src/tex2lyx/table.C

   1 /**
   2  * \file table.C
   3  * This file is part of LyX, the document processor.
   4  * Licence details can be found in the file COPYING.
   5  *
   6  * \author André Pönitz
   7  * \author Jean-Marc Lasgouttes
   8  * \author Georg Baum
   9  *
  10  * Full author contact details are available in file CREDITS.
  11  */
  12
  13 // {[(
  14
  15 #include <config.h>
  16
  17 #include "tex2lyx.h"
  18
  19 #include "support/convert.h"
  20 #include "support/lstrings.h"
  21
  22 #include <cctype>
  23 #include <fstream>
  24 #include <iostream>
  25 #include <sstream>
  26 #include <vector>
  27 #include <map>
  28
  29
  30 namespace lyx {
  31
  32 using std::cerr;
  33 using std::endl;
  34 using std::istringstream;
  35 using std::ostream;
  36 using std::ostringstream;
  37 using std::string;
  38 using std::vector;
  39
  40
  41 // filled in preamble.C
  42 std::map<char, int> special_columns;
  43
  44
  45 namespace {
  46
  47 class ColInfo {
  48 public:
  49         ColInfo() : align('n'), valign('n'), rightlines(0), leftlines(0) {}
  50         /// column alignment
  51         char align;
  52         /// vertical alignment
  53         char valign;
  54         /// column width
  55         string width;
  56         /// special column alignment
  57         string special;
  58         /// number of lines on the right
  59         int rightlines;
  60         /// number of lines on the left
  61         int leftlines;
  62 };
  63
  64
  65 /// row type for longtables
  66 enum LTRowType
  67 {
  68         /// normal row
  69         LT_NORMAL,
  70         /// part of head
  71         LT_HEAD,
  72         /// part of head on first page
  73         LT_FIRSTHEAD,
  74         /// part of foot
  75         LT_FOOT,
  76         /// part of foot on last page
  77         LT_LASTFOOT
  78 };
  79
  80
  81 class RowInfo {
  82 public:
  83         RowInfo() : topline(false), bottomline(false), type(LT_NORMAL),
  84                     newpage(false) {}
  85         /// horizontal line above
  86         bool topline;
  87         /// horizontal line below
  88         bool bottomline;
  89         /// These are for longtabulars only
  90         /// row type (head, foot, firsthead etc.)
  91         LTRowType type;
  92         /// row for a pagebreak
  93         bool newpage;
  94 };
  95
  96
  97 enum Multicolumn {
  98         /// A normal cell
  99         CELL_NORMAL = 0,
 100         /// A multicolumn cell. The number of columns is <tt>1 + number
 101         /// of CELL_PART_OF_MULTICOLUMN cells</tt> that follow directly
 102         CELL_BEGIN_OF_MULTICOLUMN,
 103         /// This is a dummy cell (part of a multicolumn cell)
 104         CELL_PART_OF_MULTICOLUMN
 105 };
 106
 107
 108 class CellInfo {
 109 public:
 110         CellInfo() : multi(CELL_NORMAL), align('n'), valign('n'),
 111                      leftlines(0), rightlines(0), topline(false),
 112                      bottomline(false), rotate(false) {}
 113         /// cell content
 114         string content;
 115         /// multicolumn flag
 116         Multicolumn multi;
 117         /// cell alignment
 118         char align;
 119         /// vertical cell alignment
 120         char valign;
 121         /// number of lines on the left
 122         int leftlines;
 123         /// number of lines on the right
 124         int rightlines;
 125         /// do we have a line above?
 126         bool topline;
 127         /// do we have a line below?
 128         bool bottomline;
 129         /// is the cell rotated?
 130         bool rotate;
 131         /// width for multicolumn cells
 132         string width;
 133         /// special formatting for multicolumn cells
 134         string special;
 135 };
 136
 137
 138 /// translate a horizontal alignment (as stored in ColInfo and CellInfo) to LyX
 139 inline char const * verbose_align(char c)
 140 {
 141         switch (c) {
 142         case 'c':
 143                 return "center";
 144         case 'r':
 145                 return "right";
 146         case 'l':
 147                 return "left";
 148         default:
 149                 return "none";
 150         }
 151 }
 152
 153
 154 /// translate a vertical alignment (as stored in ColInfo and CellInfo) to LyX
 155 inline char const * verbose_valign(char c)
 156 {
 157         // The default value for no special alignment is "top".
