]> git.lyx.org Git - lyx.git/blob - src/mathed/MathParser.cpp
851828c9481523126bb0af4c9bb43224040b5c6c
[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 "InsetMathEnv.h"
50 #include "InsetMathFrac.h"
51 #include "InsetMathKern.h"
52 #include "MathMacro.h"
53 #include "InsetMathPar.h"
54 #include "InsetMathRef.h"
55 #include "InsetMathRoot.h"
56 #include "InsetMathScript.h"
57 #include "InsetMathSpace.h"
58 #include "InsetMathSplit.h"
59 #include "InsetMathSqrt.h"
60 #include "InsetMathTabular.h"
61 #include "MathMacroTemplate.h"
62 #include "MathFactory.h"
63 #include "MathMacroArgument.h"
64 #include "MathSupport.h"
65
66 #include "Encoding.h"
67 #include "Lexer.h"
68
69 #include "support/debug.h"
70 #include "support/convert.h"
71 #include "support/docstream.h"
72 #include "support/lstrings.h"
73
74 #include <sstream>
75
76 //#define FILEDEBUG
77
78 using namespace std;
79
80 namespace lyx {
81
82 using support::subst;
83
84 namespace {
85
86 InsetMath::mode_type asMode(InsetMath::mode_type oldmode, docstring const & str)
87 {
88         //lyxerr << "handling mode: '" << str << "'" << endl;
89         if (str == "mathmode")
90                 return InsetMath::MATH_MODE;
91         if (str == "textmode" || str == "forcetext")
92                 return InsetMath::TEXT_MODE;
93         return oldmode;
94 }
95
96
97 bool stared(docstring const & s)
98 {
99         size_t const n = s.size();
100         return n && s[n - 1] == '*';
101 }
102
103
104 docstring escapeSpecialChars(docstring const & str, bool textmode)
105 {
106         docstring const backslash = textmode ? from_ascii("\\textbackslash ")
107                                              : from_ascii("\\backslash ");
108         docstring const caret = textmode ? from_ascii("\\textasciicircum ")
109                                          : from_ascii("\\mathcircumflex ");
110
111         return subst(subst(subst(subst(subst(subst(subst(subst(subst(str,
112                         from_ascii("\\"), backslash),
113                         from_ascii("^"), caret),
114                         from_ascii("_"), from_ascii("\\_")),
115                         from_ascii("$"), from_ascii("\\$")),
116                         from_ascii("#"), from_ascii("\\#")),
117                         from_ascii("&"), from_ascii("\\&")),
118                         from_ascii("%"), from_ascii("\\%")),
119                         from_ascii("{"), from_ascii("\\{")),
120                         from_ascii("}"), from_ascii("\\}"));
121 }
122
123
124 /*!
125  * Add the row \p cellrow to \p grid.
126  * \returns wether the row could be added. Adding a row can fail for
127  * environments like "equation" that have a fixed number of rows.
128  */
129 bool addRow(InsetMathGrid & grid, InsetMathGrid::row_type & cellrow,
130             docstring const & vskip, bool allow_newpage_ = true)
131 {
132         ++cellrow;
133         if (cellrow == grid.nrows()) {
134                 //lyxerr << "adding row " << cellrow << endl;
135                 grid.addRow(cellrow - 1);
136                 if (cellrow == grid.nrows()) {
137                         // We can't add a row to this grid, so let's
138                         // append the content of this cell to the previous
139                         // one.
140                         // This does not happen in well formed .lyx files,
141                         // but LyX versions 1.3.x and older could create
142                         // such files and tex2lyx can still do that.
143                         --cellrow;
144                         lyxerr << "ignoring extra row";
145                         if (!vskip.empty())
146                                 lyxerr << " with extra space " << to_utf8(vskip);
147                         if (!allow_newpage_)
148                                 lyxerr << " with no page break allowed";
149                         lyxerr << '.' << endl;
150                         return false;
151                 }
152         }
153         grid.vcrskip(Length(to_utf8(vskip)), cellrow - 1);
154         grid.rowinfo(cellrow - 1).allow_newpage_ = allow_newpage_;
155         return true;
156 }
157
158
159 /*!
160  * Add the column \p cellcol to \p grid.
161  * \returns wether the column could be added. Adding a column can fail for
162  * environments like "eqnarray" that have a fixed number of columns.
163  */
164 bool addCol(InsetMathGrid & grid, InsetMathGrid::col_type & cellcol)
165 {
166         ++cellcol;
167         if (cellcol == grid.ncols()) {
168                 //lyxerr << "adding column " << cellcol << endl;
169                 grid.addCol(cellcol);
170                 if (cellcol == grid.ncols()) {
171                         // We can't add a column to this grid, so let's
172                         // append the content of this cell to the previous
173                         // one.
174                         // This does not happen in well formed .lyx files,
175                         // but LyX versions 1.3.x and older could create
176                         // such files and tex2lyx can still do that.
177                         --cellcol;
178                         lyxerr << "ignoring extra column." << endl;
179                         return false;
180                 }
181         }
182         return true;
183 }
184
185
186 /*!
187  * Check wether the last row is empty and remove it if yes.
188  * Otherwise the following code
189  * \verbatim
190 \begin{array}{|c|c|}
191 \hline
192 1 & 2 \\ \hline
193 3 & 4 \\ \hline
194 \end{array}
195  * \endverbatim
196  * will result in a grid with 3 rows (+ the dummy row that is always present),
197  * because the last '\\' opens a new row.
198  */
199 void delEmptyLastRow(InsetMathGrid & grid)
200 {
201         InsetMathGrid::row_type const row = grid.nrows() - 1;
202         for (InsetMathGrid::col_type col = 0; col < grid.ncols(); ++col) {
203                 if (!grid.cell(grid.index(row, col)).empty())
204                         return;
205         }
206         // Copy the row information of the empty row (which would contain the
207         // last hline in the example above) to the dummy row and delete the
208         // empty row.
209         grid.rowinfo(row + 1) = grid.rowinfo(row);
210         grid.delRow(row);
211 }
212
213
214 // These are TeX's catcodes
215 enum CatCode {
216         catEscape,     // 0    backslash
217         catBegin,      // 1    {
218         catEnd,        // 2    }
219         catMath,       // 3    $
220         catAlign,      // 4    &
221         catNewline,    // 5    ^^M
222         catParameter,  // 6    #
223         catSuper,      // 7    ^
224         catSub,        // 8    _
225         catIgnore,     // 9
226         catSpace,      // 10   space
227         catLetter,     // 11   a-zA-Z
228         catOther,      // 12   none of the above
229         catActive,     // 13   ~
230         catComment,    // 14   %
231         catInvalid     // 15   <delete>
232 };
233
234 CatCode theCatcode[128];
235
236
237 inline CatCode catcode(char_type c)
238 {
239         /* The only characters that are not catOther lie in the pure ASCII
240          * range. Therefore theCatcode has only 128 entries.
241          * TeX itself deals with 8bit characters, so if needed this table
242          * could be enlarged to 256 entries.
243          * Any larger value does not make sense, since the fact that we use
244          * unicode internally does not change Knuth's TeX engine.
