src/mathed/math_parser.C

   1 /*
   2  *  File:        math_parser.C
   3  *  Purpose:     Parser for mathed
   4  *  Author:      Alejandro Aguilar Sierra <asierra@servidor.unam.mx>
   5  *  Created:     January 1996
   6  *  Description: Parse LaTeX2e math mode code.
   7  *
   8  *  Dependencies: Xlib, XForms
   9  *
  10  *  Copyright: 1996, Alejandro Aguilar Sierra
  11  *
  12  *   Version: 0.8beta.
  13  *
  14  *   You are free to use and modify this code under the terms of
  15  *   the GNU General Public Licence version 2 or later.
  16  */
  17
  18 /*
  19
  20 If someone desperately needs partial "structures" (such as a few cells of
  21 an array inset or similar) (s)he could uses the following hack as starting
  22 point to write some macros:
  23
  24   \newif\ifcomment
  25   \commentfalse
  26   \ifcomment
  27           \def\makeamptab{\catcode`\&=4\relax}
  28           \def\makeampletter{\catcode`\&=11\relax}
  29     \def\b{\makeampletter\expandafter\makeamptab\bi}
  30     \long\def\bi#1\e{}
  31   \else
  32     \def\b{}\def\e{}
  33   \fi
  34
  35   ...
  36
  37   \[\begin{array}{ccc}
  38    1 & 2\b & 3^2\\
  39    4 & 5\e & 6\\
  40    7 & 8 & 9
  41   \end{array}\]
  42
  43 */
  44
  45
  46 #include <config.h>
  47
  48 #ifdef __GNUG__
  49 #pragma implementation
  50 #endif
  51
  52 #include "math_parser.h"
  53 #include "math_inset.h"
  54 #include "math_arrayinset.h"
  55 #include "math_braceinset.h"
  56 #include "math_boxinset.h"
  57 #include "math_charinset.h"
  58 #include "math_deliminset.h"
  59 #include "math_factory.h"
  60 #include "math_funcinset.h"
  61 #include "math_kerninset.h"
  62 #include "math_macro.h"
  63 #include "math_macrotable.h"
  64 #include "math_macrotemplate.h"
  65 #include "math_hullinset.h"
  66 #include "math_rootinset.h"
  67 #include "math_sizeinset.h"
  68 #include "math_sqrtinset.h"
  69 #include "math_scriptinset.h"
  70 #include "math_specialcharinset.h"
  71 #include "math_sqrtinset.h"
  72 #include "math_support.h"
  73 #include "math_xyarrowinset.h"
  74
  75 #include "lyxlex.h"
  76 #include "debug.h"
  77 #include "support/LAssert.h"
  78 #include "support/lstrings.h"
  79
  80 #include <cctype>
  81 #include <stack>
  82 #include <algorithm>
  83
  84 using std::istream;
  85 using std::ostream;
  86 using std::ios;
  87 using std::endl;
  88 using std::stack;
  89 using std::fill;
  90 using std::vector;
  91
  92 //#define FILEDEBUG
  93
  94
  95 namespace {
  96
  97 bool stared(string const & s)
  98 {
  99         string::size_type const n = s.size();
 100         return n && s[n - 1] == '*';
 101 }
 102
 103
 104 void add(MathArray & ar, char c, MathTextCodes code)
 105 {
 106         ar.push_back(MathAtom(new MathCharInset(c, code)));
 107 }
 108
 109
 110 // These are TeX's catcodes
 111 enum CatCode {
 112         catEscape,     // 0    backslash
 113         catBegin,      // 1    {
 114         catEnd,        // 2    }
 115         catMath,       // 3    $
 116         catAlign,      // 4    &
 117         catNewline,    // 5    ^^M
 118         catParameter,  // 6    #
 119         catSuper,      // 7    ^
 120         catSub,        // 8    _
 121         catIgnore,     // 9
 122         catSpace,      // 10   space
 123         catLetter,     // 11   a-zA-Z
 124         catOther,      // 12   none of the above
 125         catActive,     // 13   ~
 126         catComment,    // 14   %
 127         catInvalid     // 15   <delete>
 128 };
 129
 130 CatCode theCatcode[256];
 131
 132
 133 inline CatCode catcode(unsigned char c)
 134 {
 135         return theCatcode[c];
 136 }
 137
 138
 139 enum {
 140         FLAG_BRACE_LAST = 1 << 1,  //  last closing brace ends the parsing process
 141         FLAG_RIGHT      = 1 << 2,  //  next \\right ends the parsing process
 142         FLAG_END        = 1 << 3,  //  next \\end ends the parsing process
 143         FLAG_BRACK_END  = 1 << 4,  //  next closing bracket ends the parsing process
 144         FLAG_BOX        = 1 << 5,  //  we are in a box
 145         FLAG_ITEM       = 1 << 6,  //  read a (possibly braced token)
 146         FLAG_BLOCK      = 1 << 7,  //  next block ends the parsing process
 147         FLAG_BLOCK2     = 1 << 8,  //  next block2 ends the parsing process
 148         FLAG_LEAVE      = 1 << 9   //  leave the loop at the end
 149 };
 150
 151
 152 void catInit()
 153 {
 154         fill(theCatcode, theCatcode + 256, catOther);
 155         fill(theCatcode + 'a', theCatcode + 'z' + 1, catLetter);
 156         fill(theCatcode + 'A', theCatcode + 'Z' + 1, catLetter);
 157
 158         theCatcode['\\'] = catEscape;
 159         theCatcode['{']  = catBegin;
 160         theCatcode['}']  = catEnd;
 161         theCatcode['$']  = catMath;
 162         theCatcode['&']  = catAlign;
 163         theCatcode['\n'] = catNewline;
 164         theCatcode['#']  = catParameter;
 165         theCatcode['^']  = catSuper;
 166         theCatcode['_']  = catSub;
 167         theCatcode['\7f'] = catIgnore;
 168         theCatcode[' ']  = catSpace;
 169         theCatcode['\t'] = catSpace;
 170         theCatcode['\r'] = catSpace;
 171         theCatcode['~']  = catActive;
 172         theCatcode['%']  = catComment;
 173 }
 174
 175
 176
 177 //
 178 // Helper class for parsing
 179 //
 180
 181 class Token {
 182 public:
 183         ///
 184         Token() : cs_(), char_(0), cat_(catIgnore) {}
 185         ///
