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                         char_type c = t.character();
 912                         if (c < 0x80 || mode_ & Parse::VERBATIM
 913                             || !(mode_ & Parse::USETEXT)
 914                             || mode == InsetMath::TEXT_MODE
 915                             || Encodings::isMathAlpha(c)) {
 916                                 cell->push_back(MathAtom(new InsetMathChar(c)));
 917                         } else {
 918                                 MathAtom at = createInsetMath("text");
 919                                 at.nucleus()->cell(0).push_back(MathAtom(new InsetMathChar(c)));
 920                                 while (nextToken().cat() == catOther
 921                                        && nextToken().character() >= 0x80
 922                                        && !Encodings::isMathAlpha(nextToken().character())) {
 923                                         c = getToken().character();
 924                                         at.nucleus()->cell(0).push_back(MathAtom(new InsetMathChar(c)));
 925                                 }
 926                                 cell->push_back(at);
 927                         }
 928                 }
 929
 930                 else if (t.cat() == catComment) {
 931                         docstring s;
 932                         while (good()) {
 933                                 Token const & t = getToken();
 934                                 if (t.cat() == catNewline)
 935                                         break;
 936                                 s += t.asString();
 937                         }
 938                         cell->push_back(MathAtom(new InsetMathComment(s)));
 939                         skipSpaces();
 940                 }
 941
 942                 //
 943                 // control sequences
 944                 //
 945
 946                 else if (t.cs() == "lyxlock") {
 947                         if (cell->size())
 948                                 cell->back().nucleus()->lock(true);
 949                 }
 950
 951                 else if ((t.cs() == "global" && nextToken().cs() == "def") ||
 952                          t.cs() == "def") {
 953                         if (t.cs() == "global")
 954                                 getToken();
 955
 956                         // get name
 957                         docstring name = getToken().cs();
 958
 959                         // read parameters
 960                         int nargs = 0;
 961                         docstring pars;
 962                         while (good() && nextToken().cat() != catBegin) {
 963                                 pars += getToken().cs();
 964                                 ++nargs;
 965                         }
 966                         nargs /= 2;
 967
 968                         // read definition
 969                         MathData def;
 970                         parse(def, FLAG_ITEM, InsetMath::UNDECIDED_MODE);
 971
 972                         // is a version for display attached?
 973                         skipSpaces();
 974                         MathData display;
 975                         if (nextToken().cat() == catBegin)
 976                                 parse(display, FLAG_ITEM, InsetMath::MATH_MODE);
 977
 978                         cell->push_back(MathAtom(new MathMacroTemplate(name, nargs,
 979                                0, MacroTypeDef, vector<MathData>(), def, display)));
 980                 }
 981
 982                 else if (t.cs() == "newcommand" ||
 983                          t.cs() == "renewcommand" ||
 984                          t.cs() == "newlyxcommand") {
 985                         // get name
 986                         if (getToken().cat() != catBegin) {
 987                                 error("'{' in \\newcommand expected (1) ");
 988                                 return success_;
 989                         }
 990                         docstring name = getToken().cs();
 991                         if (getToken().cat() != catEnd) {
 992                                 error("'}' in \\newcommand expected");
 993                                 return success_;
 994                         }
 995
 996                         // get arity
 997                         docstring const arg = getArg('[', ']');
 998                         int nargs = 0;
 999                         if (!arg.empty())
1000                                 nargs = convert<int>(arg);
1001
1002                         // optional argument given?
1003                         skipSpaces();
1004                         int optionals = 0;
1005                         vector<MathData> optionalValues;
1006                         while (nextToken().character() == '[') {
1007                                 getToken();
1008                                 optionalValues.push_back(MathData());
1009                                 parse(optionalValues[optionals], FLAG_BRACK_LAST, mode);
1010                                 ++optionals;
1011                         }
1012
1013                         MathData def;
1014                         parse(def, FLAG_ITEM, InsetMath::UNDECIDED_MODE);
1015
1016                         // is a version for display attached?
1017                         skipSpaces();
1018                         MathData display;
1019                         if (nextToken().cat() == catBegin)
1020                                 parse(display, FLAG_ITEM, InsetMath::MATH_MODE);
1021
1022                         cell->push_back(MathAtom(new MathMacroTemplate(name, nargs,
1023                                 optionals, MacroTypeNewcommand, optionalValues, def, display)));
1024
1025                 }
1026
1027                 else if (t.cs() == "newcommandx" ||
1028                          t.cs() == "renewcommandx") {
1029                         // \newcommandx{\foo}[2][usedefault, addprefix=\global,1=default]{#1,#2}
1030                         // get name
1031                         docstring name;
1032                         if (nextToken().cat() == catBegin) {
1033                                 getToken();
1034                                 name = getToken().cs();
1035                                 if (getToken().cat() != catEnd) {
1036                                         error("'}' in \\newcommandx expected");
1037                                         return success_;
1038                                 }
1039                         } else
1040                                 name = getToken().cs();
1041
1042                         // get arity
1043                         docstring const arg = getArg('[', ']');
1044                         if (arg.empty()) {
1045                                 error("[num] in \\newcommandx expected");
1046                                 return success_;
1047                         }
1048                         int nargs = convert<int>(arg);
1049
1050                         // get options
1051                         int optionals = 0;
1052                         vector<MathData> optionalValues;
1053                         if (nextToken().character() == '[') {
1054                                 // skip '['
1055                                 getToken();
1056
1057                                 // handle 'opt=value' options, separated by ','.
