src/tex2lyx/table.cpp

   1 /**
   2  * \file table.cpp
   3  * This file is part of LyX, the document processor.
   4  * Licence details can be found in the file COPYING.
   5  *
   6  * \author André Pönitz
   7  * \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/lassert.h"
  20 #include "support/convert.h"
  21 #include "support/lstrings.h"
  22
  23 #include <iostream>
  24 #include <sstream>
  25 #include <vector>
  26 #include <map>
  27
  28 using namespace std;
  29
  30 namespace lyx {
  31
  32 // filled in preamble.cpp
  33 map<char, int> special_columns;
  34
  35
  36 namespace {
  37
  38 class ColInfo {
  39 public:
  40         ColInfo() : align('n'), valign('n'), rightlines(0), leftlines(0) {}
  41         /// column alignment
  42         char align;
  43         /// vertical alignment
  44         char valign;
  45         /// column width
  46         string width;
  47         /// special column alignment
  48         string special;
  49         /// number of lines on the right
  50         int rightlines;
  51         /// number of lines on the left
  52         int leftlines;
  53 };
  54
  55
  56 /// row type for longtables
  57 enum LTRowType
  58 {
  59         /// normal row
  60         LT_NORMAL,
  61         /// part of head
  62         LT_HEAD,
  63         /// part of head on first page
  64         LT_FIRSTHEAD,
  65         /// part of foot
  66         LT_FOOT,
  67         /// part of foot on last page
  68         LT_LASTFOOT
  69 };
  70
  71
  72 class RowInfo {
  73 public:
  74         RowInfo() : topline(false), bottomline(false), type(LT_NORMAL),
  75                     caption(false), newpage(false) {}
  76         /// horizontal line above
  77         bool topline;
  78         /// horizontal line below
  79         bool bottomline;
  80         /// These are for longtabulars only
  81         /// row type (head, foot, firsthead etc.)
  82         LTRowType type;
  83         /// row for a caption
  84         bool caption;
  85         /// row for a newpage
  86         bool newpage;
  87 };
  88
  89
  90 /// the numeric values are part of the file format!
  91 enum Multicolumn {
  92         /// A normal cell
  93         CELL_NORMAL = 0,
  94         /// A multicolumn cell. The number of columns is <tt>1 + number
  95         /// of CELL_PART_OF_MULTICOLUMN cells</tt> that follow directly
  96         CELL_BEGIN_OF_MULTICOLUMN = 1,
  97         /// This is a dummy cell (part of a multicolumn cell)
  98         CELL_PART_OF_MULTICOLUMN = 2
  99 };
 100
 101
 102 class CellInfo {
 103 public:
 104         CellInfo() : multi(CELL_NORMAL), align('n'), valign('n'),
 105                      leftlines(0), rightlines(0), topline(false),
 106                      bottomline(false), rotate(false) {}
 107         /// cell content
 108         string content;
 109         /// multicolumn flag
 110         Multicolumn multi;
 111         /// cell alignment
 112         char align;
 113         /// vertical cell alignment
 114         char valign;
 115         /// number of lines on the left
 116         int leftlines;
 117         /// number of lines on the right
 118         int rightlines;
 119         /// do we have a line above?
 120         bool topline;
 121         /// do we have a line below?
 122         bool bottomline;
 123         /// is the cell rotated?
 124         bool rotate;
 125         /// width for multicolumn cells
 126         string width;
 127         /// special formatting for multicolumn cells
 128         string special;
 129 };
 130
 131
 132 /// translate a horizontal alignment (as stored in ColInfo and CellInfo) to LyX
 133 inline char const * verbose_align(char c)
 134 {
 135         switch (c) {
 136         case 'c':
 137                 return "center";
 138         case 'r':
 139                 return "right";
 140         case 'l':
 141                 return "left";
 142         default:
 143                 return "none";
 144         }
 145 }
 146
 147
 148 /// translate a vertical alignment (as stored in ColInfo and CellInfo) to LyX
 149 inline char const * verbose_valign(char c)
 150 {
 151         // The default value for no special alignment is "top".
