src/tex2lyx/table.C

   1 /**
   2  * \file table.C
   3  * This file is part of LyX, the document processor.
   4  * Licence details can be found in the file COPYING.
   5  *
   6  * \author André Pönitz
   7  * \author Jean-Marc Lasgouttes
   8  * \author Georg Baum
   9  *
  10  * Full author contact details are available in file CREDITS.
  11  */
  12
  13 // {[(
  14
  15 #include <config.h>
  16
  17 #include "tex2lyx.h"
  18
  19 #include "support/convert.h"
  20 #include "support/lstrings.h"
  21
  22 #include <cctype>
  23 #include <fstream>
  24 #include <iostream>
  25 #include <sstream>
  26 #include <vector>
  27 #include <map>
  28
  29 using std::cerr;
  30 using std::endl;
  31 using std::istringstream;
  32 using std::ostream;
  33 using std::ostringstream;
  34 using std::string;
  35 using std::vector;
  36
  37
  38 // filled in preamble.C
  39 std::map<char, int> special_columns;
  40
  41
  42 namespace {
  43
  44 class ColInfo {
  45 public:
  46         ColInfo() : align('n'), valign('n'), rightlines(0), leftlines(0) {}
  47         /// column alignment
  48         char align;
  49         /// vertical alignment
  50         char valign;
  51         /// column width
  52         string width;
  53         /// special column alignment
  54         string special;
  55         /// number of lines on the right
  56         int rightlines;
  57         /// number of lines on the left
  58         int leftlines;
  59 };
  60
  61
  62 /// row type for longtables
  63 enum LTRowType
  64 {
  65         /// normal row
  66         LT_NORMAL,
  67         /// part of head
  68         LT_HEAD,
  69         /// part of head on first page
  70         LT_FIRSTHEAD,
  71         /// part of foot
  72         LT_FOOT,
  73         /// part of foot on last page
  74         LT_LASTFOOT
  75 };
  76
  77
  78 class RowInfo {
  79 public:
  80         RowInfo() : topline(false), bottomline(false), type(LT_NORMAL),
  81                     newpage(false) {}
  82         /// horizontal line above
  83         bool topline;
  84         /// horizontal line below
  85         bool bottomline;
  86         /// These are for longtabulars only
  87         /// row type (head, foot, firsthead etc.)
  88         LTRowType type;
  89         /// row for a pagebreak
  90         bool newpage;
  91 };
  92
  93
  94 enum Multicolumn {
  95         /// A normal cell
  96         CELL_NORMAL = 0,
  97         /// A multicolumn cell. The number of columns is <tt>1 + number
  98         /// of CELL_PART_OF_MULTICOLUMN cells</tt> that follow directly
  99         CELL_BEGIN_OF_MULTICOLUMN,
 100         /// This is a dummy cell (part of a multicolumn cell)
 101         CELL_PART_OF_MULTICOLUMN
 102 };
 103
 104
 105 class CellInfo {
 106 public:
 107         CellInfo() : multi(CELL_NORMAL), align('n'), valign('n'),
 108                      leftlines(0), rightlines(0), topline(false),
 109                      bottomline(false), rotate(false) {}
 110         /// cell content
 111         string content;
 112         /// multicolumn flag
 113         Multicolumn multi;
 114         /// cell alignment
 115         char align;
 116         /// vertical cell alignment
 117         char valign;
 118         /// number of lines on the left
 119         int leftlines;
 120         /// number of lines on the right
 121         int rightlines;
 122         /// do we have a line above?
 123         bool topline;
 124         /// do we have a line below?
 125         bool bottomline;
 126         /// is the cell rotated?
 127         bool rotate;
 128         /// width for multicolumn cells
 129         string width;
 130         /// special formatting for multicolumn cells
 131         string special;
 132 };
 133
 134
 135 /// translate a horizontal alignment (as stored in ColInfo and CellInfo) to LyX
 136 inline char const * verbose_align(char c)
 137 {
 138         switch (c) {
 139         case 'c':
 140                 return "center";
 141         case 'r':
 142                 return "right";
 143         case 'l':
 144                 return "left";
 145         default:
 146                 return "none";
 147         }
 148 }
 149
 150
 151 /// translate a vertical alignment (as stored in ColInfo and CellInfo) to LyX
 152 inline char const * verbose_valign(char c)
 153 {
 154         // The default value for no special alignment is "top".
 155         switch (c) {
 156         case 'm':
 157                 return "middle";
 158         case 'b':
 159                 return "bottom";
 160         case 'p':
 161         default:
 162                 return "top";
 163         }
 164 }
 165
 166
 167 // stripped down from tabluar.C. We use it currently only for bools and
 168 // strings
 169 string const write_attribute(string const & name, bool const & b)
 170 {
 171         // we write only true attribute values so we remove a bit of the
 172         // file format bloat for tabulars.