 158         switch (c) {
 159         case 'm':
 160                 return "middle";
 161         case 'b':
 162                 return "bottom";
 163         case 'p':
 164         default:
 165                 return "top";
 166         }
 167 }
 168
 169
 170 // stripped down from tabluar.C. We use it currently only for bools and
 171 // strings
 172 string const write_attribute(string const & name, bool const & b)
 173 {
 174         // we write only true attribute values so we remove a bit of the
 175         // file format bloat for tabulars.
 176         return b ? ' ' + name + "=\"true\"" : string();
 177 }
 178
 179
 180 string const write_attribute(string const & name, string const & s)
 181 {
 182         return s.empty() ? string() : ' ' + name + "=\"" + s + '"';
 183 }
 184
 185
 186 /*! rather brutish way to code table structure in a string:
 187
 188 \verbatim
 189   \begin{tabular}{ccc}
 190     1 & 2 & 3\\ \hline
 191     \multicolumn{2}{c}{4} & 5 //
 192     6 & 7 \\
 193     8 \endhead
 194   \end{tabular}
 195 \endverbatim
 196
 197  gets "translated" to:
 198
 199 \verbatim
 200          HLINE 1                     TAB 2 TAB 3 HLINE          HLINE LINE
 201   \hline HLINE \multicolumn{2}{c}{4} TAB 5       HLINE          HLINE LINE
 202          HLINE 6                     TAB 7       HLINE          HLINE LINE
 203          HLINE 8                                 HLINE \endhead HLINE LINE
 204 \endverbatim
 205  */
 206
 207 char const TAB   = '\001';
 208 char const LINE  = '\002';
 209 char const HLINE = '\004';
 210
 211
 212 /*!
 213  * Move the information in leftlines, rightlines, align and valign to the
 214  * special field. This is necessary if the special field is not empty,
 215  * because LyX ignores leftlines, rightlines, align and valign in this case.
 216  */
 217 void ci2special(ColInfo & ci)
 218 {
 219         if (ci.width.empty() && ci.align == 'n')
 220                 // The alignment setting is already in special, since
 221                 // handle_colalign() never stores ci with these settings
 222                 // and ensures that leftlines == 0 and rightlines == 0 in
 223                 // this case.
 224                 return;
 225
 226         if (!ci.width.empty()) {
 227                 switch (ci.align) {
 228                 case 'l':
 229                         ci.special += ">{\\raggedright}";
 230                         break;
 231                 case 'r':
 232                         ci.special += ">{\\raggedleft}";
 233                         break;
 234                 case 'c':
 235                         ci.special += ">{\\centering}";
 236                         break;
 237                 }
 238                 if (ci.valign == 'n')
 239                         ci.special += 'p';
 240                 else
 241                         ci.special += ci.valign;
 242                 ci.special += '{' + ci.width + '}';
 243                 ci.width.erase();
 244         } else
 245                 ci.special += ci.align;
 246
 247         for (int i = 0; i < ci.leftlines; ++i)
 248                 ci.special.insert(0, "|");
 249         for (int i = 0; i < ci.rightlines; ++i)
 250                 ci.special += '|';
 251         ci.leftlines = 0;
 252         ci.rightlines = 0;
 253         ci.align = 'n';
 254         ci.valign = 'n';
 255 }
 256
 257
 258 /*!
 259  * Handle column specifications for tabulars and multicolumns.
 260  * The next token of the parser \p p must be an opening brace, and we read
 261  * everything until the matching closing brace.
 262  * The resulting column specifications are filled into \p colinfo. This is
 263  * in an intermediate form. fix_colalign() makes it suitable for LyX output.