 152         switch (c) {
 153         case 'm':
 154                 return "middle";
 155         case 'b':
 156                 return "bottom";
 157         case 'p':
 158         default:
 159                 return "top";
 160         }
 161 }
 162
 163
 164 // stripped down from tabluar.C. We use it currently only for bools and
 165 // strings
 166 string const write_attribute(string const & name, bool const & b)
 167 {
 168         // we write only true attribute values so we remove a bit of the
 169         // file format bloat for tabulars.
 170         return b ? ' ' + name + "=\"true\"" : string();
 171 }
 172
 173
 174 string const write_attribute(string const & name, string const & s)
 175 {
 176         return s.empty() ? string() : ' ' + name + "=\"" + s + '"';
 177 }
 178
 179
 180 /*! rather brutish way to code table structure in a string:
 181
 182 \verbatim
 183   \begin{tabular}{ccc}
 184     1 & 2 & 3\\ \hline
 185     \multicolumn{2}{c}{4} & 5 //
 186     6 & 7 \\
 187     8 \endhead
 188   \end{tabular}
 189 \endverbatim
 190
 191  gets "translated" to:
 192
 193 \verbatim
 194          HLINE 1                     TAB 2 TAB 3 HLINE          HLINE LINE
 195   \hline HLINE \multicolumn{2}{c}{4} TAB 5       HLINE          HLINE LINE
 196          HLINE 6                     TAB 7       HLINE          HLINE LINE
 197          HLINE 8                                 HLINE \endhead HLINE LINE
 198 \endverbatim
 199  */
 200
 201 char const TAB   = '\001';
 202 char const LINE  = '\002';
 203 char const HLINE = '\004';
 204
 205
 206 /*!
 207  * Move the information in leftlines, rightlines, align and valign to the
 208  * special field. This is necessary if the special field is not empty,
 209  * because LyX ignores leftlines > 1, rightlines > 1, align and valign in
 210  * this case.
 211  */
 212 void ci2special(ColInfo & ci)
 213 {
 214         if (ci.width.empty() && ci.align == 'n')
 215                 // The alignment setting is already in special, since
 216                 // handle_colalign() never stores ci with these settings
 217                 // and ensures that leftlines == 0 and rightlines == 0 in
 218                 // this case.
 219                 return;
 220
 221         if (!ci.width.empty()) {
 222                 switch (ci.align) {
 223                 case 'l':
 224                         ci.special += ">{\\raggedright}";
 225                         break;
 226                 case 'r':
 227                         ci.special += ">{\\raggedleft}";
 228                         break;
 229                 case 'c':
 230                         ci.special += ">{\\centering}";
 231                         break;
 232                 }
 233                 if (ci.valign == 'n')
 234                         ci.special += 'p';
 235                 else
 236                         ci.special += ci.valign;
 237                 ci.special += '{' + ci.width + '}';
 238                 ci.width.erase();
 239         } else
 240                 ci.special += ci.align;
 241
 242         // LyX can only have one left and one right line.
 243         for (int i = 1; i < ci.leftlines; ++i)
 244                 ci.special.insert(0, "|");
 245         for (int i = 1; i < ci.rightlines; ++i)
 246                 ci.special += '|';
 247         ci.leftlines = min(ci.leftlines, 1);
 248         ci.rightlines = min(ci.rightlines, 1);
 249         ci.align = 'n';
 250         ci.valign = 'n';
 251 }
 252
 253
 254 /*!
 255  * Handle column specifications for tabulars and multicolumns.
 256  * The next token of the parser \p p must be an opening brace, and we read
 257  * everything until the matching closing brace.
 258  * The resulting column specifications are filled into \p colinfo. This is
 259  * in an intermediate form. fix_colalign() makes it suitable for LyX output.