 173         return b ? ' ' + name + "=\"true\"" : string();
 174 }
 175
 176
 177 string const write_attribute(string const & name, string const & s)
 178 {
 179         return s.empty() ? string() : ' ' + name + "=\"" + s + '"';
 180 }
 181
 182
 183 /*! rather brutish way to code table structure in a string:
 184
 185 \verbatim
 186   \begin{tabular}{ccc}
 187     1 & 2 & 3\\ \hline
 188     \multicolumn{2}{c}{4} & 5 //
 189     6 & 7 \\
 190     8 \endhead
 191   \end{tabular}
 192 \endverbatim
 193
 194  gets "translated" to:
 195
 196 \verbatim
 197          HLINE 1                     TAB 2 TAB 3 HLINE          HLINE LINE
 198   \hline HLINE \multicolumn{2}{c}{4} TAB 5       HLINE          HLINE LINE
 199          HLINE 6                     TAB 7       HLINE          HLINE LINE
 200          HLINE 8                                 HLINE \endhead HLINE LINE
 201 \endverbatim
 202  */
 203
 204 char const TAB   = '\001';
 205 char const LINE  = '\002';
 206 char const HLINE = '\004';
 207
 208
 209 /*!
 210  * Move the information in leftlines, rightlines, align and valign to the
 211  * special field. This is necessary if the special field is not empty,
 212  * because LyX ignores leftlines, rightlines, align and valign in this case.
 213  */
 214 void ci2special(ColInfo & ci)
 215 {
 216         if (ci.width.empty() && ci.align == 'n')
 217                 // The alignment setting is already in special, since
 218                 // handle_colalign() never stores ci with these settings
 219                 // and ensures that leftlines == 0 and rightlines == 0 in
 220                 // this case.
 221                 return;
 222
 223         if (!ci.width.empty()) {
 224                 switch (ci.align) {
 225                 case 'l':
 226                         ci.special += ">{\\raggedright}";
 227                         break;
 228                 case 'r':
 229                         ci.special += ">{\\raggedleft}";
 230                         break;
 231                 case 'c':
 232                         ci.special += ">{\\centering}";
 233                         break;
 234                 }
 235                 if (ci.valign == 'n')
 236                         ci.special += 'p';
 237                 else
 238                         ci.special += ci.valign;
 239                 ci.special += '{' + ci.width + '}';
 240                 ci.width.erase();
 241         } else
 242                 ci.special += ci.align;
 243
 244         for (int i = 0; i < ci.leftlines; ++i)
 245                 ci.special.insert(0, "|");
 246         for (int i = 0; i < ci.rightlines; ++i)
 247                 ci.special += '|';
 248         ci.leftlines = 0;
 249         ci.rightlines = 0;
 250         ci.align = 'n';
 251         ci.valign = 'n';
 252 }
 253
 254
 255 /*!
 256  * Handle column specifications for tabulars and multicolumns.
 257  * The next token of the parser \p p must be an opening brace, and we read
 258  * everything until the matching closing brace.
 259  * The resulting column specifications are filled into \p colinfo. This is
 260  * in an intermediate form. fix_colalign() makes it suitable for LyX output.
 261  */
 262 void handle_colalign(Parser & p, vector<ColInfo> & colinfo,
 263                      ColInfo const & start)
 264 {
 265         if (p.get_token().cat() != catBegin)
 266                 cerr << "Wrong syntax for table column alignment.\n"
 267                         "Expected '{', got '" << p.curr_token().asInput()
 268                      << "'.\n";
 269
 270         ColInfo next = start;
 271         for (Token t = p.get_token(); p.good() && t.cat() != catEnd;
 272              t = p.get_token()) {
 273 #ifdef FILEDEBUG
 274                 cerr << "t: " << t << "  c: '" << t.character() << "'\n";
 275 #endif
 276
 277                 // We cannot handle comments here
 278                 if (t.cat() == catComment) {
 279                         if (t.cs().empty()) {
 280                                 // "%\n" combination
 281                                 p.skip_spaces();
 282                         } else
 283                                 cerr << "Ignoring comment: " << t.asInput();
 284                         continue;
 285                 }
 286
 287                 switch (t.character()) {
 288                         case 'c':
 289                         case 'l':
 290                         case 'r':
 291                                 // new column, horizontal aligned
 292                                 next.align = t.character();
 293                                 if (!next.special.empty())
 294                                         ci2special(next);
 295                                 colinfo.push_back(next);
 296                                 next = ColInfo();
 297                                 break;
 298                         case 'p':
 299                         case 'b':
 300                         case 'm':
 301                                 // new column, vertical aligned box
 302                                 next.valign = t.character();
 303                                 next.width = p.verbatim_item();
 304                                 if (!next.special.empty())
 305                                         ci2special(next);
 306                                 colinfo.push_back(next);
 307                                 next = ColInfo();
 308                                 break;
 309                         case '|':
 310                                 // vertical rule
 311                                 if (colinfo.empty()) {
 312                                         if (next.special.empty())
 313                                                 ++next.leftlines;
 314                                         else
 315                                                 next.special += '|';
 316                                 } else if (colinfo.back().special.empty())
 317                                         ++colinfo.back().rightlines;
 318                                 else if (next.special.empty())
 319                                         ++next.leftlines;
 320                                 else
 321                                         colinfo.back().special += '|';
 322                                 break;
 323                         case '>': {
 324                                 // text before the next column
 325                                 string const s = trim(p.verbatim_item());
 326                                 if (next.special.empty() &&
 327                                     next.align == 'n') {
 328                                         // Maybe this can be converted to a
 329                                         // horizontal alignment setting for
 330                                         // fixed width columns
 331                                         if (s == "\\raggedleft")
 332                                                 next.align = 'r';
 333                                         else if (s == "\\raggedright")
 334                                                 next.align = 'l';
 335                                         else if (s == "\\centering")
 336                                                 next.align = 'c';
 337                                         else
 338                                                 next.special = ">{" + s + '}';
 339                                 } else
 340                                         next.special += ">{" + s + '}';
 341                                 break;
 342                         }
 343                         case '<': {
 344                                 // text after the last column
 345                                 string const s = trim(p.verbatim_item());
 346                                 if (colinfo.empty())
 347                                         // This is not possible in LaTeX.