245          * Apart from that a table for the full 21bit UCS4 range would waste
246          * too much memory. */
247         if (c >= 128)
248                 return catOther;
249
250         return theCatcode[c];
251 }
252
253
254 enum {
255         FLAG_ALIGN      = 1 << 0,  //  next & or \\ ends the parsing process
256         FLAG_BRACE_LAST = 1 << 1,  //  next closing brace ends the parsing
257         FLAG_RIGHT      = 1 << 2,  //  next \\right ends the parsing process
258         FLAG_END        = 1 << 3,  //  next \\end ends the parsing process
259         FLAG_BRACK_LAST = 1 << 4,  //  next closing bracket ends the parsing
260         FLAG_TEXTMODE   = 1 << 5,  //  we are in a box
261         FLAG_ITEM       = 1 << 6,  //  read a (possibly braced) token
262         FLAG_LEAVE      = 1 << 7,  //  leave the loop at the end
263         FLAG_SIMPLE     = 1 << 8,  //  next $ leaves the loop
264         FLAG_EQUATION   = 1 << 9,  //  next \] leaves the loop
265         FLAG_SIMPLE2    = 1 << 10, //  next \) leaves the loop
266         FLAG_OPTION     = 1 << 11, //  read [...] style option
267         FLAG_BRACED     = 1 << 12  //  read {...} style argument
268 };
269
270
271 //
272 // Helper class for parsing
273 //
274
275 class Token {
276 public:
277         ///
278         Token() : cs_(), char_(0), cat_(catIgnore) {}
279         ///
280         Token(char_type c, CatCode cat) : cs_(), char_(c), cat_(cat) {}
281         ///
282         explicit Token(docstring const & cs) : cs_(cs), char_(0), cat_(catIgnore) {}
283
284         ///
285         docstring const & cs() const { return cs_; }
286         ///
287         CatCode cat() const { return cat_; }
288         ///
289         char_type character() const { return char_; }
290         ///
291         docstring asString() const { return cs_.size() ? cs_ : docstring(1, char_); }
292         ///
293         docstring asInput() const { return cs_.size() ? '\\' + cs_ : docstring(1, char_); }
294
295 private:
296         ///
297         docstring cs_;
298         ///
299         char_type char_;
300         ///
301         CatCode cat_;
302 };
303
304
305 ostream & operator<<(ostream & os, Token const & t)
306 {
307         if (t.cs().size()) {
308                 docstring const & cs = t.cs();
309                 // FIXME: For some strange reason, the stream operator instanciate
310                 // a new Token before outputting the contents of t.cs().
311                 // Because of this the line
312                 //     os << '\\' << cs;
313                 // below becomes recursive.
314                 // In order to avoid that we return early:
315                 if (cs == "\\")
316                         return os;
317                 os << '\\' << to_utf8(cs);
318         }
319         else if (t.cat() == catLetter)
320                 os << t.character();
321         else
322                 os << '[' << t.character() << ',' << t.cat() << ']';
323         return os;
324 }
325
326
327 class Parser {
328 public:
329         ///
330         typedef  InsetMath::mode_type mode_type;
331         ///
332         typedef  Parse::flags parse_mode;
333
334         ///
335         Parser(Lexer & lex, parse_mode mode);
336         /// Only use this for reading from .lyx file format, for the reason
337         /// see Parser::tokenize(istream &).
338         Parser(istream & is, parse_mode mode);
339         ///
340         Parser(docstring const & str, parse_mode mode);
341
342         ///
343         bool parse(MathAtom & at);
344         ///
345         bool parse(MathData & array, unsigned flags, mode_type mode);
346         ///
347         bool parse1(InsetMathGrid & grid, unsigned flags, mode_type mode,
348                 bool numbered);
349         ///
350         MathData parse(unsigned flags, mode_type mode);
351         ///
352         int lineno() const { return lineno_; }
353         ///
354         void putback();
355
356 private:
357         ///
358         void parse2(MathAtom & at, unsigned flags, mode_type mode, bool numbered);
359         /// get arg delimited by 'left' and 'right'
360         docstring getArg(char_type left, char_type right);
361         ///
362         char_type getChar();
363         ///
364         void error(string const & msg);
365         void error(docstring const & msg) { error(to_utf8(msg)); }
366         /// dump contents to screen
367         void dump() const;
368         /// Only use this for reading from .lyx file format (see
369         /// implementation for reason)
370         void tokenize(istream & is);
371         ///
372         void tokenize(docstring const & s);
373         ///
374         void skipSpaceTokens(idocstream & is, char_type c);
375         ///
376         void push_back(Token const & t);
377         ///
378         void pop_back();
379         ///
380         Token const & prevToken() const;
381         ///
382         Token const & nextToken() const;
383         ///
384         Token const & getToken();
385         /// skips spaces if any
386         void skipSpaces();
387         ///
388         void lex(docstring const & s);
389         ///
390         bool good() const;
391         ///
392         docstring parse_verbatim_item();
393         ///
394         docstring parse_verbatim_option();
395
396         ///
397         int lineno_;
398         ///
399         vector<Token> tokens_;
400         ///
401         unsigned pos_;
402         /// Stack of active environments
403         vector<docstring> environments_;
404         ///
405         parse_mode mode_;
406         ///
407         bool success_;
408 };
409
410
411 Parser::Parser(Lexer & lexer, parse_mode mode)
412         : lineno_(lexer.lineNumber()), pos_(0), mode_(mode), success_(true)
413 {
414         tokenize(lexer.getStream());
415         lexer.eatLine();
416 }
417
418
419 Parser::Parser(istream & is, parse_mode mode)
420         : lineno_(0), pos_(0), mode_(mode), success_(true)
421 {
422         tokenize(is);
423 }
424
425
426 Parser::Parser(docstring const & str, parse_mode mode)
427         : lineno_(0), pos_(0), mode_(mode), success_(true)
428 {
429         tokenize(str);
430 }
431
432
433 void Parser::push_back(Token const & t)
434 {
435         tokens_.push_back(t);
436 }
437
438
439 void Parser::pop_back()
440 {
441         tokens_.pop_back();
442 }
443
444
445 Token const & Parser::prevToken() const
446 {
447         static const Token dummy;
448         return pos_ > 0 ? tokens_[pos_ - 1] : dummy;
449 }
450
451
452 Token const & Parser::nextToken() const
453 {
454         static const Token dummy;
455         return good() ? tokens_[pos_] : dummy;
456 }
457
458
459 Token const & Parser::getToken()
460 {
461         static const Token dummy;
462         //lyxerr << "looking at token " << tokens_[pos_] << " pos: " << pos_ << endl;
463         return good() ? tokens_[pos_++] : dummy;
464 }
465
466
467 void Parser::skipSpaces()
468 {
469         while (nextToken().cat() == catSpace || nextToken().cat() == catNewline)
470                 getToken();
471 }
472
473
474 void Parser::putback()
475 {
476         --pos_;
477 }
478
479
480 bool Parser::good() const
481 {
482         return pos_ < tokens_.size();
483 }
484
485
486 char_type Parser::getChar()
487 {
488         if (!good()) {
489                 error("The input stream is not well...");
490                 putback();
491                 return 0;
492         }
493         return tokens_[pos_++].character();
494 }
495
496
497 docstring Parser::getArg(char_type left, char_type right)
498 {
499         skipSpaces();
500
501         docstring result;
502         char_type c = getChar();
503
504         if (c != left)
505                 putback();
506         else
507                 while ((c = getChar()) != right && good())
508                         result += c;
509
510         return result;
511 }
512
513
514 void Parser::skipSpaceTokens(idocstream & is, char_type c)
515 {
516         // skip trailing spaces
517         while (catcode(c) == catSpace || catcode(c) == catNewline)
518                 if (!is.get(c))
519                         break;
520         //lyxerr << "putting back: " << c << endl;
521         is.putback(c);
522 }
523
524
525 void Parser::tokenize(istream & is)
526 {
527         // eat everything up to the next \end_inset or end of stream
528         // and store it in s for further tokenization
529         string s;
530         char c;
531         while (is.get(c)) {
532                 s += c;
533                 if (s.size() >= 10 && s.substr(s.size() - 10) == "\\end_inset") {
534                         s = s.substr(0, s.size() - 10);
535                         break;
536                 }
537         }
538         // Remove the space after \end_inset
539         if (is.get(c) && c != ' ')
540                 is.unget();
541
542         // tokenize buffer
543         tokenize(from_utf8(s));
544 }
545
546
547 void Parser::tokenize(docstring const & buffer)
548 {
549         idocstringstream is(mode_ & Parse::VERBATIM
550                         ? escapeSpecialChars(buffer, mode_ & Parse::TEXTMODE)
551                         : buffer, ios::in | ios::binary);
552
553         char_type c;
554         while (is.get(c)) {
555                 //lyxerr << "reading c: " << c << endl;
556
557                 switch (catcode(c)) {
558                         case catNewline: {
559                                 ++lineno_;
560                                 is.get(c);
561                                 if (catcode(c) == catNewline)
562                                         ; //push_back(Token("par"));
563                                 else {
564                                         push_back(Token('\n', catNewline));
565                                         is.putback(c);
566                                 }
567                                 break;
568                         }
569
570 /*
571                         case catComment: {
572                                 while (is.get(c) && catcode(c) != catNewline)
573                                         ;
574                                 ++lineno_;
575                                 break;
576                         }
577 */
578
579                         case catEscape: {
580                                 is.get(c);
581                                 if (!is) {
582                                         error("unexpected end of input");
583                                 } else {
584                                         docstring s(1, c);
585                                         if (catcode(c) == catLetter) {
586                                                 // collect letters
587                                                 while (is.get(c) && catcode(c) == catLetter)
588                                                         s += c;
589                                                 skipSpaceTokens(is, c);
590                                         }
591                                         push_back(Token(s));
592                                 }
593                                 break;
594                         }
595
596                         case catSuper:
597                         case catSub: {
598                                 push_back(Token(c, catcode(c)));
599                                 is.get(c);
600                                 skipSpaceTokens(is, c);
601                                 break;
602                         }
603
604                         case catIgnore: {
605                                 if (!(mode_ & Parse::QUIET))
606                                         lyxerr << "ignoring a char: " << int(c) << endl;
607                                 break;