 186         Token(char c, CatCode cat) : cs_(), char_(c), cat_(cat) {}
 187         ///
 188         Token(string const & cs) : cs_(cs), char_(0), cat_(catIgnore) {}
 189
 190         ///
 191         string const & cs() const { return cs_; }
 192         ///
 193         CatCode cat() const { return cat_; }
 194         ///
 195         char character() const { return char_; }
 196         ///
 197         string asString() const;
 198         ///
 199         bool isCR() const;
 200
 201 private:
 202         ///
 203         string cs_;
 204         ///
 205         char char_;
 206         ///
 207         CatCode cat_;
 208 };
 209
 210 bool Token::isCR() const
 211 {
 212         return cs_ == "\\" || cs_ == "cr" || cs_ == "crcr";
 213 }
 214
 215 string Token::asString() const
 216 {
 217         return cs_.size() ? cs_ : string(1, char_);
 218 }
 219
 220 // Angus' compiler says these are not needed
 221 //bool operator==(Token const & s, Token const & t)
 222 //{
 223 //      return s.character() == t.character()
 224 //              && s.cat() == t.cat() && s.cs() == t.cs();
 225 //}
 226 //
 227 //bool operator!=(Token const & s, Token const & t)
 228 //{
 229 //      return !(s == t);
 230 //}
 231
 232 ostream & operator<<(ostream & os, Token const & t)
 233 {
 234         if (t.cs().size())
 235                 os << "\\" << t.cs();
 236         else
 237                 os << "[" << t.character() << "," << t.cat() << "]";
 238         return os;
 239 }
 240
 241
 242 class Parser {
 243
 244 public:
 245         ///
 246         Parser(LyXLex & lex);
 247         ///
 248         Parser(istream & is);
 249
 250         ///
 251         bool parse_macro(string & name);
 252         ///
 253         bool parse_normal(MathAtom &);
 254         ///
 255         void parse_into(MathArray & array, unsigned flags, MathTextCodes = LM_TC_MIN);
 256         ///
 257         int lineno() const { return lineno_; }
 258         ///
 259         void putback();
 260
 261 private:
 262         ///
 263         void parse_into1(MathArray & array, unsigned flags, MathTextCodes);
 264         ///
 265         string getArg(char lf, char rf);
 266         ///
 267         char getChar();
 268         ///
 269         void error(string const & msg);
 270         ///
 271         bool parse_lines(MathAtom & t, bool numbered, bool outmost);
 272         /// parses {... & ... \\ ... & ... }
 273         bool parse_lines2(MathAtom & t, bool braced);
 274         /// dump contents to screen
 275         void dump() const;
 276
 277 private:
 278         ///
 279         void tokenize(istream & is);
 280         ///
 281         void tokenize(string const & s);
 282         ///
 283         void skipSpaceTokens(istream & is, char c);
 284         ///
 285         void push_back(Token const & t);
 286         ///
 287         void pop_back();
 288         ///
 289         Token const & prevToken() const;
 290         ///
 291         Token const & nextToken() const;
 292         ///
 293         Token const & getToken();
 294         /// skips spaces if any
 295         void skipSpaces();
 296         /// skips opening brace
 297         void skipBegin();
 298         /// skips closing brace
 299         void skipEnd();
 300         /// counts a sequence of hlines
 301         int readHLines();
 302         ///
 303         void lex(string const & s);
 304         ///
 305         bool good() const;
 306
 307         ///
 308         int lineno_;
 309         ///
 310         vector<Token> tokens_;
 311         ///
 312         unsigned pos_;
 313         ///
 314         bool   curr_num_;
 315         ///
 316         string curr_label_;
 317         ///
 318         string curr_skip_;
 319 };
 320
 321
 322 Parser::Parser(LyXLex & lexer)
 323         : lineno_(lexer.getLineNo()), pos_(0), curr_num_(false)
 324 {
 325         tokenize(lexer.getStream());
 326         lexer.eatLine();
 327 }
 328
 329
 330 Parser::Parser(istream & is)
 331         : lineno_(0), pos_(0), curr_num_(false)
 332 {
 333         tokenize(is);
 334 }
 335
 336
 337 void Parser::push_back(Token const & t)
 338 {
 339         tokens_.push_back(t);
 340 }
 341
 342
 343 void Parser::pop_back()
 344 {
 345         tokens_.pop_back();
 346 }
 347
 348
 349 Token const & Parser::prevToken() const
 350 {
 351         static const Token dummy;
 352         return pos_ > 0 ? tokens_[pos_ - 1] : dummy;
 353 }
 354
 355
 356 Token const & Parser::nextToken() const
 357 {
 358         static const Token dummy;
 359         return good() ? tokens_[pos_] : dummy;
 360 }
 361
 362
 363 Token const & Parser::getToken()
 364 {
 365         static const Token dummy;
 366         //lyxerr << "looking at token " << tokens_[pos_] << " pos: " << pos_ << '\n';
 367         return good() ? tokens_[pos_++] : dummy;
 368 }
 369
 370
 371 void Parser::skipSpaces()
 372 {
 373         while (nextToken().cat() == catSpace)
 374                 getToken();
 375 }
 376
 377
 378 void Parser::skipBegin()
 379 {
 380         if (nextToken().cat() == catBegin)
 381                 getToken();
 382         else
 383                 lyxerr << "'{' expected\n";
 384 }
 385
 386
 387 void Parser::skipEnd()
 388 {
 389         if (nextToken().cat() == catEnd)
 390                 getToken();
 391         else
 392                 lyxerr << "'}' expected\n";
 393 }
 394
 395
 396 int Parser::readHLines()
 397 {
 398         int num = 0;