 348                                         cerr << "Ignoring separator '<{"
 349                                              << s << "}'." << endl;
 350                                 else {
 351                                         ColInfo & ci = colinfo.back();
 352                                         ci2special(ci);
 353                                         ci.special += "<{" + s + '}';
 354                                 }
 355                                 break;
 356                         }
 357                         case '*': {
 358                                 // *{n}{arg} means 'n' columns of type 'arg'
 359                                 string const num = p.verbatim_item();
 360                                 string const arg = p.verbatim_item();
 361                                 size_t const n = convert<unsigned int>(num);
 362                                 if (!arg.empty() && n > 0) {
 363                                         string s("{");
 364                                         for (size_t i = 0; i < n; ++i)
 365                                                 s += arg;
 366                                         s += '}';
 367                                         Parser p2(s);
 368                                         handle_colalign(p2, colinfo, next);
 369                                         next = ColInfo();
 370                                 } else {
 371                                         cerr << "Ignoring column specification"
 372                                                 " '*{" << num << "}{"
 373                                              << arg << "}'." << endl;
 374                                 }
 375                                 break;
 376                         }
 377                         case '@':
 378                                 // text instead of the column spacing
 379                         case '!':
 380                                 // text in addition to the column spacing
 381                                 next.special += t.character();
 382                                 next.special += '{' + p.verbatim_item() + '}';
 383                                 break;
 384                         default:
 385                                 // try user defined column types
 386                                 if (special_columns.find(t.character()) !=
 387                                     special_columns.end()) {
 388                                         ci2special(next);
 389                                         next.special += t.character();
 390                                         int const nargs =
 391                                                 special_columns[t.character()];
 392                                         for (int i = 0; i < nargs; ++i)
 393                                                 next.special += '{' +
 394                                                         p.verbatim_item() +
 395                                                         '}';
 396                                         colinfo.push_back(next);
 397                                         next = ColInfo();
 398                                 } else
 399                                         cerr << "Ignoring column specification"
 400                                                 " '" << t << "'." << endl;
 401                                 break;
 402                         }
 403         }
 404
 405         // Maybe we have some column separators that need to be added to the
 406         // last column?
 407         ci2special(next);
 408         if (!next.special.empty()) {
 409                 ColInfo & ci = colinfo.back();
 410                 ci2special(ci);
 411                 ci.special += next.special;
 412                 next.special.erase();
 413         }
 414 }
 415
 416
 417 /*!
 418  * Move the left and right lines and alignment settings of the column \p ci
 419  * to the special field if necessary.
 420  */
 421 void fix_colalign(ColInfo & ci)
 422 {
 423         if (ci.leftlines > 1 || ci.rightlines > 1)
 424                 ci2special(ci);
 425 }
 426
 427
 428 /*!
 429  * LyX can't handle more than one vertical line at the left or right side
 430  * of a column.
 431  * This function moves the left and right lines and alignment settings of all
 432  * columns in \p colinfo to the special field if necessary.
 433  */
 434 void fix_colalign(vector<ColInfo> & colinfo)
 435 {
 436         // Try to move extra leftlines to the previous column.
 437         // We do this only if both special fields are empty, otherwise we
 438         // can't tell wether the result will be the same.
 439         for (size_t col = 0; col < colinfo.size(); ++col) {
 440                 if (colinfo[col].leftlines > 1 &&
 441                     colinfo[col].special.empty() && col > 0 &&
 442                     colinfo[col - 1].rightlines == 0 &&
 443                     colinfo[col - 1].special.empty()) {
 444                         ++colinfo[col - 1].rightlines;
 445                         --colinfo[col].leftlines;
 446                 }
 447         }
 448         // Try to move extra rightlines to the next column
 449         for (size_t col = 0; col < colinfo.size(); ++col) {
 450                 if (colinfo[col].rightlines > 1 &&
 451                     colinfo[col].special.empty() &&
 452                     col < colinfo.size() - 1 &&
 453                     colinfo[col + 1].leftlines == 0 &&
 454                     colinfo[col + 1].special.empty()) {
 455                         ++colinfo[col + 1].leftlines;
 456                         --colinfo[col].rightlines;
 457                 }
 458         }
 459         // Move the lines and alignment settings to the special field if
 460         // necessary
 461         for (size_t col = 0; col < colinfo.size(); ++col)
 462                 fix_colalign(colinfo[col]);
 463 }
 464
 465
 466 /*!