 260  */
 261 void handle_colalign(Parser & p, vector<ColInfo> & colinfo,
 262                      ColInfo const & start)
 263 {
 264         if (p.get_token().cat() != catBegin)
 265                 cerr << "Wrong syntax for table column alignment.\n"
 266                         "Expected '{', got '" << p.curr_token().asInput()
 267                      << "'.\n";
 268
 269         ColInfo next = start;
 270         for (Token t = p.get_token(); p.good() && t.cat() != catEnd;
 271              t = p.get_token()) {
 272 #ifdef FILEDEBUG
 273                 cerr << "t: " << t << "  c: '" << t.character() << "'\n";
 274 #endif
 275
 276                 // We cannot handle comments here
 277                 if (t.cat() == catComment) {
 278                         if (t.cs().empty()) {
 279                                 // "%\n" combination
 280                                 p.skip_spaces();
 281                         } else
 282                                 cerr << "Ignoring comment: " << t.asInput();
 283                         continue;
 284                 }
 285
 286                 switch (t.character()) {
 287                         case 'c':
 288                         case 'l':
 289                         case 'r':
 290                                 // new column, horizontal aligned
 291                                 next.align = t.character();
 292                                 if (!next.special.empty())
 293                                         ci2special(next);
 294                                 colinfo.push_back(next);
 295                                 next = ColInfo();
 296                                 break;
 297                         case 'p':
 298                         case 'b':
 299                         case 'm':
 300                                 // new column, vertical aligned box
 301                                 next.valign = t.character();
 302                                 next.width = p.verbatim_item();
 303                                 if (!next.special.empty())
 304                                         ci2special(next);
 305                                 colinfo.push_back(next);
 306                                 next = ColInfo();
 307                                 break;
 308                         case '|':
 309                                 // vertical rule
 310                                 if (colinfo.empty()) {
 311                                         if (next.special.empty())
 312                                                 ++next.leftlines;
 313                                         else
 314                                                 next.special += '|';
 315                                 } else if (colinfo.back().special.empty())
 316                                         ++colinfo.back().rightlines;
 317                                 else if (next.special.empty())
 318                                         ++next.leftlines;
 319                                 else
 320                                         colinfo.back().special += '|';
 321                                 break;
 322                         case '>': {
 323                                 // text before the next column
 324                                 string const s = trim(p.verbatim_item());
 325                                 if (next.special.empty() &&
 326                                     next.align == 'n') {
 327                                         // Maybe this can be converted to a
 328                                         // horizontal alignment setting for
 329                                         // fixed width columns
 330                                         if (s == "\\raggedleft")
 331                                                 next.align = 'r';
 332                                         else if (s == "\\raggedright")
 333                                                 next.align = 'l';
 334                                         else if (s == "\\centering")
 335                                                 next.align = 'c';
 336                                         else
 337                                                 next.special = ">{" + s + '}';
 338                                 } else
 339                                         next.special += ">{" + s + '}';
 340                                 break;
 341                         }
 342                         case '<': {
 343                                 // text after the last column
 344                                 string const s = trim(p.verbatim_item());
 345                                 if (colinfo.empty())
 346                                         // This is not possible in LaTeX.
 347                                         cerr << "Ignoring separator '<{"
 348                                              << s << "}'." << endl;
 349                                 else {
 350                                         ColInfo & ci = colinfo.back();
 351                                         ci2special(ci);
 352                                         ci.special += "<{" + s + '}';
 353                                 }
 354                                 break;
 355                         }
 356                         case '*': {
 357                                 // *{n}{arg} means 'n' columns of type 'arg'
 358                                 string const num = p.verbatim_item();
 359                                 string const arg = p.verbatim_item();
 360                                 size_t const n = convert<unsigned int>(num);
 361                                 if (!arg.empty() && n > 0) {
 362                                         string s("{");
 363                                         for (size_t i = 0; i < n; ++i)
 364                                                 s += arg;
 365                                         s += '}';
 366                                         Parser p2(s);
 367                                         handle_colalign(p2, colinfo, next);
 368                                         next = ColInfo();
 369                                 } else {
 370                                         cerr << "Ignoring column specification"
 371                                                 " '*{" << num << "}{"
 372                                              << arg << "}'." << endl;
 373                                 }
 374                                 break;
 375                         }
 376                         case '@':
 377                                 // text instead of the column spacing
 378                         case '!':
 379                                 // text in addition to the column spacing
 380                                 next.special += t.character();
 381                                 next.special += '{' + p.verbatim_item() + '}';
 382                                 break;
 383                         default:
 384                                 // try user defined column types
 385                                 if (special_columns.find(t.character()) !=
 386                                     special_columns.end()) {
 387                                         ci2special(next);
 388                                         next.special += t.character();
 389                                         int const nargs =
 390                                                 special_columns[t.character()];
 391                                         for (int i = 0; i < nargs; ++i)
 392                                                 next.special += '{' +
 393                                                         p.verbatim_item() +
 394                                                         '}';
 395                                         colinfo.push_back(next);
 396                                         next = ColInfo();
 397                                 } else
 398                                         cerr << "Ignoring column specification"
 399                                                 " '" << t << "'." << endl;
 400                                 break;
 401                         }
 402         }
 403
 404         // Maybe we have some column separators that need to be added to the
 405         // last column?
