src/mathed/MathParser.cpp

   1 /**
   2  * \file MathParser.cpp
   3  * This file is part of LyX, the document processor.
   4  * Licence details can be found in the file COPYING.
   5  *
   6  * \author André Pönitz
   7  *
   8  * Full author contact details are available in file CREDITS.
   9  */
  10
  11 /*
  12
  13 If someone desperately needs partial "structures" (such as a few
  14 cells of an array inset or similar) (s)he could uses the
  15 following hack as starting point to write some macros:
  16
  17   \newif\ifcomment
  18   \commentfalse
  19   \ifcomment
  20           \def\makeamptab{\catcode`\&=4\relax}
  21           \def\makeampletter{\catcode`\&=11\relax}
  22     \def\b{\makeampletter\expandafter\makeamptab\bi}
  23     \long\def\bi#1\e{}
  24   \else
  25     \def\b{}\def\e{}
  26   \fi
  27   ...
  28
  29   \[\begin{array}{ccc}
  30 1
  31 &
  32
  33   \end{array}\]
  34
  35 */
  36
  37
  38 #include <config.h>
  39
  40 #include "MathParser.h"
  41
  42 #include "InsetMathArray.h"
  43 #include "InsetMathBig.h"
  44 #include "InsetMathBrace.h"
  45 #include "InsetMathChar.h"
  46 #include "InsetMathColor.h"
  47 #include "InsetMathComment.h"
  48 #include "InsetMathDelim.h"
  49 #include "InsetMathEnv.h"
  50 #include "InsetMathFrac.h"
  51 #include "InsetMathKern.h"
  52 #include "MathMacro.h"
  53 #include "InsetMathPar.h"
  54 #include "InsetMathRef.h"
  55 #include "InsetMathRoot.h"
  56 #include "InsetMathScript.h"
  57 #include "InsetMathSplit.h"
  58 #include "InsetMathSqrt.h"
  59 #include "InsetMathTabular.h"
  60 #include "MathMacroTemplate.h"
  61 #include "MathFactory.h"
  62 #include "MathMacroArgument.h"
  63 #include "MathSupport.h"
  64
  65 #include "Lexer.h"
  66
  67 #include "support/debug.h"
  68 #include "support/convert.h"
  69 #include "support/docstream.h"
  70
  71 #include <sstream>
  72
  73 //#define FILEDEBUG
  74
  75 using namespace std;
  76
  77 namespace lyx {
  78
  79 namespace {
  80
  81 InsetMath::mode_type asMode(InsetMath::mode_type oldmode, docstring const & str)
  82 {
  83         //lyxerr << "handling mode: '" << str << "'" << endl;
  84         if (str == "mathmode")
  85                 return InsetMath::MATH_MODE;
  86         if (str == "textmode" || str == "forcetext")
  87                 return InsetMath::TEXT_MODE;
  88         return oldmode;
  89 }
  90
  91
  92 bool stared(docstring const & s)
  93 {
  94         size_t const n = s.size();
  95         return n && s[n - 1] == '*';
  96 }
  97
  98
  99 /*!
 100  * Add the row \p cellrow to \p grid.
 101  * \returns wether the row could be added. Adding a row can fail for
 102  * environments like "equation" that have a fixed number of rows.
 103  */
 104 bool addRow(InsetMathGrid & grid, InsetMathGrid::row_type & cellrow,
 105             docstring const & vskip, bool allow_newpage_ = true)
 106 {
 107         ++cellrow;
 108         if (cellrow == grid.nrows()) {
 109                 //lyxerr << "adding row " << cellrow << endl;
 110                 grid.addRow(cellrow - 1);
 111                 if (cellrow == grid.nrows()) {
 112                         // We can't add a row to this grid, so let's
 113                         // append the content of this cell to the previous
 114                         // one.
 115                         // This does not happen in well formed .lyx files,
 116                         // but LyX versions 1.3.x and older could create
 117                         // such files and tex2lyx can still do that.
 118                         --cellrow;
 119                         lyxerr << "ignoring extra row";
 120                         if (!vskip.empty())
 121                                 lyxerr << " with extra space " << to_utf8(vskip);
 122                         if (!allow_newpage_)
 123                                 lyxerr << " with no page break allowed";
 124                         lyxerr << '.' << endl;
 125                         return false;
 126                 }
 127         }
 128         grid.vcrskip(Length(to_utf8(vskip)), cellrow - 1);
 129         grid.rowinfo(cellrow - 1).allow_newpage_ = allow_newpage_;
 130         return true;
 131 }
 132
 133
 134 /*!
 135  * Add the column \p cellcol to \p grid.
 136  * \returns wether the column could be added. Adding a column can fail for
 137  * environments like "eqnarray" that have a fixed number of columns.
 138  */
 139 bool addCol(InsetMathGrid & grid, InsetMathGrid::col_type & cellcol)
 140 {
 141         ++cellcol;
 142         if (cellcol == grid.ncols()) {
 143                 //lyxerr << "adding column " << cellcol << endl;
 144                 grid.addCol(cellcol);
 145                 if (cellcol == grid.ncols()) {
 146                         // We can't add a column to this grid, so let's
 147                         // append the content of this cell to the previous
 148                         // one.
 149                         // This does not happen in well formed .lyx files,
 150                         // but LyX versions 1.3.x and older could create
 151                         // such files and tex2lyx can still do that.
 152                         --cellcol;
 153                         lyxerr << "ignoring extra column." << endl;
 154                         return false;
 155                 }
 156         }
 157         return true;
 158 }
 159
 160
 161 /*!
 162  * Check wether the last row is empty and remove it if yes.
 163  * Otherwise the following code
 164  * \verbatim
 165 \begin{array}{|c|c|}
 166 \hline
 167 1 & 2 \\ \hline
 168 3 & 4 \\ \hline
 169 \end{array}
 170  * \endverbatim
 171  * will result in a grid with 3 rows (+ the dummy row that is always present),
 172  * because the last '\\' opens a new row.
 173  */
 174 void delEmptyLastRow(InsetMathGrid & grid)
 175 {
 176         InsetMathGrid::row_type const row = grid.nrows() - 1;
 177         for (InsetMathGrid::col_type col = 0; col < grid.ncols(); ++col) {
 178                 if (!grid.cell(grid.index(row, col)).empty())
 179                         return;
 180         }
 181         // Copy the row information of the empty row (which would contain the
 182         // last hline in the example above) to the dummy row and delete the
 183         // empty row.
 184         grid.rowinfo(row + 1) = grid.rowinfo(row);
 185         grid.delRow(row);
 186 }
 187
 188
 189 // These are TeX's catcodes
 190 enum CatCode {
 191         catEscape,     // 0    backslash
 192         catBegin,      // 1    {
 193         catEnd,        // 2    }
 194         catMath,       // 3    $
 195         catAlign,      // 4    &
 196         catNewline,    // 5    ^^M
 197         catParameter,  // 6    #
 198         catSuper,      // 7    ^
 199         catSub,        // 8    _
 200         catIgnore,     // 9
 201         catSpace,      // 10   space
 202         catLetter,     // 11   a-zA-Z
 203         catOther,      // 12   none of the above
 204         catActive,     // 13   ~
 205         catComment,    // 14   %
 206         catInvalid     // 15   <delete>
 207 };
 208
 209 CatCode theCatcode[128];
 210
 211
 212 inline CatCode catcode(char_type c)
 213 {
 214         /* The only characters that are not catOther lie in the pure ASCII
 215          * range. Therefore theCatcode has only 128 entries.
 216          * TeX itself deals with 8bit characters, so if needed this table
 217          * could be enlarged to 256 entries.
