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/convert.h"
  20 #include "support/lstrings.h"
  21
  22 #include <iostream>
  23 #include <sstream>
  24 #include <vector>
  25 #include <map>
  26
  27
  28 namespace lyx {
  29
  30 using std::cerr;
  31 using std::endl;
  32 using std::istringstream;
  33 using std::ostream;
  34 using std::ostringstream;
  35 using std::string;
  36 using std::vector;
  37
  38
  39 // filled in preamble.cpp
  40 std::map<char, int> special_columns;
  41
  42
  43 namespace {
  44
  45 class ColInfo {
  46 public:
  47         ColInfo() : align('n'), valign('n'), rightlines(0), leftlines(0) {}
  48         /// column alignment
  49         char align;
  50         /// vertical alignment
  51         char valign;
  52         /// column width
  53         string width;
  54         /// special column alignment
  55         string special;
  56         /// number of lines on the right
  57         int rightlines;
  58         /// number of lines on the left
  59         int leftlines;
  60 };
  61
  62
  63 /// row type for longtables
  64 enum LTRowType
  65 {
  66         /// normal row
  67         LT_NORMAL,
  68         /// part of head
  69         LT_HEAD,
  70         /// part of head on first page
  71         LT_FIRSTHEAD,
  72         /// part of foot
  73         LT_FOOT,
  74         /// part of foot on last page
  75         LT_LASTFOOT
  76 };
  77
  78
  79 class RowInfo {
  80 public:
  81         RowInfo() : topline(false), bottomline(false), type(LT_NORMAL),
  82                     newpage(false) {}
  83         /// horizontal line above
  84         bool topline;
  85         /// horizontal line below
  86         bool bottomline;
  87         /// These are for longtabulars only
  88         /// row type (head, foot, firsthead etc.)
  89         LTRowType type;
  90         /// row for a pagebreak
  91         bool newpage;
  92 };
  93
  94
  95 enum Multicolumn {
  96         /// A normal cell
  97         CELL_NORMAL = 0,
  98         /// A multicolumn cell. The number of columns is <tt>1 + number
  99         /// of CELL_PART_OF_MULTICOLUMN cells</tt> that follow directly
 100         CELL_BEGIN_OF_MULTICOLUMN,
 101         /// This is a dummy cell (part of a multicolumn cell)
 102         CELL_PART_OF_MULTICOLUMN
 103 };
 104
 105
 106 class CellInfo {
 107 public:
 108         CellInfo() : multi(CELL_NORMAL), align('n'), valign('n'),
 109                      leftlines(0), rightlines(0), topline(false),
 110                      bottomline(false), rotate(false) {}
 111         /// cell content
 112         string content;
 113         /// multicolumn flag
 114         Multicolumn multi;
 115         /// cell alignment
 116         char align;
 117         /// vertical cell alignment
 118         char valign;
 119         /// number of lines on the left
 120         int leftlines;
 121         /// number of lines on the right
 122         int rightlines;
 123         /// do we have a line above?
 124         bool topline;
 125         /// do we have a line below?
 126         bool bottomline;
 127         /// is the cell rotated?
 128         bool rotate;
 129         /// width for multicolumn cells
 130         string width;
 131         /// special formatting for multicolumn cells
 132         string special;
 133 };
 134
 135
 136 /// translate a horizontal alignment (as stored in ColInfo and CellInfo) to LyX
 137 inline char const * verbose_align(char c)
 138 {
 139         switch (c) {
 140         case 'c':
 141                 return "center";
 142         case 'r':
 143                 return "right";
 144         case 'l':
 145                 return "left";
 146         default:
 147                 return "none";
 148         }
 149 }
 150
 151
 152 /// translate a vertical alignment (as stored in ColInfo and CellInfo) to LyX
 153 inline char const * verbose_valign(char c)
 154 {
 155         // The default value for no special alignment is "top".
 156         switch (c) {
 157         case 'm':
 158                 return "middle";
 159         case 'b':
 160                 return "bottom";
 161         case 'p':
 162         default:
 163                 return "top";
 164         }
 165 }
 166
 167
 168 // stripped down from tabluar.C. We use it currently only for bools and
 169 // strings
 170 string const write_attribute(string const & name, bool const & b)
 171 {
 172         // we write only true attribute values so we remove a bit of the
 173         // file format bloat for tabulars.
 174         return b ? ' ' + name + "=\"true\"" : string();
 175 }
 176
 177
 178 string const write_attribute(string const & name, string const & s)
 179 {
 180         return s.empty() ? string() : ' ' + name + "=\"" + s + '"';
 181 }
 182
 183
 184 /*! rather brutish way to code table structure in a string:
 185
 186 \verbatim
 187   \begin{tabular}{ccc}
 188     1 & 2 & 3\\ \hline
 189     \multicolumn{2}{c}{4} & 5 //
 190     6 & 7 \\
 191     8 \endhead
 192   \end{tabular}
 193 \endverbatim
 194
 195  gets "translated" to:
 196
 197 \verbatim
 198          HLINE 1                     TAB 2 TAB 3 HLINE          HLINE LINE
 199   \hline HLINE \multicolumn{2}{c}{4} TAB 5       HLINE          HLINE LINE
 200          HLINE 6                     TAB 7       HLINE          HLINE LINE
 201          HLINE 8                                 HLINE \endhead HLINE LINE
 202 \endverbatim
 203  */
 204
 205 char const TAB   = '\001';
 206 char const LINE  = '\002';
 207 char const HLINE = '\004';
 208
 209
 210 /*!