1058                                 skipSpaces();
1059                                 while (nextToken().character() != ']' && good()) {
1060                                         if (nextToken().character() >= '1'
1061                                             && nextToken().character() <= '9') {
1062                                                 // optional value -> get parameter number
1063                                                 int n = getChar() - '0';
1064                                                 if (n > nargs) {
1065                                                         error("Arity of \\newcommandx too low "
1066                                                               "for given optional parameter.");
1067                                                         return success_;
1068                                                 }
1069
1070                                                 // skip '='
1071                                                 if (getToken().character() != '=') {
1072                                                         error("'=' and optional parameter value "
1073                                                               "expected for \\newcommandx");
1074                                                         return success_;
1075                                                 }
1076
1077                                                 // get value
1078                                                 int optNum = max(size_t(n), optionalValues.size());
1079                                                 optionalValues.resize(optNum);
1080                                                 optionalValues[n - 1].clear();
1081                                                 while (nextToken().character() != ']'
1082                                                        && nextToken().character() != ',') {
1083                                                         MathData data;
1084                                                         parse(data, FLAG_ITEM, InsetMath::UNDECIDED_MODE);
1085                                                         optionalValues[n - 1].append(data);
1086                                                 }
1087                                                 optionals = max(n, optionals);
1088                                         } else if (nextToken().cat() == catLetter) {
1089                                                 // we in fact ignore every non-optional
1090                                                 // parameter
1091
1092                                                 // get option name
1093                                                 docstring opt;
1094                                                 while (nextToken().cat() == catLetter)
1095                                                         opt += getChar();
1096
1097                                                 // value?
1098                                                 skipSpaces();
1099                                                 MathData value;
1100                                                 if (nextToken().character() == '=') {
1101                                                         getToken();
1102                                                         while (nextToken().character() != ']'
1103                                                                 && nextToken().character() != ',')
1104                                                                 parse(value, FLAG_ITEM,
1105                                                                       InsetMath::UNDECIDED_MODE);
1106                                                 }
1107                                         } else {
1108                                                 error("option for \\newcommandx expected");
1109                                                 return success_;
1110                                         }
1111
1112                                         // skip komma
1113                                         skipSpaces();
1114                                         if (nextToken().character() == ',') {
1115                                                 getChar();
1116                                                 skipSpaces();
1117                                         } else if (nextToken().character() != ']') {
1118                                                 error("Expecting ',' or ']' in options "
1119                                                       "of \\newcommandx");
1120                                                 return success_;
1121                                         }
1122                                 }
1123
1124                                 // skip ']'
1125                                 if (!good())
1126                                         return success_;
1127                                 getToken();
1128                         }
1129
1130                         // get definition
1131                         MathData def;
1132                         parse(def, FLAG_ITEM, InsetMath::UNDECIDED_MODE);
1133
1134                         // is a version for display attached?
1135                         skipSpaces();
1136                         MathData display;
1137                         if (nextToken().cat() == catBegin)
1138                                 parse(display, FLAG_ITEM, InsetMath::MATH_MODE);
1139
1140                         cell->push_back(MathAtom(new MathMacroTemplate(name, nargs,
1141                                 optionals, MacroTypeNewcommandx, optionalValues, def,
1142                                 display)));
1143                 }
1144
1145                 else if (t.cs() == "(") {
1146                         if (mode == InsetMath::MATH_MODE) {
1147                                 error("bad math environment");
1148                                 break;
1149                         }
1150                         cell->push_back(MathAtom(new InsetMathHull(hullSimple)));
1151                         parse2(cell->back(), FLAG_SIMPLE2, InsetMath::MATH_MODE, false);
1152                 }
1153
1154                 else if (t.cs() == "[") {
1155                         if (mode != InsetMath::UNDECIDED_MODE) {
1156                                 error("bad math environment");
1157                                 break;
1158                         }
1159                         cell->push_back(MathAtom(new InsetMathHull(hullEquation)));
1160                         parse2(cell->back(), FLAG_EQUATION, InsetMath::MATH_MODE, false);
1161                 }
1162
1163                 else if (t.cs() == "protect")
1164                         // ignore \\protect, will hopefully be re-added during output
1165                         ;
1166
1167                 else if (t.cs() == "end") {
1168                         if (flags & FLAG_END) {
1169                                 // eat environment name
1170                                 docstring const name = getArg('{', '}');
1171                                 if (environments_.empty())
1172                                         error("'found \\end{" + name +
1173                                               "}' without matching '\\begin{" +
1174                                               name + "}'");
1175                                 else if (name != environments_.back())
1176                                         error("'\\end{" + name +
1177                                               "}' does not match '\\begin{" +
1178                                               environments_.back() + "}'");
1179                                 else {
1180                                         environments_.pop_back();
1181                                         // Delete empty last row in matrix
1182                                         // like insets.