 264  */
 265 void handle_colalign(Parser & p, vector<ColInfo> & colinfo,
 266                      ColInfo const & start)
 267 {
 268         if (p.get_token().cat() != catBegin)
 269                 cerr << "Wrong syntax for table column alignment.\n"
 270                         "Expected '{', got '" << p.curr_token().asInput()
 271                      << "'.\n";
 272
 273         ColInfo next = start;
 274         for (Token t = p.get_token(); p.good() && t.cat() != catEnd;
 275              t = p.get_token()) {
 276 #ifdef FILEDEBUG
 277                 cerr << "t: " << t << "  c: '" << t.character() << "'\n";
 278 #endif
 279
 280                 // We cannot handle comments here
 281                 if (t.cat() == catComment) {
 282                         if (t.cs().empty()) {
 283                                 // "%\n" combination
 284                                 p.skip_spaces();
 285                         } else
 286                                 cerr << "Ignoring comment: " << t.asInput();
 287                         continue;
 288                 }
 289
 290                 switch (t.character()) {
 291                         case 'c':
 292                         case 'l':
 293                         case 'r':
 294                                 // new column, horizontal aligned
 295                                 next.align = t.character();
 296                                 if (!next.special.empty())
 297                                         ci2special(next);
 298                                 colinfo.push_back(next);
 299                                 next = ColInfo();
 300                                 break;
 301                         case 'p':
 302                         case 'b':
 303                         case 'm':
 304                                 // new column, vertical aligned box
 305                                 next.valign = t.character();
 306                                 next.width = p.verbatim_item();
 307                                 if (!next.special.empty())
 308                                         ci2special(next);
 309                                 colinfo.push_back(next);
 310                                 next = ColInfo();
 311                                 break;
 312                         case '|':
 313                                 // vertical rule
 314                                 if (colinfo.empty()) {
 315                                         if (next.special.empty())
 316                                                 ++next.leftlines;
 317                                         else
 318                                                 next.special += '|';
 319                                 } else if (colinfo.back().special.empty())
 320                                         ++colinfo.back().rightlines;
 321                                 else if (next.special.empty())
 322                                         ++next.leftlines;
 323                                 else
 324                                         colinfo.back().special += '|';
 325                                 break;
 326                         case '>': {
 327                                 // text before the next column
 328                                 string const s = trim(p.verbatim_item());
 329                                 if (next.special.empty() &&
 330                                     next.align == 'n') {
 331                                         // Maybe this can be converted to a
 332                                         // horizontal alignment setting for
 333                                         // fixed width columns
 334                                         if (s == "\\raggedleft")
 335                                                 next.align = 'r';
 336                                         else if (s == "\\raggedright")
 337                                                 next.align = 'l';
 338                                         else if (s == "\\centering")
 339                                                 next.align = 'c';
 340                                         else
 341                                                 next.special = ">{" + s + '}';
 342                                 } else
 343                                         next.special += ">{" + s + '}';
 344                                 break;
 345                         }
 346                         case '<': {
 347                                 // text after the last column
 348                                 string const s = trim(p.verbatim_item());
 349                                 if (colinfo.empty())
 350                                         // This is not possible in LaTeX.
 351                                         cerr << "Ignoring separator '<{"
 352                                              << s << "}'." << endl;
 353                                 else {
 354                                         ColInfo & ci = colinfo.back();
 355                                         ci2special(ci);
 356                                         ci.special += "<{" + s + '}';
 357                                 }
 358                                 break;
 359                         }
 360                         case '*': {
 361                                 // *{n}{arg} means 'n' columns of type 'arg'
 362                                 string const num = p.verbatim_item();
 363                                 string const arg = p.verbatim_item();
 364                                 size_t const n = convert<unsigned int>(num);
 365                                 if (!arg.empty() && n > 0) {
 366                                         string s("{");
 367                                         for (size_t i = 0; i < n; ++i)
 368                                                 s += arg;
 369                                         s += '}';
 370                                         Parser p2(s);
 371                                         handle_colalign(p2, colinfo, next);
 372                                         next = ColInfo();
 373                                 } else {
 374                                         cerr << "Ignoring column specification"
 375                                                 " '*{" << num << "}{"
 376                                              << arg << "}'." << endl;
 377                                 }
 378                                 break;
 379                         }
 380                         case '@':
 381                                 // text instead of the column spacing
 382                         case '!':
 383                                 // text in addition to the column spacing
 384                                 next.special += t.character();
 385                                 next.special += '{' + p.verbatim_item() + '}';
 386                                 break;
 387                         default:
 388                                 // try user defined column types
 389                                 if (special_columns.find(t.character()) !=
 390                                     special_columns.end()) {
 391                                         ci2special(next);
 392                                         next.special += t.character();
 393                                         int const nargs =
 394                                                 special_columns[t.character()];
 395                                         for (int i = 0; i < nargs; ++i)
 396                                                 next.special += '{' +
 397                                                         p.verbatim_item() +
 398                                                         '}';
 399                                         colinfo.push_back(next);
 400                                         next = ColInfo();
 401                                 } else
 402                                         cerr << "Ignoring column specification"
 403                                                 " '" << t << "'." << endl;
 404                                 break;
 405                         }
 406         }
 407
 408         // Maybe we have some column separators that need to be added to the
 409         // last column?