 211  * Move the information in leftlines, rightlines, align and valign to the
 212  * special field. This is necessary if the special field is not empty,
 213  * because LyX ignores leftlines, rightlines, align and valign in this case.
 214  */
 215 void ci2special(ColInfo & ci)
 216 {
 217         if (ci.width.empty() && ci.align == 'n')
 218                 // The alignment setting is already in special, since
 219                 // handle_colalign() never stores ci with these settings
 220                 // and ensures that leftlines == 0 and rightlines == 0 in
 221                 // this case.
 222                 return;
 223
 224         if (!ci.width.empty()) {
 225                 switch (ci.align) {
 226                 case 'l':
 227                         ci.special += ">{\\raggedright}";
 228                         break;
 229                 case 'r':
 230                         ci.special += ">{\\raggedleft}";
 231                         break;
 232                 case 'c':
 233                         ci.special += ">{\\centering}";
 234                         break;
 235                 }
 236                 if (ci.valign == 'n')
 237                         ci.special += 'p';
 238                 else
 239                         ci.special += ci.valign;
 240                 ci.special += '{' + ci.width + '}';
 241                 ci.width.erase();
 242         } else
 243                 ci.special += ci.align;
 244
 245         for (int i = 0; i < ci.leftlines; ++i)
 246                 ci.special.insert(0, "|");
 247         for (int i = 0; i < ci.rightlines; ++i)
 248                 ci.special += '|';
 249         ci.leftlines = 0;
 250         ci.rightlines = 0;
 251         ci.align = 'n';
 252         ci.valign = 'n';
 253 }
 254
 255
 256 /*!
 257  * Handle column specifications for tabulars and multicolumns.
 258  * The next token of the parser \p p must be an opening brace, and we read
 259  * everything until the matching closing brace.
 260  * The resulting column specifications are filled into \p colinfo. This is
 261  * in an intermediate form. fix_colalign() makes it suitable for LyX output.
 262  */
 263 void handle_colalign(Parser & p, vector<ColInfo> & colinfo,
 264                      ColInfo const & start)
 265 {
 266         if (p.get_token().cat() != catBegin)
 267                 cerr << "Wrong syntax for table column alignment.\n"
 268                         "Expected '{', got '" << p.curr_token().asInput()
 269                      << "'.\n";
 270
 271         ColInfo next = start;
 272         for (Token t = p.get_token(); p.good() && t.cat() != catEnd;
 273              t = p.get_token()) {
 274 #ifdef FILEDEBUG
 275                 cerr << "t: " << t << "  c: '" << t.character() << "'\n";
 276 #endif
 277
 278                 // We cannot handle comments here
 279                 if (t.cat() == catComment) {
 280                         if (t.cs().empty()) {
 281                                 // "%\n" combination
 282                                 p.skip_spaces();
 283                         } else
 284                                 cerr << "Ignoring comment: " << t.asInput();
 285                         continue;
 286                 }
 287
 288                 switch (t.character()) {
 289                         case 'c':
 290                         case 'l':
 291                         case 'r':
 292                                 // new column, horizontal aligned
 293                                 next.align = t.character();
 294                                 if (!next.special.empty())
 295                                         ci2special(next);
 296                                 colinfo.push_back(next);
 297                                 next = ColInfo();
 298                                 break;
 299                         case 'p':
 300                         case 'b':
 301                         case 'm':
 302                                 // new column, vertical aligned box
 303                                 next.valign = t.character();
 304                                 next.width = p.verbatim_item();
 305                                 if (!next.special.empty())
 306                                         ci2special(next);
 307                                 colinfo.push_back(next);
 308                                 next = ColInfo();
 309                                 break;
 310                         case '|':
 311                                 // vertical rule
 312                                 if (colinfo.empty()) {
 313                                         if (next.special.empty())
 314                                                 ++next.leftlines;
 315                                         else
 316                                                 next.special += '|';
 317                                 } else if (colinfo.back().special.empty())
 318                                         ++colinfo.back().rightlines;
 319                                 else if (next.special.empty())
 320                                         ++next.leftlines;
 321                                 else
 322                                         colinfo.back().special += '|';
 323                                 break;
 324                         case '>': {
 325                                 // text before the next column
 326                                 string const s = trim(p.verbatim_item());
 327                                 if (next.special.empty() &&
 328                                     next.align == 'n') {
 329                                         // Maybe this can be converted to a
 330                                         // horizontal alignment setting for
 331                                         // fixed width columns
 332                                         if (s == "\\raggedleft")
 333                                                 next.align = 'r';
 334                                         else if (s == "\\raggedright")
 335                                                 next.align = 'l';
 336                                         else if (s == "\\centering")
 337                                                 next.align = 'c';
 338                                         else
 339                                                 next.special = ">{" + s + '}';
 340                                 } else
 341                                         next.special += ">{" + s + '}';
 342                                 break;
 343                         }
 344                         case '<': {
 345                                 // text after the last column
 346                                 string const s = trim(p.verbatim_item());
 347                                 if (colinfo.empty())
 348                                         // This is not possible in LaTeX.
