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