 399         skipSpaces();
 400         while (nextToken().cs() == "hline") {
 401                 getToken();
 402                 ++num;
 403                 skipSpaces();
 404         }
 405         return num;
 406 }
 407
 408
 409 void Parser::putback()
 410 {
 411         --pos_;
 412 }
 413
 414
 415 bool Parser::good() const
 416 {
 417         return pos_ < tokens_.size();
 418 }
 419
 420
 421 char Parser::getChar()
 422 {
 423         if (!good())
 424                 lyxerr << "The input stream is not well..." << endl;
 425         return tokens_[pos_++].character();
 426 }
 427
 428
 429 string Parser::getArg(char lf, char rg)
 430 {
 431         skipSpaces();
 432
 433         string result;
 434         char c = getChar();
 435
 436         if (c != lf)
 437                 putback();
 438         else
 439                 while ((c = getChar()) != rg && good())
 440                         result += c;
 441
 442         return result;
 443 }
 444
 445
 446 void Parser::tokenize(istream & is)
 447 {
 448         // eat everything up to the next \end_inset or end of stream
 449         // and store it in s for further tokenization
 450         string s;
 451         char c;
 452         while (is.get(c)) {
 453                 s += c;
 454                 if (s.size() >= 10 && s.substr(s.size() - 10) == "\\end_inset") {
 455                         s = s.substr(0, s.size() - 10);
 456                         break;
 457                 }
 458         }
 459
 460         // tokenize buffer
 461         tokenize(s);
 462 }
 463
 464
 465 void Parser::skipSpaceTokens(istream & is, char c)
 466 {
 467         // skip trailing spaces
 468         while (catcode(c) == catSpace || catcode(c) == catNewline)
 469                 if (!is.get(c))
 470                         break;
 471         //lyxerr << "putting back: " << c << "\n";
 472         is.putback(c);
 473 }
 474
 475
 476 void Parser::tokenize(string const & buffer)
 477 {
 478         static bool init_done = false;
 479
 480         if (!init_done) {
 481                 catInit();
 482                 init_done = true;
 483         }
 484
 485         istringstream is(buffer.c_str(), ios::in | ios::binary);
 486
 487         char c;
 488         while (is.get(c)) {
 489                 //lyxerr << "reading c: " << c << "\n";
 490
 491                 switch (catcode(c)) {
 492                         case catNewline: {
 493                                 ++lineno_;
 494                                 is.get(c);
 495                                 if (catcode(c) == catNewline)
 496                                         ; //push_back(Token("par"));
 497                                 else {
 498                                         push_back(Token(' ', catSpace));
 499                                         is.putback(c);
 500                                 }
 501                                 break;
 502                         }
 503
 504                         case catComment: {
 505                                 while (is.get(c) && catcode(c) != catNewline)
 506                                         ;
 507                                 ++lineno_;
 508                                 break;
 509                         }
 510
 511                         case catEscape: {
 512                                 is.get(c);
 513                                 if (!is) {
 514                                         error("unexpected end of input");
 515                                 } else {
 516                                         string s(1, c);
 517                                         if (catcode(c) == catLetter) {
 518                                                 // collect letters
 519                                                 while (is.get(c) && catcode(c) == catLetter)
 520                                                         s += c;
 521                                                 skipSpaceTokens(is, c);
 522                                         }
 523                                         push_back(Token(s));
 524                                 }
 525                                 break;
 526                         }
 527
 528                         case catSuper:
 529                         case catSub: {
 530                                 push_back(Token(c, catcode(c)));
 531                                 is.get(c);
 532                                 skipSpaceTokens(is, c);
 533                                 break;
 534                         }
 535
 536                         case catIgnore: {
 537                                 lyxerr << "ignoring a char: " << int(c) << "\n";
 538                                 break;
 539                         }
 540
 541                         default:
 542                                 push_back(Token(c, catcode(c)));
 543                 }
 544         }
 545
 546 #ifdef FILEDEBUG
 547         dump();
 548 #endif
 549 }
 550
 551
 552 void Parser::dump() const
 553 {
 554         lyxerr << "\nTokens: ";
 555         for (unsigned i = 0; i < tokens_.size(); ++i)
 556                 lyxerr << tokens_[i];
 557         lyxerr << "\n";
 558 }
 559
 560
 561 void Parser::error(string const & msg)
 562 {
 563         lyxerr << "Line ~" << lineno_ << ": Math parse error: " << msg << endl;
 564         dump();
 565         //exit(1);
 566 }
 567
 568
 569