1183                                         // If you abuse InsetMathGrid for
1184                                         // non-matrix like structures you
1185                                         // probably need to refine this test.
1186                                         // Right now we only have to test for
1187                                         // single line hull insets.
1188                                         if (grid.nrows() > 1)
1189                                                 delEmptyLastRow(grid);
1190                                         return success_;
1191                                 }
1192                         } else
1193                                 error("found 'end' unexpectedly");
1194                 }
1195
1196                 else if (t.cs() == ")") {
1197                         if (flags & FLAG_SIMPLE2)
1198                                 return success_;
1199                         error("found '\\)' unexpectedly");
1200                 }
1201
1202                 else if (t.cs() == "]") {
1203                         if (flags & FLAG_EQUATION)
1204                                 return success_;
1205                         error("found '\\]' unexpectedly");
1206                 }
1207
1208                 else if (t.cs() == "\\") {
1209                         if (flags & FLAG_ALIGN)
1210                                 return success_;
1211                         bool added = false;
1212                         if (nextToken().asInput() == "*") {
1213                                 getToken();
1214                                 added = addRow(grid, cellrow, docstring(), false);
1215                         } else if (good())
1216                                 added = addRow(grid, cellrow, getArg('[', ']'));
1217                         else
1218                                 error("missing token after \\\\");
1219                         if (added) {
1220                                 cellcol = 0;
1221                                 if (grid.asHullInset())
1222                                         grid.asHullInset()->numbered(
1223                                                         cellrow, numbered);
1224                                 cell = &grid.cell(grid.index(cellrow,
1225                                                              cellcol));
1226                         }
1227                 }
1228
1229 #if 0
1230                 else if (t.cs() == "multicolumn") {
1231                         // extract column count and insert dummy cells
1232                         MathData count;
1233                         parse(count, FLAG_ITEM, mode);
1234                         int cols = 1;
1235                         if (!extractNumber(count, cols)) {
1236                                 success_ = false;
1237                                 lyxerr << " can't extract number of cells from " << count << endl;
1238                         }
1239                         // resize the table if necessary
1240                         for (int i = 0; i < cols; ++i) {
1241                                 if (addCol(grid, cellcol)) {
1242                                         cell = &grid.cell(grid.index(
1243                                                         cellrow, cellcol));
1244                                         // mark this as dummy
1245                                         grid.cellinfo(grid.index(
1246                                                 cellrow, cellcol)).dummy_ = true;
1247                                 }
1248                         }
1249                         // the last cell is the real thing, not a dummy
1250                         grid.cellinfo(grid.index(cellrow, cellcol)).dummy_ = false;
1251
1252                         // read special alignment
1253                         MathData align;
1254                         parse(align, FLAG_ITEM, mode);
1255                         //grid.cellinfo(grid.index(cellrow, cellcol)).align_ = extractString(align);
1256
1257                         // parse the remaining contents into the "real" cell
1258                         parse(*cell, FLAG_ITEM, mode);
1259                 }
1260 #endif
1261
1262                 else if (t.cs() == "limits")
1263                         limits = 1;
1264
1265                 else if (t.cs() == "nolimits")
1266                         limits = -1;
1267
1268                 else if (t.cs() == "nonumber") {
1269                         if (grid.asHullInset())
1270                                 grid.asHullInset()->numbered(cellrow, false);
1271                 }
1272
1273                 else if (t.cs() == "number") {
1274                         if (grid.asHullInset())
1275                                 grid.asHullInset()->numbered(cellrow, true);
1276                 }
1277
1278                 else if (t.cs() == "hline") {
1279                         grid.rowinfo(cellrow).lines_ ++;
1280                 }
1281
1282                 else if (t.cs() == "sqrt") {
1283                         MathData ar;
1284                         parse(ar, FLAG_OPTION, mode);
1285                         if (ar.size()) {
1286                                 cell->push_back(MathAtom(new InsetMathRoot));
1287                                 cell->back().nucleus()->cell(0) = ar;
1288                                 parse(cell->back().nucleus()->cell(1), FLAG_ITEM, mode);
1289                         } else {
1290                                 cell->push_back(MathAtom(new InsetMathSqrt));
1291                                 parse(cell->back().nucleus()->cell(0), FLAG_ITEM, mode);
1292                         }
1293                 }
1294
1295                 else if (t.cs() == "unit") {
1296                         // Allowed formats \unit[val]{unit}
1297                         MathData ar;
1298                         parse(ar, FLAG_OPTION, mode);
1299                         if (ar.size()) {
1300                                 cell->push_back(MathAtom(new InsetMathFrac(InsetMathFrac::UNIT)));
1301                                 cell->back().nucleus()->cell(0) = ar;
1302                                 parse(cell->back().nucleus()->cell(1), FLAG_ITEM, mode);
1303                         } else {
1304                                 cell->push_back(MathAtom(new InsetMathFrac(InsetMathFrac::UNIT, 1)));
1305                                 parse(cell->back().nucleus()->cell(0), FLAG_ITEM, mode);
1306                         }
1307                 }
1308
1309                 else if (t.cs() == "unitfrac") {
1310                         // Here allowed formats are \unitfrac[val]{num}{denom}
1311                         MathData ar;
1312                         parse(ar, FLAG_OPTION, mode);
1313                         if (ar.size()) {
1314                                 cell->push_back(MathAtom(new InsetMathFrac(InsetMathFrac::UNITFRAC, 3)));