 406         ci2special(next);
 407         if (!next.special.empty()) {
 408                 ColInfo & ci = colinfo.back();
 409                 ci2special(ci);
 410                 ci.special += next.special;
 411                 next.special.erase();
 412         }
 413 }
 414
 415
 416 /*!
 417  * Move the left and right lines and alignment settings of the column \p ci
 418  * to the special field if necessary.
 419  */
 420 void fix_colalign(ColInfo & ci)
 421 {
 422         if (ci.leftlines > 1 || ci.rightlines > 1)
 423                 ci2special(ci);
 424 }
 425
 426
 427 /*!
 428  * LyX can't handle more than one vertical line at the left or right side
 429  * of a column.
 430  * This function moves the left and right lines and alignment settings of all
 431  * columns in \p colinfo to the special field if necessary.
 432  */
 433 void fix_colalign(vector<ColInfo> & colinfo)
 434 {
 435         // Try to move extra leftlines to the previous column.
 436         // We do this only if both special fields are empty, otherwise we
 437         // can't tell wether the result will be the same.
 438         for (size_t col = 0; col < colinfo.size(); ++col) {
 439                 if (colinfo[col].leftlines > 1 &&
 440                     colinfo[col].special.empty() && col > 0 &&
 441                     colinfo[col - 1].rightlines == 0 &&
 442                     colinfo[col - 1].special.empty()) {
 443                         ++colinfo[col - 1].rightlines;
 444                         --colinfo[col].leftlines;
 445                 }
 446         }
 447         // Try to move extra rightlines to the next column
 448         for (size_t col = 0; col < colinfo.size(); ++col) {
 449                 if (colinfo[col].rightlines > 1 &&
 450                     colinfo[col].special.empty() &&
 451                     col < colinfo.size() - 1 &&
 452                     colinfo[col + 1].leftlines == 0 &&
 453                     colinfo[col + 1].special.empty()) {
 454                         ++colinfo[col + 1].leftlines;
 455                         --colinfo[col].rightlines;
 456                 }
 457         }
 458         // Move the lines and alignment settings to the special field if
 459         // necessary
 460         for (size_t col = 0; col < colinfo.size(); ++col)
 461                 fix_colalign(colinfo[col]);
 462 }
 463
 464
 465 /*!
 466  * Parse hlines and similar stuff.
 467  * \returns wether the token \p t was parsed
 468  */
 469 bool parse_hlines(Parser & p, Token const & t, string & hlines,
 470                   bool is_long_tabular)
 471 {
 472         LASSERT(t.cat() == catEscape, return false);
 473
 474         if (t.cs() == "hline")
 475                 hlines += "\\hline";
 476
 477         else if (t.cs() == "cline")
 478                 hlines += "\\cline{" + p.verbatim_item() + '}';
 479
 480         else if (is_long_tabular && t.cs() == "newpage")
 481                 hlines += "\\newpage";
 482
 483         else
 484                 return false;
 485
 486         return true;
 487 }
 488
 489
 490 /// Position in a row
 491 enum RowPosition {
 492         /// At the very beginning, before the first token
 493         ROW_START,
 494         /// After the first token and before any column token
 495         IN_HLINES_START,
 496         /// After the first column token. Comments and whitespace are only
 497         /// treated as tokens in this position
 498         IN_COLUMNS,
 499         /// After the first non-column token at the end
 500         IN_HLINES_END
 501 };
 502
 503
 504 /*!
 505  * Parse table structure.
