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