 349                                         cerr << "Ignoring separator '<{"
 350                                              << s << "}'." << endl;
 351                                 else {
 352                                         ColInfo & ci = colinfo.back();
 353                                         ci2special(ci);
 354                                         ci.special += "<{" + s + '}';
 355                                 }
 356                                 break;
 357                         }
 358                         case '*': {
 359                                 // *{n}{arg} means 'n' columns of type 'arg'
 360                                 string const num = p.verbatim_item();
 361                                 string const arg = p.verbatim_item();
 362                                 size_t const n = convert<unsigned int>(num);
 363                                 if (!arg.empty() && n > 0) {
 364                                         string s("{");
 365                                         for (size_t i = 0; i < n; ++i)
 366                                                 s += arg;
 367                                         s += '}';
 368                                         Parser p2(s);
 369                                         handle_colalign(p2, colinfo, next);
 370                                         next = ColInfo();
 371                                 } else {
 372                                         cerr << "Ignoring column specification"
 373                                                 " '*{" << num << "}{"
 374                                              << arg << "}'." << endl;
 375                                 }
 376                                 break;
 377                         }
 378                         case '@':
 379                                 // text instead of the column spacing
 380                         case '!':
 381                                 // text in addition to the column spacing
 382                                 next.special += t.character();
 383                                 next.special += '{' + p.verbatim_item() + '}';
 384                                 break;
 385                         default:
 386                                 // try user defined column types
 387                                 if (special_columns.find(t.character()) !=
 388                                     special_columns.end()) {
 389                                         ci2special(next);
 390                                         next.special += t.character();
 391                                         int const nargs =
 392                                                 special_columns[t.character()];
 393                                         for (int i = 0; i < nargs; ++i)
 394                                                 next.special += '{' +
 395                                                         p.verbatim_item() +
 396                                                         '}';
 397                                         colinfo.push_back(next);
 398                                         next = ColInfo();
 399                                 } else
 400                                         cerr << "Ignoring column specification"
 401                                                 " '" << t << "'." << endl;
 402                                 break;
 403                         }
 404         }
 405
 406         // Maybe we have some column separators that need to be added to the
 407         // last column?
 408         ci2special(next);
 409         if (!next.special.empty()) {
 410                 ColInfo & ci = colinfo.back();
 411                 ci2special(ci);
 412                 ci.special += next.special;
 413                 next.special.erase();
 414         }
 415 }
 416
 417
 418 /*!
 419  * Move the left and right lines and alignment settings of the column \p ci
 420  * to the special field if necessary.
 421  */
 422 void fix_colalign(ColInfo & ci)
 423 {
 424         if (ci.leftlines > 1 || ci.rightlines > 1)
 425                 ci2special(ci);
 426 }
 427
 428
 429 /*!
 430  * LyX can't handle more than one vertical line at the left or right side
 431  * of a column.
 432  * This function moves the left and right lines and alignment settings of all
 433  * columns in \p colinfo to the special field if necessary.
 434  */
 435 void fix_colalign(vector<ColInfo> & colinfo)
 436 {
 437         // Try to move extra leftlines to the previous column.
 438         // We do this only if both special fields are empty, otherwise we
 439         // can't tell wether the result will be the same.
 440         for (size_t col = 0; col < colinfo.size(); ++col) {
 441                 if (colinfo[col].leftlines > 1 &&
 442                     colinfo[col].special.empty() && col > 0 &&
 443                     colinfo[col - 1].rightlines == 0 &&
 444                     colinfo[col - 1].special.empty()) {
 445                         ++colinfo[col - 1].rightlines;
 446                         --colinfo[col].leftlines;
 447                 }
 448         }
 449         // Try to move extra rightlines to the next column
 450         for (size_t col = 0; col < colinfo.size(); ++col) {
 451                 if (colinfo[col].rightlines > 1 &&
 452                     colinfo[col].special.empty() &&
 453                     col < colinfo.size() - 1 &&
 454                     colinfo[col + 1].leftlines == 0 &&
 455                     colinfo[col + 1].special.empty()) {
 456                         ++colinfo[col + 1].leftlines;
 457                         --colinfo[col].rightlines;
 458                 }
 459         }
 460         // Move the lines and alignment settings to the special field if
 461         // necessary
 462         for (size_t col = 0; col < colinfo.size(); ++col)
 463                 fix_colalign(colinfo[col]);
 464 }
 465
 466
 467 /*!
 468  * Parse hlines and similar stuff.