 570 bool Parser::parse_lines(MathAtom & t, bool numbered, bool outmost)
 571 {
 572         MathGridInset * p = t->asGridInset();
 573         if (!p) {
 574                 dump();
 575                 lyxerr << "error in Parser::parse_lines() 1\n";
 576                 return false;
 577         }
 578
 579         // save global variables
 580         bool   const saved_num   = curr_num_;
 581         string const saved_label = curr_label_;
 582
 583         // read initial hlines
 584         p->rowinfo(0).lines_ = readHLines();
 585
 586         for (int row = 0; true; ++row) {
 587                 // reset global variables
 588                 curr_num_   = numbered;
 589                 curr_label_.erase();
 590
 591                 // reading a row
 592                 for (MathInset::col_type col = 0; true; ++col) {
 593                         //lyxerr << "reading cell " << row << " " << col << " "
 594                         // << p->ncols() << "\n";
 595                         //lyxerr << "ncols: " << p->ncols() << "\n";
 596
 597                         if (col >= p->ncols()) {
 598                                 //lyxerr << "adding col " << col << "\n";
 599                                 p->addCol(p->ncols());
 600                         }
 601
 602                         MathArray & ar = p->cell(col + row * p->ncols());
 603                         parse_into(ar, FLAG_BLOCK);
 604                         // remove 'unnecessary' braces:
 605                         if (ar.size() == 1 && ar.back()->asBraceInset())
 606                                 ar = ar.back()->asBraceInset()->cell(0);
 607                         //lyxerr << "ar: " << ar << "\n";
 608
 609                         // break if cell is not followed by an ampersand
 610                         if (nextToken().cat() != catAlign) {
 611                                 //lyxerr << "less cells read than normal in row/col: "
 612                                 //      << row << " " << col << "\n";
 613                                 break;
 614                         }
 615
 616                         // skip the ampersand
 617                         getToken();
 618                 }
 619
 620                 if (outmost) {
 621                         MathHullInset * m = t->asHullInset();
 622                         if (!m) {
 623                                 lyxerr << "error in Parser::parse_lines() 2\n";
 624                                 return false;
 625                         }
 626                         m->numbered(row, curr_num_);
 627                         m->label(row, curr_label_);
 628                         if (curr_skip_.size()) {
 629                                 m->vcrskip(LyXLength(curr_skip_), row);
 630                                 curr_skip_.erase();
 631                         }
 632                 }
 633
 634                 // is a \\ coming?
 635                 if (nextToken().isCR()) {
 636                         // skip the cr-token
 637                         getToken();
 638
 639                         // try to read a length
 640                         //get
 641
 642                         // read hlines for next row
 643                         p->rowinfo(row + 1).lines_ = readHLines();
 644                 }
 645
 646                 // we are finished if the next token is an 'end'
 647                 if (nextToken().cs() == "end") {
 648                         // skip the end-token
 649                         getToken();
 650                         getArg('{','}');
 651
 652                         // leave the 'read a line'-loop
 653                         break;
 654                 }
 655
 656                 // otherwise, we have to start a new row
 657                 p->appendRow();
 658         }
 659
 660         // restore "global" variables
 661         curr_num_   = saved_num;
 662         curr_label_ = saved_label;
 663
 664         return true;
 665 }
 666
 667
 668 bool Parser::parse_lines2(MathAtom & t, bool braced)
 669 {
 670         MathGridInset * p = t->asGridInset();
 671         if (!p) {
 672                 lyxerr << "error in Parser::parse_lines() 1\n";
 673                 return false;
 674         }
 675
 676         for (int row = 0; true; ++row) {
 677                 // reading a row
 678                 for (MathInset::col_type col = 0; true; ++col) {
 679                         //lyxerr << "reading cell " << row << " " << col << " " << p->ncols() << "\n";
 680
 681                         if (col >= p->ncols()) {
 682                                 //lyxerr << "adding col " << col << "\n";
 683                                 p->addCol(p->ncols());
 684                         }
 685
 686                         parse_into(p->cell(col + row * p->ncols()), FLAG_BLOCK2);
 687                         //lyxerr << "read cell: " << p->cell(col + row * p->ncols()) << "\n";
 688
 689                         // break if cell is not followed by an ampersand
 690                         if (nextToken().cat() != catAlign) {
 691                                 //lyxerr << "less cells read than normal in row/col: " << row << " " << col << "\n";
 692                                 break;
 693                         }
 694
 695                         // skip the ampersand
 696                         getToken();
 697                 }
 698
 699                 // is a \\ coming?