 218          * Any larger value does not make sense, since the fact that we use
 219          * unicode internally does not change Knuth's TeX engine.
 220          * Apart from that a table for the full 21bit UCS4 range would waste
 221          * too much memory. */
 222         if (c >= 128)
 223                 return catOther;
 224
 225         return theCatcode[c];
 226 }
 227
 228
 229 enum {
 230         FLAG_ALIGN      = 1 << 0,  //  next & or \\ ends the parsing process
 231         FLAG_BRACE_LAST = 1 << 1,  //  next closing brace ends the parsing
 232         FLAG_RIGHT      = 1 << 2,  //  next \\right ends the parsing process
 233         FLAG_END        = 1 << 3,  //  next \\end ends the parsing process
 234         FLAG_BRACK_LAST = 1 << 4,  //  next closing bracket ends the parsing
 235         FLAG_TEXTMODE   = 1 << 5,  //  we are in a box
 236         FLAG_ITEM       = 1 << 6,  //  read a (possibly braced) token
 237         FLAG_LEAVE      = 1 << 7,  //  leave the loop at the end
 238         FLAG_SIMPLE     = 1 << 8,  //  next $ leaves the loop
 239         FLAG_EQUATION   = 1 << 9,  //  next \] leaves the loop
 240         FLAG_SIMPLE2    = 1 << 10, //  next \) leaves the loop
 241         FLAG_OPTION     = 1 << 11, //  read [...] style option
 242         FLAG_BRACED     = 1 << 12  //  read {...} style argument
 243 };
 244
 245
 246 //
 247 // Helper class for parsing
 248 //
 249
 250 class Token {
 251 public:
 252         ///
 253         Token() : cs_(), char_(0), cat_(catIgnore) {}
 254         ///
 255         Token(char_type c, CatCode cat) : cs_(), char_(c), cat_(cat) {}
 256         ///
 257         explicit Token(docstring const & cs) : cs_(cs), char_(0), cat_(catIgnore) {}
 258
 259         ///
 260         docstring const & cs() const { return cs_; }
 261         ///
 262         CatCode cat() const { return cat_; }
 263         ///
 264         char_type character() const { return char_; }
 265         ///
 266         docstring asString() const { return cs_.size() ? cs_ : docstring(1, char_); }
 267         ///
 268         docstring asInput() const { return cs_.size() ? '\\' + cs_ : docstring(1, char_); }
 269
 270 private:
 271         ///
 272         docstring cs_;
 273         ///
 274         char_type char_;
 275         ///
 276         CatCode cat_;
 277 };
 278
 279
 280 ostream & operator<<(ostream & os, Token const & t)
 281 {
 282         if (t.cs().size()) {
 283                 docstring const & cs = t.cs();
 284                 // FIXME: For some strange reason, the stream operator instanciate
 285                 // a new Token before outputting the contents of t.cs().
 286                 // Because of this the line
 287                 //     os << '\\' << cs;
 288                 // below becomes recursive.
 289                 // In order to avoid that we return early:
 290                 if (cs == "\\")
 291                         return os;
 292                 os << '\\' << to_utf8(cs);
 293         }
 294         else if (t.cat() == catLetter)
 295                 os << t.character();
 296         else
 297                 os << '[' << t.character() << ',' << t.cat() << ']';
 298         return os;
 299 }
 300
 301
 302 class Parser {
 303 public:
 304         ///
 305         typedef  InsetMath::mode_type mode_type;
 306
 307         ///
 308         Parser(Lexer & lex);
 309         /// Only use this for reading from .lyx file format, for the reason
 310         /// see Parser::tokenize(istream &).
 311         Parser(istream & is);
 312         ///
 313         Parser(docstring const & str);
 314
 315         ///
 316         bool parse(MathAtom & at);
 317         ///
 318         void parse(MathData & array, unsigned flags, mode_type mode);
 319         ///
 320         void parse1(InsetMathGrid & grid, unsigned flags, mode_type mode,
 321                 bool numbered);
 322         ///
 323         MathData parse(unsigned flags, mode_type mode);
 324         ///
 325         int lineno() const { return lineno_; }
 326         ///
 327         void putback();
 328
 329 private:
 330         ///
 331         void parse2(MathAtom & at, unsigned flags, mode_type mode, bool numbered);
 332         /// get arg delimited by 'left' and 'right'
 333         docstring getArg(char_type left, char_type right);
 334         ///
 335         char_type getChar();
 336         ///
 337         void error(string const & msg);
 338         void error(docstring const & msg) { error(to_utf8(msg)); }
 339         /// dump contents to screen
 340         void dump() const;
 341         /// Only use this for reading from .lyx file format (see
 342         /// implementation for reason)
 343         void tokenize(istream & is);
 344         ///
 345         void tokenize(docstring const & s);
 346         ///
 347         void skipSpaceTokens(idocstream & is, char_type c);
 348         ///
 349         void push_back(Token const & t);
 350         ///
 351         void pop_back();
 352         ///
 353         Token const & prevToken() const;
 354         ///
 355         Token const & nextToken() const;
 356         ///
 357         Token const & getToken();
 358         /// skips spaces if any
 359         void skipSpaces();
 360         ///
 361         void lex(docstring const & s);
 362         ///
 363         bool good() const;
 364         ///
 365         docstring parse_verbatim_item();
 366         ///
 367         docstring parse_verbatim_option();
 368
 369         ///
 370         int lineno_;
 371         ///
 372         vector<Token> tokens_;
 373         ///
 374         unsigned pos_;
 375         /// Stack of active environments
 376         vector<docstring> environments_;
 377 };
 378
 379
 380 Parser::Parser(Lexer & lexer)
 381         : lineno_(lexer.getLineNo()), pos_(0)
 382 {
 383         tokenize(lexer.getStream());
 384         lexer.eatLine();
 385 }
 386
 387
 388 Parser::Parser(istream & is)
 389         : lineno_(0), pos_(0)
 390 {
 391         tokenize(is);
 392 }
 393
 394
 395 Parser::Parser(docstring const & str)
 396         : lineno_(0), pos_(0)
 397 {
 398         tokenize(str);
 399 }
 400
 401
 402 void Parser::push_back(Token const & t)
 403 {
 404         tokens_.push_back(t);
 405 }
 406
 407
 408 void Parser::pop_back()
 409 {
 410         tokens_.pop_back();
 411 }
 412
 413
 414 Token const & Parser::prevToken() const
 415 {
 416         static const Token dummy;
 417         return pos_ > 0 ? tokens_[pos_ - 1] : dummy;
 418 }
 419
 420
 421 Token const & Parser::nextToken() const
 422 {
 423         static const Token dummy;
 424         return good() ? tokens_[pos_] : dummy;
 425 }
 426
 427
 428 Token const & Parser::getToken()
 429 {
 430         static const Token dummy;
 431         //lyxerr << "looking at token " << tokens_[pos_] << " pos: " << pos_ << endl;
 432         return good() ? tokens_[pos_++] : dummy;
 433 }
 434
 435
 436 void Parser::skipSpaces()
 437 {
 438         while (nextToken().cat() == catSpace || nextToken().cat() == catNewline)
 439                 getToken();
 440 }
 441
 442
 443 void Parser::putback()
 444 {
 445         --pos_;
 446 }
 447
 448
 449 bool Parser::good() const
 450 {
 451         return pos_ < tokens_.size();
 452 }
 453
 454
 455 char_type Parser::getChar()
 456 {
 457         if (!good()) {
 458                 error("The input stream is not well...");
 459                 putback();
 460                 return 0;
 461         }
 462         return tokens_[pos_++].character();
 463 }
 464
 465
 466 docstring Parser::getArg(char_type left, char_type right)
 467 {
 468         skipSpaces();
 469
 470         docstring result;
 471         char_type c = getChar();
 472
 473         if (c != left)
 474                 putback();
 475         else
 476                 while ((c = getChar()) != right && good())
 477                         result += c;
 478
 479         return result;
 480 }
 481
 482
 483 void Parser::skipSpaceTokens(idocstream & is, char_type c)
