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