 700                 if (nextToken().isCR()) {
 701                         // skip the cr-token
 702                         getToken();
 703                 }
 704
 705                 // we are finished if the next token is the one we expected
 706                 // skip the end-token
 707                 // leave the 'read a line'-loop
 708                 if (braced) {
 709                         if (nextToken().cat() == catEnd) {
 710                                 getToken();
 711                                 break;
 712                         }
 713                 } else {
 714                         if (nextToken().cs() == "end") {
 715                                 getToken();
 716                                 getArg('{','}');
 717                                 break;
 718                         }
 719                 }
 720
 721                 // otherwise, we have to start a new row
 722                 p->appendRow();
 723         }
 724
 725         return true;
 726 }
 727
 728
 729
 730 bool Parser::parse_macro(string & name)
 731 {
 732         name = "{error}";
 733         skipSpaces();
 734
 735         if (getToken().cs() != "newcommand") {
 736                 lyxerr << "\\newcommand expected\n";
 737                 return false;
 738         }
 739
 740         if (getToken().cat() != catBegin) {
 741                 lyxerr << "'{' in \\newcommand expected (1)\n";
 742                 return false;
 743         }
 744
 745         name = getToken().cs();
 746
 747         if (getToken().cat() != catEnd) {
 748                 lyxerr << "'}' expected\n";
 749                 return false;
 750         }
 751
 752         string    arg  = getArg('[', ']');
 753         int       narg = arg.empty() ? 0 : atoi(arg.c_str());
 754
 755         if (getToken().cat() != catBegin) {
 756                 lyxerr << "'{' in \\newcommand expected (2)\n";
 757                 return false;
 758         }
 759
 760         MathArray ar;
 761         parse_into(ar, FLAG_BRACE_LAST);
 762
 763         // we cannot handle recursive stuff at all
 764         MathArray test;
 765         test.push_back(createMathInset(name));
 766         if (ar.contains(test)) {
 767                 lyxerr << "we cannot handle recursive macros at all.\n";
 768                 return false;
 769         }
 770
 771         MathMacroTable::create(name, narg, ar);
 772         return true;
 773 }
 774
 775
 776 bool Parser::parse_normal(MathAtom & matrix)
 777 {
 778         skipSpaces();
 779         Token const & t = getToken();
 780
 781         if (t.cs() == "(") {
 782                 matrix = MathAtom(new MathHullInset(LM_OT_SIMPLE));
 783                 parse_into(matrix->cell(0), 0);
 784                 return true;
 785         }
 786
 787         if (t.cat() == catMath) {
 788                 Token const & n = getToken();
 789                 if (n.cat() == catMath) {
 790                         // TeX's $$...$$ syntax for displayed math
 791                         matrix = MathAtom(new MathHullInset(LM_OT_EQUATION));
 792                         MathHullInset * p = matrix->asHullInset();
 793                         parse_into(p->cell(0), 0);
 794                         p->numbered(0, curr_num_);
 795                         p->label(0, curr_label_);
 796                 } else {
 797                         // simple $...$  stuff
 798                         putback();
 799                         matrix = MathAtom(new MathHullInset(LM_OT_SIMPLE));
 800                         parse_into(matrix->cell(0), 0);
 801                 }
 802                 return true;
 803         }
 804
 805         if (!t.cs().size()) {
 806                 lyxerr << "start of math expected, got '" << t << "'\n";
 807                 return false;
 808         }
 809
 810         string const & cs = t.cs();
 811
 812         if (cs == "[") {
 813                 curr_num_ = 0;
 814                 curr_label_.erase();
 815                 matrix = MathAtom(new MathHullInset(LM_OT_EQUATION));
 816                 MathHullInset * p = matrix->asHullInset();
 817                 parse_into(p->cell(0), 0);
 818                 p->numbered(0, curr_num_);
 819                 p->label(0, curr_label_);
 820                 return true;
 821         }
 822
 823         if (cs != "begin") {
 824                 lyxerr << "'begin' of un-simple math expected, got '" << cs << "'\n";
 825                 return false;
 826         }
 827
 828         string const name = getArg('{', '}');
 829
 830         if (name == "math") {
 831                 matrix = MathAtom(new MathHullInset(LM_OT_SIMPLE));
 832                 parse_into(matrix->cell(0), 0);
 833                 return true;
 834         }
 835
 836         if (name == "equation" || name == "equation*" || name == "displaymath") {
 837                 curr_num_ = (name == "equation");
 838                 curr_label_.erase();
 839                 matrix = MathAtom(new MathHullInset(LM_OT_EQUATION));
 840                 MathHullInset * p = matrix->asHullInset();
 841                 parse_into(p->cell(0), FLAG_END);
 842                 p->numbered(0, curr_num_);
 843                 p->label(0, curr_label_);
 844                 return true;
 845         }
 846
 847         if (name == "eqnarray" || name == "eqnarray*") {
 848                 matrix = MathAtom(new MathHullInset(LM_OT_EQNARRAY));