 484 {
 485         // skip trailing spaces
 486         while (catcode(c) == catSpace || catcode(c) == catNewline)
 487                 if (!is.get(c))
 488                         break;
 489         //lyxerr << "putting back: " << c << endl;
 490         is.putback(c);
 491 }
 492
 493
 494 void Parser::tokenize(istream & is)
 495 {
 496         // eat everything up to the next \end_inset or end of stream
 497         // and store it in s for further tokenization
 498         string s;
 499         char c;
 500         while (is.get(c)) {
 501                 s += c;
 502                 if (s.size() >= 10 && s.substr(s.size() - 10) == "\\end_inset") {
 503                         s = s.substr(0, s.size() - 10);
 504                         break;
 505                 }
 506         }
 507         // Remove the space after \end_inset
 508         if (is.get(c) && c != ' ')
 509                 is.unget();
 510
 511         // tokenize buffer
 512         tokenize(from_utf8(s));
 513 }
 514
 515
 516 void Parser::tokenize(docstring const & buffer)
 517 {
 518         idocstringstream is(buffer, ios::in | ios::binary);
 519
 520         char_type c;
 521         while (is.get(c)) {
 522                 //lyxerr << "reading c: " << c << endl;
 523
 524                 switch (catcode(c)) {
 525                         case catNewline: {
 526                                 ++lineno_;
 527                                 is.get(c);
 528                                 if (catcode(c) == catNewline)
 529                                         ; //push_back(Token("par"));
 530                                 else {
 531                                         push_back(Token('\n', catNewline));
 532                                         is.putback(c);
 533                                 }
 534                                 break;
 535                         }
 536
 537 /*
 538                         case catComment: {
 539                                 while (is.get(c) && catcode(c) != catNewline)
 540                                         ;
 541                                 ++lineno_;
 542                                 break;
 543                         }
 544 */
 545
 546                         case catEscape: {
 547                                 is.get(c);
 548                                 if (!is) {
 549                                         error("unexpected end of input");
 550                                 } else {
 551                                         docstring s(1, c);
 552                                         if (catcode(c) == catLetter) {
 553                                                 // collect letters
 554                                                 while (is.get(c) && catcode(c) == catLetter)
 555                                                         s += c;
 556                                                 skipSpaceTokens(is, c);
 557                                         }
 558                                         push_back(Token(s));
 559                                 }
 560                                 break;
 561                         }
 562
 563                         case catSuper:
 564                         case catSub: {
 565                                 push_back(Token(c, catcode(c)));
 566                                 is.get(c);
 567                                 skipSpaceTokens(is, c);
 568                                 break;
 569                         }
 570
 571                         case catIgnore: {
 572                                 lyxerr << "ignoring a char: " << int(c) << endl;
 573                                 break;
 574                         }
 575
 576                         default:
 577                                 push_back(Token(c, catcode(c)));
 578                 }
 579         }
 580
 581 #ifdef FILEDEBUG
 582         dump();
 583 #endif
 584 }
 585
 586
 587 void Parser::dump() const
 588 {
 589         lyxerr << "\nTokens: ";
 590         for (unsigned i = 0; i < tokens_.size(); ++i) {
 591                 if (i == pos_)
 592                         lyxerr << " <#> ";
 593                 lyxerr << tokens_[i];
 594         }
 595         lyxerr << " pos: " << pos_ << endl;
 596 }
 597
 598
 599 void Parser::error(string const & msg)
 600 {
 601         lyxerr << "Line ~" << lineno_ << ": Math parse error: " << msg << endl;
 602         dump();
 603         //exit(1);
 604 }
 605
 606
 607 bool Parser::parse(MathAtom & at)
 608 {
 609         skipSpaces();
 610         MathData ar;
 611         parse(ar, false, InsetMath::UNDECIDED_MODE);
 612         if (ar.size() != 1 || ar.front()->getType() == hullNone) {
 613                 lyxerr << "unusual contents found: " << ar << endl;
 614                 at = MathAtom(new InsetMathPar(ar));
 615                 //if (at->nargs() > 0)
 616                 //      at.nucleus()->cell(0) = ar;
 617                 //else
 618                 //      lyxerr << "unusual contents found: " << ar << endl;
 619                 return true;
 620         }
 621         at = ar[0];
 622         return true;
 623 }
 624
 625
 626 docstring Parser::parse_verbatim_option()
 627 {
 628         skipSpaces();
 629         docstring res;
 630         if (nextToken().character() == '[') {
 631                 Token t = getToken();
 632                 for (Token t = getToken(); t.character() != ']' && good(); t = getToken()) {
 633                         if (t.cat() == catBegin) {
 634                                 putback();
 635                                 res += '{' + parse_verbatim_item() + '}';
 636                         } else
 637                                 res += t.asString();
 638                 }
 639         }
 640         return res;
 641 }
 642
 643
 644 docstring Parser::parse_verbatim_item()
 645 {
 646         skipSpaces();
 647         docstring res;
 648         if (nextToken().cat() == catBegin) {
 649                 Token t = getToken();
 650                 for (Token t = getToken(); t.cat() != catEnd && good(); t = getToken()) {
 651                         if (t.cat() == catBegin) {
 652                                 putback();
 653                                 res += '{' + parse_verbatim_item() + '}';
 654                         }
 655                         else
 656                                 res += t.asString();
 657                 }
 658         }
 659         return res;
 660 }
 661
 662
 663 MathData Parser::parse(unsigned flags, mode_type mode)
 664 {
 665         MathData ar;
 666         parse(ar, flags, mode);
 667         return ar;
 668 }
 669
 670
 671 void Parser::parse(MathData & array, unsigned flags, mode_type mode)
 672 {
 673         InsetMathGrid grid(1, 1);
 674         parse1(grid, flags, mode, false);
 675         array = grid.cell(0);
 676 }
 677
 678
 679 void Parser::parse2(MathAtom & at, const unsigned flags, const mode_type mode,
 680         const bool numbered)
 681 {
 682         parse1(*(at.nucleus()->asGridInset()), flags, mode, numbered);
 683 }
 684
 685
 686 void Parser::parse1(InsetMathGrid & grid, unsigned flags,
 687         const mode_type mode, const bool numbered)
 688 {
 689         int limits = 0;
 690         InsetMathGrid::row_type cellrow = 0;
 691         InsetMathGrid::col_type cellcol = 0;
 692         MathData * cell = &grid.cell(grid.index(cellrow, cellcol));
 693
 694         if (grid.asHullInset())
 695                 grid.asHullInset()->numbered(cellrow, numbered);
 696
 697         //dump();
 698         //lyxerr << " flags: " << flags << endl;
 699         //lyxerr << " mode: " << mode  << endl;
 700         //lyxerr << "grid: " << grid << endl;
 701
 702         while (good()) {
 703                 Token const & t = getToken();
 704
 705 #ifdef FILEDEBUG
 706                 lyxerr << "t: " << t << " flags: " << flags << endl;
 707                 lyxerr << "mode: " << mode  << endl;
 708                 cell->dump();