 469  * \returns wether the token \p t was parsed
 470  */
 471 bool parse_hlines(Parser & p, Token const & t, string & hlines,
 472                   bool is_long_tabular)
 473 {
 474         BOOST_ASSERT(t.cat() == catEscape);
 475
 476         if (t.cs() == "hline")
 477                 hlines += "\\hline";
 478
 479         else if (t.cs() == "cline")
 480                 hlines += "\\cline{" + p.verbatim_item() + '}';
 481
 482         else if (is_long_tabular && t.cs() == "newpage")
 483                 hlines += "\\newpage";
 484
 485         else
 486                 return false;
 487
 488         return true;
 489 }
 490
 491
 492 /// Position in a row
 493 enum RowPosition {
 494         /// At the very beginning, before the first token
 495         ROW_START,
 496         /// After the first token and before any column token
 497         IN_HLINES_START,
 498         /// After the first column token. Comments and whitespace are only
 499         /// treated as tokens in this position
 500         IN_COLUMNS,
 501         /// After the first non-column token at the end
 502         IN_HLINES_END
 503 };
 504
 505
 506 /*!
 507  * Parse table structure.
 508  * We parse tables in a two-pass process: This function extracts the table
 509  * structure (rows, columns, hlines etc.), but does not change the cell
 510  * content. The cell content is parsed in a second step in handle_tabular().
 511  */
 512 void parse_table(Parser & p, ostream & os, bool is_long_tabular,
 513                  RowPosition & pos, unsigned flags)
 514 {
 515         // table structure commands such as \hline
 516         string hlines;
 517
 518         // comments that occur at places where we can't handle them
 519         string comments;
 520
 521         while (p.good()) {
 522                 Token const & t = p.get_token();
 523
 524 #ifdef FILEDEBUG
 525                 cerr << "t: " << t << " flags: " << flags << "\n";
 526 #endif
 527
 528                 // comments and whitespace in hlines
 529                 switch (pos) {
 530                 case ROW_START:
 531                 case IN_HLINES_START:
 532                 case IN_HLINES_END:
 533                         if (t.cat() == catComment) {
 534                                 if (t.cs().empty())
 535                                         // line continuation
 536                                         p.skip_spaces();
 537                                 else
 538                                         // We can't handle comments here,
 539                                         // store them for later use
 540                                         comments += t.asInput();
 541                                 continue;
 542                         } else if (t.cat() == catSpace ||
 543                                    t.cat() == catNewline) {
 544                                 // whitespace is irrelevant here, we
 545                                 // need to recognize hline stuff
 546                                 p.skip_spaces();
 547                                 continue;
 548                         }
 549                         break;
 550                 case IN_COLUMNS:
 551                         break;
 552                 }
 553
 554                 // We need to handle structure stuff first in order to
 555                 // determine wether we need to output a HLINE separator
 556                 // before the row or not.
 557                 if (t.cat() == catEscape) {
 558                         if (parse_hlines(p, t, hlines, is_long_tabular)) {
 559                                 switch (pos) {
 560                                 case ROW_START:
 561                                         pos = IN_HLINES_START;
 562                                         break;
 563                                 case IN_COLUMNS:
 564                                         pos = IN_HLINES_END;
 565                                         break;
 566                                 case IN_HLINES_START:
 567                                 case IN_HLINES_END:
 568                                         break;
 569                                 }
 570                                 continue;
 571                         }
 572
 573                         else if (t.cs() == "tabularnewline" ||
 574                                  t.cs() == "\\" ||
 575                                  t.cs() == "cr") {
 576                                 if (t.cs() == "cr")
 577                                         cerr << "Warning: Converting TeX "
 578                                                 "'\\cr' to LaTeX '\\\\'."
 579                                              << endl;
 580                                 // stuff before the line break
 581                                 os << comments << HLINE << hlines << HLINE
 582                                    << LINE;
 583                                 //cerr << "hlines: " << hlines << endl;
 584                                 hlines.erase();
 585                                 comments.erase();
 586                                 pos = ROW_START;
 587                                 continue;
 588                         }
 589
 590                         else if (is_long_tabular &&
 591                                  (t.cs() == "endhead" ||
 592                                   t.cs() == "endfirsthead" ||
 593                                   t.cs() == "endfoot" ||
 594                                   t.cs() == "endlastfoot")) {
 595                                 hlines += t.asInput();
 596                                 switch (pos) {
 597                                 case IN_COLUMNS:
 598                                 case IN_HLINES_END:
 599                                         // these commands are implicit line
 600                                         // breaks
 601                                         os << comments << HLINE << hlines
 602                                            << HLINE << LINE;
 603                                         hlines.erase();
 604                                         comments.erase();
 605                                         pos = ROW_START;
 606                                         break;
 607                                 case ROW_START:
 608                                         pos = IN_HLINES_START;
 609                                         break;
 610                                 case IN_HLINES_START:
 611                                         break;
 612                                 }
 613                                 continue;
 614                         }
 615
 616                 }
 617
 618                 // We need a HLINE separator if we either have no hline
 619                 // stuff at all and are just starting a row or if we just
 620                 // got the first non-hline token.
 621                 switch (pos) {
 622                 case ROW_START:
 623                         // no hline tokens exist, first token at row start
 624                 case IN_HLINES_START:
 625                         // hline tokens exist, first non-hline token at row
 626                         // start
 627                         os << hlines << HLINE << comments;
 628                         hlines.erase();
 629                         comments.erase();
 630                         pos = IN_COLUMNS;
 631                         break;
 632                 case IN_HLINES_END:
 633                         // Oops, there is still cell content after hline
 634                         // stuff. This does not work in LaTeX, so we ignore
 635                         // the hlines.