 849                 return parse_lines(matrix, !stared(name), true);
 850         }
 851
 852         if (name == "align" || name == "align*") {
 853                 matrix = MathAtom(new MathHullInset(LM_OT_ALIGN));
 854                 return parse_lines(matrix, !stared(name), true);
 855         }
 856
 857         if (name == "alignat" || name == "alignat*") {
 858                 int nc = 2 * atoi(getArg('{', '}').c_str());
 859                 matrix = MathAtom(new MathHullInset(LM_OT_ALIGNAT, nc));
 860                 return parse_lines(matrix, !stared(name), true);
 861         }
 862
 863         if (name == "xalignat" || name == "xalignat*") {
 864                 int nc = 2 * atoi(getArg('{', '}').c_str());
 865                 matrix = MathAtom(new MathHullInset(LM_OT_XALIGNAT, nc));
 866                 return parse_lines(matrix, !stared(name), true);
 867         }
 868
 869         if (name == "xxalignat") {
 870                 int nc = 2 * atoi(getArg('{', '}').c_str());
 871                 matrix = MathAtom(new MathHullInset(LM_OT_XXALIGNAT, nc));
 872                 return parse_lines(matrix, !stared(name), true);
 873         }
 874
 875         if (name == "multline" || name == "multline*") {
 876                 matrix = MathAtom(new MathHullInset(LM_OT_MULTLINE));
 877                 return parse_lines(matrix, !stared(name), true);
 878         }
 879
 880         if (name == "gather" || name == "gather*") {
 881                 matrix = MathAtom(new MathHullInset(LM_OT_GATHER));
 882                 return parse_lines(matrix, !stared(name), true);
 883         }
 884
 885         lyxerr[Debug::MATHED] << "1: unknown math environment: " << name << "\n";
 886         lyxerr << "1: unknown math environment: " << name << "\n";
 887         return false;
 888 }
 889
 890
 891 void Parser::parse_into(MathArray & array, unsigned flags, MathTextCodes code)
 892 {
 893         parse_into1(array, flags, code);
 894         // remove 'unnecessary' braces:
 895         if (array.size() == 1 && array.back()->asBraceInset()) {
 896                 lyxerr << "extra braces removed\n";
 897                 array = array.back()->asBraceInset()->cell(0);
 898         }
 899 }
 900
 901
 902 void Parser::parse_into1(MathArray & array, unsigned flags, MathTextCodes code)
 903 {
 904         bool panic  = false;
 905         int  limits = 0;
 906
 907         while (good()) {
 908                 Token const & t = getToken();
 909
 910 #ifdef FILEDEBUG
 911                 lyxerr << "t: " << t << " flags: " << flags << "\n";
 912                 //array.dump();
 913                 lyxerr << "\n";
 914 #endif
 915
 916                 if (flags & FLAG_ITEM) {
 917                         flags &= ~FLAG_ITEM;
 918                         if (t.cat() == catBegin) {
 919                                 // skip the brace and collect everything to the next matching
 920                                 // closing brace
 921                                 flags |= FLAG_BRACE_LAST;
 922                                 continue;
 923                         } else {
 924                                 // handle only this single token, leave the loop if done
 925                                 flags |= FLAG_LEAVE;
 926                         }
 927                 }
 928
 929                 if (flags & FLAG_BLOCK) {
 930                         if (t.cat() == catAlign || t.isCR() || t.cs() == "end") {
 931                                 putback();
 932                                 return;
 933                         }
 934                 }
 935
 936                 if (flags & FLAG_BLOCK2) {
 937                         if (t.cat() == catAlign || t.isCR() || t.cs() == "end"
 938                                         || t.cat() == catEnd) {
 939                                 putback();
 940                                 return;
 941                         }
 942                 }
 943
 944                 //
 945                 // cat codes
 946                 //
 947                 if (t.cat() == catMath) {
 948                         if (flags & FLAG_BOX) {
 949                                 // we are inside an mbox, so opening new math is allowed
 950                                 array.push_back(MathAtom(new MathHullInset(LM_OT_SIMPLE)));
 951                                 parse_into(array.back()->cell(0), 0);
 952                         } else {
 953                                 // otherwise this is the end of the formula
 954                                 break;
 955                         }
 956                 }
 957
 958                 else if (t.cat() == catLetter)
 959                         add(array, t.character(), code);
 960
 961                 else if (t.cat() == catSpace && code == LM_TC_TEXTRM)
 962                         add(array, t.character(), code);
 963
 964                 else if (t.cat() == catParameter) {
 965                         Token const & n = getToken();
 966                         array.push_back(MathAtom(new MathMacroArgument(n.character()-'0', code)));
 967                 }
 968
 969                 else if (t.cat() == catBegin) {
 970                         MathArray ar;
 971                         parse_into(ar, FLAG_BRACE_LAST);
 972 #ifndef WITH_WARNINGS
 973 #warning this might be wrong in general!