1315                                 cell->back().nucleus()->cell(2) = ar;
1316                         } else {
1317                                 cell->push_back(MathAtom(new InsetMathFrac(InsetMathFrac::UNITFRAC)));
1318                         }
1319                         parse(cell->back().nucleus()->cell(0), FLAG_ITEM, mode);
1320                         parse(cell->back().nucleus()->cell(1), FLAG_ITEM, mode);
1321                 }
1322
1323                 else if (t.cs() == "cfrac") {
1324                         // Here allowed formats are \cfrac[pos]{num}{denom}
1325                         MathData ar;
1326                         parse(ar, FLAG_OPTION, mode);
1327                         if (ar.size()) {
1328                                 cell->push_back(MathAtom(new InsetMathFrac(InsetMathFrac::CFRAC, 3)));
1329                                 cell->back().nucleus()->cell(2) = ar;
1330                         } else {
1331                                 cell->push_back(MathAtom(new InsetMathFrac(InsetMathFrac::CFRAC)));
1332                         }
1333                         parse(cell->back().nucleus()->cell(0), FLAG_ITEM, mode);
1334                         parse(cell->back().nucleus()->cell(1), FLAG_ITEM, mode);
1335                 }
1336
1337                 else if (t.cs() == "xrightarrow" || t.cs() == "xleftarrow") {
1338                         cell->push_back(createInsetMath(t.cs()));
1339                         parse(cell->back().nucleus()->cell(1), FLAG_OPTION, mode);
1340                         parse(cell->back().nucleus()->cell(0), FLAG_ITEM, mode);
1341                 }
1342
1343                 else if (t.cs() == "ref" || t.cs() == "eqref" || t.cs() == "prettyref"
1344                           || t.cs() == "pageref" || t.cs() == "vpageref" || t.cs() == "vref") {
1345                         cell->push_back(MathAtom(new InsetMathRef(t.cs())));
1346                         parse(cell->back().nucleus()->cell(1), FLAG_OPTION, mode);
1347                         parse(cell->back().nucleus()->cell(0), FLAG_ITEM, mode);
1348                 }
1349
1350                 else if (t.cs() == "left") {
1351                         skipSpaces();
1352                         Token const & tl = getToken();
1353                         // \| and \Vert are equivalent, and InsetMathDelim
1354                         // can't handle \|
1355                         // FIXME: fix this in InsetMathDelim itself!
1356                         docstring const l = tl.cs() == "|" ? from_ascii("Vert") : tl.asString();
1357                         MathData ar;
1358                         parse(ar, FLAG_RIGHT, mode);
1359                         if (!good())
1360                                 break;
1361                         skipSpaces();
1362                         Token const & tr = getToken();
1363                         docstring const r = tr.cs() == "|" ? from_ascii("Vert") : tr.asString();
1364                         cell->push_back(MathAtom(new InsetMathDelim(l, r, ar)));
1365                 }
1366
1367                 else if (t.cs() == "right") {
1368                         if (flags & FLAG_RIGHT)
1369                                 return success_;
1370                         //lyxerr << "got so far: '" << cell << "'" << endl;
1371                         error("Unmatched right delimiter");
1372                         return success_;
1373                 }
1374
1375                 else if (t.cs() == "begin") {
1376                         docstring const name = getArg('{', '}');
1377                         environments_.push_back(name);
1378
1379                         if (name == "array" || name == "subarray") {
1380                                 docstring const valign = parse_verbatim_option() + 'c';
1381                                 docstring const halign = parse_verbatim_item();
1382                                 cell->push_back(MathAtom(new InsetMathArray(name,
1383                                         InsetMathGrid::guessColumns(halign), 1, (char)valign[0], halign)));
1384                                 parse2(cell->back(), FLAG_END, mode, false);
1385                         }
1386
1387                         else if (name == "tabular") {
1388                                 docstring const valign = parse_verbatim_option() + 'c';
1389                                 docstring const halign = parse_verbatim_item();
1390                                 cell->push_back(MathAtom(new InsetMathTabular(name,
1391                                         InsetMathGrid::guessColumns(halign), 1, (char)valign[0], halign)));
1392                                 parse2(cell->back(), FLAG_END, InsetMath::TEXT_MODE, false);
1393                         }
1394
1395                         else if (name == "split" || name == "cases") {
1396                                 cell->push_back(createInsetMath(name));
1397                                 parse2(cell->back(), FLAG_END, mode, false);
1398                         }
1399
1400                         else if (name == "alignedat") {
1401                                 docstring const valign = parse_verbatim_option() + 'c';
1402                                 // ignore this for a while
1403                                 getArg('{', '}');
1404                                 cell->push_back(MathAtom(new InsetMathSplit(name, (char)valign[0])));
1405                                 parse2(cell->back(), FLAG_END, mode, false);
1406                         }
1407
1408                         else if (name == "math") {
1409                                 if (mode == InsetMath::MATH_MODE) {
1410                                         error("bad math environment");
1411                                         break;
1412                                 }
1413                                 cell->push_back(MathAtom(new InsetMathHull(hullSimple)));
1414                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, true);
1415                         }
1416
1417                         else if (name == "equation" || name == "equation*"
1418                                         || name == "displaymath") {
1419                                 if (mode != InsetMath::UNDECIDED_MODE) {
1420                                         error("bad math environment");
1421                                         break;
1422                                 }
1423                                 cell->push_back(MathAtom(new InsetMathHull(hullEquation)));
1424                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, (name == "equation"));
1425                         }
1426
1427                         else if (name == "eqnarray" || name == "eqnarray*") {
1428                                 if (mode != InsetMath::UNDECIDED_MODE) {
1429                                         error("bad math environment");