608                         }
609
610                         default:
611                                 push_back(Token(c, catcode(c)));
612                 }
613         }
614
615 #ifdef FILEDEBUG
616         dump();
617 #endif
618 }
619
620
621 void Parser::dump() const
622 {
623         lyxerr << "\nTokens: ";
624         for (unsigned i = 0; i < tokens_.size(); ++i) {
625                 if (i == pos_)
626                         lyxerr << " <#> ";
627                 lyxerr << tokens_[i];
628         }
629         lyxerr << " pos: " << pos_ << endl;
630 }
631
632
633 void Parser::error(string const & msg)
634 {
635         success_ = false;
636         if (!(mode_ & Parse::QUIET)) {
637                 lyxerr << "Line ~" << lineno_ << ": Math parse error: "
638                        << msg << endl;
639                 dump();
640         }
641 }
642
643
644 bool Parser::parse(MathAtom & at)
645 {
646         skipSpaces();
647         MathData ar;
648         parse(ar, false, InsetMath::UNDECIDED_MODE);
649         if (ar.size() != 1 || ar.front()->getType() == hullNone) {
650                 if (!(mode_ & Parse::QUIET))
651                         lyxerr << "unusual contents found: " << ar << endl;
652                 at = MathAtom(new InsetMathPar(ar));
653                 //if (at->nargs() > 0)
654                 //      at.nucleus()->cell(0) = ar;
655                 //else
656                 //      lyxerr << "unusual contents found: " << ar << endl;
657                 success_ = false;
658         } else
659                 at = ar[0];
660         return success_;
661 }
662
663
664 docstring Parser::parse_verbatim_option()
665 {
666         skipSpaces();
667         docstring res;
668         if (nextToken().character() == '[') {
669                 Token t = getToken();
670                 for (Token t = getToken(); t.character() != ']' && good(); t = getToken()) {
671                         if (t.cat() == catBegin) {
672                                 putback();
673                                 res += '{' + parse_verbatim_item() + '}';
674                         } else
675                                 res += t.asInput();
676                 }
677         }
678         return res;
679 }
680
681
682 docstring Parser::parse_verbatim_item()
683 {
684         skipSpaces();
685         docstring res;
686         if (nextToken().cat() == catBegin) {
687                 Token t = getToken();
688                 for (Token t = getToken(); t.cat() != catEnd && good(); t = getToken()) {
689                         if (t.cat() == catBegin) {
690                                 putback();
691                                 res += '{' + parse_verbatim_item() + '}';
692                         }
693                         else
694                                 res += t.asInput();
695                 }
696         }
697         return res;
698 }
699
700
701 MathData Parser::parse(unsigned flags, mode_type mode)
702 {
703         MathData ar;
704         parse(ar, flags, mode);
705         return ar;
706 }
707
708
709 bool Parser::parse(MathData & array, unsigned flags, mode_type mode)
710 {
711         InsetMathGrid grid(1, 1);
712         parse1(grid, flags, mode, false);
713         array = grid.cell(0);
714         return success_;
715 }
716
717
718 void Parser::parse2(MathAtom & at, const unsigned flags, const mode_type mode,
719         const bool numbered)
720 {
721         parse1(*(at.nucleus()->asGridInset()), flags, mode, numbered);
722 }
723
724
725 bool Parser::parse1(InsetMathGrid & grid, unsigned flags,
726         const mode_type mode, const bool numbered)
727 {
728         int limits = 0;
729         InsetMathGrid::row_type cellrow = 0;
730         InsetMathGrid::col_type cellcol = 0;
731         MathData * cell = &grid.cell(grid.index(cellrow, cellcol));
732
733         if (grid.asHullInset())
734                 grid.asHullInset()->numbered(cellrow, numbered);
735
736         //dump();
737         //lyxerr << " flags: " << flags << endl;
738         //lyxerr << " mode: " << mode  << endl;
739         //lyxerr << "grid: " << grid << endl;
740
741         while (good()) {
742                 Token const & t = getToken();
743
744 #ifdef FILEDEBUG
745                 lyxerr << "t: " << t << " flags: " << flags << endl;
746                 lyxerr << "mode: " << mode  << endl;
747                 cell->dump();
748                 lyxerr << endl;
749 #endif
750
751                 if (flags & FLAG_ITEM) {
752
753                         if (t.cat() == catBegin) {
754                                 // skip the brace and collect everything to the next matching
755                                 // closing brace
756                                 parse1(grid, FLAG_BRACE_LAST, mode, numbered);
757                                 return success_;
758                         }
759
760                         // handle only this single token, leave the loop if done
761                         flags = FLAG_LEAVE;
762                 }
763
764
765                 if (flags & FLAG_BRACED) {
766                         if (t.cat() == catSpace)
767                                 continue;
768
769                         if (t.cat() != catBegin) {
770                                 error("opening brace expected");
771                                 return success_;
772                         }
773
774                         // skip the brace and collect everything to the next matching
775                         // closing brace
776                         flags = FLAG_BRACE_LAST;
777                 }
778
779
780                 if (flags & FLAG_OPTION) {
781                         if (t.cat() == catOther && t.character() == '[') {
782                                 MathData ar;
783                                 parse(ar, FLAG_BRACK_LAST, mode);
784                                 cell->append(ar);
785                         } else {
786                                 // no option found, put back token and we are done
787                                 putback();
788                         }
789                         return success_;
790                 }
791
792                 //
793                 // cat codes
794                 //
795                 if (t.cat() == catMath) {
796                         if (mode != InsetMath::MATH_MODE) {
797                                 // we are inside some text mode thingy, so opening new math is allowed
798                                 Token const & n = getToken();
799                                 if (n.cat() == catMath) {
800                                         // TeX's $$...$$ syntax for displayed math
801                                         cell->push_back(MathAtom(new InsetMathHull(hullEquation)));
802                                         parse2(cell->back(), FLAG_SIMPLE, InsetMath::MATH_MODE, false);
803                                         getToken(); // skip the second '$' token
804                                 } else {
805                                         // simple $...$  stuff
806                                         putback();
807                                         cell->push_back(MathAtom(new InsetMathHull(hullSimple)));
808                                         parse2(cell->back(), FLAG_SIMPLE, InsetMath::MATH_MODE, false);
809                                 }
810                         }
811
812                         else if (flags & FLAG_SIMPLE) {
813                                 // this is the end of the formula
814                                 return success_;
815                         }
816
817                         else {
818                                 error("something strange in the parser");
819                                 break;
820                         }
821                 }
822
823                 else if (t.cat() == catLetter)
824                         cell->push_back(MathAtom(new InsetMathChar(t.character())));
825
826                 else if (t.cat() == catSpace && mode != InsetMath::MATH_MODE) {
827                         if (cell->empty() || cell->back()->getChar() != ' ')
828                                 cell->push_back(MathAtom(new InsetMathChar(t.character())));
829                 }
830
831                 else if (t.cat() == catNewline && mode != InsetMath::MATH_MODE) {
832                         if (cell->empty() || cell->back()->getChar() != ' ')
833                                 cell->push_back(MathAtom(new InsetMathChar(' ')));
834                 }
835
836                 else if (t.cat() == catParameter) {
837                         Token const & n = getToken();
838                         cell->push_back(MathAtom(new MathMacroArgument(n.character()-'0')));
839                 }
840
841                 else if (t.cat() == catActive)
842                         cell->push_back(MathAtom(new InsetMathChar(t.character())));
843
844                 else if (t.cat() == catBegin) {
845                         MathData ar;
846                         parse(ar, FLAG_BRACE_LAST, mode);
847                         // do not create a BraceInset if they were written by LyX
848                         // this helps to keep the annoyance of  "a choose b"  to a minimum
849                         if (ar.size() == 1 && ar[0]->extraBraces())
850                                 cell->append(ar);
851                         else
852                                 cell->push_back(MathAtom(new InsetMathBrace(ar)));
853                 }
854
855                 else if (t.cat() == catEnd) {
856                         if (flags & FLAG_BRACE_LAST)
857                                 return success_;
858                         error("found '}' unexpectedly");
859                         //LASSERT(false, /**/);
860                         //add(cell, '}', LM_TC_TEX);
861                 }
862
863                 else if (t.cat() == catAlign) {
864                         //lyxerr << " column now " << (cellcol + 1)
865                         //       << " max: " << grid.ncols() << endl;
866                         if (flags & FLAG_ALIGN)
867                                 return success_;
868                         if (addCol(grid, cellcol))
869                                 cell = &grid.cell(grid.index(cellrow, cellcol));
870                 }
871
872                 else if (t.cat() == catSuper || t.cat() == catSub) {
873                         bool up = (t.cat() == catSuper);
874                         // we need no new script inset if the last thing was a scriptinset,
875                         // which has that script already not the same script already
876                         if (!cell->size())
877                                 cell->push_back(MathAtom(new InsetMathScript(up)));
878                         else if (cell->back()->asScriptInset() &&
879                                         !cell->back()->asScriptInset()->has(up))
880                                 cell->back().nucleus()->asScriptInset()->ensure(up);
881                         else if (cell->back()->asScriptInset())
882                                 cell->push_back(MathAtom(new InsetMathScript(up)));
883                         else
884                                 cell->back() = MathAtom(new InsetMathScript(cell->back(), up));
885                         InsetMathScript * p = cell->back().nucleus()->asScriptInset();
886                         // special handling of {}-bases
887                         // Here we could remove the brace inset for things
888                         // like {a'}^2 and add the braces back in
889                         // InsetMathScript::write().