 410         ci2special(next);
 411         if (!next.special.empty()) {
 412                 ColInfo & ci = colinfo.back();
 413                 ci2special(ci);
 414                 ci.special += next.special;
 415                 next.special.erase();
 416         }
 417 }
 418
 419
 420 /*!
 421  * Move the left and right lines and alignment settings of the column \p ci
 422  * to the special field if necessary.
 423  */
 424 void fix_colalign(ColInfo & ci)
 425 {
 426         if (ci.leftlines > 1 || ci.rightlines > 1)
 427                 ci2special(ci);
 428 }
 429
 430
 431 /*!
 432  * LyX can't handle more than one vertical line at the left or right side
 433  * of a column.
 434  * This function moves the left and right lines and alignment settings of all
 435  * columns in \p colinfo to the special field if necessary.
 436  */
 437 void fix_colalign(vector<ColInfo> & colinfo)
 438 {
 439         // Try to move extra leftlines to the previous column.
 440         // We do this only if both special fields are empty, otherwise we
 441         // can't tell wether the result will be the same.
 442         for (size_t col = 0; col < colinfo.size(); ++col) {
 443                 if (colinfo[col].leftlines > 1 &&
 444                     colinfo[col].special.empty() && col > 0 &&
 445                     colinfo[col - 1].rightlines == 0 &&
 446                     colinfo[col - 1].special.empty()) {
 447                         ++colinfo[col - 1].rightlines;
 448                         --colinfo[col].leftlines;
 449                 }
 450         }
 451         // Try to move extra rightlines to the next column
 452         for (size_t col = 0; col < colinfo.size(); ++col) {
 453                 if (colinfo[col].rightlines > 1 &&
 454                     colinfo[col].special.empty() &&
 455                     col < colinfo.size() - 1 &&
 456                     colinfo[col + 1].leftlines == 0 &&
 457                     colinfo[col + 1].special.empty()) {
 458                         ++colinfo[col + 1].leftlines;
 459                         --colinfo[col].rightlines;
 460                 }
 461         }
 462         // Move the lines and alignment settings to the special field if
 463         // necessary
 464         for (size_t col = 0; col < colinfo.size(); ++col)
 465                 fix_colalign(colinfo[col]);
 466 }
 467
 468
 469 /*!
 470  * Parse hlines and similar stuff.
 471  * \returns wether the token \p t was parsed
 472  */
 473 bool parse_hlines(Parser & p, Token const & t, string & hlines,
 474                   bool is_long_tabular)
 475 {
 476         BOOST_ASSERT(t.cat() == catEscape);
 477
 478         if (t.cs() == "hline")
 479                 hlines += "\\hline";
 480
 481         else if (t.cs() == "cline")
 482                 hlines += "\\cline{" + p.verbatim_item() + '}';
 483
 484         else if (is_long_tabular && t.cs() == "newpage")
 485                 hlines += "\\newpage";
 486
 487         else
 488                 return false;
 489
 490         return true;
 491 }
 492
 493
 494 /// Position in a row
 495 enum RowPosition {
 496         /// At the very beginning, before the first token
 497         ROW_START,
 498         /// After the first token and before any column token
 499         IN_HLINES_START,
 500         /// After the first column token. Comments and whitespace are only
 501         /// treated as tokens in this position
 502         IN_COLUMNS,
 503         /// After the first non-column token at the end
 504         IN_HLINES_END
 505 };
 506
 507
 508 /*!
 509  * Parse table structure.
 510  * We parse tables in a two-pass process: This function extracts the table
 511  * structure (rows, columns, hlines etc.), but does not change the cell
 512  * content. The cell content is parsed in a second step in handle_tabular().