 974 #endif
 975                         // ignore braces around simple items
 976                         if ((ar.size() == 1 && !ar.front()->needsBraces()
 977        || (ar.size() == 2 && !ar.front()->needsBraces()
 978                                             && ar.back()->asScriptInset()))
 979        || (ar.size() == 0 && array.size() == 0))
 980                         {
 981                                 array.push_back(ar);
 982                         } else {
 983                                 array.push_back(MathAtom(new MathBraceInset));
 984                                 array.back()->cell(0).swap(ar);
 985                         }
 986                 }
 987
 988                 else if (t.cat() == catEnd) {
 989                         if (flags & FLAG_BRACE_LAST)
 990                                 return;
 991                         lyxerr << "found '}' unexpectedly, array: '" << array << "'\n";
 992                         //lyxerr << "found '}' unexpectedly\n";
 993                         lyx::Assert(0);
 994                         add(array, '}', LM_TC_TEX);
 995                 }
 996
 997                 else if (t.cat() == catAlign) {
 998                         lyxerr << "found tab unexpectedly, array: '" << array << "'\n";
 999                         //lyxerr << "found tab unexpectedly\n";
1000                         add(array, '&', LM_TC_TEX);
1001                 }
1002
1003                 else if (t.cat() == catSuper || t.cat() == catSub) {
1004                         bool up = (t.cat() == catSuper);
1005                         MathScriptInset * p = 0;
1006                         if (array.size())
1007                                 p = array.back()->asScriptInset();
1008                         if (!p || p->has(up)) {
1009                                 array.push_back(MathAtom(new MathScriptInset(up)));
1010                                 p = array.back()->asScriptInset();
1011                         }
1012                         p->ensure(up);
1013                         parse_into(p->cell(up), FLAG_ITEM);
1014                         p->limits(limits);
1015                         limits = 0;
1016                 }
1017
1018                 else if (t.character() == ']' && (flags & FLAG_BRACK_END))
1019                         return;
1020
1021                 else if (t.cat() == catOther)
1022                         add(array, t.character(), code);
1023
1024                 //
1025                 // control sequences
1026                 //
1027                 else if (t.cs() == "protect")
1028                         // ignore \\protect, will be re-added during output
1029                         ;
1030
1031                 else if (t.cs() == "end")
1032                         break;
1033
1034                 else if (t.cs() == ")")
1035                         break;
1036
1037                 else if (t.cs() == "]")
1038                         break;
1039
1040                 else if (t.cs() == "\\") {
1041                         curr_skip_ = getArg('[', ']');
1042                         //lyxerr << "found newline unexpectedly, array: '" << array << "'\n";
1043                         lyxerr << "found newline unexpectedly\n";
1044                         array.push_back(createMathInset("\\"));
1045                 }
1046
1047                 else if (t.cs() == "limits")
1048                         limits = 1;
1049
1050                 else if (t.cs() == "nolimits")
1051                         limits = -1;
1052
1053                 else if (t.cs() == "nonumber")
1054                         curr_num_ = false;
1055
1056                 else if (t.cs() == "number")
1057                         curr_num_ = true;
1058
1059                 else if (t.cs() == "sqrt") {
1060                         char c = getChar();
1061                         if (c == '[') {
1062                                 array.push_back(MathAtom(new MathRootInset));
1063                                 parse_into(array.back()->cell(0), FLAG_BRACK_END);
1064                                 parse_into(array.back()->cell(1), FLAG_ITEM);
1065                         } else {
1066                                 putback();
1067                                 array.push_back(MathAtom(new MathSqrtInset));
1068                                 parse_into(array.back()->cell(0), FLAG_ITEM);
1069                         }
1070                 }
1071
1072                 else if (t.cs() == "left") {
1073                         string l = getToken().asString();
1074                         MathArray ar;
1075                         parse_into(ar, FLAG_RIGHT);
1076                         string r = getToken().asString();
1077                         MathAtom dl(new MathDelimInset(l, r));
1078                         dl->cell(0) = ar;
1079                         array.push_back(dl);
1080                 }
1081
1082                 else if (t.cs() == "right") {
1083                         if (!(flags & FLAG_RIGHT)) {
1084                                 //lyxerr << "got so far: '" << array << "'\n";
1085                                 error("Unmatched right delimiter");
1086                         }
1087                         return;
1088                 }
1089
1090                 else if (t.cs() == "begin") {
1091                         string const name = getArg('{', '}');
1092                         if (name == "array" || name == "subarray") {
1093                                 string const valign = getArg('[', ']') + 'c';
1094                                 string const halign = getArg('{', '}');
1095                                 array.push_back(MathAtom(new MathArrayInset(name, valign[0], halign)));
1096                                 parse_lines(array.back(), false, false);
1097                         } else if (name == "split" || name == "cases" ||
1098                                          name == "gathered" || name == "aligned") {
1099                                 array.push_back(createMathInset(name));
1100                                 parse_lines(array.back(), false, false);
1101                         } else if (name == "matrix"  || name == "pmatrix" || name == "bmatrix" ||
1102                                          name == "vmatrix" || name == "Vmatrix") {
1103                                 array.push_back(createMathInset(name));
1104                                 parse_lines2(array.back(), false);
1105                         } else
1106                                 lyxerr << "unknow math inset begin '" << name << "'\n";
1107                 }
1108
1109                 else if (t.cs() == "kern") {
1110 #ifdef WITH_WARNINGS
1111 #warning A hack...