 709                 lyxerr << endl;
 710 #endif
 711
 712                 if (flags & FLAG_ITEM) {
 713
 714                         if (t.cat() == catBegin) {
 715                                 // skip the brace and collect everything to the next matching
 716                                 // closing brace
 717                                 parse1(grid, FLAG_BRACE_LAST, mode, numbered);
 718                                 return;
 719                         }
 720
 721                         // handle only this single token, leave the loop if done
 722                         flags = FLAG_LEAVE;
 723                 }
 724
 725
 726                 if (flags & FLAG_BRACED) {
 727                         if (t.cat() == catSpace)
 728                                 continue;
 729
 730                         if (t.cat() != catBegin) {
 731                                 error("opening brace expected");
 732                                 return;
 733                         }
 734
 735                         // skip the brace and collect everything to the next matching
 736                         // closing brace
 737                         flags = FLAG_BRACE_LAST;
 738                 }
 739
 740
 741                 if (flags & FLAG_OPTION) {
 742                         if (t.cat() == catOther && t.character() == '[') {
 743                                 MathData ar;
 744                                 parse(ar, FLAG_BRACK_LAST, mode);
 745                                 cell->append(ar);
 746                         } else {
 747                                 // no option found, put back token and we are done
 748                                 putback();
 749                         }
 750                         return;
 751                 }
 752
 753                 //
 754                 // cat codes
 755                 //
 756                 if (t.cat() == catMath) {
 757                         if (mode != InsetMath::MATH_MODE) {
 758                                 // we are inside some text mode thingy, so opening new math is allowed
 759                                 Token const & n = getToken();
 760                                 if (n.cat() == catMath) {
 761                                         // TeX's $$...$$ syntax for displayed math
 762                                         cell->push_back(MathAtom(new InsetMathHull(hullEquation)));
 763                                         parse2(cell->back(), FLAG_SIMPLE, InsetMath::MATH_MODE, false);
 764                                         getToken(); // skip the second '$' token
 765                                 } else {
 766                                         // simple $...$  stuff
 767                                         putback();
 768                                         cell->push_back(MathAtom(new InsetMathHull(hullSimple)));
 769                                         parse2(cell->back(), FLAG_SIMPLE, InsetMath::MATH_MODE, false);
 770                                 }
 771                         }
 772
 773                         else if (flags & FLAG_SIMPLE) {
 774                                 // this is the end of the formula
 775                                 return;
 776                         }
 777
 778                         else {
 779                                 error("something strange in the parser");
 780                                 break;
 781                         }
 782                 }
 783
 784                 else if (t.cat() == catLetter)
 785                         cell->push_back(MathAtom(new InsetMathChar(t.character())));
 786
 787                 else if (t.cat() == catSpace && mode != InsetMath::MATH_MODE) {
 788                         if (cell->empty() || cell->back()->getChar() != ' ')
 789                                 cell->push_back(MathAtom(new InsetMathChar(t.character())));
 790                 }
 791
 792                 else if (t.cat() == catNewline && mode != InsetMath::MATH_MODE) {
 793                         if (cell->empty() || cell->back()->getChar() != ' ')
 794                                 cell->push_back(MathAtom(new InsetMathChar(' ')));
 795                 }
 796
 797                 else if (t.cat() == catParameter) {
 798                         Token const & n = getToken();
 799                         cell->push_back(MathAtom(new MathMacroArgument(n.character()-'0')));
 800                 }
 801
 802                 else if (t.cat() == catActive)
 803                         cell->push_back(MathAtom(new InsetMathChar(t.character())));
 804
 805                 else if (t.cat() == catBegin) {
 806                         MathData ar;
 807                         parse(ar, FLAG_BRACE_LAST, mode);
 808                         // do not create a BraceInset if they were written by LyX
 809                         // this helps to keep the annoyance of  "a choose b"  to a minimum
 810                         if (ar.size() == 1 && ar[0]->extraBraces())
 811                                 cell->append(ar);
 812                         else
 813                                 cell->push_back(MathAtom(new InsetMathBrace(ar)));
 814                 }
 815
 816                 else if (t.cat() == catEnd) {
 817                         if (flags & FLAG_BRACE_LAST)
 818                                 return;
 819                         error("found '}' unexpectedly");
 820                         //BOOST_ASSERT(false);
 821                         //add(cell, '}', LM_TC_TEX);
 822                 }
 823
 824                 else if (t.cat() == catAlign) {
 825                         //lyxerr << " column now " << (cellcol + 1)
 826                         //       << " max: " << grid.ncols() << endl;
 827                         if (flags & FLAG_ALIGN)
 828                                 return;
 829                         if (addCol(grid, cellcol))
 830                                 cell = &grid.cell(grid.index(cellrow, cellcol));
 831                 }
 832
 833                 else if (t.cat() == catSuper || t.cat() == catSub) {
 834                         bool up = (t.cat() == catSuper);
 835                         // we need no new script inset if the last thing was a scriptinset,
 836                         // which has that script already not the same script already
 837                         if (!cell->size())
 838                                 cell->push_back(MathAtom(new InsetMathScript(up)));
 839                         else if (cell->back()->asScriptInset() &&
 840                                         !cell->back()->asScriptInset()->has(up))
 841                                 cell->back().nucleus()->asScriptInset()->ensure(up);
 842                         else if (cell->back()->asScriptInset())
 843                                 cell->push_back(MathAtom(new InsetMathScript(up)));
 844                         else
 845                                 cell->back() = MathAtom(new InsetMathScript(cell->back(), up));
 846                         InsetMathScript * p = cell->back().nucleus()->asScriptInset();
 847                         // special handling of {}-bases
 848                         // Here we could remove the brace inset for things
 849                         // like {a'}^2 and add the braces back in
 850                         // InsetMathScript::write().
 851                         // We do not do it, since it is not possible to detect
 852                         // reliably whether the braces are needed because the
 853                         // nucleus contains more than one symbol, or whether
 854                         // they are needed for unknown commands like \xx{a}_0
 855                         // or \yy{a}{b}_0. This was done in revision 14819
 856                         // in an unreliable way. See this thread
 857                         // http://www.mail-archive.com/lyx-devel%40lists.lyx.org/msg104917.html
 858                         // for more details.