1430                                         break;
1431                                 }
1432                                 cell->push_back(MathAtom(new InsetMathHull(hullEqnArray)));
1433                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1434                         }
1435
1436                         else if (name == "align" || name == "align*") {
1437                                 if (mode != InsetMath::UNDECIDED_MODE) {
1438                                         error("bad math environment");
1439                                         break;
1440                                 }
1441                                 cell->push_back(MathAtom(new InsetMathHull(hullAlign)));
1442                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1443                         }
1444
1445                         else if (name == "flalign" || name == "flalign*") {
1446                                 if (mode != InsetMath::UNDECIDED_MODE) {
1447                                         error("bad math environment");
1448                                         break;
1449                                 }
1450                                 cell->push_back(MathAtom(new InsetMathHull(hullFlAlign)));
1451                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1452                         }
1453
1454                         else if (name == "alignat" || name == "alignat*") {
1455                                 if (mode != InsetMath::UNDECIDED_MODE) {
1456                                         error("bad math environment");
1457                                         break;
1458                                 }
1459                                 // ignore this for a while
1460                                 getArg('{', '}');
1461                                 cell->push_back(MathAtom(new InsetMathHull(hullAlignAt)));
1462                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1463                         }
1464
1465                         else if (name == "xalignat" || name == "xalignat*") {
1466                                 if (mode != InsetMath::UNDECIDED_MODE) {
1467                                         error("bad math environment");
1468                                         break;
1469                                 }
1470                                 // ignore this for a while
1471                                 getArg('{', '}');
1472                                 cell->push_back(MathAtom(new InsetMathHull(hullXAlignAt)));
1473                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1474                         }
1475
1476                         else if (name == "xxalignat") {
1477                                 if (mode != InsetMath::UNDECIDED_MODE) {
1478                                         error("bad math environment");
1479                                         break;
1480                                 }
1481                                 // ignore this for a while
1482                                 getArg('{', '}');
1483                                 cell->push_back(MathAtom(new InsetMathHull(hullXXAlignAt)));
1484                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1485                         }
1486
1487                         else if (name == "multline" || name == "multline*") {
1488                                 if (mode != InsetMath::UNDECIDED_MODE) {
1489                                         error("bad math environment");
1490                                         break;
1491                                 }
1492                                 cell->push_back(MathAtom(new InsetMathHull(hullMultline)));
1493                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1494                         }
1495
1496                         else if (name == "gather" || name == "gather*") {
1497                                 if (mode != InsetMath::UNDECIDED_MODE) {
1498                                         error("bad math environment");
1499                                         break;
1500                                 }
1501                                 cell->push_back(MathAtom(new InsetMathHull(hullGather)));
1502                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1503                         }
1504
1505                         else if (latexkeys const * l = in_word_set(name)) {
1506                                 if (l->inset == "matrix") {
1507                                         cell->push_back(createInsetMath(name));
1508                                         parse2(cell->back(), FLAG_END, mode, false);
1509                                 } else if (l->inset == "split") {
1510                                         docstring const valign = parse_verbatim_option() + 'c';
1511                                         cell->push_back(MathAtom(new InsetMathSplit(name, (char)valign[0])));
1512                                         parse2(cell->back(), FLAG_END, mode, false);
1513                                 } else {
1514                                         success_ = false;
1515                                         if (!(mode_ & Parse::QUIET)) {
1516                                                 dump();
1517                                                 lyxerr << "found math environment `"
1518                                                        << to_utf8(name)
1519                                                        << "' in symbols file with unsupported inset `"
1520                                                        << to_utf8(l->inset)
1521                                                        << "'." << endl;
1522                                         }
1523                                         // create generic environment inset
1524                                         cell->push_back(MathAtom(new InsetMathEnv(name)));
1525                                         parse(cell->back().nucleus()->cell(0), FLAG_END, mode);
1526                                 }
1527                         }
1528
1529                         else {
1530                                 success_ = false;
1531                                 if (!(mode_ & Parse::QUIET)) {
1532                                         dump();
1533                                         lyxerr << "found unknown math environment '"
1534                                                << to_utf8(name) << "'" << endl;
1535                                 }
1536                                 // create generic environment inset
1537                                 cell->push_back(MathAtom(new InsetMathEnv(name)));
1538                                 parse(cell->back().nucleus()->cell(0), FLAG_END, mode);
1539                         }
1540                 }
1541
1542                 else if (t.cs() == "kern") {
1543                         // FIXME: A hack...