890                         // We do not do it, since it is not possible to detect
891                         // reliably whether the braces are needed because the
892                         // nucleus contains more than one symbol, or whether
893                         // they are needed for unknown commands like \xx{a}_0
894                         // or \yy{a}{b}_0. This was done in revision 14819
895                         // in an unreliable way. See this thread
896                         // http://www.mail-archive.com/lyx-devel%40lists.lyx.org/msg104917.html
897                         // for more details.
898                         parse(p->cell(p->idxOfScript(up)), FLAG_ITEM, mode);
899                         if (limits) {
900                                 p->limits(limits);
901                                 limits = 0;
902                         }
903                 }
904
905                 else if (t.character() == ']' && (flags & FLAG_BRACK_LAST)) {
906                         //lyxerr << "finished reading option" << endl;
907                         return success_;
908                 }
909
910                 else if (t.cat() == catOther)
911                         cell->push_back(MathAtom(new InsetMathChar(t.character())));
912
913                 else if (t.cat() == catComment) {
914                         docstring s;
915                         while (good()) {
916                                 Token const & t = getToken();
917                                 if (t.cat() == catNewline)
918                                         break;
919                                 s += t.asString();
920                         }
921                         cell->push_back(MathAtom(new InsetMathComment(s)));
922                         skipSpaces();
923                 }
924
925                 //
926                 // control sequences
927                 //
928
929                 else if (t.cs() == "lyxlock") {
930                         if (cell->size())
931                                 cell->back().nucleus()->lock(true);
932                 }
933
934                 else if ((t.cs() == "global" && nextToken().cs() == "def") ||
935                          t.cs() == "def") {
936                         if (t.cs() == "global")
937                                 getToken();
938                         
939                         // get name
940                         docstring name = getToken().cs();
941                         
942                         // read parameters
943                         int nargs = 0;
944                         docstring pars;
945                         while (good() && nextToken().cat() != catBegin) {
946                                 pars += getToken().cs();
947                                 ++nargs;
948                         }
949                         nargs /= 2;
950                         
951                         // read definition
952                         MathData def;
953                         parse(def, FLAG_ITEM, InsetMath::UNDECIDED_MODE);
954                         
955                         // is a version for display attached?
956                         skipSpaces();
957                         MathData display;
958                         if (nextToken().cat() == catBegin)
959                                 parse(display, FLAG_ITEM, InsetMath::MATH_MODE);
960                         
961                         cell->push_back(MathAtom(new MathMacroTemplate(name, nargs,
962                                0, MacroTypeDef, vector<MathData>(), def, display)));
963                 }
964                 
965                 else if (t.cs() == "newcommand" ||
966                          t.cs() == "renewcommand" ||
967                          t.cs() == "newlyxcommand") {
968                         // get name
969                         if (getToken().cat() != catBegin) {
970                                 error("'{' in \\newcommand expected (1) ");
971                                 return success_;
972                         }
973                         docstring name = getToken().cs();
974                         if (getToken().cat() != catEnd) {
975                                 error("'}' in \\newcommand expected");
976                                 return success_;
977                         }
978                                 
979                         // get arity
980                         docstring const arg = getArg('[', ']');
981                         int nargs = 0;
982                         if (!arg.empty())
983                                 nargs = convert<int>(arg);
984                                 
985                         // optional argument given?
986                         skipSpaces();
987                         int optionals = 0;
988                         vector<MathData> optionalValues;
989                         while (nextToken().character() == '[') {
990                                 getToken();
991                                 optionalValues.push_back(MathData());
992                                 parse(optionalValues[optionals], FLAG_BRACK_LAST, mode);
993                                 ++optionals;
994                         }
995                         
996                         MathData def;
997                         parse(def, FLAG_ITEM, InsetMath::UNDECIDED_MODE);
998                         
999                         // is a version for display attached?
1000                         skipSpaces();
1001                         MathData display;
1002                         if (nextToken().cat() == catBegin)
1003                                 parse(display, FLAG_ITEM, InsetMath::MATH_MODE);
1004                         
1005                         cell->push_back(MathAtom(new MathMacroTemplate(name, nargs,
1006                                 optionals, MacroTypeNewcommand, optionalValues, def, display)));
1007                         
1008                 }
1009                 
1010                 else if (t.cs() == "newcommandx" ||
1011                          t.cs() == "renewcommandx") {
1012                         // \newcommandx{\foo}[2][usedefault, addprefix=\global,1=default]{#1,#2}
1013                         // get name
1014                         docstring name;
1015                         if (nextToken().cat() == catBegin) {
1016                                 getToken();
1017                                 name = getToken().cs();
1018                                 if (getToken().cat() != catEnd) {
1019                                         error("'}' in \\newcommandx expected");
1020                                         return success_;
1021                                 }
1022                         } else
1023                                 name = getToken().cs();
1024                                 
1025                         // get arity
1026                         docstring const arg = getArg('[', ']');
1027                         if (arg.empty()) {
1028                                 error("[num] in \\newcommandx expected");
1029                                 return success_;
1030                         }
1031                         int nargs = convert<int>(arg);
1032                         
1033                         // get options
1034                         int optionals = 0;
1035                         vector<MathData> optionalValues;
1036                         if (nextToken().character() == '[') {
1037                                 // skip '['
1038                                 getToken();
1039                                         
1040                                 // handle 'opt=value' options, separated by ','.
1041                                 skipSpaces();
1042                                 while (nextToken().character() != ']' && good()) {
1043                                         if (nextToken().character() >= '1'
1044                                             && nextToken().character() <= '9') {
1045                                                 // optional value -> get parameter number
1046                                                 int n = getChar() - '0';
1047                                                 if (n > nargs) {
1048                                                         error("Arity of \\newcommandx too low "
1049                                                               "for given optional parameter.");
1050                                                         return success_;
1051                                                 }
1052                                                 
1053                                                 // skip '='
1054                                                 if (getToken().character() != '=') {
1055                                                         error("'=' and optional parameter value "
1056                                                               "expected for \\newcommandx");
1057                                                         return success_;
1058                                                 }
1059                                                 
1060                                                 // get value
1061                                                 int optNum = max(size_t(n), optionalValues.size());
1062                                                 optionalValues.resize(optNum);
1063                                                 optionalValues[n - 1].clear();
1064                                                 while (nextToken().character() != ']'
1065                                                        && nextToken().character() != ',') {
1066                                                         MathData data;
1067                                                         parse(data, FLAG_ITEM, InsetMath::UNDECIDED_MODE);
1068                                                         optionalValues[n - 1].append(data);
1069                                                 }
1070                                                 optionals = max(n, optionals);
1071                                         } else if (nextToken().cat() == catLetter) {
1072                                                 // we in fact ignore every non-optional
1073                                                 // parameter
1074                                                 
1075                                                 // get option name
1076                                                 docstring opt;
1077                                                 while (nextToken().cat() == catLetter)
1078                                                         opt += getChar();
1079                                         
1080                                                 // value?