 859                         parse(p->cell(p->idxOfScript(up)), FLAG_ITEM, mode);
 860                         if (limits) {
 861                                 p->limits(limits);
 862                                 limits = 0;
 863                         }
 864                 }
 865
 866                 else if (t.character() == ']' && (flags & FLAG_BRACK_LAST)) {
 867                         //lyxerr << "finished reading option" << endl;
 868                         return;
 869                 }
 870
 871                 else if (t.cat() == catOther)
 872                         cell->push_back(MathAtom(new InsetMathChar(t.character())));
 873
 874                 else if (t.cat() == catComment) {
 875                         docstring s;
 876                         while (good()) {
 877                                 Token const & t = getToken();
 878                                 if (t.cat() == catNewline)
 879                                         break;
 880                                 s += t.asString();
 881                         }
 882                         cell->push_back(MathAtom(new InsetMathComment(s)));
 883                         skipSpaces();
 884                 }
 885
 886                 //
 887                 // control sequences
 888                 //
 889
 890                 else if (t.cs() == "lyxlock") {
 891                         if (cell->size())
 892                                 cell->back().nucleus()->lock(true);
 893                 }
 894
 895                 else if ((t.cs() == "global" && nextToken().cs() == "def") ||
 896                         t.cs() == "def" ||
 897                         t.cs() == "newcommand" ||
 898                         t.cs() == "renewcommand" ||
 899                         t.cs() == "newlyxcommand" ||
 900                         t.cs() == "newcommandx" ||
 901                         t.cs() == "renewcommandx")
 902                 {
 903                         MacroType type = MacroTypeNewcommand;
 904                         if (t.cs() == "global") {
 905                                 getToken();
 906                                 type = MacroTypeDef;
 907                         }
 908                         if (t.cs() == "def")
 909                                 type = MacroTypeDef;
 910                         if (t.cs() == "newcommandx" || t.cs() == "renewcommandx")
 911                                 type = MacroTypeNewcommandx;
 912                         docstring name;
 913                         int nargs = 0;
 914                         int optionals = 0;
 915                         vector<MathData> optionalValues;
 916                         if (type == MacroTypeDef) {
 917                                 // get name
 918                                 name = getToken().cs();
 919
 920                                 // read parameter
 921                                 docstring pars;
 922                                 while (good() && nextToken().cat() != catBegin) {
 923                                         pars += getToken().cs();
 924                                         ++nargs;
 925                                 }
 926                                 nargs /= 2;
 927                                 //lyxerr << "read \\def parameter list '" << pars << "'" << endl;
 928
 929                         } else if (type == MacroTypeNewcommandx) {
 930                                 // \newcommandx{\foo}[2][usedefault, addprefix=\global,1=default]{#1,#2}
 931                                 // name
 932                                 if (nextToken().cat() == catBegin) {
 933                                         getToken();
 934                                         name = getToken().cs();
 935                                         if (getToken().cat() != catEnd) {
 936                                                 error("'}' in \\newcommandx expected");
 937                                                 return;
 938                                         }
 939                                 } else
 940                                         name = getToken().cs();
 941
 942                                 // arity
 943                                 docstring const arg = getArg('[', ']');
 944                                 if (arg.empty()) {
 945                                         error("[num] in \\newcommandx expected");
 946                                         return;
 947                                 }
 948                                 nargs = convert<int>(arg);
 949
 950                                 // options
 951                                 if (nextToken().character() == '[') {
 952                                         // skip '['
 953                                         getToken();
 954
 955                                         // handle 'opt=value' options, separated by ','.
 956                                         skipSpaces();
 957                                         while (nextToken().character() != ']' && good()) {
 958                                                 if (nextToken().character() >= '1'
 959                                                     && nextToken().character() <= '9') {
 960                                                         // optional value -> get parameter number
 961                                                         int n = getChar() - '0';
 962                                                         if (n > nargs) {
 963                                                                 error("Arity of \\newcommandx too low for given optional parameter.");
 964                                                                 return;
 965                                                         }
 966
 967                                                         // skip '='
 968                                                         if (getToken().character() != '=') {
 969                                                                 error("'=' and optional parameter value expected for \\newcommandx");
 970                                                                 return;
 971                                                         }
 972
 973                                                         // get value
 974                                                         optionalValues.resize(max(size_t(n), optionalValues.size()));
 975                                                         optionalValues[n - 1].clear();
 976                                                         while (nextToken().character() != ']'
 977                                                                && nextToken().character() != ',') {
 978                                                                 MathData data;
 979                                                                 parse(data, FLAG_ITEM, InsetMath::UNDECIDED_MODE);
 980                                                                 optionalValues[n - 1].append(data);
 981                                                         }
 982                                                         optionals = max(n, optionals);
 983                                                 } else if (nextToken().cat() == catLetter) {
 984                                                         // we in fact ignore every non-optional
 985                                                         // parameters
 986
 987                                                         // get option name
 988                                                         docstring opt;
 989                                                         while (nextToken().cat() == catLetter)
 990                                                                 opt += getChar();
 991
 992                                                         // value?
 993                                                         skipSpaces();
 994                                                         MathData value;
 995                                                         if (nextToken().character() == '=') {
 996                                                                 getToken();
 997                                                                 while (nextToken().character() != ']'
 998                                                                         && nextToken().character() != ',')
 999                                                                         parse(value, FLAG_ITEM, InsetMath::UNDECIDED_MODE);
1000                                                         }
1001                                                 } else {
1002                                                         error("option for \\newcommandx expected");
1003                                                         return;
1004                                                 }
1005
1006                                                 // skip komma
1007                                                 skipSpaces();
1008                                                 if (nextToken().character() == ',') {
1009                                                         getChar();
1010                                                         skipSpaces();
1011                                                 } else if (nextToken().character() != ']') {
1012                                                         error("Expecting ',' or ']' in options of \\newcommandx");
1013                                                         return;
1014                                                 }
1015                                         }
1016
1017                                         // skip ']'
1018                                         if (!good())
1019                                                 return;
1020                                         getToken();
1021                                 }
1022                         } else {
1023                                 if (getToken().cat() != catBegin) {
1024                                         error("'{' in \\newcommand expected (1) ");
1025                                         return;
1026                                 }
1027
1028                                 name = getToken().cs();
1029
1030                                 if (getToken().cat() != catEnd) {
1031                                         error("'}' in \\newcommand expected");
1032                                         return;
1033                                 }
1034
1035                                 docstring const arg = getArg('[', ']');
1036                                 if (!arg.empty())
1037                                         nargs = convert<int>(arg);
1038
1039                                 // optional argument given?
1040                                 skipSpaces();
1041                                 while (nextToken().character() == '[') {
1042                                         getToken();
1043                                         optionalValues.push_back(MathData());
1044                                         parse(optionalValues[optionals], FLAG_BRACK_LAST, mode);
1045                                         ++optionals;
1046                                 }
1047                         }
1048
1049                         MathData def;
1050                         parse(def, FLAG_ITEM, InsetMath::UNDECIDED_MODE);
1051
1052                         // is a version for display attached?
1053                         skipSpaces();
1054                         MathData display;
1055                         if (nextToken().cat() == catBegin)
1056                                 parse(display, FLAG_ITEM, InsetMath::MATH_MODE);
1057
1058                         cell->push_back(MathAtom(new MathMacroTemplate(name, nargs,
1059                                 optionals, type, optionalValues, def, display)));
1060                 }
1061
1062                 else if (t.cs() == "(") {
1063                         cell->push_back(MathAtom(new InsetMathHull(hullSimple)));
1064                         parse2(cell->back(), FLAG_SIMPLE2, InsetMath::MATH_MODE, false);
1065                 }
1066
1067                 else if (t.cs() == "[") {
1068                         cell->push_back(MathAtom(new InsetMathHull(hullEquation)));
1069                         parse2(cell->back(), FLAG_EQUATION, InsetMath::MATH_MODE, false);
1070                 }
1071
1072                 else if (t.cs() == "protect")
1073                         // ignore \\protect, will hopefully be re-added during output
1074                         ;
1075
1076                 else if (t.cs() == "end") {
1077                         if (flags & FLAG_END) {
1078                                 // eat environment name
1079                                 docstring const name = getArg('{', '}');
1080                                 if (environments_.empty())
1081                                         error("'found \\end{" + name +
1082                                               "}' without matching '\\begin{" +
1083                                               name + "}'");
1084                                 else if (name != environments_.back())
1085                                         error("'\\end{" + name +
1086                                               "}' does not match '\\begin{" +
1087                                               environments_.back() + "}'");
1088                                 else {
1089                                         environments_.pop_back();
1090                                         // Delete empty last row in matrix
1091                                         // like insets.
1092                                         // If you abuse InsetMathGrid for
1093                                         // non-matrix like structures you
1094                                         // probably need to refine this test.
1095                                         // Right now we only have to test for
1096                                         // single line hull insets.