 636                         cerr << "Ignoring '" << hlines << "' in a cell"
 637                              << endl;
 638                         os << comments;
 639                         hlines.erase();
 640                         comments.erase();
 641                         pos = IN_COLUMNS;
 642                         break;
 643                 case IN_COLUMNS:
 644                         break;
 645                 }
 646
 647                 // If we come here we have normal cell content
 648                 //
 649                 // cat codes
 650                 //
 651                 if (t.cat() == catMath) {
 652                         // we are inside some text mode thingy, so opening new math is allowed
 653                         Token const & n = p.get_token();
 654                         if (n.cat() == catMath) {
 655                                 // TeX's $$...$$ syntax for displayed math
 656                                 os << "\\[";
 657                                 // This does only work because parse_math outputs TeX
 658                                 parse_math(p, os, FLAG_SIMPLE, MATH_MODE);
 659                                 os << "\\]";
 660                                 p.get_token(); // skip the second '$' token
 661                         } else {
 662                                 // simple $...$  stuff
 663                                 p.putback();
 664                                 os << '$';
 665                                 // This does only work because parse_math outputs TeX
 666                                 parse_math(p, os, FLAG_SIMPLE, MATH_MODE);
 667                                 os << '$';
 668                         }
 669                 }
 670
 671                 else if (t.cat() == catSpace || t.cat() == catNewline)
 672                                 os << t.cs();
 673
 674                 else if (t.cat() == catLetter ||
 675                                t.cat() == catSuper ||
 676                                t.cat() == catSub ||
 677                                t.cat() == catOther ||
 678                                t.cat() == catActive ||
 679                                t.cat() == catParameter)
 680                         os << t.character();
 681
 682                 else if (t.cat() == catBegin) {
 683                         os << '{';
 684                         parse_table(p, os, is_long_tabular, pos,
 685                                     FLAG_BRACE_LAST);
 686                         os << '}';
 687                 }
 688
 689                 else if (t.cat() == catEnd) {
 690                         if (flags & FLAG_BRACE_LAST)
 691                                 return;
 692                         cerr << "unexpected '}'\n";
 693                 }
 694
 695                 else if (t.cat() == catAlign) {
 696                         os << TAB;
 697                         p.skip_spaces();
 698                 }
 699
 700                 else if (t.cat() == catComment)
 701                         os << t.asInput();
 702
 703                 else if (t.cs() == "(") {
 704                         os << "\\(";
 705                         // This does only work because parse_math outputs TeX
 706                         parse_math(p, os, FLAG_SIMPLE2, MATH_MODE);
 707                         os << "\\)";
 708                 }
 709
 710                 else if (t.cs() == "[") {
 711                         os << "\\[";
 712                         // This does only work because parse_math outputs TeX
 713                         parse_math(p, os, FLAG_EQUATION, MATH_MODE);
 714                         os << "\\]";
 715                 }
 716
 717                 else if (t.cs() == "begin") {
 718                         string const name = p.getArg('{', '}');
 719                         active_environments.push_back(name);
 720                         os << "\\begin{" << name << '}';
 721                         // treat the nested environment as a block, don't
 722                         // parse &, \\ etc, because they don't belong to our
 723                         // table if they appear.
 724                         os << p.verbatimEnvironment(name);
 725                         os << "\\end{" << name << '}';
 726                         active_environments.pop_back();
 727                 }
 728
 729                 else if (t.cs() == "end") {
 730                         if (flags & FLAG_END) {
 731                                 // eat environment name
 732                                 string const name = p.getArg('{', '}');
 733                                 if (name != active_environment())
 734                                         p.error("\\end{" + name + "} does not match \\begin{"
 735                                                 + active_environment() + "}");
 736                                 return;
 737                         }
 738                         p.error("found 'end' unexpectedly");
 739                 }
 740
 741                 else
 742                         os << t.asInput();
 743         }
 744
 745         // We can have comments if the last line is incomplete
 746         os << comments;
 747
 748         // We can have hline stuff if the last line is incomplete
 749         if (!hlines.empty()) {
 750                 // this does not work in LaTeX, so we ignore it
 751                 cerr << "Ignoring '" << hlines << "' at end of tabular"
 752                      << endl;
 753         }
 754 }
 755
 756
 757 void handle_hline_above(RowInfo & ri, vector<CellInfo> & ci)
 758 {
 759         ri.topline = true;
 760         for (size_t col = 0; col < ci.size(); ++col)
 761                 ci[col].topline = true;
 762 }
 763
 764
 765 void handle_hline_below(RowInfo & ri, vector<CellInfo> & ci)
 766 {
 767         ri.bottomline = true;
 768         for (size_t col = 0; col < ci.size(); ++col)
 769                 ci[col].bottomline = true;
 770 }
 771
 772
 773 } // anonymous namespace
 774
 775
 776 void handle_tabular(Parser & p, ostream & os, bool is_long_tabular,
 777                     Context & context)
 778 {
 779         string posopts = p.getOpt();
 780         if (!posopts.empty()) {
 781                 // FIXME: Convert this to ERT
 782                 if (is_long_tabular)
 783                         cerr << "horizontal longtable";
 784                 else
 785                         cerr << "vertical tabular";
 786                 cerr << " positioning '" << posopts << "' ignored\n";
 787         }
 788
 789         vector<ColInfo> colinfo;
 790
 791         // handle column formatting
 792         handle_colalign(p, colinfo, ColInfo());
 793         fix_colalign(colinfo);
 794
 795         // first scan of cells
 796         // use table mode to keep it minimal-invasive
 797         // not exactly what's TeX doing...