 506  * We parse tables in a two-pass process: This function extracts the table
 507  * structure (rows, columns, hlines etc.), but does not change the cell
 508  * content. The cell content is parsed in a second step in handle_tabular().
 509  */
 510 void parse_table(Parser & p, ostream & os, bool is_long_tabular,
 511                  RowPosition & pos, unsigned flags)
 512 {
 513         // table structure commands such as \hline
 514         string hlines;
 515
 516         // comments that occur at places where we can't handle them
 517         string comments;
 518
 519         while (p.good()) {
 520                 Token const & t = p.get_token();
 521
 522 #ifdef FILEDEBUG
 523                 debugToken(cerr, t, flags);
 524 #endif
 525
 526                 // comments and whitespace in hlines
 527                 switch (pos) {
 528                 case ROW_START:
 529                 case IN_HLINES_START:
 530                 case IN_HLINES_END:
 531                         if (t.cat() == catComment) {
 532                                 if (t.cs().empty())
 533                                         // line continuation
 534                                         p.skip_spaces();
 535                                 else
 536                                         // We can't handle comments here,
 537                                         // store them for later use
 538                                         comments += t.asInput();
 539                                 continue;
 540                         } else if (t.cat() == catSpace ||
 541                                    t.cat() == catNewline) {
 542                                 // whitespace is irrelevant here, we
 543                                 // need to recognize hline stuff
 544                                 p.skip_spaces();
 545                                 continue;
 546                         }
 547                         break;
 548                 case IN_COLUMNS:
 549                         break;
 550                 }
 551
 552                 // We need to handle structure stuff first in order to
 553                 // determine wether we need to output a HLINE separator
 554                 // before the row or not.
 555                 if (t.cat() == catEscape) {
 556                         if (parse_hlines(p, t, hlines, is_long_tabular)) {
 557                                 switch (pos) {
 558                                 case ROW_START:
 559                                         pos = IN_HLINES_START;
 560                                         break;
 561                                 case IN_COLUMNS:
 562                                         pos = IN_HLINES_END;
 563                                         break;
 564                                 case IN_HLINES_START:
 565                                 case IN_HLINES_END:
 566                                         break;
 567                                 }
 568                                 continue;
 569                         }
 570
 571                         else if (t.cs() == "tabularnewline" ||
 572                                  t.cs() == "\\" ||
 573                                  t.cs() == "cr") {
 574                                 if (t.cs() == "cr")
 575                                         cerr << "Warning: Converting TeX "
 576                                                 "'\\cr' to LaTeX '\\\\'."
 577                                              << endl;
 578                                 // stuff before the line break
 579                                 os << comments << HLINE << hlines << HLINE
 580                                    << LINE;
 581                                 //cerr << "hlines: " << hlines << endl;
 582                                 hlines.erase();
 583                                 comments.erase();
 584                                 pos = ROW_START;
 585                                 continue;
 586                         }
 587
 588                         else if (is_long_tabular &&
 589                                  (t.cs() == "endhead" ||
 590                                   t.cs() == "endfirsthead" ||
 591                                   t.cs() == "endfoot" ||
 592                                   t.cs() == "endlastfoot")) {
 593                                 hlines += t.asInput();
 594                                 switch (pos) {
 595                                 case IN_COLUMNS:
 596                                 case IN_HLINES_END:
 597                                         // these commands are implicit line
 598                                         // breaks
 599                                         os << comments << HLINE << hlines
 600                                            << HLINE << LINE;
 601                                         hlines.erase();
 602                                         comments.erase();
 603                                         pos = ROW_START;
 604                                         break;
 605                                 case ROW_START:
 606                                         pos = IN_HLINES_START;
 607                                         break;
 608                                 case IN_HLINES_START:
 609                                         break;
 610                                 }
 611                                 continue;
 612                         }
 613
 614                 }
 615
 616                 // We need a HLINE separator if we either have no hline
 617                 // stuff at all and are just starting a row or if we just
 618                 // got the first non-hline token.