1097                                         if (grid.nrows() > 1)
1098                                                 delEmptyLastRow(grid);
1099                                         return;
1100                                 }
1101                         } else
1102                                 error("found 'end' unexpectedly");
1103                 }
1104
1105                 else if (t.cs() == ")") {
1106                         if (flags & FLAG_SIMPLE2)
1107                                 return;
1108                         error("found '\\)' unexpectedly");
1109                 }
1110
1111                 else if (t.cs() == "]") {
1112                         if (flags & FLAG_EQUATION)
1113                                 return;
1114                         error("found '\\]' unexpectedly");
1115                 }
1116
1117                 else if (t.cs() == "\\") {
1118                         if (flags & FLAG_ALIGN)
1119                                 return;
1120                         bool added = false;
1121                         if (nextToken().asInput() == "*") {
1122                                 getToken();
1123                                 added = addRow(grid, cellrow, docstring(), false);
1124                         } else if (good())
1125                                 added = addRow(grid, cellrow, getArg('[', ']'));
1126                         else
1127                                 error("missing token after \\\\");
1128                         if (added) {
1129                                 cellcol = 0;
1130                                 if (grid.asHullInset())
1131                                         grid.asHullInset()->numbered(
1132                                                         cellrow, numbered);
1133                                 cell = &grid.cell(grid.index(cellrow,
1134                                                              cellcol));
1135                         }
1136                 }
1137
1138 #if 0
1139                 else if (t.cs() == "multicolumn") {
1140                         // extract column count and insert dummy cells
1141                         MathData count;
1142                         parse(count, FLAG_ITEM, mode);
1143                         int cols = 1;
1144                         if (!extractNumber(count, cols)) {
1145                                 lyxerr << " can't extract number of cells from " << count << endl;
1146                         }
1147                         // resize the table if necessary
1148                         for (int i = 0; i < cols; ++i) {
1149                                 if (addCol(grid, cellcol)) {
1150                                         cell = &grid.cell(grid.index(
1151                                                         cellrow, cellcol));
1152                                         // mark this as dummy
1153                                         grid.cellinfo(grid.index(
1154                                                 cellrow, cellcol)).dummy_ = true;
1155                                 }
1156                         }
1157                         // the last cell is the real thing, not a dummy
1158                         grid.cellinfo(grid.index(cellrow, cellcol)).dummy_ = false;
1159
1160                         // read special alignment
1161                         MathData align;
1162                         parse(align, FLAG_ITEM, mode);
1163                         //grid.cellinfo(grid.index(cellrow, cellcol)).align_ = extractString(align);
1164
1165                         // parse the remaining contents into the "real" cell
1166                         parse(*cell, FLAG_ITEM, mode);
1167                 }
1168 #endif
1169
1170                 else if (t.cs() == "limits")
1171                         limits = 1;
1172
1173                 else if (t.cs() == "nolimits")
1174                         limits = -1;
1175
1176                 else if (t.cs() == "nonumber") {
1177                         if (grid.asHullInset())
1178                                 grid.asHullInset()->numbered(cellrow, false);
1179                 }
1180
1181                 else if (t.cs() == "number") {
1182                         if (grid.asHullInset())
1183                                 grid.asHullInset()->numbered(cellrow, true);
1184                 }
1185
1186                 else if (t.cs() == "hline") {
1187                         grid.rowinfo(cellrow).lines_ ++;
1188                 }
1189
1190                 else if (t.cs() == "sqrt") {
1191                         MathData ar;
1192                         parse(ar, FLAG_OPTION, mode);
1193                         if (ar.size()) {
1194                                 cell->push_back(MathAtom(new InsetMathRoot));
1195                                 cell->back().nucleus()->cell(0) = ar;
1196                                 parse(cell->back().nucleus()->cell(1), FLAG_ITEM, mode);
1197                         } else {
1198                                 cell->push_back(MathAtom(new InsetMathSqrt));
1199                                 parse(cell->back().nucleus()->cell(0), FLAG_ITEM, mode);
1200                         }
1201                 }
1202
1203                 else if (t.cs() == "unit") {
1204                         // Allowed formats \unit[val]{unit}
1205                         MathData ar;
1206                         parse(ar, FLAG_OPTION, mode);
1207                         if (ar.size()) {
1208                                 cell->push_back(MathAtom(new InsetMathFrac(InsetMathFrac::UNIT)));
1209                                 cell->back().nucleus()->cell(0) = ar;
1210                                 parse(cell->back().nucleus()->cell(1), FLAG_ITEM, mode);
1211                         } else {
1212                                 cell->push_back(MathAtom(new InsetMathFrac(InsetMathFrac::UNIT, 1)));
1213                                 parse(cell->back().nucleus()->cell(0), FLAG_ITEM, mode);
1214                         }
1215                 }
1216                 else if (t.cs() == "unitfrac") {
1217                         // Here allowed formats are \unitfrac[val]{num}{denom}
1218                         MathData ar;
1219                         parse(ar, FLAG_OPTION, mode);
1220                         if (ar.size()) {
1221                                 cell->push_back(MathAtom(new InsetMathFrac(InsetMathFrac::UNITFRAC, 3)));
1222                                 cell->back().nucleus()->cell(2) = ar;
1223                         } else {
1224                                 cell->push_back(MathAtom(new InsetMathFrac(InsetMathFrac::UNITFRAC)));
1225                         }
1226                         parse(cell->back().nucleus()->cell(0), FLAG_ITEM, mode);
1227                         parse(cell->back().nucleus()->cell(1), FLAG_ITEM, mode);
1228                 }
1229
1230                 else if (t.cs() == "xrightarrow" || t.cs() == "xleftarrow") {
1231                         cell->push_back(createInsetMath(t.cs()));
1232                         parse(cell->back().nucleus()->cell(1), FLAG_OPTION, mode);
1233                         parse(cell->back().nucleus()->cell(0), FLAG_ITEM, mode);
1234                 }
1235
1236                 else if (t.cs() == "ref" || t.cs() == "prettyref" ||
1237                                 t.cs() == "pageref" || t.cs() == "vpageref" || t.cs() == "vref") {
1238                         cell->push_back(MathAtom(new InsetMathRef(t.cs())));
1239                         parse(cell->back().nucleus()->cell(1), FLAG_OPTION, mode);
1240                         parse(cell->back().nucleus()->cell(0), FLAG_ITEM, mode);
1241                 }
1242
1243                 else if (t.cs() == "left") {
1244                         skipSpaces();
1245                         Token const & tl = getToken();
1246                         // \| and \Vert are equivalent, and InsetMathDelim
1247                         // can't handle \|
1248                         // FIXME: fix this in InsetMathDelim itself!
