "matrix") {
1595                                         cell->push_back(createInsetMath(name, buf));
1596                                         parse2(cell->back(), FLAG_END, mode, false);
1597                                 } else if (l->inset == "split") {
1598                                         docstring const valign = parse_verbatim_option() + 'c';
1599                                         cell->push_back(MathAtom(
1600                                                 new InsetMathSplit(buf, name, (char)valign[0])));
1601                                         parse2(cell->back(), FLAG_END, mode, false);
1602                                 } else {
1603                                         success_ = false;
1604                                         if (!(mode_ & Parse::QUIET)) {
1605                                                 dump();
1606                                                 lyxerr << "found math environment `"
1607                                                        << to_utf8(name)
1608                                                        << "' in symbols file with unsupported inset `"
1609                                                        << to_utf8(l->inset)
1610                                                        << "'." << endl;
1611                                         }
1612                                         // create generic environment inset
1613                                         cell->push_back(MathAtom(new InsetMathEnv(buf, name)));
1614                                         parse(cell->back().nucleus()->cell(0), FLAG_END, mode);
1615                                 }
1616                         }
1617
1618                         else {
1619                                 success_ = false;
1620                                 if (!(mode_ & Parse::QUIET)) {
1621                                         dump();
1622                                         lyxerr << "found unknown math environment '"
1623                                                << to_utf8(name) << "'" << endl;
1624                                 }
1625                                 // create generic environment inset
1626                                 cell->push_back(MathAtom(new InsetMathEnv(buf, name)));
1627                                 parse(cell->back().nucleus()->cell(0), FLAG_END, mode);
1628                         }
1629                 }
1630
1631                 else if (t.cs() == "kern") {
1632                         // FIXME: A hack...
1633                         docstring s;
1634                         int num_tokens = 0;
1635                         while (true) {
1636                                 Token const & t = getToken();
1637                                 ++num_tokens;
1638                                 if (!good()) {
1639                                         s.clear();
1640                                         while (num_tokens--)
1641                                                 putback();
1642                                         break;
1643                                 }
1644                                 s += t.character();
1645                                 if (isValidLength(to_utf8(s)))
1646                                         break;
1647                         }
1648                         if (s.empty())
1649                                 cell->push_back(MathAtom(new InsetMathKern));
1650                         else
1651                                 cell->push_back(MathAtom(new InsetMathKern(s)));
1652                 }
1653
1654                 else if (t.cs() == "label") {
1655                         // FIXME: This is swallowed in inline formulas
1656                         docstring label = parse_verbatim_item();
1657                         MathData ar;
1658                         asArray(label, ar);
1659                         if (grid.asHullInset()) {
1660                                 grid.asHullInset()->label(cellrow, label);
1661                         } else {
1662                                 cell->push_back(createInsetMath(t.cs(), buf));
1663                                 cell->push_back(MathAtom(new InsetMathBrace(ar)));
1664                         }
1665                 }
1666
1667                 else if (t.cs() == "choose" || t.cs() == "over"
1668                                 || t.cs() == "atop" || t.cs() == "brace"
1669                                 || t.cs() == "brack") {
1670                         MathAtom at = createInsetMath(t.cs(), buf);
1671                         at.nucleus()->cell(0) = *cell;
1672                         cell->clear();
1673                         parse(at.nucleus()->cell(1), flags, mode);
1674                         cell->push_back(at);
1675                         return success_;
1676                 }
1677
1678                 else if (t.cs() == "color") {
1679                         docstring const color = parse_verbatim_item();
1680                         cell->push_back(MathAtom(new InsetMathColor(buf, true, color)));
1681                         parse(cell->back().nucleus()->cell(0), flags, mode);
1682                         return success_;
1683                 }
1684
1685                 else if (t.cs() == "textcolor") {
1686                         docstring const color = parse_verbatim_item();
1687                         cell->push_back(MathAtom(new InsetMathColor(buf, false, color)));
1688                         parse(cell->back().nucleus()->cell(0), FLAG_ITEM, InsetMath::TEXT_MODE);
1689                 }
1690
1691                 else if (t.cs() == "normalcolor") {
1692                         cell->push_back(createInsetMath(t.cs(), buf));
1693                         parse(cell->back().nucleus()->cell(0), flags, mode);
1694                         return success_;
1695                 }
1696
1697                 else if (t.cs() == "substack") {
1698                         cell->push_back(createInsetMath(t.cs(), buf));
1699                         parse2(cell->back(), FLAG_ITEM, mode, false);
1700                         // Delete empty last row if present
1701                         InsetMathGrid & subgrid =
1702                                 *(cell->back().nucleus()->asGridInset());
