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