1112 #endif
1113                         string s;
1114                         while (1) {
1115                                 Token const & t = getToken();
1116                                 if (!good()) {
1117                                         putback();
1118                                         break;
1119                                 }
1120                                 s += t.character();
1121                                 if (isValidLength(s))
1122                                         break;
1123                         }
1124                         array.push_back(MathAtom(new MathKernInset(s)));
1125                 }
1126
1127                 else if (t.cs() == "label") {
1128                         curr_label_ = getArg('{', '}');
1129                 }
1130
1131                 else if (t.cs() == "choose" || t.cs() == "over" || t.cs() == "atop") {
1132                         MathAtom p = createMathInset(t.cs());
1133                         array.swap(p->cell(0));
1134                         parse_into(p->cell(1), flags, code);
1135                         array.push_back(p);
1136                         return;
1137                 }
1138
1139                 else if (t.cs() == "substack") {
1140                         array.push_back(createMathInset(t.cs()));
1141                         skipBegin();
1142                         parse_lines2(array.back(), true);
1143                 }
1144
1145                 else if (t.cs() == "xymatrix") {
1146                         array.push_back(createMathInset(t.cs()));
1147                         skipBegin();
1148                         parse_lines2(array.back(), true);
1149                 }
1150
1151 #if 0
1152                 // Disabled
1153                 else if (1 && t.cs() == "ar") {
1154                         MathXYArrowInset * p = new MathXYArrowInset;
1155
1156                         // try to read target
1157                         char c = getChar();
1158                         if (c == '[') {
1159                                 parse_into(p->cell(0), FLAG_BRACK_END);
1160                                 //lyxerr << "read target: " << p->cell(0) << "\n";
1161                         } else {
1162                                 putback();
1163                         }
1164
1165                         // try to read label
1166                         if (nextToken().cat() == catSuper || nextToken().cat() == catSub) {
1167                                 p->up_ = nextToken().cat() == catSuper;
1168                                 getToken();
1169                                 parse_into(p->cell(1), FLAG_ITEM);
1170                                 //lyxerr << "read label: " << p->cell(1) << "\n";
1171                         }
1172
1173                         array.push_back(MathAtom(p));
1174                         //lyxerr << "read array: " << array << "\n";
1175                 }
1176 #endif
1177
1178 #if 0
1179                 else if (t.cs() == "mbox" || t.cs() == "text") {
1180                         //array.push_back(createMathInset(t.cs()));
1181                         array.push_back(MathAtom(new MathBoxInset(t.cs())));
1182                         // slurp in the argument of mbox
1183
1184                         MathBoxInset * p = array.back()->asBoxInset();
1185                         //lyx::assert(p);
1186                 }
1187 #endif
1188
1189
1190                 else if (t.cs().size()) {
1191                         latexkeys const * l = in_word_set(t.cs());
1192                         if (l) {
1193                                 if (l->token == LM_TK_FONT) {
1194                                         //lyxerr << "starting font\n";
1195                                         //CatCode catSpaceSave = theCatcode[' '];
1196                                         //if (l->id == LM_TC_TEXTRM) {
1197                                         //      // temporarily change catcode
1198                                         //      theCatcode[' '] = catLetter;
1199                                         //}
1200
1201                                         MathArray ar;
1202                                         parse_into(ar, FLAG_ITEM, static_cast<MathTextCodes>(l->id));
1203                                         array.push_back(ar);
1204
1205                                         // undo catcode changes
1206                                         ////theCatcode[' '] = catSpaceSave;
1207                                         //lyxerr << "ending font\n";
1208                                 }
1209
1210                                 else if (l->token == LM_TK_OLDFONT) {
1211                                         code = static_cast<MathTextCodes>(l->id);
1212                                 }
1213
1214                                 else if (l->token == LM_TK_BOX) {
1215                                         MathAtom p = createMathInset(t.cs());
1216                                         parse_into(p->cell(0), FLAG_ITEM | FLAG_BOX, LM_TC_BOX);
1217                                         array.push_back(p);
1218                                 }
1219
1220                                 else if (l->token == LM_TK_STY) {
1221                                         MathAtom p = createMathInset(t.cs());
1222                                         parse_into(p->cell(0), flags, code);
1223                                         array.push_back(p);
1224                                         return;
1225                                 }
1226
1227                                 else {
1228                                         MathAtom p = createMathInset(t.cs());
1229                                         for (MathInset::idx_type i = 0; i < p->nargs(); ++i)
1230                                                 parse_into(p->cell(i), FLAG_ITEM);
1231                                         array.push_back(p);
1232                                 }
1233                         }
1234
1235                         else {
1236                                 MathAtom p = createMathInset(t.cs());
1237                                 for (MathInset::idx_type i = 0; i < p->nargs(); ++i)
1238                                         parse_into(p->cell(i), FLAG_ITEM);
1239                                 array.push_back(p);
1240                         }
1241                 }
1242
1243
1244                 if (flags & FLAG_LEAVE) {
1245                         flags &= ~FLAG_LEAVE;
1246                         break;
1247                 }
1248         }
1249
1250         if (panic) {
1251                 lyxerr << " Math Panic, expect problems!\n";
1252                 //   Search for the end command.
1253                 Token t;
1254                 do {
1255                         t = getToken();
1256                 } while (good() && t.cs() != "end");
1257         }
1258 }
1259
1260
1261
1262 } // anonymous namespace
1263
1264
1265 void mathed_parse_cell(MathArray & ar, string const & str)
1266 {
1267         istringstream is(str.c_str());
1268         mathed_parse_cell(ar, is);
1269 }
1270
1271
1272 void mathed_parse_cell(MathArray & ar, istream & is)
1273 {
1274         Parser(is).parse_into(ar, 0);
1275 }
1276
1277
1278
1279 bool mathed_parse_macro(string & name, string const & str)
1280 {
1281         istringstream is(str.c_str());
1282         Parser parser(is);
1283         return parser.parse_macro(name);
1284 }
1285
1286 bool mathed_parse_macro(string & name, istream & is)
1287 {
1288         Parser parser(is);
1289         return parser.parse_macro(name);
1290 }
1291
1292 bool mathed_parse_macro(string & name, LyXLex & lex)
1293 {
1294         Parser parser(lex);
1295         return parser.parse_macro(name);
1296 }
1297
1298
1299
1300 bool mathed_parse_normal(MathAtom & t, string const & str)
1301 {
1302         istringstream is(str.c_str());
1303         Parser parser(is);
1304         return parser.parse_normal(t);
1305 }
1306
1307 bool mathed_parse_normal(MathAtom & t, istream & is)
1308 {
1309         Parser parser(is);
1310         return parser.parse_normal(t);
1311 }
1312
1313 bool mathed_parse_normal(MathAtom & t, LyXLex & lex)
1314 {
1315         Parser parser(lex);
1316         return parser.parse_normal(t);
1317 }