1081                                                 skipSpaces();
1082                                                 MathData value;
1083                                                 if (nextToken().character() == '=') {
1084                                                         getToken();
1085                                                         while (nextToken().character() != ']'
1086                                                                 && nextToken().character() != ',')
1087                                                                 parse(value, FLAG_ITEM, 
1088                                                                       InsetMath::UNDECIDED_MODE);
1089                                                 }
1090                                         } else {
1091                                                 error("option for \\newcommandx expected");
1092                                                 return success_;
1093                                         }
1094                                         
1095                                         // skip komma
1096                                         skipSpaces();
1097                                         if (nextToken().character() == ',') {
1098                                                 getChar();
1099                                                 skipSpaces();
1100                                         } else if (nextToken().character() != ']') {
1101                                                 error("Expecting ',' or ']' in options "
1102                                                       "of \\newcommandx");
1103                                                 return success_;
1104                                         }
1105                                 }
1106                                 
1107                                 // skip ']'
1108                                 if (!good())
1109                                         return success_;
1110                                 getToken();
1111                         }
1112
1113                         // get definition
1114                         MathData def;
1115                         parse(def, FLAG_ITEM, InsetMath::UNDECIDED_MODE);
1116
1117                         // is a version for display attached?
1118                         skipSpaces();
1119                         MathData display;
1120                         if (nextToken().cat() == catBegin)
1121                                 parse(display, FLAG_ITEM, InsetMath::MATH_MODE);
1122
1123                         cell->push_back(MathAtom(new MathMacroTemplate(name, nargs,
1124                                 optionals, MacroTypeNewcommandx, optionalValues, def, 
1125                                 display)));
1126                 }
1127
1128                 else if (t.cs() == "(") {
1129                         if (mode == InsetMath::MATH_MODE) {
1130                                 error("bad math environment");
1131                                 break;
1132                         }
1133                         cell->push_back(MathAtom(new InsetMathHull(hullSimple)));
1134                         parse2(cell->back(), FLAG_SIMPLE2, InsetMath::MATH_MODE, false);
1135                 }
1136
1137                 else if (t.cs() == "[") {
1138                         if (mode != InsetMath::UNDECIDED_MODE) {
1139                                 error("bad math environment");
1140                                 break;
1141                         }
1142                         cell->push_back(MathAtom(new InsetMathHull(hullEquation)));
1143                         parse2(cell->back(), FLAG_EQUATION, InsetMath::MATH_MODE, false);
1144                 }
1145
1146                 else if (t.cs() == "protect")
1147                         // ignore \\protect, will hopefully be re-added during output
1148                         ;
1149
1150                 else if (t.cs() == "end") {
1151                         if (flags & FLAG_END) {
1152                                 // eat environment name
1153                                 docstring const name = getArg('{', '}');
1154                                 if (environments_.empty())
1155                                         error("'found \\end{" + name +
1156                                               "}' without matching '\\begin{" +
1157                                               name + "}'");
1158                                 else if (name != environments_.back())
1159                                         error("'\\end{" + name +
1160                                               "}' does not match '\\begin{" +
1161                                               environments_.back() + "}'");
1162                                 else {
1163                                         environments_.pop_back();
1164                                         // Delete empty last row in matrix
1165                                         // like insets.
1166                                         // If you abuse InsetMathGrid for
1167                                         // non-matrix like structures you
1168                                         // probably need to refine this test.
1169                                         // Right now we only have to test for
1170                                         // single line hull insets.
1171                                         if (grid.nrows() > 1)
1172                                                 delEmptyLastRow(grid);
1173                                         return success_;
1174                                 }
1175                         } else
1176                                 error("found 'end' unexpectedly");
1177                 }
1178
1179                 else if (t.cs() == ")") {
1180                         if (flags & FLAG_SIMPLE2)
1181                                 return success_;
1182                         error("found '\\)' unexpectedly");
1183                 }
1184
1185                 else if (t.cs() == "]") {
1186                         if (flags & FLAG_EQUATION)
1187                                 return success_;
1188                         error("found '\\]' unexpectedly");
1189                 }
1190
1191                 else if (t.cs() == "\\") {
1192                         if (flags & FLAG_ALIGN)
1193                                 return success_;
1194                         bool added = false;
1195                         if (nextToken().asInput() == "*") {
1196                                 getToken();
1197                                 added = addRow(grid, cellrow, docstring(), false);
1198                         } else if (good())
1199                                 added = addRow(grid, cellrow, getArg('[', ']'));
1200                         else
1201                                 error("missing token after \\\\");
1202                         if (added) {
1203                                 cellcol = 0;
1204                                 if (grid.asHullInset())
1205                                         grid.asHullInset()->numbered(
1206                                                         cellrow, numbered);
1207                                 cell = &grid.cell(grid.index(cellrow,
1208                                                              cellcol));
1209                         }
1210                 }
1211
1212 #if 0
1213                 else if (t.cs() == "multicolumn") {
1214                         // extract column count and insert dummy cells
1215                         MathData count;
1216                         parse(count, FLAG_ITEM, mode);
1217                         int cols = 1;
1218                         if (!extractNumber(count, cols)) {
1219                                 success_ = false;
1220                                 lyxerr << " can't extract number of cells from " << count << endl;
1221                         }
1222                         // resize the table if necessary
1223                         for (int i = 0; i < cols; ++i) {
1224                                 if (addCol(grid, cellcol)) {
1225                                         cell = &grid.cell(grid.index(
1226                                                         cellrow, cellcol));
1227                                         // mark this as dummy
1228                                         grid.cellinfo(grid.index(
1229                                                 cellrow, cellcol)).dummy_ = true;
1230                                 }
1231                         }
1232                         // the last cell is the real thing, not a dummy
1233                         grid.cellinfo(grid.index(cellrow, cellcol)).dummy_ = false;
1234
1235                         // read special alignment
1236                         MathData align;
1237                         parse(align, FLAG_ITEM, mode);
1238                         //grid.cellinfo(grid.index(cellrow, cellcol)).align_ = extractString(align);
1239
1240                         // parse the remaining contents into the "real" cell
1241                         parse(*cell, FLAG_ITEM, mode);
1242                 }
1243 #endif
1244
1245                 else if (t.cs() == "limits")
1246                         limits = 1;
1247
1248                 else if (t.cs() == "nolimits")
1249                         limits = -1;
1250
1251                 else if (t.cs() == "nonumber") {
1252                         if (grid.asHullInset())
1253                                 grid.asHullInset()->numbered(cellrow, false);
1254                 }
1255
1256                 else if (t.cs() == "number") {
1257                         if (grid.asHullInset())
1258                                 grid.asHullInset()->numbered(cellrow, true);
1259                 }
1260
1261                 else if (t.cs() == "hline") {
1262                         grid.rowinfo(cellrow).lines_ ++;
1263                 }
1264
1265                 else if (t.cs() == "sqrt") {
1266                         MathData ar;
1267                         parse(ar, FLAG_OPTION, mode);
1268                         if (ar.size()) {
1269                                 cell->push_back(MathAtom(new InsetMathRoot));
1270                                 cell->back().nucleus()->cell(0) = ar;
1271                                 parse(cell->back().nucleus()->cell(1), FLAG_ITEM, mode);
1272                         } else {
1273                                 cell->push_back(MathAtom(new InsetMathSqrt));
1274                                 parse(cell->back().nucleus()->cell(0), FLAG_ITEM, mode);
1275                         }
1276                 }
1277
1278                 else if (t.cs() == "unit") {
1279                         // Allowed formats \unit[val]{unit}
1280                         MathData ar;
1281                         parse(ar, FLAG_OPTION, mode);
1282                         if (ar.size()) {
1283                                 cell->push_back(MathAtom(new InsetMathFrac(InsetMathFrac::UNIT)));
1284                                 cell->back().nucleus()->cell(0) = ar;
1285                                 parse(cell->back().nucleus()->cell(1), FLAG_ITEM, mode);
1286                         } else {
1287                                 cell->push_back(MathAtom(new InsetMathFrac(InsetMathFrac::UNIT, 1)));
1288                                 parse(cell->back().nucleus()->cell(0), FLAG_ITEM, mode);
1289                         }
1290                 }
1291                 else if (t.cs() == "unitfrac") {
1292                         // Here allowed formats are \unitfrac[val]{num}{denom}
1293                         MathData ar;
1294                         parse(ar, FLAG_OPTION, mode);
1295                         if (ar.size()) {
1296                                 cell->push_back(MathAtom(new InsetMathFrac(InsetMathFrac::UNITFRAC, 3)));
1297                                 cell->back().nucleus()->cell(2) = ar;
1298                         } else {
1299                                 cell->push_back(MathAtom(new InsetMathFrac(InsetMathFrac::UNITFRAC)));
1300                         }
1301                         parse(cell->back().nucleus()->cell(0), FLAG_ITEM, mode);
1302                         parse(cell->back().nucleus()->cell(1), FLAG_ITEM, mode);
1303                 }
1304
1305                 else if (t.cs() == "xrightarrow" || t.cs() == "xleftarrow") {