1544                         docstring s;
1545                         while (true) {
1546                                 Token const & t = getToken();
1547                                 if (!good()) {
1548                                         putback();
1549                                         break;
1550                                 }
1551                                 s += t.character();
1552                                 if (isValidLength(to_utf8(s)))
1553                                         break;
1554                         }
1555                         cell->push_back(MathAtom(new InsetMathKern(s)));
1556                 }
1557
1558                 else if (t.cs() == "label") {
1559                         // FIXME: This is swallowed in inline formulas
1560                         docstring label = parse_verbatim_item();
1561                         MathData ar;
1562                         asArray(label, ar);
1563                         if (grid.asHullInset()) {
1564                                 grid.asHullInset()->label(cellrow, label);
1565                         } else {
1566                                 cell->push_back(createInsetMath(t.cs()));
1567                                 cell->push_back(MathAtom(new InsetMathBrace(ar)));
1568                         }
1569                 }
1570
1571                 else if (t.cs() == "choose" || t.cs() == "over"
1572                                 || t.cs() == "atop" || t.cs() == "brace"
1573                                 || t.cs() == "brack") {
1574                         MathAtom at = createInsetMath(t.cs());
1575                         at.nucleus()->cell(0) = *cell;
1576                         cell->clear();
1577                         parse(at.nucleus()->cell(1), flags, mode);
1578                         cell->push_back(at);
1579                         return success_;
1580                 }
1581
1582                 else if (t.cs() == "color") {
1583                         docstring const color = parse_verbatim_item();
1584                         cell->push_back(MathAtom(new InsetMathColor(true, color)));
1585                         parse(cell->back().nucleus()->cell(0), flags, mode);
1586                         return success_;
1587                 }
1588
1589                 else if (t.cs() == "textcolor") {
1590                         docstring const color = parse_verbatim_item();
1591                         cell->push_back(MathAtom(new InsetMathColor(false, color)));
1592                         parse(cell->back().nucleus()->cell(0), FLAG_ITEM, InsetMath::TEXT_MODE);
1593                 }
1594
1595                 else if (t.cs() == "normalcolor") {
1596                         cell->push_back(createInsetMath(t.cs()));
1597                         parse(cell->back().nucleus()->cell(0), flags, mode);
1598                         return success_;
1599                 }
1600
1601                 else if (t.cs() == "substack") {
1602                         cell->push_back(createInsetMath(t.cs()));
1603                         parse2(cell->back(), FLAG_ITEM, mode, false);
1604                 }
1605
1606                 else if (t.cs() == "xymatrix") {
1607                         odocstringstream os;
1608                         while (good() && nextToken().cat() != catBegin)
1609                                 os << getToken().asInput();
1610                         cell->push_back(createInsetMath(t.cs() + os.str()));
1611                         parse2(cell->back(), FLAG_ITEM, mode, false);
1612                 }
1613
1614                 else if (t.cs() == "framebox" || t.cs() == "makebox") {
1615                         cell->push_back(createInsetMath(t.cs()));
1616                         parse(cell->back().nucleus()->cell(0), FLAG_OPTION, InsetMath::TEXT_MODE);
1617                         parse(cell->back().nucleus()->cell(1), FLAG_OPTION, InsetMath::TEXT_MODE);
1618                         parse(cell->back().nucleus()->cell(2), FLAG_ITEM, InsetMath::TEXT_MODE);
1619                 }
1620
1621                 else if (t.cs() == "tag") {
1622                         if (nextToken().character() == '*') {
1623                                 getToken();
1624                                 cell->push_back(createInsetMath(t.cs() + '*'));
1625                         } else
1626                                 cell->push_back(createInsetMath(t.cs()));
1627                         parse(cell->back().nucleus()->cell(0), FLAG_ITEM, InsetMath::TEXT_MODE);
1628                 }
1629
1630                 else if (t.cs() == "hspace" && nextToken().character() != '*') {
1631                         docstring const name = t.cs();
1632                         docstring const arg = parse_verbatim_item();
1633                         Length length;
1634                         if (isValidLength(to_utf8(arg), &length))
1635                                 cell->push_back(MathAtom(new InsetMathSpace(length)));
1636                         else {
1637                                 // Since the Length class cannot use length variables
1638                                 // we must not create an InsetMathSpace.