1703                         if (subgrid.nrows() > 1)
1704                                 delEmptyLastRow(subgrid);
1705                 }
1706
1707                 else if (t.cs() == "xymatrix") {
1708                         odocstringstream os;
1709                         while (good() && nextToken().cat() != catBegin)
1710                                 os << getToken().asInput();
1711                         cell->push_back(createInsetMath(t.cs() + os.str(), buf));
1712                         parse2(cell->back(), FLAG_ITEM, mode, false);
1713                         // Delete empty last row if present
1714                         InsetMathGrid & subgrid =
1715                                 *(cell->back().nucleus()->asGridInset());
1716                         if (subgrid.nrows() > 1)
1717                                 delEmptyLastRow(subgrid);
1718                 }
1719
1720                 else if (t.cs() == "framebox" || t.cs() == "makebox") {
1721                         cell->push_back(createInsetMath(t.cs(), buf));
1722                         parse(cell->back().nucleus()->cell(0), FLAG_OPTION, InsetMath::TEXT_MODE);
1723                         parse(cell->back().nucleus()->cell(1), FLAG_OPTION, InsetMath::TEXT_MODE);
1724                         parse(cell->back().nucleus()->cell(2), FLAG_ITEM, InsetMath::TEXT_MODE);
1725                 }
1726
1727                 else if (t.cs() == "tag") {
1728                         if (nextToken().character() == '*') {
1729                                 getToken();
1730                                 cell->push_back(createInsetMath(t.cs() + '*', buf));
1731                         } else
1732                                 cell->push_back(createInsetMath(t.cs(), buf));
1733                         parse(cell->back().nucleus()->cell(0), FLAG_ITEM, InsetMath::TEXT_MODE);
1734                 }
1735
1736                 else if (t.cs() == "hspace" && nextToken().character() != '*') {
1737                         docstring const name = t.cs();
1738                         docstring const arg = parse_verbatim_item();
1739                         Length length;
1740                         if (isValidLength(to_utf8(arg), &length))
1741                                 cell->push_back(MathAtom(new InsetMathSpace(length)));
1742                         else {
1743                                 // Since the Length class cannot use length variables
1744                                 // we must not create an InsetMathSpace.
1745                                 cell->push_back(MathAtom(new MathMacro(buf, name)));
1746                                 MathData ar;
1747                                 mathed_parse_cell(ar, '{' + arg + '}', mode_);
1748                                 cell->append(ar);
1749                         }
1750                 }
1751
1752 #if 0
1753                 else if (t.cs() == "infer") {
1754                         MathData ar;
1755                         parse(ar, FLAG_OPTION, mode);
1756                         cell->push_back(createInsetMath(t.cs(), buf));
1757                         parse2(cell->back(), FLAG_ITEM, mode, false);
1758                 }
1759
1760                 // Disabled
1761                 else if (1 && t.cs() == "ar") {
1762                         auto_ptr<InsetMathXYArrow> p(new InsetMathXYArrow);
1763                         // try to read target
1764                         parse(p->cell(0), FLAG_OTPTION, mode);
1765                         // try to read label
1766                         if (nextToken().cat() == catSuper || nextToken().cat() == catSub) {
1767                                 p->up_ = nextToken().cat() == catSuper;
1768                                 getToken();
1769                                 parse(p->cell(1), FLAG_ITEM, mode);
1770                                 //lyxerr << "read label: " << p->cell(1) << endl;
1771                         }
1772
1773                         cell->push_back(MathAtom(p.release()));
1774                         //lyxerr << "read cell: " << cell << endl;
1775                 }
1776 #endif
1777
1778                 else if (t.cs() == "lyxmathsym") {
1779                         skipSpaces();
1780                         if (getToken().cat() != catBegin) {
1781                                 error("'{' expected in \\" + t.cs());
1782                                 return success_;
1783                         }
1784                         int count = 0;
1785                         docstring cmd;
1786                         CatCode cat = nextToken().cat();
1787                         while (good() && (count || cat != catEnd)) {
1788                                 if (cat == catBegin)
1789                                         ++count;
1790                                 else if (cat == catEnd)
1791                                         --count;
1792                                 cmd += getToken().asInput();
1793                                 cat = nextToken().cat();
1794                         }
1795                         if (getToken().cat() != catEnd) {
1796                                 error("'}' expected in \\" + t.cs());
1797                                 return success_;
1798                         }
1799                         docstring rem;
1800                         do {
1801                                 cmd = Encodings::fromLaTeXCommand(cmd, rem);
1802                                 for (size_t i = 0; i < cmd.size(); ++i)
1803                                         cell->push_back(MathAtom(new InsetMathChar(cmd[i])));
1804                                 if (rem.size()) {
1805                                         char_type c = rem[0];
1806                                         cell->push_back(MathAtom(new InsetMathChar(c)));
1807                                         cmd = rem.substr(1);
1808                                         rem.clear();
1809                                 } else
1810                                         cmd.clear();
1811                         } while (cmd.size());
1812                 }
1813
1814                 else if (t.cs().size()) {
1815                         bool const no_mhchem =