1306                         cell->push_back(createInsetMath(t.cs()));
1307                         parse(cell->back().nucleus()->cell(1), FLAG_OPTION, mode);
1308                         parse(cell->back().nucleus()->cell(0), FLAG_ITEM, mode);
1309                 }
1310
1311                 else if (t.cs() == "ref" || t.cs() == "eqref" || t.cs() == "prettyref"
1312                           || t.cs() == "pageref" || t.cs() == "vpageref" || t.cs() == "vref") {
1313                         cell->push_back(MathAtom(new InsetMathRef(t.cs())));
1314                         parse(cell->back().nucleus()->cell(1), FLAG_OPTION, mode);
1315                         parse(cell->back().nucleus()->cell(0), FLAG_ITEM, mode);
1316                 }
1317
1318                 else if (t.cs() == "left") {
1319                         skipSpaces();
1320                         Token const & tl = getToken();
1321                         // \| and \Vert are equivalent, and InsetMathDelim
1322                         // can't handle \|
1323                         // FIXME: fix this in InsetMathDelim itself!
1324                         docstring const l = tl.cs() == "|" ? from_ascii("Vert") : tl.asString();
1325                         MathData ar;
1326                         parse(ar, FLAG_RIGHT, mode);
1327                         if (!good())
1328                                 break;
1329                         skipSpaces();
1330                         Token const & tr = getToken();
1331                         docstring const r = tr.cs() == "|" ? from_ascii("Vert") : tr.asString();
1332                         cell->push_back(MathAtom(new InsetMathDelim(l, r, ar)));
1333                 }
1334
1335                 else if (t.cs() == "right") {
1336                         if (flags & FLAG_RIGHT)
1337                                 return success_;
1338                         //lyxerr << "got so far: '" << cell << "'" << endl;
1339                         error("Unmatched right delimiter");
1340                         return success_;
1341                 }
1342
1343                 else if (t.cs() == "begin") {
1344                         docstring const name = getArg('{', '}');
1345                         environments_.push_back(name);
1346
1347                         if (name == "array" || name == "subarray") {
1348                                 docstring const valign = parse_verbatim_option() + 'c';
1349                                 docstring const halign = parse_verbatim_item();
1350                                 cell->push_back(MathAtom(new InsetMathArray(name,
1351                                         InsetMathGrid::guessColumns(halign), 1, (char)valign[0], halign)));
1352                                 parse2(cell->back(), FLAG_END, mode, false);
1353                         }
1354
1355                         else if (name == "tabular") {
1356                                 docstring const valign = parse_verbatim_option() + 'c';
1357                                 docstring const halign = parse_verbatim_item();
1358                                 cell->push_back(MathAtom(new InsetMathTabular(name,
1359                                         InsetMathGrid::guessColumns(halign), 1, (char)valign[0], halign)));
1360                                 parse2(cell->back(), FLAG_END, InsetMath::TEXT_MODE, false);
1361                         }
1362
1363                         else if (name == "split" || name == "cases") {
1364                                 cell->push_back(createInsetMath(name));
1365                                 parse2(cell->back(), FLAG_END, mode, false);
1366                         }
1367
1368                         else if (name == "alignedat") {
1369                                 docstring const valign = parse_verbatim_option() + 'c';
1370                                 // ignore this for a while
1371                                 getArg('{', '}');
1372                                 cell->push_back(MathAtom(new InsetMathSplit(name, (char)valign[0])));
1373                                 parse2(cell->back(), FLAG_END, mode, false);
1374                         }
1375
1376                         else if (name == "math") {
1377                                 if (mode == InsetMath::MATH_MODE) {
1378                                         error("bad math environment");
1379                                         break;
1380                                 }
1381                                 cell->push_back(MathAtom(new InsetMathHull(hullSimple)));
1382                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, true);
1383                         }
1384
1385                         else if (name == "equation" || name == "equation*"
1386                                         || name == "displaymath") {
1387                                 if (mode != InsetMath::UNDECIDED_MODE) {
1388                                         error("bad math environment");
1389                                         break;
1390                                 }
1391                                 cell->push_back(MathAtom(new InsetMathHull(hullEquation)));
1392                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, (name == "equation"));
1393                         }
1394
1395                         else if (name == "eqnarray" || name == "eqnarray*") {
1396                                 if (mode != InsetMath::UNDECIDED_MODE) {
1397                                         error("bad math environment");
1398                                         break;
1399                                 }
1400                                 cell->push_back(MathAtom(new InsetMathHull(hullEqnArray)));
1401                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1402                         }
1403
1404                         else if (name == "align" || name == "align*") {
1405                                 if (mode != InsetMath::UNDECIDED_MODE) {
1406                                         error("bad math environment");
1407                                         break;
1408                                 }
1409                                 cell->push_back(MathAtom(new InsetMathHull(hullAlign)));
1410                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1411                         }
1412
1413                         else if (name == "flalign" || name == "flalign*") {
1414                                 if (mode != InsetMath::UNDECIDED_MODE) {
1415                                         error("bad math environment");
1416                                         break;
1417                                 }
1418                                 cell->push_back(MathAtom(new InsetMathHull(hullFlAlign)));
1419                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1420                         }
1421
1422                         else if (name == "alignat" || name == "alignat*") {
1423                                 if (mode != InsetMath::UNDECIDED_MODE) {
1424                                         error("bad math environment");
1425                                         break;
1426                                 }
1427                                 // ignore this for a while
1428                                 getArg('{', '}');
1429                                 cell->push_back(MathAtom(new InsetMathHull(hullAlignAt)));
1430                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1431                         }
1432
1433                         else if (name == "xalignat" || name == "xalignat*") {
1434                                 if (mode != InsetMath::UNDECIDED_MODE) {
1435                                         error("bad math environment");
1436                                         break;
1437                                 }
1438                                 // ignore this for a while
1439                                 getArg('{', '}');
1440                                 cell->push_back(MathAtom(new InsetMathHull(hullXAlignAt)));
1441                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1442                         }
1443
1444                         else if (name == "xxalignat") {
1445                                 if (mode != InsetMath::UNDECIDED_MODE) {
1446                                         error("bad math environment");
1447                                         break;
1448                                 }
1449                                 // ignore this for a while
1450                                 getArg('{', '}');
1451                                 cell->push_back(MathAtom(new InsetMathHull(hullXXAlignAt)));
1452                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1453                         }
1454
1455                         else if (name == "multline" || name == "multline*") {
1456                                 if (mode != InsetMath::UNDECIDED_MODE) {
1457                                         error("bad math environment");
1458                                         break;
1459                                 }
1460                                 cell->push_back(MathAtom(new InsetMathHull(hullMultline)));
1461                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1462                         }
1463
1464                         else if (name == "gather" || name == "gather*") {
1465                                 if (mode != InsetMath::UNDECIDED_MODE) {
1466                                         error("bad math environment");
1467                                         break;
1468                                 }
1469                                 cell->push_back(MathAtom(new InsetMathHull(hullGather)));
1470                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1471                         }
1472
1473                         else if (latexkeys const * l = in_word_set(name)) {
1474                                 if (l->inset == "matrix") {
1475                                         cell->push_back(createInsetMath(name));
1476                                         parse2(cell->back(), FLAG_END, mode, false);
1477                                 } else if (l->inset == "split") {
1478                                         docstring const valign = parse_verbatim_option() + 'c';
1479                                         cell->push_back(MathAtom(new InsetMathSplit(name, (char)valign[0])));
1480                                         parse2(cell->back(), FLAG_END, mode, false);
1481                                 } else {
1482                                         success_ = false;
1483                                         if (!(mode_ & Parse::QUIET)) {
1484                                                 dump();
1485                                                 lyxerr << "found math environment `"
1486                                                        << to_utf8(name)
1487                                                        << "' in symbols file with unsupported inset `"
1488                                                        << to_utf8(l->inset)
1489                                                        << "'." << endl;
1490                                         }
1491                                         // create generic environment inset
1492                                         cell->push_back(MathAtom(new InsetMathEnv(name)));
1493                                         parse(cell->back().nucleus()->cell(0), FLAG_END, mode);
1494                                 }
1495                         }
1496
1497                         else {
1498                                 success_ = false;
1499                                 if (!(mode_ & Parse::QUIET)) {
1500                                         dump();
1501                                         lyxerr << "found unknown math environment '"
1502                                                << to_utf8(name) << "'" << endl;
1503                                 }
1504                                 // create generic environment inset
1505                                 cell->push_back(MathAtom(new InsetMathEnv(name)));
1506                                 parse(cell->back().nucleus()->cell(0), FLAG_END, mode);
1507                         }
1508                 }
1509
1510                 else if (t.cs() == "kern") {
1511                         // FIXME: A hack...