1639                                 cell->push_back(MathAtom(new MathMacro(name)));
1640                                 MathData ar;
1641                                 mathed_parse_cell(ar, '{' + arg + '}');
1642                                 cell->append(ar);
1643                         }
1644                 }
1645
1646 #if 0
1647                 else if (t.cs() == "infer") {
1648                         MathData ar;
1649                         parse(ar, FLAG_OPTION, mode);
1650                         cell->push_back(createInsetMath(t.cs()));
1651                         parse2(cell->back(), FLAG_ITEM, mode, false);
1652                 }
1653
1654                 // Disabled
1655                 else if (1 && t.cs() == "ar") {
1656                         auto_ptr<InsetMathXYArrow> p(new InsetMathXYArrow);
1657                         // try to read target
1658                         parse(p->cell(0), FLAG_OTPTION, mode);
1659                         // try to read label
1660                         if (nextToken().cat() == catSuper || nextToken().cat() == catSub) {
1661                                 p->up_ = nextToken().cat() == catSuper;
1662                                 getToken();
1663                                 parse(p->cell(1), FLAG_ITEM, mode);
1664                                 //lyxerr << "read label: " << p->cell(1) << endl;
1665                         }
1666
1667                         cell->push_back(MathAtom(p.release()));
1668                         //lyxerr << "read cell: " << cell << endl;
1669                 }
1670 #endif
1671
1672                 else if (t.cs() == "lyxmathsym") {
1673                         skipSpaces();
1674                         if (getToken().cat() != catBegin) {
1675                                 error("'{' expected in \\" + t.cs());
1676                                 return success_;
1677                         }
1678                         int count = 0;
1679                         docstring cmd;
1680                         CatCode cat = nextToken().cat();
1681                         while (good() && (count || cat != catEnd)) {
1682                                 if (cat == catBegin)
1683                                         ++count;
1684                                 else if (cat == catEnd)
1685                                         --count;
1686                                 cmd += getToken().asInput();
1687                                 cat = nextToken().cat();
1688                         }
1689                         if (getToken().cat() != catEnd) {
1690                                 error("'}' expected in \\" + t.cs());
1691                                 return success_;
1692                         }
1693                         docstring rem;
1694                         do {
1695                                 cmd = Encodings::fromLaTeXCommand(cmd, rem);
1696                                 for (size_t i = 0; i < cmd.size(); ++i)
1697                                         cell->push_back(MathAtom(new InsetMathChar(cmd[i])));
1698                                 if (rem.size()) {
1699                                         char_type c = rem[0];
1700                                         cell->push_back(MathAtom(new InsetMathChar(c)));
1701                                         cmd = rem.substr(1);
1702                                         rem.clear();
1703                                 } else
1704                                         cmd.clear();
1705                         } while (cmd.size());
1706                 }
1707
1708                 else if (t.cs().size()) {
1709                         latexkeys const * l = in_word_set(t.cs());
1710                         if (l) {
1711                                 if (l->inset == "big") {
1712                                         skipSpaces();
1713                                         docstring const delim = getToken().asInput();
1714                                         if (InsetMathBig::isBigInsetDelim(delim))
1715                                                 cell->push_back(MathAtom(
1716                                                         new InsetMathBig(t.cs(), delim)));
1717                                         else {
1718                                                 cell->push_back(createInsetMath(t.cs()));
1719                                                 putback();
1720                                         }
1721                                 }
1722
1723                                 else if (l->inset == "font") {
1724                                         cell->push_back(createInsetMath(t.cs()));
1725                                         parse(cell->back().nucleus()->cell(0),
1726                                                 FLAG_ITEM, asMode(mode, l->extra));
1727                                 }
1728
1729                                 else if (l->inset == "oldfont") {
1730                                         cell->push_back(createInsetMath(t.cs()));
1731                                         parse(cell->back().nucleus()->cell(0),
1732                                                 flags | FLAG_ALIGN, asMode(mode, l->extra));
1733                                         if (prevToken().cat() != catAlign &&
1734                                             prevToken().cs() != "\\")
1735                                                 return success_;
1736                                         putback();
1737                                 }
1738
1739                                 else if (l->inset == "style") {
1740                                         cell->push_back(createInsetMath(t.cs()));
1741                                         parse(cell->back().nucleus()->cell(0),
1742                                                 flags | FLAG_ALIGN, mode);
1743                                         if (prevToken().cat() != catAlign &&
1744                                             prevToken().cs() != "\\")
1745                                                 return success_;
1746                                         putback();
1747                                 }
1748
1749                                 else {
1750                                         MathAtom at = createInsetMath(t.cs());
1751                                         for (InsetMath::idx_type i = 0; i < at->nargs(); ++i)
1752                                                 parse(at.nucleus()->cell(i),
1753                                                         FLAG_ITEM, asMode(mode, l->extra));
1754                                         cell->push_back(at);
1755                                 }
1756                         }
1757
1758                         else {