1249                         docstring const l = tl.cs() == "|" ? from_ascii("Vert") : tl.asString();
1250                         MathData ar;
1251                         parse(ar, FLAG_RIGHT, mode);
1252                         if (!good())
1253                                 break;
1254                         skipSpaces();
1255                         Token const & tr = getToken();
1256                         docstring const r = tr.cs() == "|" ? from_ascii("Vert") : tr.asString();
1257                         cell->push_back(MathAtom(new InsetMathDelim(l, r, ar)));
1258                 }
1259
1260                 else if (t.cs() == "right") {
1261                         if (flags & FLAG_RIGHT)
1262                                 return;
1263                         //lyxerr << "got so far: '" << cell << "'" << endl;
1264                         error("Unmatched right delimiter");
1265                         return;
1266                 }
1267
1268                 else if (t.cs() == "begin") {
1269                         docstring const name = getArg('{', '}');
1270                         environments_.push_back(name);
1271
1272                         if (name == "array" || name == "subarray") {
1273                                 docstring const valign = parse_verbatim_option() + 'c';
1274                                 docstring const halign = parse_verbatim_item();
1275                                 cell->push_back(MathAtom(new InsetMathArray(name, (char)valign[0], halign)));
1276                                 parse2(cell->back(), FLAG_END, mode, false);
1277                         }
1278
1279                         else if (name == "tabular") {
1280                                 docstring const valign = parse_verbatim_option() + 'c';
1281                                 docstring const halign = parse_verbatim_item();
1282                                 cell->push_back(MathAtom(new InsetMathTabular(name, (char)valign[0], halign)));
1283                                 parse2(cell->back(), FLAG_END, InsetMath::TEXT_MODE, false);
1284                         }
1285
1286                         else if (name == "split" || name == "cases") {
1287                                 cell->push_back(createInsetMath(name));
1288                                 parse2(cell->back(), FLAG_END, mode, false);
1289                         }
1290
1291                         else if (name == "alignedat") {
1292                                 docstring const valign = parse_verbatim_option() + 'c';
1293                                 // ignore this for a while
1294                                 getArg('{', '}');
1295                                 cell->push_back(MathAtom(new InsetMathSplit(name, (char)valign[0])));
1296                                 parse2(cell->back(), FLAG_END, mode, false);
1297                         }
1298
1299                         else if (name == "math") {
1300                                 cell->push_back(MathAtom(new InsetMathHull(hullSimple)));
1301                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, true);
1302                         }
1303
1304                         else if (name == "equation" || name == "equation*"
1305                                         || name == "displaymath") {
1306                                 cell->push_back(MathAtom(new InsetMathHull(hullEquation)));
1307                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, (name == "equation"));
1308                         }
1309
1310                         else if (name == "eqnarray" || name == "eqnarray*") {
1311                                 cell->push_back(MathAtom(new InsetMathHull(hullEqnArray)));
1312                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1313                         }
1314
1315                         else if (name == "align" || name == "align*") {
1316                                 cell->push_back(MathAtom(new InsetMathHull(hullAlign)));
1317                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1318                         }
1319
1320                         else if (name == "flalign" || name == "flalign*") {
1321                                 cell->push_back(MathAtom(new InsetMathHull(hullFlAlign)));
1322                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1323                         }
1324
1325                         else if (name == "alignat" || name == "alignat*") {
1326                                 // ignore this for a while
1327                                 getArg('{', '}');
1328                                 cell->push_back(MathAtom(new InsetMathHull(hullAlignAt)));
1329                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1330                         }
1331
1332                         else if (name == "xalignat" || name == "xalignat*") {
1333                                 // ignore this for a while
1334                                 getArg('{', '}');
1335                                 cell->push_back(MathAtom(new InsetMathHull(hullXAlignAt)));
1336                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1337                         }
1338
1339                         else if (name == "xxalignat") {
1340                                 // ignore this for a while
1341                                 getArg('{', '}');
1342                                 cell->push_back(MathAtom(new InsetMathHull(hullXXAlignAt)));
1343                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1344                         }
1345
1346                         else if (name == "multline" || name == "multline*") {
1347                                 cell->push_back(MathAtom(new InsetMathHull(hullMultline)));
1348                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1349                         }
1350
1351                         else if (name == "gather" || name == "gather*") {
1352                                 cell->push_back(MathAtom(new InsetMathHull(hullGather)));
1353                                 parse2(cell->back(), FLAG_END, InsetMath::MATH_MODE, !stared(name));
1354                         }
1355
1356                         else if (latexkeys const * l = in_word_set(name)) {
1357                                 if (l->inset == "matrix") {
1358                                         cell->push_back(createInsetMath(name));
1359                                         parse2(cell->back(), FLAG_END, mode, false);
1360                                 } else if (l->inset == "split") {
1361                                         docstring const valign = parse_verbatim_option() + 'c';
1362                                         cell->push_back(MathAtom(new InsetMathSplit(name, (char)valign[0])));
1363                                         parse2(cell->back(), FLAG_END, mode, false);
1364                                 } else {
1365                                         dump();
1366                                         lyxerr << "found math environment `" << to_utf8(name)
1367                                                << "' in symbols file with unsupported inset `"
1368                                                << to_utf8(l->inset) << "'." << endl;
1369                                         // create generic environment inset
1370                                         cell->push_back(MathAtom(new InsetMathEnv(name)));
1371                                         parse(cell->back().nucleus()->cell(0), FLAG_END, mode);
1372                                 }
1373                         }
1374
1375                         else {
1376                                 dump();
1377                                 lyxerr << "found unknown math environment '" << to_utf8(name)
1378                                         << "'" << endl;
1379                                 // create generic environment inset
1380                                 cell->push_back(MathAtom(new InsetMathEnv(name)));
1381                                 parse(cell->back().nucleus()->cell(0), FLAG_END, mode);
1382                         }
1383                 }
1384
1385                 else if (t.cs() == "kern") {
1386                         // FIXME: A hack...
1387                         docstring s;
1388                         while (true) {
1389                                 Token const & t = getToken();
1390                                 if (!good()) {
1391                                         putback();
1392                                         break;
1393                                 }
1394                                 s += t.character();
1395                                 if (isValidLength(to_utf8(s)))
1396                                         break;
1397                         }
1398                         cell->push_back(MathAtom(new InsetMathKern(s)));
1399                 }
1400
1401                 else if (t.cs() == "label") {
1402                         // FIXME: This is swallowed in inline formulas
1403                         docstring label = parse_verbatim_item();
1404                         MathData ar;
1405                         asArray(label, ar);
1406                         if (grid.asHullInset()) {
1407                                 grid.asHullInset()->label(cellrow, label);
1408                         } else {
1409                                 cell->push_back(createInsetMath(t.cs()));
1410                                 cell->push_back(MathAtom(new InsetMathBrace(ar)));
1411                         }
1412                 }
1413
1414                 else if (t.cs() == "choose" || t.cs() == "over" || t.cs() == "atop") {
1415                         MathAtom at = createInsetMath(t.cs());
1416                         at.nucleus()->cell(0) = *cell;
1417                         cell->clear();
1418                         parse(at.nucleus()->cell(1), flags, mode);
1419                         cell->push_back(at);
1420                         return;
1421                 }
1422
1423                 else if (t.cs() == "color") {
1424                         docstring const color = parse_verbatim_item();
1425                         cell->push_back(MathAtom(new InsetMathColor(true, color)));