 619                 switch (pos) {
 620                 case ROW_START:
 621                         // no hline tokens exist, first token at row start
 622                 case IN_HLINES_START:
 623                         // hline tokens exist, first non-hline token at row
 624                         // start
 625                         os << hlines << HLINE << comments;
 626                         hlines.erase();
 627                         comments.erase();
 628                         pos = IN_COLUMNS;
 629                         break;
 630                 case IN_HLINES_END:
 631                         // Oops, there is still cell content after hline
 632                         // stuff. This does not work in LaTeX, so we ignore
 633                         // the hlines.
 634                         cerr << "Ignoring '" << hlines << "' in a cell"
 635                              << endl;
 636                         os << comments;
 637                         hlines.erase();
 638                         comments.erase();
 639                         pos = IN_COLUMNS;
 640                         break;
 641                 case IN_COLUMNS:
 642                         break;
 643                 }
 644
 645                 // If we come here we have normal cell content
 646                 //
 647                 // cat codes
 648                 //
 649                 if (t.cat() == catMath) {
 650                         // we are inside some text mode thingy, so opening new math is allowed
 651                         Token const & n = p.get_token();
 652                         if (n.cat() == catMath) {
 653                                 // TeX's $$...$$ syntax for displayed math
 654                                 os << "\\[";
 655                                 // This does only work because parse_math outputs TeX
 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 << '$';
 663                                 // This does only work because parse_math outputs TeX
 664                                 parse_math(p, os, FLAG_SIMPLE, MATH_MODE);
 665                                 os << '$';
 666                         }
 667                 }
 668
 669                 else if (t.cat() == catSpace
 670                          || t.cat() == catNewline
 671                          || 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.cs();
 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                         // This does only work because parse_math outputs TeX
 703                         parse_math(p, os, FLAG_SIMPLE2, MATH_MODE);
 704                         os << "\\)";
 705                 }
 706
 707                 else if (t.cs() == "[") {
 708                         os << "\\[";
 709                         // This does only work because parse_math outputs TeX
 710                         parse_math(p, os, FLAG_EQUATION, MATH_MODE);
 711                         os << "\\]";
 712                 }
 713
 714                 else if (t.cs() == "begin") {
 715                         string const name = p.getArg('{', '}');
 716                         active_environments.push_back(name);
 717                         os << "\\begin{" << name << '}';
 718                         // treat the nested environment as a block, don't
 719                         // parse &, \\ etc, because they don't belong to our
 720                         // table if they appear.
 721                         os << p.verbatimEnvironment(name);
 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                                 p.skip_spaces(true);
 972                                 ColInfo & ci = t.front();
 973
 974                                 // The logic of LyX for multicolumn vertical
 975                                 // lines is too complicated to reproduce it
 976                                 // here (see LyXTabular::TeXCellPreamble()).
 977                                 // Therefore we simply put everything in the
 978                                 // special field.
 979                                 ci2special(ci);
 980
 981                                 cellinfo[row][col].multi      = CELL_BEGIN_OF_MULTICOLUMN;
 982                                 cellinfo[row][col].align      = ci.align;
 983                                 cellinfo[row][col].special    = ci.special;
 984                                 cellinfo[row][col].leftlines  = ci.leftlines;
 985                                 cellinfo[row][col].rightlines = ci.rightlines;
 986                                 ostringstream os;
 987                                 parse_text_in_inset(p, os, FLAG_ITEM, false, context);
 988                                 if (!cellinfo[row][col].content.empty()) {
 989                                         // This may or may not work in LaTeX,
 990                                         // but it does not work in LyX.
 991                                         // FIXME: Handle it correctly!
 992                                         cerr << "Moving cell content '"
 993                                              << cells[cell]
 994                                              << "' into a multicolumn cell. "
 995                                                 "This will probably not work."
 996                                              << endl;
 997                                 }
 998                                 cellinfo[row][col].content += os.str();
 999
1000                                 // add dummy cells for multicol
1001                                 for (size_t i = 0; i < ncells - 1 && col < colinfo.size(); ++i) {
1002                                         ++col;
1003                                         cellinfo[row][col].multi = CELL_PART_OF_MULTICOLUMN;
1004                                         cellinfo[row][col].align = 'c';
1005                                 }
1006
1007                         } else if (col == 0 && is_long_tabular &&
1008                                    p.next_token().cs() == "caption") {
1009                                 // longtable caption support in LyX is a hack:
1010                                 // Captions require a row of their own with
1011                                 // the caption flag set to true, having only
1012                                 // one multicolumn cell. The contents of that
1013                                 // cell must contain exactly one caption inset
1014                                 // and nothing else.