 467  * Parse hlines and similar stuff.
 468  * \returns wether the token \p t was parsed
 469  */
 470 bool parse_hlines(Parser & p, Token const & t, string & hlines,
 471                   bool is_long_tabular)
 472 {
 473         BOOST_ASSERT(t.cat() == catEscape);
 474
 475         if (t.cs() == "hline")
 476                 hlines += "\\hline";
 477
 478         else if (t.cs() == "cline")
 479                 hlines += "\\cline{" + p.verbatim_item() + '}';
 480
 481         else if (is_long_tabular && t.cs() == "newpage")
 482                 hlines += "\\newpage";
 483
 484         else
 485                 return false;
 486
 487         return true;
 488 }
 489
 490
 491 /// Position in a row
 492 enum RowPosition {
 493         /// At the very beginning, before the first token
 494         ROW_START,
 495         /// After the first token and before any column token
 496         IN_HLINES_START,
 497         /// After the first column token. Comments and whitespace are only
 498         /// treated as tokens in this position
 499         IN_COLUMNS,
 500         /// After the first non-column token at the end
 501         IN_HLINES_END
 502 };
 503
 504
 505 /*!
 506  * Parse table structure.
 507  * We parse tables in a two-pass process: This function extracts the table
 508  * structure (rows, columns, hlines etc.), but does not change the cell
 509  * content. The cell content is parsed in a second step in handle_tabular().
 510  */
 511 void parse_table(Parser & p, ostream & os, bool is_long_tabular,
 512                  RowPosition & pos, unsigned flags)
 513 {
 514         // table structure commands such as \hline
 515         string hlines;
 516
 517         // comments that occur at places where we can't handle them
 518         string comments;
 519
 520         while (p.good()) {
 521                 Token const & t = p.get_token();
 522
 523 #ifdef FILEDEBUG
 524                 cerr << "t: " << t << " flags: " << flags << "\n";
 525 #endif
 526
 527                 // comments and whitespace in hlines
 528                 switch (pos) {
 529                 case ROW_START:
 530                 case IN_HLINES_START:
 531                 case IN_HLINES_END:
 532                         if (t.cat() == catComment) {
 533                                 if (t.cs().empty())
 534                                         // line continuation
 535                                         p.skip_spaces();
 536                                 else
 537                                         // We can't handle comments here,
 538                                         // store them for later use
 539                                         comments += t.asInput();
 540                                 continue;
 541                         } else if (t.cat() == catSpace ||
 542                                    t.cat() == catNewline) {
 543                                 // whitespace is irrelevant here, we
 544                                 // need to recognize hline stuff
 545                                 p.skip_spaces();
 546                                 continue;
 547                         }
 548                         break;
 549                 case IN_COLUMNS:
 550                         break;
 551                 }
 552
 553                 // We need to handle structure stuff first in order to
 554                 // determine wether we need to output a HLINE separator
 555                 // before the row or not.
 556                 if (t.cat() == catEscape) {
 557                         if (parse_hlines(p, t, hlines, is_long_tabular)) {
 558                                 switch (pos) {
 559                                 case ROW_START:
 560                                         pos = IN_HLINES_START;
 561                                         break;
 562                                 case IN_COLUMNS:
 563                                         pos = IN_HLINES_END;
 564                                         break;
 565                                 case IN_HLINES_START:
 566                                 case IN_HLINES_END:
 567                                         break;
 568                                 }
 569                                 continue;
 570                         }
 571
 572                         else if (t.cs() == "tabularnewline" ||
 573                                  t.cs() == "\\" ||
 574                                  t.cs() == "cr") {
 575                                 if (t.cs() == "cr")
 576                                         cerr << "Warning: Converting TeX "
 577                                                 "'\\cr' to LaTeX '\\\\'."
 578                                              << endl;
 579                                 // stuff before the line break
 580                                 os << comments << HLINE << hlines << HLINE
 581                                    << LINE;
 582                                 //cerr << "hlines: " << hlines << endl;
 583                                 hlines.erase();
 584                                 comments.erase();
 585                                 pos = ROW_START;
 586                                 continue;
 587                         }
 588
 589                         else if (is_long_tabular &&
 590                                  (t.cs() == "endhead" ||
 591                                   t.cs() == "endfirsthead" ||
 592                                   t.cs() == "endfoot" ||
 593                                   t.cs() == "endlastfoot")) {
 594                                 hlines += t.asInput();
 595                                 switch (pos) {
 596                                 case IN_COLUMNS:
 597                                 case IN_HLINES_END:
 598                                         // these commands are implicit line
 599                                         // breaks
 600                                         os << comments << HLINE << hlines
 601                                            << HLINE << LINE;
 602                                         hlines.erase();
 603                                         comments.erase();
 604                                         pos = ROW_START;
 605                                         break;
 606                                 case ROW_START:
 607                                         pos = IN_HLINES_START;
 608                                         break;
 609                                 case IN_HLINES_START:
 610                                         break;
 611                                 }
 612                                 continue;
 613                         }
 614
 615                 }
 616
 617                 // We need a HLINE separator if we either have no hline
 618                 // stuff at all and are just starting a row or if we just
 619                 // got the first non-hline token.