 798         vector<string> lines;
 799         ostringstream ss;
 800         RowPosition rowpos = ROW_START;
 801         parse_table(p, ss, is_long_tabular, rowpos, FLAG_END);
 802         split(ss.str(), lines, LINE);
 803
 804         vector< vector<CellInfo> > cellinfo(lines.size());
 805         vector<RowInfo> rowinfo(lines.size());
 806
 807         // split into rows
 808         //cerr << "// split into rows\n";
 809         for (size_t row = 0; row < rowinfo.size(); ++row) {
 810
 811                 // init row
 812                 cellinfo[row].resize(colinfo.size());
 813
 814                 // split row
 815                 vector<string> dummy;
 816                 //cerr << "\n########### LINE: " << lines[row] << "########\n";
 817                 split(lines[row], dummy, HLINE);
 818
 819                 // handle horizontal line fragments
 820                 // we do only expect this for a last line without '\\'
 821                 if (dummy.size() != 3) {
 822                         if ((dummy.size() != 1 && dummy.size() != 2) ||
 823                             row != rowinfo.size() - 1)
 824                                 cerr << "unexpected dummy size: " << dummy.size()
 825                                         << " content: " << lines[row] << "\n";
 826                         dummy.resize(3);
 827                 }
 828                 lines[row] = dummy[1];
 829
 830                 //cerr << "line: " << row << " above 0: " << dummy[0] << "\n";
 831                 //cerr << "line: " << row << " below 2: " << dummy[2] <<  "\n";
 832                 //cerr << "line: " << row << " cells 1: " << dummy[1] <<  "\n";
 833
 834                 for (int i = 0; i <= 2; i += 2) {
 835                         //cerr << "   reading from line string '" << dummy[i] << "'\n";
 836                         Parser p1(dummy[i]);
 837                         while (p1.good()) {
 838                                 Token t = p1.get_token();
 839                                 //cerr << "read token: " << t << "\n";
 840                                 if (t.cs() == "hline") {
 841                                         if (i == 0) {
 842                                                 if (rowinfo[row].topline) {
 843                                                         if (row > 0) // extra bottomline above
 844                                                                 handle_hline_below(rowinfo[row - 1], cellinfo[row - 1]);
 845                                                         else
 846                                                                 cerr << "dropping extra hline\n";
 847                                                         //cerr << "below row: " << row-1 << endl;
 848                                                 } else {
 849                                                         handle_hline_above(rowinfo[row], cellinfo[row]);
 850                                                         //cerr << "above row: " << row << endl;
 851                                                 }
 852                                         } else {
 853                                                 //cerr << "below row: " << row << endl;
 854                                                 handle_hline_below(rowinfo[row], cellinfo[row]);
 855                                         }
 856                                 } else if (t.cs() == "cline") {
 857                                         string arg = p1.verbatim_item();
 858                                         //cerr << "read cline arg: '" << arg << "'\n";
 859                                         vector<string> t;
 860                                         split(arg, t, '-');
 861                                         t.resize(2);
 862                                         size_t from = convert<unsigned int>(t[0]);
 863                                         if (from == 0)
 864                                                 cerr << "Could not parse "
 865                                                         "cline start column."
 866                                                      << endl;
 867                                         else
 868                                                 // 1 based index -> 0 based
 869                                                 --from;
 870                                         if (from >= colinfo.size()) {
 871                                                 cerr << "cline starts at non "
 872                                                         "existing column "
 873                                                      << (from + 1) << endl;
 874                                                 from = colinfo.size() - 1;
 875                                         }
 876                                         size_t to = convert<unsigned int>(t[1]);
 877                                         if (to == 0)
 878                                                 cerr << "Could not parse "
 879                                                         "cline end column."