1816                                 (t.cs() == "ce" || t.cs() == "cf") && buf
1817                                 && buf->params().use_mhchem == BufferParams::package_off;
1818                         bool const is_user_macro = no_mhchem ||
1819                                 (buf && (mode_ & Parse::TRACKMACRO
1820                                         ? buf->usermacros.count(t.cs()) != 0
1821                                         : buf->getMacro(t.cs(), false) != 0));
1822                         latexkeys const * l = in_word_set(t.cs());
1823                         if (l && !is_user_macro) {
1824                                 if (l->inset == "big") {
1825                                         skipSpaces();
1826                                         docstring const delim = getToken().asInput();
1827                                         if (InsetMathBig::isBigInsetDelim(delim))
1828                                                 cell->push_back(MathAtom(
1829                                                         new InsetMathBig(t.cs(), delim)));
1830                                         else {
1831                                                 cell->push_back(createInsetMath(t.cs(), buf));
1832                                                 putback();
1833                                         }
1834                                 }
1835
1836                                 else if (l->inset == "font") {
1837                                         cell->push_back(createInsetMath(t.cs(), buf));
1838                                         parse(cell->back().nucleus()->cell(0),
1839                                                 FLAG_ITEM, asMode(mode, l->extra));
1840                                 }
1841
1842                                 else if (l->inset == "oldfont") {
1843                                         cell->push_back(createInsetMath(t.cs(), buf));
1844                                         parse(cell->back().nucleus()->cell(0),
1845                                                 flags | FLAG_ALIGN, asMode(mode, l->extra));
1846                                         if (prevToken().cat() != catAlign &&
1847                                             prevToken().cs() != "\\")
1848                                                 return success_;
1849                                         putback();
1850                                 }
1851
1852                                 else if (l->inset == "style") {
1853                                         cell->push_back(createInsetMath(t.cs(), buf));
1854                                         parse(cell->back().nucleus()->cell(0),
1855                                                 flags | FLAG_ALIGN, mode);
1856                                         if (prevToken().cat() != catAlign &&
1857                                             prevToken().cs() != "\\")
1858                                                 return success_;
1859                                         putback();
1860                                 }
1861
1862                                 else {
1863                                         MathAtom at = createInsetMath(t.cs(), buf);
1864                                         for (InsetMath::idx_type i = 0; i < at->nargs(); ++i)
1865                                                 parse(at.nucleus()->cell(i),
1866                                                         FLAG_ITEM, asMode(mode, l->extra));
1867                                         cell->push_back(at);
1868                                 }
1869                         }
1870
1871                         else {
1872                                 bool is_unicode_symbol = false;
1873                                 if (mode == InsetMath::TEXT_MODE && !is_user_macro) {
1874                                         int num_tokens = 0;
1875                                         docstring cmd = prevToken().asInput();
1876                                         CatCode cat = nextToken().cat();
1877                                         if (cat == catBegin) {
1878                                                 int count = 0;
1879                                                 while (good() && (count || cat != catEnd)) {
1880                                                         cat = nextToken().cat();
1881                                                         cmd += getToken().asInput();
1882                                                         ++num_tokens;
1883                                                         if (cat == catBegin)
1884                                                                 ++count;
1885                                                         else if (cat == catEnd)
1886                                                                 --count;
1887                                                 }
1888                                         }
1889                                         bool is_combining;
1890                                         char_type c =
1891                                                 Encodings::fromLaTeXCommand(cmd, is_combining);
1892                                         if (is_combining) {
1893                                                 if (cat == catLetter)
1894                                                         cmd += '{';
1895                                                 cmd += getToken().asInput();
1896                                                 ++num_tokens;
1897                                                 if (cat == catLetter)
1898                                                         cmd += '}';
1899                                                 c = Encodings::fromLaTeXCommand(cmd, is_combining);
1900                                         }
1901                                         if (c) {
1902                                                 is_unicode_symbol = true;
1903                                                 cell->push_back(MathAtom(new InsetMathChar(c)));
1904                                         } else {
1905                                                 while (num_tokens--)
1906                                                         putback();