1512                         docstring s;
1513                         while (true) {
1514                                 Token const & t = getToken();
1515                                 if (!good()) {
1516                                         putback();
1517                                         break;
1518                                 }
1519                                 s += t.character();
1520                                 if (isValidLength(to_utf8(s)))
1521                                         break;
1522                         }
1523                         cell->push_back(MathAtom(new InsetMathKern(s)));
1524                 }
1525
1526                 else if (t.cs() == "label") {
1527                         // FIXME: This is swallowed in inline formulas
1528                         docstring label = parse_verbatim_item();
1529                         MathData ar;
1530                         asArray(label, ar);
1531                         if (grid.asHullInset()) {
1532                                 grid.asHullInset()->label(cellrow, label);
1533                         } else {
1534                                 cell->push_back(createInsetMath(t.cs()));
1535                                 cell->push_back(MathAtom(new InsetMathBrace(ar)));
1536                         }
1537                 }
1538
1539                 else if (t.cs() == "choose" || t.cs() == "over"
1540                                 || t.cs() == "atop" || t.cs() == "brace"
1541                                 || t.cs() == "brack") {
1542                         MathAtom at = createInsetMath(t.cs());
1543                         at.nucleus()->cell(0) = *cell;
1544                         cell->clear();
1545                         parse(at.nucleus()->cell(1), flags, mode);
1546                         cell->push_back(at);
1547                         return success_;
1548                 }
1549
1550                 else if (t.cs() == "color") {
1551                         docstring const color = parse_verbatim_item();
1552                         cell->push_back(MathAtom(new InsetMathColor(true, color)));
1553                         parse(cell->back().nucleus()->cell(0), flags, mode);
1554                         return success_;
1555                 }
1556
1557                 else if (t.cs() == "textcolor") {
1558                         docstring const color = parse_verbatim_item();
1559                         cell->push_back(MathAtom(new InsetMathColor(false, color)));
1560                         parse(cell->back().nucleus()->cell(0), FLAG_ITEM, InsetMath::TEXT_MODE);
1561                 }
1562
1563                 else if (t.cs() == "normalcolor") {
1564                         cell->push_back(createInsetMath(t.cs()));
1565                         parse(cell->back().nucleus()->cell(0), flags, mode);
1566                         return success_;
1567                 }
1568
1569                 else if (t.cs() == "substack") {
1570                         cell->push_back(createInsetMath(t.cs()));
1571                         parse2(cell->back(), FLAG_ITEM, mode, false);
1572                 }
1573
1574                 else if (t.cs() == "xymatrix") {
1575                         odocstringstream os;
1576                         while (good() && nextToken().cat() != catBegin)
1577                                 os << getToken().asInput();
1578                         cell->push_back(createInsetMath(t.cs() + os.str()));
1579                         parse2(cell->back(), FLAG_ITEM, mode, false);
1580                 }
1581
1582                 else if (t.cs() == "framebox" || t.cs() == "makebox") {
1583                         cell->push_back(createInsetMath(t.cs()));
1584                         parse(cell->back().nucleus()->cell(0), FLAG_OPTION, InsetMath::TEXT_MODE);
1585                         parse(cell->back().nucleus()->cell(1), FLAG_OPTION, InsetMath::TEXT_MODE);
1586                         parse(cell->back().nucleus()->cell(2), FLAG_ITEM, InsetMath::TEXT_MODE);
1587                 }
1588
1589                 else if (t.cs() == "tag") {
1590                         if (nextToken().character() == '*') {
1591                                 getToken();
1592                                 cell->push_back(createInsetMath(t.cs() + '*'));
1593                         } else
1594                                 cell->push_back(createInsetMath(t.cs()));
1595                         parse(cell->back().nucleus()->cell(0), FLAG_ITEM, InsetMath::TEXT_MODE);
1596                 }
1597
1598                 else if (t.cs() == "hspace" && nextToken().character() != '*') {
1599                         docstring const name = t.cs();
1600                         docstring const arg = parse_verbatim_item();
1601                         Length length;
1602                         if (isValidLength(to_utf8(arg), &length))
1603                                 cell->push_back(MathAtom(new InsetMathSpace(length)));
1604                         else {
1605                                 // Since the Length class cannot use length variables
1606                                 // we must not create an InsetMathSpace.
