src/mathed/MathParser.cpp

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