 880                                                      << endl;
 881                                         else
 882                                                 // 1 based index -> 0 based
 883                                                 --to;
 884                                         if (to >= colinfo.size()) {
 885                                                 cerr << "cline ends at non "
 886                                                         "existing column "
 887                                                      << (to + 1) << endl;
 888                                                 to = colinfo.size() - 1;
 889                                         }
 890                                         for (size_t col = from; col <= to; ++col) {
 891                                                 //cerr << "row: " << row << " col: " << col << " i: " << i << endl;
 892                                                 if (i == 0) {
 893                                                         rowinfo[row].topline = true;
 894                                                         cellinfo[row][col].topline = true;
 895                                                 } else {
 896                                                         rowinfo[row].bottomline = true;
 897                                                         cellinfo[row][col].bottomline = true;
 898                                                 }
 899                                         }
 900                                 } else if (t.cs() == "endhead") {
 901                                         if (i > 0)
 902                                                 rowinfo[row].type = LT_HEAD;
 903                                         for (int r = row - 1; r >= 0; --r) {
 904                                                 if (rowinfo[r].type != LT_NORMAL)
 905                                                         break;
 906                                                 rowinfo[r].type = LT_HEAD;
 907                                         }
 908                                 } else if (t.cs() == "endfirsthead") {
 909                                         if (i > 0)
 910                                                 rowinfo[row].type = LT_FIRSTHEAD;
 911                                         for (int r = row - 1; r >= 0; --r) {
 912                                                 if (rowinfo[r].type != LT_NORMAL)
 913                                                         break;
 914                                                 rowinfo[r].type = LT_FIRSTHEAD;
 915                                         }
 916                                 } else if (t.cs() == "endfoot") {
 917                                         if (i > 0)
 918                                                 rowinfo[row].type = LT_FOOT;
 919                                         for (int r = row - 1; r >= 0; --r) {
 920                                                 if (rowinfo[r].type != LT_NORMAL)
 921                                                         break;
 922                                                 rowinfo[r].type = LT_FOOT;
 923                                         }
 924                                 } else if (t.cs() == "endlastfoot") {
 925                                         if (i > 0)
 926                                                 rowinfo[row].type = LT_LASTFOOT;
 927                                         for (int r = row - 1; r >= 0; --r) {
 928                                                 if (rowinfo[r].type != LT_NORMAL)
 929                                                         break;
 930                                                 rowinfo[r].type = LT_LASTFOOT;
 931                                         }
 932                                 } else if (t.cs() == "newpage") {
 933                                         if (i == 0) {
 934                                                 if (row > 0)
 935                                                         rowinfo[row - 1].newpage = true;
 936                                                 else
 937                                                         // This does not work in LaTeX
 938                                                         cerr << "Ignoring "
 939                                                                 "'\\newpage' "
 940                                                                 "before rows."
 941                                                              << endl;
 942                                         } else
 943                                                 rowinfo[row].newpage = true;
 944                                 } else {
 945                                         cerr << "unexpected line token: " << t << endl;
 946                                 }
 947                         }
 948                 }
 949
 950                 // split into cells
 951                 vector<string> cells;
 952                 split(lines[row], cells, TAB);
 953                 for (size_t col = 0, cell = 0; cell < cells.size();
 954                      ++col, ++cell) {
 955                         //cerr << "cell content: '" << cells[cell] << "'\n";
 956                         if (col >= colinfo.size()) {
 957                                 // This does not work in LaTeX
 958                                 cerr << "Ignoring extra cell '"
 959                                      << cells[cell] << "'." << endl;
 960                                 continue;
 961                         }
 962                         Parser p(cells[cell]);
 963                         p.skip_spaces();
 964                         //cells[cell] << "'\n";
 965                         if (p.next_token().cs() == "multicolumn") {
 966                                 // how many cells?
 967                                 p.get_token();
 968                                 size_t const ncells =
 969                                         convert<unsigned int>(p.verbatim_item());
 970
 971                                 // special cell properties alignment
 972                                 vector<ColInfo> t;
 973                                 handle_colalign(p, t, ColInfo());
 974                                 ColInfo & ci = t.front();
 975
 976                                 // The logic of LyX for multicolumn vertical
 977                                 // lines is too complicated to reproduce it
 978                                 // here (see LyXTabular::TeXCellPreamble()).
 979                                 // Therefore we simply put everything in the
 980                                 // special field.
 981                                 ci2special(ci);
 982
 983                                 cellinfo[row][col].multi      = CELL_BEGIN_OF_MULTICOLUMN;
 984                                 cellinfo[row][col].align      = ci.align;
 985                                 cellinfo[row][col].special    = ci.special;
 986                                 cellinfo[row][col].leftlines  = ci.leftlines;
 987                                 cellinfo[row][col].rightlines = ci.rightlines;
 988                                 ostringstream os;
 989                                 parse_text_in_inset(p, os, FLAG_ITEM, false, context);
 990                                 if (!cellinfo[row][col].content.empty()) {
 991                                         // This may or may not work in LaTeX,
 992                                         // but it does not work in LyX.
 993                                         // FIXME: Handle it correctly!
 994                                         cerr << "Moving cell content '"
 995                                              << cells[cell]
 996                                              << "' into a multicolumn cell. "
 997                                                 "This will probably not work."