1426                         parse(cell->back().nucleus()->cell(0), flags, mode);
1427                         return;
1428                 }
1429
1430                 else if (t.cs() == "textcolor") {
1431                         docstring const color = parse_verbatim_item();
1432                         cell->push_back(MathAtom(new InsetMathColor(false, color)));
1433                         parse(cell->back().nucleus()->cell(0), FLAG_ITEM, InsetMath::TEXT_MODE);
1434                 }
1435
1436                 else if (t.cs() == "normalcolor") {
1437                         cell->push_back(createInsetMath(t.cs()));
1438                         parse(cell->back().nucleus()->cell(0), flags, mode);
1439                         return;
1440                 }
1441
1442                 else if (t.cs() == "substack") {
1443                         cell->push_back(createInsetMath(t.cs()));
1444                         parse2(cell->back(), FLAG_ITEM, mode, false);
1445                 }
1446
1447                 else if (t.cs() == "xymatrix") {
1448                         odocstringstream os;
1449                         while (good() && nextToken().cat() != catBegin)
1450                                 os << getToken().asInput();
1451                         cell->push_back(createInsetMath(t.cs() + os.str()));
1452                         parse2(cell->back(), FLAG_ITEM, mode, false);
1453                 }
1454
1455                 else if (t.cs() == "framebox" || t.cs() == "makebox") {
1456                         cell->push_back(createInsetMath(t.cs()));
1457                         parse(cell->back().nucleus()->cell(0), FLAG_OPTION, InsetMath::TEXT_MODE);
1458                         parse(cell->back().nucleus()->cell(1), FLAG_OPTION, InsetMath::TEXT_MODE);
1459                         parse(cell->back().nucleus()->cell(2), FLAG_ITEM, InsetMath::TEXT_MODE);
1460                 }
1461
1462                 else if (t.cs() == "tag") {
1463                         if (nextToken().character() == '*') {
1464                                 getToken();
1465                                 cell->push_back(createInsetMath(t.cs() + '*'));
1466                         } else
1467                                 cell->push_back(createInsetMath(t.cs()));
1468                         parse(cell->back().nucleus()->cell(0), FLAG_ITEM, InsetMath::TEXT_MODE);
1469                 }
1470
1471 #if 0
1472                 else if (t.cs() == "infer") {
1473                         MathData ar;
1474                         parse(ar, FLAG_OPTION, mode);
1475                         cell->push_back(createInsetMath(t.cs()));
1476                         parse2(cell->back(), FLAG_ITEM, mode, false);
1477                 }
1478
1479                 // Disabled
1480                 else if (1 && t.cs() == "ar") {
1481                         auto_ptr<InsetMathXYArrow> p(new InsetMathXYArrow);
1482                         // try to read target
1483                         parse(p->cell(0), FLAG_OTPTION, mode);
1484                         // try to read label
1485                         if (nextToken().cat() == catSuper || nextToken().cat() == catSub) {
1486                                 p->up_ = nextToken().cat() == catSuper;
1487                                 getToken();
1488                                 parse(p->cell(1), FLAG_ITEM, mode);
1489                                 //lyxerr << "read label: " << p->cell(1) << endl;
1490                         }
1491
1492                         cell->push_back(MathAtom(p.release()));
1493                         //lyxerr << "read cell: " << cell << endl;
1494                 }
1495 #endif
1496
1497                 else if (t.cs().size()) {
1498                         latexkeys const * l = in_word_set(t.cs());
1499                         if (l) {
1500                                 if (l->inset == "big") {
1501                                         skipSpaces();
1502                                         docstring const delim = getToken().asInput();
1503                                         if (InsetMathBig::isBigInsetDelim(delim))
1504                                                 cell->push_back(MathAtom(
1505                                                         new InsetMathBig(t.cs(), delim)));
1506                                         else {
1507                                                 cell->push_back(createInsetMath(t.cs()));
1508                                                 putback();
1509                                         }
1510                                 }
1511
1512                                 else if (l->inset == "font") {
1513                                         cell->push_back(createInsetMath(t.cs()));
1514                                         parse(cell->back().nucleus()->cell(0),
1515                                                 FLAG_ITEM, asMode(mode, l->extra));
1516                                 }
1517
1518                                 else if (l->inset == "oldfont") {
1519                                         cell->push_back(createInsetMath(t.cs()));
1520                                         parse(cell->back().nucleus()->cell(0),
1521                                                 flags | FLAG_ALIGN, asMode(mode, l->extra));
1522                                         if (prevToken().cat() != catAlign &&
1523                                             prevToken().cs() != "\\")
1524                                                 return;
1525                                         putback();
1526                                 }
1527
1528                                 else if (l->inset == "style") {
1529                                         cell->push_back(createInsetMath(t.cs()));
1530                                         parse(cell->back().nucleus()->cell(0),
1531                                                 flags | FLAG_ALIGN, mode);
1532                                         if (prevToken().cat() != catAlign &&
1533                                             prevToken().cs() != "\\")
1534                                                 return;
1535                                         putback();
1536                                 }
1537
1538                                 else {
1539                                         MathAtom at = createInsetMath(t.cs());
1540                                         for (InsetMath::idx_type i = 0; i < at->nargs(); ++i)
1541                                                 parse(at.nucleus()->cell(i),
1542                                                         FLAG_ITEM, asMode(mode, l->extra));
1543                                         cell->push_back(at);
1544                                 }
1545                         }
1546
1547                         else {
1548                                 MathAtom at = createInsetMath(t.cs());
1549                                 InsetMath::mode_type m = mode;
1550                                 //if (m == InsetMath::UNDECIDED_MODE)
1551                                 //lyxerr << "default creation: m1: " << m << endl;
1552                                 if (at->currentMode() != InsetMath::UNDECIDED_MODE)
1553                                         m = at->currentMode();
1554                                 //lyxerr << "default creation: m2: " << m << endl;
1555                                 InsetMath::idx_type start = 0;
1556                                 // this fails on \bigg[...\bigg]
1557                                 //MathData opt;
1558                                 //parse(opt, FLAG_OPTION, InsetMath::VERBATIM_MODE);
1559                                 //if (opt.size()) {
1560                                 //      start = 1;
1561                                 //      at.nucleus()->cell(0) = opt;
1562                                 //}
1563                                 for (InsetMath::idx_type i = start; i < at->nargs(); ++i) {
1564                                         parse(at.nucleus()->cell(i), FLAG_ITEM, m);
1565                                         skipSpaces();
1566                                 }
1567                                 cell->push_back(at);
1568                         }
1569                 }
1570
1571
1572                 if (flags & FLAG_LEAVE) {
1573                         flags &= ~FLAG_LEAVE;
1574                         break;
1575                 }
1576         }
1577 }
1578
1579
1580
1581 } // anonymous namespace
1582
1583
1584 void mathed_parse_cell(MathData & ar, docstring const & str)
1585 {
1586         Parser(str).parse(ar, 0, InsetMath::MATH_MODE);
1587 }
1588
1589
1590 void mathed_parse_cell(MathData & ar, istream & is)
1591 {
1592         Parser(is).parse(ar, 0, InsetMath::MATH_MODE);
1593 }
1594
1595
1596 bool mathed_parse_normal(MathAtom & t, docstring const & str)
1597 {
1598         return Parser(str).parse(t);
1599 }
1600
1601
1602 bool mathed_parse_normal(MathAtom & t, Lexer & lex)
1603 {
1604         return Parser(lex).parse(t);
1605 }
1606
1607
1608 void mathed_parse_normal(InsetMathGrid & grid, docstring const & str)
1609 {
1610         Parser(str).parse1(grid, 0, InsetMath::MATH_MODE, false);
1611 }
1612
1613
1614 void initParser()
1615 {
1616         fill(theCatcode, theCatcode + 128, catOther);
1617         fill(theCatcode + 'a', theCatcode + 'z' + 1, catLetter);
1618         fill(theCatcode + 'A', theCatcode + 'Z' + 1, catLetter);
1619
1620         theCatcode[int('\\')] = catEscape;
1621         theCatcode[int('{')]  = catBegin;
1622         theCatcode[int('}')]  = catEnd;
1623         theCatcode[int('$')]  = catMath;
1624         theCatcode[int('&')]  = catAlign;
1625         theCatcode[int('\n')] = catNewline;
1626         theCatcode[int('#')]  = catParameter;
1627         theCatcode[int('^')]  = catSuper;
1628         theCatcode[int('_')]  = catSub;
1629         theCatcode[int(0x7f)] = catIgnore;
1630         theCatcode[int(' ')]  = catSpace;
1631         theCatcode[int('\t')] = catSpace;
1632         theCatcode[int('\r')] = catNewline;
1633         theCatcode[int('~')]  = catActive;
1634         theCatcode[int('%')]  = catComment;
1635 }
1636
1637
1638 } // namespace lyx