1015                                 rowinfo[row].caption = true;
1016                                 for (size_t c = 1; c < cells.size(); ++c) {
1017                                         if (!cells[c].empty()) {
1018                                                 cerr << "Moving cell content '"
1019                                                      << cells[c]
1020                                                      << "' into the caption cell. "
1021                                                         "This will probably not work."
1022                                                      << endl;
1023                                                 cells[0] += cells[c];
1024                                         }
1025                                 }
1026                                 cells.resize(1);
1027                                 cellinfo[row][col].align      = colinfo[col].align;
1028                                 cellinfo[row][col].multi      = CELL_BEGIN_OF_MULTICOLUMN;
1029                                 ostringstream os;
1030                                 parse_text_in_inset(p, os, FLAG_CELL, false, context);
1031                                 cellinfo[row][col].content += os.str();
1032                                 // add dummy multicolumn cells
1033                                 for (size_t c = 1; c < colinfo.size(); ++c)
1034                                         cellinfo[row][c].multi = CELL_PART_OF_MULTICOLUMN;
1035
1036                         } else {
1037                                 cellinfo[row][col].leftlines  = colinfo[col].leftlines;
1038                                 cellinfo[row][col].rightlines = colinfo[col].rightlines;
1039                                 cellinfo[row][col].align      = colinfo[col].align;
1040                                 ostringstream os;
1041                                 parse_text_in_inset(p, os, FLAG_CELL, false, context);
1042                                 cellinfo[row][col].content += os.str();
1043                         }
1044                 }
1045
1046                 //cerr << "//  handle almost empty last row what we have\n";
1047                 // handle almost empty last row
1048                 if (row && lines[row].empty() && row + 1 == rowinfo.size()) {
1049                         //cerr << "remove empty last line\n";
1050                         if (rowinfo[row].topline)
1051                                 rowinfo[row - 1].bottomline = true;
1052                         for (size_t col = 0; col < colinfo.size(); ++col)
1053                                 if (cellinfo[row][col].topline)
1054                                         cellinfo[row - 1][col].bottomline = true;
1055                         rowinfo.pop_back();
1056                 }
1057         }
1058
1059         // Now we have the table structure and content in rowinfo, colinfo
1060         // and cellinfo.
1061         // Unfortunately LyX has some limitations that we need to work around.
1062
1063         // Convert cells with special content to multicolumn cells
1064         // (LyX ignores the special field for non-multicolumn cells).
1065         for (size_t row = 0; row < rowinfo.size(); ++row) {
1066                 for (size_t col = 0; col < cellinfo[row].size(); ++col) {
1067                         if (cellinfo[row][col].multi == CELL_NORMAL &&
1068                             !cellinfo[row][col].special.empty())
1069                                 cellinfo[row][col].multi = CELL_BEGIN_OF_MULTICOLUMN;
1070                 }
1071         }
1072
1073         // Distribute lines from rows/columns to cells
1074         // The code was stolen from convert_tablines() in lyx2lyx/lyx_1_6.py.
1075         // Each standard cell inherits the settings of the corresponding
1076         // rowinfo/colinfo. This works because all cells with individual
1077         // settings were converted to multicolumn cells above.
1078         // Each multicolumn cell inherits the settings of the rowinfo/colinfo
1079         // corresponding to the first column of the multicolumn cell (default
1080         // of the multicol package). This works because the special field
1081         // overrides the line fields.