1607                                 cell->push_back(MathAtom(new MathMacro(name)));
1608                                 MathData ar;
1609                                 mathed_parse_cell(ar, '{' + arg + '}');
1610                                 cell->append(ar);
1611                         }
1612                 }
1613
1614 #if 0
1615                 else if (t.cs() == "infer") {
1616                         MathData ar;
1617                         parse(ar, FLAG_OPTION, mode);
1618                         cell->push_back(createInsetMath(t.cs()));
1619                         parse2(cell->back(), FLAG_ITEM, mode, false);
1620                 }
1621
1622                 // Disabled
1623                 else if (1 && t.cs() == "ar") {
1624                         auto_ptr<InsetMathXYArrow> p(new InsetMathXYArrow);
1625                         // try to read target
1626                         parse(p->cell(0), FLAG_OTPTION, mode);
1627                         // try to read label
1628                         if (nextToken().cat() == catSuper || nextToken().cat() == catSub) {
1629                                 p->up_ = nextToken().cat() == catSuper;
1630                                 getToken();
1631                                 parse(p->cell(1), FLAG_ITEM, mode);
1632                                 //lyxerr << "read label: " << p->cell(1) << endl;
1633                         }
1634
1635                         cell->push_back(MathAtom(p.release()));
1636                         //lyxerr << "read cell: " << cell << endl;
1637                 }
1638 #endif
1639
1640                 else if (t.cs() == "lyxmathsym") {
1641                         skipSpaces();
1642                         if (getToken().cat() != catBegin) {
1643                                 error("'{' expected in \\" + t.cs());
1644                                 return success_;
1645                         }
1646                         int count = 0;
1647                         docstring cmd;
1648                         CatCode cat = nextToken().cat();
1649                         while (good() && (count || cat != catEnd)) {
1650                                 if (cat == catBegin)
1651                                         ++count;
1652                                 else if (cat == catEnd)
1653                                         --count;
1654                                 cmd += getToken().asInput();
1655                                 cat = nextToken().cat();
1656                         }
1657                         if (getToken().cat() != catEnd) {
1658                                 error("'}' expected in \\" + t.cs());
1659                                 return success_;
1660                         }
1661                         docstring rem;
1662                         do {
1663                                 cmd = Encodings::fromLaTeXCommand(cmd, rem);
1664                                 for (size_t i = 0; i < cmd.size(); ++i)
1665                                         cell->push_back(MathAtom(new InsetMathChar(cmd[i])));
1666                                 if (rem.size()) {
1667                                         char_type c = rem[0];
1668                                         cell->push_back(MathAtom(new InsetMathChar(c)));
1669                                         cmd = rem.substr(1);
1670                                         rem.clear();
1671                                 } else
1672                                         cmd.clear();
1673                         } while (cmd.size());
1674                 }
1675
1676                 else if (t.cs().size()) {
1677                         latexkeys const * l = in_word_set(t.cs());
1678                         if (l) {
1679                                 if (l->inset == "big") {
1680                                         skipSpaces();
1681                                         docstring const delim = getToken().asInput();
1682                                         if (InsetMathBig::isBigInsetDelim(delim))
1683                                                 cell->push_back(MathAtom(
1684                                                         new InsetMathBig(t.cs(), delim)));
1685                                         else {
1686                                                 cell->push_back(createInsetMath(t.cs()));
1687                                                 putback();
1688                                         }
1689                                 }
1690
1691                                 else if (l->inset == "font") {
1692                                         cell->push_back(createInsetMath(t.cs()));
1693                                         parse(cell->back().nucleus()->cell(0),
1694                                                 FLAG_ITEM, asMode(mode, l->extra));
1695                                 }
1696
1697                                 else if (l->inset == "oldfont") {
1698                                         cell->push_back(createInsetMath(t.cs()));
1699                                         parse(cell->back().nucleus()->cell(0),
1700                                                 flags | FLAG_ALIGN, asMode(mode, l->extra));
1701                                         if (prevToken().cat() != catAlign &&
1702                                             prevToken().cs() != "\\")
1703                                                 return success_;
1704                                         putback();
1705                                 }
1706
1707                                 else if (l->inset == "style") {
1708                                         cell->push_back(createInsetMath(t.cs()));
1709                                         parse(cell->back().nucleus()->cell(0),
1710                                                 flags | FLAG_ALIGN, mode);
1711                                         if (prevToken().cat() != catAlign &&
1712                                             prevToken().cs() != "\\")
1713                                                 return success_;
1714                                         putback();
1715                                 }
1716
1717                                 else {
1718                                         MathAtom at = createInsetMath(t.cs());
1719                                         for (InsetMath::idx_type i = 0; i < at->nargs(); ++i)
1720                                                 parse(at.nucleus()->cell(i),
1721                                                         FLAG_ITEM, asMode(mode, l->extra));
1722                                         cell->push_back(at);
1723                                 }
1724                         }
1725
1726                         else {
1727                                 bool is_unicode_symbol = false;
1728                                 if (mode == InsetMath::TEXT_MODE) {
1729                                         int num_tokens = 0;
1730                                         docstring cmd = prevToken().asInput();
1731                                         skipSpaces();
1732                                         CatCode cat = nextToken().cat();
1733                                         if (cat == catBegin) {
1734                                                 int count = 0;
1735                                                 while (good() && (count || cat != catEnd)) {
1736                                                         cat = nextToken().cat();
1737                                                         cmd += getToken().asInput();
1738                                                         ++num_tokens;
1739                                                         if (cat == catBegin)
1740                                                                 ++count;
1741                                                         else if (cat == catEnd)
1742                                                                 --count;
1743                                                 }
1744                                         }
1745                                         bool is_combining;
1746                                         char_type c =
1747                                                 Encodings::fromLaTeXCommand(cmd, is_combining);
1748                                         if (is_combining) {
1749                                                 if (cat == catLetter)
1750                                                         cmd += '{';
1751                                                 cmd += getToken().asInput();
1752                                                 ++num_tokens;
1753                                                 if (cat == catLetter)
1754                                                         cmd += '}';
1755                                                 c = Encodings::fromLaTeXCommand(cmd, is_combining);
1756                                         }
1757                                         if (c) {
1758                                                 is_unicode_symbol = true;
1759                                                 cell->push_back(MathAtom(new InsetMathChar(c)));
1760                                         } else {
1761                                                 while (num_tokens--)
1762                                                         putback();
1763                                         }
1764                                 }
1765                                 if (!is_unicode_symbol) {
1766                                         MathAtom at = createInsetMath(t.cs());
1767                                         InsetMath::mode_type m = mode;
1768                                         //if (m == InsetMath::UNDECIDED_MODE)
1769                                         //lyxerr << "default creation: m1: " << m << endl;
1770                                         if (at->currentMode() != InsetMath::UNDECIDED_MODE)
1771                                                 m = at->currentMode();
1772                                         //lyxerr << "default creation: m2: " << m << endl;
1773                                         InsetMath::idx_type start = 0;
1774                                         // this fails on \bigg[...\bigg]
1775                                         //MathData opt;
1776                                         //parse(opt, FLAG_OPTION, InsetMath::VERBATIM_MODE);
1777                                         //if (opt.size()) {
1778                                         //      start = 1;
1779                                         //      at.nucleus()->cell(0) = opt;
1780                                         //}
1781                                         for (InsetMath::idx_type i = start; i < at->nargs(); ++i) {
1782                                                 parse(at.nucleus()->cell(i), FLAG_ITEM, m);
1783                                                 skipSpaces();
1784                                         }
1785                                         cell->push_back(at);
1786                                 }
1787                         }
1788                 }
1789
1790
1791                 if (flags & FLAG_LEAVE) {
1792                         flags &= ~FLAG_LEAVE;
1793                         break;
1794                 }
1795         }
1796         return success_;
1797 }
1798
1799
1800
1801 } // anonymous namespace
1802
1803
1804 bool mathed_parse_cell(MathData & ar, docstring const & str, Parse::flags f)
1805 {
1806         return Parser(str, f).parse(ar, 0, f & Parse::TEXTMODE ?
1807                                 InsetMath::TEXT_MODE : InsetMath::MATH_MODE);
1808 }
1809
1810
1811 bool mathed_parse_cell(MathData & ar, istream & is, Parse::flags f)
1812 {
1813         return Parser(is, f).parse(ar, 0, f & Parse::TEXTMODE ?
1814                                 InsetMath::TEXT_MODE : InsetMath::MATH_MODE);
1815 }
1816
1817
1818 bool mathed_parse_normal(MathAtom & t, docstring const & str, Parse::flags f)
1819 {
1820         return Parser(str, f).parse(t);
1821 }
1822
1823
1824 bool mathed_parse_normal(MathAtom & t, Lexer & lex, Parse::flags f)
1825 {
1826         return Parser(lex, f).parse(t);
1827 }
1828
1829
1830 bool mathed_parse_normal(InsetMathGrid & grid, docstring const & str, Parse::flags f)
1831 {
1832         return Parser(str, f).parse1(grid, 0, f & Parse::TEXTMODE ?
1833                         InsetMath::TEXT_MODE : InsetMath::MATH_MODE, false);
1834 }
1835
1836
1837 void initParser()
1838 {
1839         fill(theCatcode, theCatcode + 128, catOther);
1840         fill(theCatcode + 'a', theCatcode + 'z' + 1, catLetter);
1841         fill(theCatcode + 'A', theCatcode + 'Z' + 1, catLetter);
1842
1843         theCatcode[int('\\')] = catEscape;
1844         theCatcode[int('{')]  = catBegin;
1845         theCatcode[int('}')]  = catEnd;
1846         theCatcode[int('$')]  = catMath;
1847         theCatcode[int('&')]  = catAlign;
1848         theCatcode[int('\n')] = catNewline;
1849         theCatcode[int('#')]  = catParameter;
1850         theCatcode[int('^')]  = catSuper;
1851         theCatcode[int('_')]  = catSub;
1852         theCatcode[int(0x7f)] = catIgnore;
1853         theCatcode[int(' ')]  = catSpace;
1854         theCatcode[int('\t')] = catSpace;
1855         theCatcode[int('\r')] = catNewline;
1856         theCatcode[int('~')]  = catActive;
1857         theCatcode[int('%')]  = catComment;
1858 }
1859
1860
1861 } // namespace lyx