 620                 switch (pos) {
 621                 case ROW_START:
 622                         // no hline tokens exist, first token at row start
 623                 case IN_HLINES_START:
 624                         // hline tokens exist, first non-hline token at row
 625                         // start
 626                         os << hlines << HLINE << comments;
 627                         hlines.erase();
 628                         comments.erase();
 629                         pos = IN_COLUMNS;
 630                         break;
 631                 case IN_HLINES_END:
 632                         // Oops, there is still cell content after hline
 633                         // stuff. This does not work in LaTeX, so we ignore
 634                         // the hlines.
 635                         cerr << "Ignoring '" << hlines << "' in a cell"
 636                              << endl;
 637                         os << comments;
 638                         hlines.erase();
 639                         comments.erase();
 640                         pos = IN_COLUMNS;
 641                         break;
 642                 case IN_COLUMNS:
 643                         break;
 644                 }
 645
 646                 // If we come here we have normal cell content
 647                 //
 648                 // cat codes
 649                 //
 650                 if (t.cat() == catMath) {
 651                         // we are inside some text mode thingy, so opening new math is allowed
 652                         Token const & n = p.get_token();
 653                         if (n.cat() == catMath) {
 654                                 // TeX's $$...$$ syntax for displayed math
 655                                 os << "\\[";
 656                                 parse_math(p, os, FLAG_SIMPLE, MATH_MODE);
 657                                 os << "\\]";
 658                                 p.get_token(); // skip the second '$' token
 659                         } else {
 660                                 // simple $...$  stuff
 661                                 p.putback();
 662                                 os << '$';
 663                                 parse_math(p, os, FLAG_SIMPLE, MATH_MODE);
 664                                 os << '$';
 665                         }
 666                 }
 667
 668                 else if (t.cat() == catSpace || t.cat() == catNewline)
 669                                 os << t.cs();
 670
 671                 else if (t.cat() == catLetter ||
 672                                t.cat() == catSuper ||
 673                                t.cat() == catSub ||
 674                                t.cat() == catOther ||
 675                                t.cat() == catActive ||
 676                                t.cat() == catParameter)
 677                         os << t.character();
 678
 679                 else if (t.cat() == catBegin) {
 680                         os << '{';
 681                         parse_table(p, os, is_long_tabular, pos,
 682                                     FLAG_BRACE_LAST);
 683                         os << '}';
 684                 }
 685
 686                 else if (t.cat() == catEnd) {
 687                         if (flags & FLAG_BRACE_LAST)
 688                                 return;
 689                         cerr << "unexpected '}'\n";
 690                 }
 691
 692                 else if (t.cat() == catAlign) {
 693                         os << TAB;
 694                         p.skip_spaces();
 695                 }
 696
 697                 else if (t.cat() == catComment)
 698                         os << t.asInput();
 699
 700                 else if (t.cs() == "(") {
 701                         os << "\\(";
 702                         parse_math(p, os, FLAG_SIMPLE2, MATH_MODE);
 703                         os << "\\)";
 704                 }
 705
 706                 else if (t.cs() == "[") {
 707                         os << "\\[";
 708                         parse_math(p, os, FLAG_EQUATION, MATH_MODE);
 709                         os << "\\]";
 710                 }
 711
 712                 else if (t.cs() == "begin") {
 713                         string const name = p.getArg('{', '}');
 714                         active_environments.push_back(name);
 715                         os << "\\begin{" << name << '}';
 716                         if (is_math_env(name)) {
 717                                 parse_math(p, os, FLAG_END, MATH_MODE);
 718                         } else {
 719                                 parse_table(p, os, is_long_tabular, pos,
 720                                             FLAG_END);
 721                         }
 722                         os << "\\end{" << name << '}';
 723                         active_environments.pop_back();
 724                 }
 725
 726                 else if (t.cs() == "end") {
 727                         if (flags & FLAG_END) {
 728                                 // eat environment name
 729                                 string const name = p.getArg('{', '}');
 730                                 if (name != active_environment())
 731                                         p.error("\\end{" + name + "} does not match \\begin{"
 732                                                 + active_environment() + "}");
 733                                 return;
 734                         }
 735                         p.error("found 'end' unexpectedly");
 736                 }
 737
 738                 else
 739                         os << t.asInput();
 740         }
 741
 742         // We can have comments if the last line is incomplete
 743         os << comments;
 744
 745         // We can have hline stuff if the last line is incomplete
 746         if (!hlines.empty()) {
 747                 // this does not work in LaTeX, so we ignore it
 748                 cerr << "Ignoring '" << hlines << "' at end of tabular"
 749                      << endl;
 750         }
 751 }
 752
 753
 754 void handle_hline_above(RowInfo & ri, vector<CellInfo> & ci)
 755 {
 756         ri.topline = true;
 757         for (size_t col = 0; col < ci.size(); ++col)
 758                 ci[col].topline = true;
 759 }
 760
 761
 762 void handle_hline_below(RowInfo & ri, vector<CellInfo> & ci)
 763 {
 764         ri.bottomline = true;
 765         for (size_t col = 0; col < ci.size(); ++col)
 766                 ci[col].bottomline = true;
 767 }
 768
 769
 770 } // anonymous namespace
 771
 772
 773 void handle_tabular(Parser & p, ostream & os, bool is_long_tabular,
 774                     Context & context)
 775 {
 776         string posopts = p.getOpt();
 777         if (!posopts.empty()) {
 778                 // FIXME: Convert this to ERT
 779                 if (is_long_tabular)
 780                         cerr << "horizontal longtable";
 781                 else
 782                         cerr << "vertical tabular";
 783                 cerr << " positioning '" << posopts << "' ignored\n";
 784         }
 785
 786         vector<ColInfo> colinfo;
 787
 788         // handle column formatting
 789         handle_colalign(p, colinfo, ColInfo());
 790         fix_colalign(colinfo);
 791
 792         // first scan of cells
 793         // use table mode to keep it minimal-invasive
 794         // not exactly what's TeX doing...