1907                                         }
1908                                 }
1909                                 if (!is_unicode_symbol) {
1910                                         MathAtom at = is_user_macro ?
1911                                                 MathAtom(new MathMacro(buf, t.cs()))
1912                                                 : createInsetMath(t.cs(), buf);
1913                                         InsetMath::mode_type m = mode;
1914                                         //if (m == InsetMath::UNDECIDED_MODE)
1915                                         //lyxerr << "default creation: m1: " << m << endl;
1916                                         if (at->currentMode() != InsetMath::UNDECIDED_MODE)
1917                                                 m = at->currentMode();
1918                                         //lyxerr << "default creation: m2: " << m << endl;
1919                                         InsetMath::idx_type start = 0;
1920                                         // this fails on \bigg[...\bigg]
1921                                         //MathData opt;
1922                                         //parse(opt, FLAG_OPTION, InsetMath::VERBATIM_MODE);
1923                                         //if (opt.size()) {
1924                                         //      start = 1;
1925                                         //      at.nucleus()->cell(0) = opt;
1926                                         //}
1927                                         for (InsetMath::idx_type i = start; i < at->nargs(); ++i) {
1928                                                 parse(at.nucleus()->cell(i), FLAG_ITEM, m);
1929                                                 skipSpaces();
1930                                         }
1931                                         cell->push_back(at);
1932                                 }
1933                         }
1934                 }
1935
1936
1937                 if (flags & FLAG_LEAVE) {
1938                         flags &= ~FLAG_LEAVE;
1939                         break;
1940                 }
1941         }
1942         return success_;
1943 }
1944
1945
1946
1947 } // anonymous namespace
1948
1949
1950 bool mathed_parse_cell(MathData & ar, docstring const & str, Parse::flags f)
1951 {
1952         return Parser(str, f, ar.buffer()).parse(ar, 0, f & Parse::TEXTMODE ?
1953                                 InsetMath::TEXT_MODE : InsetMath::MATH_MODE);
1954 }
1955
1956
1957 bool mathed_parse_cell(MathData & ar, istream & is, Parse::flags f)
1958 {
1959         return Parser(is, f, ar.buffer()).parse(ar, 0, f & Parse::TEXTMODE ?
1960                                 InsetMath::TEXT_MODE : InsetMath::MATH_MODE);
1961 }
1962
1963
1964 bool mathed_parse_normal(Buffer * buf, MathAtom & t, docstring const & str,
1965                          Parse::flags f)
1966 {
1967         return Parser(str, f, buf).parse(t);
1968 }
1969
1970
1971 bool mathed_parse_normal(Buffer * buf, MathAtom & t, Lexer & lex,
1972                          Parse::flags f)
1973 {
1974         return Parser(lex, f, buf).parse(t);
1975 }
1976
1977
1978 bool mathed_parse_normal(InsetMathGrid & grid, docstring const & str,
1979                          Parse::flags f)
1980 {
1981         return Parser(str, f, &grid.buffer()).parse1(grid, 0, f & Parse::TEXTMODE ?
1982                         InsetMath::TEXT_MODE : InsetMath::MATH_MODE, false);
1983 }
1984
1985
1986 void initParser()
1987 {
1988         fill(theCatcode, theCatcode + 128, catOther);
1989         fill(theCatcode + 'a', theCatcode + 'z' + 1, catLetter);
1990         fill(theCatcode + 'A', theCatcode + 'Z' + 1, catLetter);
1991
1992         theCatcode[int('\\')] = catEscape;
1993         theCatcode[int('{')]  = catBegin;
1994         theCatcode[int('}')]  = catEnd;
1995         theCatcode[int('$')]  = catMath;
1996         theCatcode[int('&')]  = catAlign;
1997         theCatcode[int('\n')] = catNewline;
1998         theCatcode[int('#')]  = catParameter;
1999         theCatcode[int('^')]  = catSuper;
2000         theCatcode[int('_')]  = catSub;
2001         theCatcode[int(0x7f)] = catIgnore;
2002         theCatcode[int(' ')]  = catSpace;
2003         theCatcode[int('\t')] = catSpace;
2004         theCatcode[int('\r')] = catNewline;
2005         theCatcode[int('~')]  = catActive;
2006         theCatcode[int('%')]  = catComment;
2007 }
2008
2009
2010 } // namespace lyx