 998                                              << endl;
 999                                 }
1000                                 cellinfo[row][col].content += os.str();
1001
1002                                 // add dummy cells for multicol
1003                                 for (size_t i = 0; i < ncells - 1 && col < colinfo.size(); ++i) {
1004                                         ++col;
1005                                         cellinfo[row][col].multi = CELL_PART_OF_MULTICOLUMN;
1006                                         cellinfo[row][col].align = 'c';
1007                                 }
1008
1009                         } else {
1010                                 cellinfo[row][col].leftlines  = colinfo[col].leftlines;
1011                                 cellinfo[row][col].rightlines = colinfo[col].rightlines;
1012                                 cellinfo[row][col].align      = colinfo[col].align;
1013                                 ostringstream os;
1014                                 parse_text_in_inset(p, os, FLAG_CELL, false, context);
1015                                 cellinfo[row][col].content += os.str();
1016                         }
1017                 }
1018
1019                 //cerr << "//  handle almost empty last row what we have\n";
1020                 // handle almost empty last row
1021                 if (row && lines[row].empty() && row + 1 == rowinfo.size()) {
1022                         //cerr << "remove empty last line\n";
1023                         if (rowinfo[row].topline)
1024                                 rowinfo[row - 1].bottomline = true;
1025                         for (size_t col = 0; col < colinfo.size(); ++col)
1026                                 if (cellinfo[row][col].topline)
1027                                         cellinfo[row - 1][col].bottomline = true;
1028                         rowinfo.pop_back();
1029                 }
1030         }
1031
1032         // Now we have the table structure and content in rowinfo, colinfo
1033         // and cellinfo.
1034         // Unfortunately LyX has some limitations that we need to work around.
1035
1036         // Convert cells with special content to multicolumn cells
1037         // (LyX ignores the special field for non-multicolumn cells).
1038         for (size_t row = 0; row < rowinfo.size(); ++row) {
1039                 for (size_t col = 0; col < cellinfo[row].size(); ++col) {
1040                         if (cellinfo[row][col].multi == CELL_NORMAL &&
1041                             !cellinfo[row][col].special.empty())
1042                                 cellinfo[row][col].multi = CELL_BEGIN_OF_MULTICOLUMN;
1043                 }
1044         }
1045
1046         //cerr << "// output what we have\n";
1047         // output what we have
1048         os << "\n<lyxtabular version=\"3\" rows=\"" << rowinfo.size()
1049            << "\" columns=\"" << colinfo.size() << "\">\n";
1050         os << "<features"
1051            << write_attribute("rotate", false)
1052            << write_attribute("islongtable", is_long_tabular)
1053            << ">\n";
1054
1055         //cerr << "// after header\n";
1056         for (size_t col = 0; col < colinfo.size(); ++col) {
1057                 os << "<column alignment=\""
1058                    << verbose_align(colinfo[col].align) << "\""
1059                    << " valignment=\""
1060                    << verbose_valign(colinfo[col].valign) << "\""
1061                    << write_attribute("leftline", colinfo[col].leftlines > 0)
1062                    << write_attribute("rightline", colinfo[col].rightlines > 0)
1063                    << write_attribute("width", translate_len(colinfo[col].width))
1064                    << write_attribute("special", colinfo[col].special)
1065                    << ">\n";
1066         }
1067         //cerr << "// after cols\n";
1068
1069         for (size_t row = 0; row < rowinfo.size(); ++row) {
1070                 os << "<row"
1071                    << write_attribute("topline", rowinfo[row].topline)
1072                    << write_attribute("bottomline", rowinfo[row].bottomline)
1073                    << write_attribute("endhead",
1074                                       rowinfo[row].type == LT_HEAD)
1075                    << write_attribute("endfirsthead",
1076                                       rowinfo[row].type == LT_FIRSTHEAD)
1077                    << write_attribute("endfoot",
1078                                       rowinfo[row].type == LT_FOOT)
1079                    << write_attribute("endlastfoot",
1080                                       rowinfo[row].type == LT_LASTFOOT)
1081                    << write_attribute("newpage", rowinfo[row].newpage)
1082                    << ">\n";
1083                 for (size_t col = 0; col < colinfo.size(); ++col) {
1084                         CellInfo const & cell = cellinfo[row][col];
1085                         os << "<cell";
1086                         if (cell.multi != CELL_NORMAL)
1087                                 os << " multicolumn=\"" << cell.multi << "\"";
1088                         os << " alignment=\"" << verbose_align(cell.align)
1089                            << "\""
1090                            << " valignment=\"" << verbose_valign(cell.valign)
1091                            << "\""
1092                            << write_attribute("topline", cell.topline)
1093                            << write_attribute("bottomline", cell.bottomline)
1094                            << write_attribute("leftline", cell.leftlines > 0)
1095                            << write_attribute("rightline", cell.rightlines > 0)
1096                            << write_attribute("rotate", cell.rotate);
1097                         //cerr << "\nrow: " << row << " col: " << col;
1098                         //if (cell.topline)
1099                         //      cerr << " topline=\"true\"";
1100                         //if (cell.bottomline)
1101                         //      cerr << " bottomline=\"true\"";
1102                         os << " usebox=\"none\""
1103                            << write_attribute("width", translate_len(cell.width));
1104                         if (cell.multi != CELL_NORMAL)
1105                                 os << write_attribute("special", cell.special);
1106                         os << ">"
1107                            << "\n\\begin_inset Text\n"
1108                            << cell.content
1109                            << "\n\\end_inset\n"
1110                            << "</cell>\n";
1111                 }
1112                 os << "</row>\n";
1113         }
1114
1115         os << "</lyxtabular>\n";
1116 }
1117
1118
1119
1120
1121 // }])
1122
1123
1124 } // namespace lyx