 795         vector<string> lines;
 796         ostringstream ss;
 797         RowPosition rowpos = ROW_START;
 798         parse_table(p, ss, is_long_tabular, rowpos, FLAG_END);
 799         split(ss.str(), lines, LINE);
 800
 801         vector< vector<CellInfo> > cellinfo(lines.size());
 802         vector<RowInfo> rowinfo(lines.size());
 803
 804         // split into rows
 805         //cerr << "// split into rows\n";
 806         for (size_t row = 0; row < rowinfo.size(); ++row) {
 807
 808                 // init row
 809                 cellinfo[row].resize(colinfo.size());
 810
 811                 // split row
 812                 vector<string> dummy;
 813                 //cerr << "\n########### LINE: " << lines[row] << "########\n";
 814                 split(lines[row], dummy, HLINE);
 815
 816                 // handle horizontal line fragments
 817                 // we do only expect this for a last line without '\\'
 818                 if (dummy.size() != 3) {
 819                         if ((dummy.size() != 1 && dummy.size() != 2) ||
 820                             row != rowinfo.size() - 1)
 821                                 cerr << "unexpected dummy size: " << dummy.size()
 822                                         << " content: " << lines[row] << "\n";
 823                         dummy.resize(3);
 824                 }
 825                 lines[row] = dummy[1];
 826
 827                 //cerr << "line: " << row << " above 0: " << dummy[0] << "\n";
 828                 //cerr << "line: " << row << " below 2: " << dummy[2] <<  "\n";
 829                 //cerr << "line: " << row << " cells 1: " << dummy[1] <<  "\n";
 830
 831                 for (int i = 0; i <= 2; i += 2) {
 832                         //cerr << "   reading from line string '" << dummy[i] << "'\n";
 833                         Parser p1(dummy[i]);
 834                         while (p1.good()) {
 835                                 Token t = p1.get_token();
 836                                 //cerr << "read token: " << t << "\n";
 837                                 if (t.cs() == "hline") {
 838                                         if (i == 0) {
 839                                                 if (rowinfo[row].topline) {
 840                                                         if (row > 0) // extra bottomline above
 841                                                                 handle_hline_below(rowinfo[row - 1], cellinfo[row - 1]);
 842                                                         else
 843                                                                 cerr << "dropping extra hline\n";
 844                                                         //cerr << "below row: " << row-1 << endl;
 845                                                 } else {
 846                                                         handle_hline_above(rowinfo[row], cellinfo[row]);
 847                                                         //cerr << "above row: " << row << endl;
 848                                                 }
 849                                         } else {
 850                                                 //cerr << "below row: " << row << endl;
 851                                                 handle_hline_below(rowinfo[row], cellinfo[row]);
 852                                         }
 853                                 } else if (t.cs() == "cline") {
 854                                         string arg = p1.verbatim_item();
 855                                         //cerr << "read cline arg: '" << arg << "'\n";
 856                                         vector<string> t;
 857                                         split(arg, t, '-');
 858                                         t.resize(2);
 859                                         size_t from = convert<unsigned int>(t[0]);
 860                                         if (from == 0)
 861                                                 cerr << "Could not parse "
 862                                                         "cline start column."
 863                                                      << endl;
 864                                         else
 865                                                 // 1 based index -> 0 based
 866                                                 --from;
 867                                         if (from >= colinfo.size()) {
 868                                                 cerr << "cline starts at non "
 869                                                         "existing column "
 870                                                      << (from + 1) << endl;
 871                                                 from = colinfo.size() - 1;
 872                                         }
 873                                         size_t to = convert<unsigned int>(t[1]);
 874                                         if (to == 0)
 875                                                 cerr << "Could not parse "
 876                                                         "cline end column."