1082         for (size_t row = 0; row < rowinfo.size(); ++row) {
1083                 for (size_t col = 0; col < cellinfo[row].size(); ++col) {
1084                         if (cellinfo[row][col].multi == CELL_NORMAL) {
1085                                 cellinfo[row][col].topline = rowinfo[row].topline;
1086                                 cellinfo[row][col].bottomline = rowinfo[row].bottomline;
1087                                 cellinfo[row][col].leftlines = colinfo[col].leftlines;
1088                                 cellinfo[row][col].rightlines = colinfo[col].rightlines;
1089                         } else if (cellinfo[row][col].multi == CELL_BEGIN_OF_MULTICOLUMN) {
1090                                 size_t s = col + 1;
1091                                 while (s < cellinfo[row].size() &&
1092                                        cellinfo[row][s].multi == CELL_PART_OF_MULTICOLUMN)
1093                                         s++;
1094                                 if (s < cellinfo[row].size() &&
1095                                     cellinfo[row][s].multi != CELL_BEGIN_OF_MULTICOLUMN)
1096                                         cellinfo[row][col].rightlines = colinfo[col].rightlines;
1097                                 if (col > 0 && cellinfo[row][col-1].multi == CELL_NORMAL)
1098                                         cellinfo[row][col].leftlines = colinfo[col].leftlines;
1099                         }
1100                 }
1101         }
1102
1103         //cerr << "// output what we have\n";
1104         // output what we have
1105         os << "\n<lyxtabular version=\"3\" rows=\"" << rowinfo.size()
1106            << "\" columns=\"" << colinfo.size() << "\">\n";
1107         os << "<features"
1108            << write_attribute("rotate", false)
1109            << write_attribute("islongtable", is_long_tabular)
1110            << ">\n";
1111
1112         //cerr << "// after header\n";
1113         for (size_t col = 0; col < colinfo.size(); ++col) {
1114                 os << "<column alignment=\""
1115                    << verbose_align(colinfo[col].align) << "\""
1116                    << " valignment=\""
1117                    << verbose_valign(colinfo[col].valign) << "\""
1118                    << write_attribute("width", translate_len(colinfo[col].width))
1119                    << write_attribute("special", colinfo[col].special)
1120                    << ">\n";
1121         }
1122         //cerr << "// after cols\n";
1123
1124         for (size_t row = 0; row < rowinfo.size(); ++row) {
1125                 os << "<row"
1126                    << write_attribute("endhead",
1127                                       rowinfo[row].type == LT_HEAD)
1128                    << write_attribute("endfirsthead",
1129                                       rowinfo[row].type == LT_FIRSTHEAD)
1130                    << write_attribute("endfoot",
1131                                       rowinfo[row].type == LT_FOOT)
1132                    << write_attribute("endlastfoot",
1133                                       rowinfo[row].type == LT_LASTFOOT)
1134                    << write_attribute("newpage", rowinfo[row].newpage)
1135                    << write_attribute("caption", rowinfo[row].caption)
1136                    << ">\n";
1137                 for (size_t col = 0; col < colinfo.size(); ++col) {
1138                         CellInfo const & cell = cellinfo[row][col];
1139                         os << "<cell";
1140                         if (cell.multi != CELL_NORMAL)
1141                                 os << " multicolumn=\"" << cell.multi << "\"";
1142                         os << " alignment=\"" << verbose_align(cell.align)
1143                            << "\""
1144                            << " valignment=\"" << verbose_valign(cell.valign)
1145                            << "\""
1146                            << write_attribute("topline", cell.topline)
1147                            << write_attribute("bottomline", cell.bottomline)
1148                            << write_attribute("leftline", cell.leftlines > 0)
1149                            << write_attribute("rightline", cell.rightlines > 0)
1150                            << write_attribute("rotate", cell.rotate);
1151                         //cerr << "\nrow: " << row << " col: " << col;
1152                         //if (cell.topline)
1153                         //      cerr << " topline=\"true\"";
1154                         //if (cell.bottomline)
1155                         //      cerr << " bottomline=\"true\"";
1156                         os << " usebox=\"none\""
1157                            << write_attribute("width", translate_len(cell.width));
1158                         if (cell.multi != CELL_NORMAL)
1159                                 os << write_attribute("special", cell.special);
1160                         os << ">"
1161                            << "\n\\begin_inset Text\n"
1162                            << cell.content
1163                            << "\n\\end_inset\n"
1164                            << "</cell>\n";
1165                 }
1166                 os << "</row>\n";
1167         }
1168
1169         os << "</lyxtabular>\n";
1170 }
1171
1172
1173
1174
1175 // }])
1176
1177
1178 } // namespace lyx