1759                                 bool is_unicode_symbol = false;
1760                                 if (mode == InsetMath::TEXT_MODE) {
1761                                         int num_tokens = 0;
1762                                         docstring cmd = prevToken().asInput();
1763                                         skipSpaces();
1764                                         CatCode cat = nextToken().cat();
1765                                         if (cat == catBegin) {
1766                                                 int count = 0;
1767                                                 while (good() && (count || cat != catEnd)) {
1768                                                         cat = nextToken().cat();
1769                                                         cmd += getToken().asInput();
1770                                                         ++num_tokens;
1771                                                         if (cat == catBegin)
1772                                                                 ++count;
1773                                                         else if (cat == catEnd)
1774                                                                 --count;
1775                                                 }
1776                                         }
1777                                         bool is_combining;
1778                                         char_type c =
1779                                                 Encodings::fromLaTeXCommand(cmd, is_combining);
1780                                         if (is_combining) {
1781                                                 if (cat == catLetter)
1782                                                         cmd += '{';
1783                                                 cmd += getToken().asInput();
1784                                                 ++num_tokens;
1785                                                 if (cat == catLetter)
1786                                                         cmd += '}';
1787                                                 c = Encodings::fromLaTeXCommand(cmd, is_combining);
1788                                         }
1789                                         if (c) {
1790                                                 is_unicode_symbol = true;
1791                                                 cell->push_back(MathAtom(new InsetMathChar(c)));
1792                                         } else {
1793                                                 while (num_tokens--)
1794                                                         putback();
1795                                         }
1796                                 }
1797                                 if (!is_unicode_symbol) {
1798                                         MathAtom at = createInsetMath(t.cs());
1799                                         InsetMath::mode_type m = mode;
1800                                         //if (m == InsetMath::UNDECIDED_MODE)
1801                                         //lyxerr << "default creation: m1: " << m << endl;
1802                                         if (at->currentMode() != InsetMath::UNDECIDED_MODE)
1803                                                 m = at->currentMode();
1804                                         //lyxerr << "default creation: m2: " << m << endl;
1805                                         InsetMath::idx_type start = 0;
1806                                         // this fails on \bigg[...\bigg]
1807                                         //MathData opt;
1808                                         //parse(opt, FLAG_OPTION, InsetMath::VERBATIM_MODE);
1809                                         //if (opt.size()) {
1810                                         //      start = 1;
1811                                         //      at.nucleus()->cell(0) = opt;
1812                                         //}
1813                                         for (InsetMath::idx_type i = start; i < at->nargs(); ++i) {
1814                                                 parse(at.nucleus()->cell(i), FLAG_ITEM, m);
1815                                                 skipSpaces();
1816                                         }
1817                                         cell->push_back(at);
1818                                 }
1819                         }
1820                 }
1821
1822
1823                 if (flags & FLAG_LEAVE) {
1824                         flags &= ~FLAG_LEAVE;
1825                         break;
1826                 }
1827         }
1828         return success_;
1829 }
1830
1831
1832
1833 } // anonymous namespace
1834
1835
1836 bool mathed_parse_cell(MathData & ar, docstring const & str, Parse::flags f)
1837 {
1838         return Parser(str, f).parse(ar, 0, f & Parse::TEXTMODE ?
1839                                 InsetMath::TEXT_MODE : InsetMath::MATH_MODE);
1840 }
1841
1842
1843 bool mathed_parse_cell(MathData & ar, istream & is, Parse::flags f)
1844 {
1845         return Parser(is, f).parse(ar, 0, f & Parse::TEXTMODE ?
1846                                 InsetMath::TEXT_MODE : InsetMath::MATH_MODE);
1847 }
1848
1849
1850 bool mathed_parse_normal(MathAtom & t, docstring const & str, Parse::flags f)
1851 {
1852         return Parser(str, f).parse(t);
1853 }
1854
1855
1856 bool mathed_parse_normal(MathAtom & t, Lexer & lex, Parse::flags f)
1857 {
1858         return Parser(lex, f).parse(t);
1859 }
1860
1861
1862 bool mathed_parse_normal(InsetMathGrid & grid, docstring const & str, Parse::flags f)
1863 {
1864         return Parser(str, f).parse1(grid, 0, f & Parse::TEXTMODE ?
1865                         InsetMath::TEXT_MODE : InsetMath::MATH_MODE, false);
1866 }
1867
1868
1869 void initParser()
1870 {
1871         fill(theCatcode, theCatcode + 128, catOther);
1872         fill(theCatcode + 'a', theCatcode + 'z' + 1, catLetter);
1873         fill(theCatcode + 'A', theCatcode + 'Z' + 1, catLetter);
1874
1875         theCatcode[int('\\')] = catEscape;
1876         theCatcode[int('{')]  = catBegin;
1877         theCatcode[int('}')]  = catEnd;
1878         theCatcode[int('$')]  = catMath;
1879         theCatcode[int('&')]  = catAlign;
1880         theCatcode[int('\n')] = catNewline;
1881         theCatcode[int('#')]  = catParameter;
1882         theCatcode[int('^')]  = catSuper;
1883         theCatcode[int('_')]  = catSub;
1884         theCatcode[int(0x7f)] = catIgnore;
1885         theCatcode[int(' ')]  = catSpace;
1886         theCatcode[int('\t')] = catSpace;
1887         theCatcode[int('\r')] = catNewline;
1888         theCatcode[int('~')]  = catActive;
1889         theCatcode[int('%')]  = catComment;
1890 }
1891
1892
1893 } // namespace lyx