 877                                                      << endl;
 878                                         else
 879                                                 // 1 based index -> 0 based
 880                                                 --to;
 881                                         if (to >= colinfo.size()) {
 882                                                 cerr << "cline ends at non "
 883                                                         "existing column "
 884                                                      << (to + 1) << endl;
 885                                                 to = colinfo.size() - 1;
 886                                         }
 887                                         for (size_t col = from; col <= to; ++col) {
 888                                                 //cerr << "row: " << row << " col: " << col << " i: " << i << endl;
 889                                                 if (i == 0) {
 890                                                         rowinfo[row].topline = true;
 891                                                         cellinfo[row][col].topline = true;
 892                                                 } else {
 893                                                         rowinfo[row].bottomline = true;
 894                                                         cellinfo[row][col].bottomline = true;
 895                                                 }
 896                                         }
 897                                 } else if (t.cs() == "endhead") {
 898                                         if (i > 0)
 899                                                 rowinfo[row].type = LT_HEAD;
 900                                         for (int r = row - 1; r >= 0; --r) {
 901                                                 if (rowinfo[r].type != LT_NORMAL)
 902                                                         break;
 903                                                 rowinfo[r].type = LT_HEAD;
 904                                         }
 905                                 } else if (t.cs() == "endfirsthead") {
 906                                         if (i > 0)
 907                                                 rowinfo[row].type = LT_FIRSTHEAD;
 908                                         for (int r = row - 1; r >= 0; --r) {
 909                                                 if (rowinfo[r].type != LT_NORMAL)
 910                                                         break;
 911                                                 rowinfo[r].type = LT_FIRSTHEAD;
 912                                         }
 913                                 } else if (t.cs() == "endfoot") {
 914                                         if (i > 0)
 915                                                 rowinfo[row].type = LT_FOOT;
 916                                         for (int r = row - 1; r >= 0; --r) {
 917                                                 if (rowinfo[r].type != LT_NORMAL)
 918                                                         break;
 919                                                 rowinfo[r].type = LT_FOOT;
 920                                         }
 921                                 } else if (t.cs() == "endlastfoot") {
 922                                         if (i > 0)
 923                                                 rowinfo[row].type = LT_LASTFOOT;
 924                                         for (int r = row - 1; r >= 0; --r) {
 925                                                 if (rowinfo[r].type != LT_NORMAL)
 926                                                         break;
 927                                                 rowinfo[r].type = LT_LASTFOOT;
 928                                         }
 929                                 } else if (t.cs() == "newpage") {
 930                                         if (i == 0) {
 931                                                 if (row > 0)
 932                                                         rowinfo[row - 1].newpage = true;
 933                                                 else
 934                                                         // This does not work in LaTeX
 935                                                         cerr << "Ignoring "
 936                                                                 "'\\newpage' "
 937                                                                 "before rows."
 938                                                              << endl;
 939                                         } else
 940                                                 rowinfo[row].newpage = true;
 941                                 } else {
 942                                         cerr << "unexpected line token: " << t << endl;
 943                                 }
 944                         }
 945                 }
 946
 947                 // split into cells
 948                 vector<string> cells;
 949                 split(lines[row], cells, TAB);
 950                 for (size_t col = 0, cell = 0; cell < cells.size();
 951                      ++col, ++cell) {
 952                         //cerr << "cell content: '" << cells[cell] << "'\n";
 953                         if (col >= colinfo.size()) {
 954                                 // This does not work in LaTeX
 955                                 cerr << "Ignoring extra cell '"
 956                                      << cells[cell] << "'." << endl;
 957                                 continue;
 958                         }
 959                         Parser p(cells[cell]);
 960                         p.skip_spaces();
 961                         //cells[cell] << "'\n";
 962                         if (p.next_token().cs() == "multicolumn") {
 963                                 // how many cells?
 964                                 p.get_token();
 965                                 size_t const ncells =
 966                                         convert<unsigned int>(p.verbatim_item());
 967
 968                                 // special cell properties alignment
 969                                 vector<ColInfo> t;
 970                                 handle_colalign(p, t, ColInfo());
 971                                 ColInfo & ci = t.front();
 972
 973                                 // The logic of LyX for multicolumn vertical
 974                                 // lines is too complicated to reproduce it
 975                                 // here (see LyXTabular::TeXCellPreamble()).
 976                                 // Therefore we simply put everything in the
 977                                 // special field.
 978                                 ci2special(ci);
 979
 980                                 cellinfo[row][col].multi      = CELL_BEGIN_OF_MULTICOLUMN;
 981                                 cellinfo[row][col].align      = ci.align;
 982                                 cellinfo[row][col].special    = ci.special;
 983                                 cellinfo[row][col].leftlines  = ci.leftlines;
 984                                 cellinfo[row][col].rightlines = ci.rightlines;
 985                                 ostringstream os;
 986                                 parse_text_in_inset(p, os, FLAG_ITEM, false, context);
 987                                 if (!cellinfo[row][col].content.empty()) {
 988                                         // This may or may not work in LaTeX,
 989                                         // but it does not work in LyX.
 990                                         // FIXME: Handle it correctly!
 991                                         cerr << "Moving cell content '"
 992                                              << cells[cell]
 993                                              << "' into a multicolumn cell. "
 994                                                 "This will probably not work."
 995                                              << endl;
 996                                 }
 997                                 cellinfo[row][col].content += os.str();
 998
 999                                 // add dummy cells for multicol
1000                                 for (size_t i = 0; i < ncells - 1 && col < colinfo.size(); ++i) {
1001                                         ++col;
1002                                         cellinfo[row][col].multi = CELL_PART_OF_MULTICOLUMN;
1003                                         cellinfo[row][col].align = 'c';
1004                                 }
1005
1006                         } else {
1007                                 cellinfo[row][col].leftlines  = colinfo[col].leftlines;
1008                                 cellinfo[row][col].rightlines = colinfo[col].rightlines;
1009                                 cellinfo[row][col].align      = colinfo[col].align;
1010                                 ostringstream os;
1011                                 parse_text_in_inset(p, os, FLAG_CELL, false, context);
1012                                 cellinfo[row][col].content += os.str();
1013                         }
1014                 }
1015
1016                 //cerr << "//  handle almost empty last row what we have\n";
1017                 // handle almost empty last row
1018                 if (row && lines[row].empty() && row + 1 == rowinfo.size()) {
1019                         //cerr << "remove empty last line\n";
1020                         if (rowinfo[row].topline)
1021                                 rowinfo[row - 1].bottomline = true;
1022                         for (size_t col = 0; col < colinfo.size(); ++col)
1023                                 if (cellinfo[row][col].topline)
1024                                         cellinfo[row - 1][col].bottomline = true;
1025                         rowinfo.pop_back();
1026                 }
1027         }
1028
1029         // Now we have the table structure and content in rowinfo, colinfo
1030         // and cellinfo.
1031         // Unfortunately LyX has some limitations that we need to work around.
1032
1033         // Convert cells with special content to multicolumn cells
1034         // (LyX ignores the special field for non-multicolumn cells).
1035         for (size_t row = 0; row < rowinfo.size(); ++row) {
1036                 for (size_t col = 0; col < cellinfo[row].size(); ++col) {
1037                         if (cellinfo[row][col].multi == CELL_NORMAL &&
1038                             !cellinfo[row][col].special.empty())
1039                                 cellinfo[row][col].multi = CELL_BEGIN_OF_MULTICOLUMN;
1040                 }
1041         }
1042
1043         //cerr << "// output what we have\n";
1044         // output what we have
1045         os << "\n<lyxtabular version=\"3\" rows=\"" << rowinfo.size()
1046            << "\" columns=\"" << colinfo.size() << "\">\n";
1047         os << "<features"
1048            << write_attribute("rotate", false)
1049            << write_attribute("islongtable", is_long_tabular)
1050            << ">\n";
1051
1052         //cerr << "// after header\n";
1053         for (size_t col = 0; col < colinfo.size(); ++col) {
1054                 os << "<column alignment=\""
1055                    << verbose_align(colinfo[col].align) << "\""
1056                    << " valignment=\""
1057                    << verbose_valign(colinfo[col].valign) << "\""
1058                    << write_attribute("leftline", colinfo[col].leftlines > 0)
1059                    << write_attribute("rightline", colinfo[col].rightlines > 0)
1060                    << write_attribute("width", colinfo[col].width)
1061                    << write_attribute("special", colinfo[col].special)
1062                    << ">\n";
1063         }
1064         //cerr << "// after cols\n";
1065
1066         for (size_t row = 0; row < rowinfo.size(); ++row) {
1067                 os << "<row"
1068                    << write_attribute("topline", rowinfo[row].topline)
1069                    << write_attribute("bottomline", rowinfo[row].bottomline)
1070                    << write_attribute("endhead",
1071                                       rowinfo[row].type == LT_HEAD)
1072                    << write_attribute("endfirsthead",
1073                                       rowinfo[row].type == LT_FIRSTHEAD)
1074                    << write_attribute("endfoot",
1075                                       rowinfo[row].type == LT_FOOT)
1076                    << write_attribute("endlastfoot",
1077                                       rowinfo[row].type == LT_LASTFOOT)
1078                    << write_attribute("newpage", rowinfo[row].newpage)
1079                    << ">\n";
1080                 for (size_t col = 0; col < colinfo.size(); ++col) {
1081                         CellInfo const & cell = cellinfo[row][col];
1082                         os << "<cell";
1083                         if (cell.multi != CELL_NORMAL)
1084                                 os << " multicolumn=\"" << cell.multi << "\"";
1085                         os << " alignment=\"" << verbose_align(cell.align)
1086                            << "\""
1087                            << " valignment=\"" << verbose_valign(cell.valign)
1088                            << "\""
1089                            << write_attribute("topline", cell.topline)
1090                            << write_attribute("bottomline", cell.bottomline)
1091                            << write_attribute("leftline", cell.leftlines > 0)
1092                            << write_attribute("rightline", cell.rightlines > 0)
1093                            << write_attribute("rotate", cell.rotate);
1094                         //cerr << "\nrow: " << row << " col: " << col;
1095                         //if (cell.topline)
1096                         //      cerr << " topline=\"true\"";
1097                         //if (cell.bottomline)
1098                         //      cerr << " bottomline=\"true\"";
1099                         os << " usebox=\"none\""
1100                            << write_attribute("width", cell.width);
1101                         if (cell.multi != CELL_NORMAL)
1102                                 os << write_attribute("special", cell.special);
1103                         os << ">"
1104                            << "\n\\begin_inset Text\n"
1105                            << cell.content
1106                            << "\n\\end_inset\n"
1107                            << "</cell>\n";
1108                 }
1109                 os << "</row>\n";
1110         }
1111
1112         os << "</lyxtabular>\n";
1113 }
1114
1115
1116
1117
1118 // }])