 513  */
 514 void parse_table(Parser & p, ostream & os, bool is_long_tabular,
 515                  RowPosition & pos, unsigned flags)
 516 {
 517         // table structure commands such as \hline
 518         string hlines;
 519
 520         // comments that occur at places where we can't handle them
 521         string comments;
 522
 523         while (p.good()) {
 524                 Token const & t = p.get_token();
 525
 526 #ifdef FILEDEBUG
 527                 cerr << "t: " << t << " flags: " << flags << "\n";
 528 #endif
 529
 530                 // comments and whitespace in hlines
 531                 switch (pos) {
 532                 case ROW_START:
 533                 case IN_HLINES_START:
 534                 case IN_HLINES_END:
 535                         if (t.cat() == catComment) {
 536                                 if (t.cs().empty())
 537                                         // line continuation
 538                                         p.skip_spaces();
 539                                 else
 540                                         // We can't handle comments here,
 541                                         // store them for later use
 542                                         comments += t.asInput();
 543                                 continue;
 544                         } else if (t.cat() == catSpace ||
 545                                    t.cat() == catNewline) {
 546                                 // whitespace is irrelevant here, we
 547                                 // need to recognize hline stuff
 548                                 p.skip_spaces();
 549                                 continue;
 550                         }
 551                         break;
 552                 case IN_COLUMNS:
 553                         break;
 554                 }
 555
 556                 // We need to handle structure stuff first in order to
 557                 // determine wether we need to output a HLINE separator
 558                 // before the row or not.
 559                 if (t.cat() == catEscape) {
 560                         if (parse_hlines(p, t, hlines, is_long_tabular)) {
 561                                 switch (pos) {
 562                                 case ROW_START:
 563                                         pos = IN_HLINES_START;
 564                                         break;
 565                                 case IN_COLUMNS:
 566                                         pos = IN_HLINES_END;
 567                                         break;
 568                                 case IN_HLINES_START:
 569                                 case IN_HLINES_END:
 570                                         break;
 571                                 }
 572                                 continue;
 573                         }
 574
 575                         else if (t.cs() == "tabularnewline" ||
 576                                  t.cs() == "\\" ||
 577                                  t.cs() == "cr") {
 578                                 if (t.cs() == "cr")
 579                                         cerr << "Warning: Converting TeX "
 580                                                 "'\\cr' to LaTeX '\\\\'."
 581                                              << endl;
 582                                 // stuff before the line break
 583                                 os << comments << HLINE << hlines << HLINE
 584                                    << LINE;
 585                                 //cerr << "hlines: " << hlines << endl;
 586                                 hlines.erase();
 587                                 comments.erase();
 588                                 pos = ROW_START;
 589                                 continue;
 590                         }
 591
 592                         else if (is_long_tabular &&
 593                                  (t.cs() == "endhead" ||
 594                                   t.cs() == "endfirsthead" ||
 595                                   t.cs() == "endfoot" ||
 596                                   t.cs() == "endlastfoot")) {
 597                                 hlines += t.asInput();
 598                                 switch (pos) {
 599                                 case IN_COLUMNS:
 600                                 case IN_HLINES_END:
 601                                         // these commands are implicit line
 602                                         // breaks
 603                                         os << comments << HLINE << hlines
 604                                            << HLINE << LINE;
 605                                         hlines.erase();
 606                                         comments.erase();
 607                                         pos = ROW_START;
 608                                         break;
 609                                 case ROW_START:
 610                                         pos = IN_HLINES_START;
 611                                         break;
 612                                 case IN_HLINES_START:
 613                                         break;
 614                                 }
 615                                 continue;
 616                         }
 617
 618                 }
 619
 620                 // We need a HLINE separator if we either have no hline
 621                 // stuff at all and are just starting a row or if we just
 622                 // got the first non-hline token.
 623                 switch (pos) {
 624                 case ROW_START:
 625                         // no hline tokens exist, first token at row start
 626                 case IN_HLINES_START:
 627                         // hline tokens exist, first non-hline token at row
 628                         // start
 629                         os << hlines << HLINE << comments;
 630                         hlines.erase();
 631                         comments.erase();
 632                         pos = IN_COLUMNS;
 633                         break;
 634                 case IN_HLINES_END:
 635                         // Oops, there is still cell content after hline
 636                         // stuff. This does not work in LaTeX, so we ignore
 637                         // the hlines.
 638                         cerr << "Ignoring '" << hlines << "' in a cell"
 639                              << endl;
 640                         os << comments;
 641                         hlines.erase();
 642                         comments.erase();
 643                         pos = IN_COLUMNS;
 644                         break;
 645                 case IN_COLUMNS:
 646                         break;
 647                 }
 648
 649                 // If we come here we have normal cell content
 650                 //
 651                 // cat codes
 652                 //
 653                 if (t.cat() == catMath) {
 654                         // we are inside some text mode thingy, so opening new math is allowed
 655                         Token const & n = p.get_token();
 656                         if (n.cat() == catMath) {
 657                                 // TeX's $$...$$ syntax for displayed math
 658                                 os << "\\[";
 659                                 // This does only work because parse_math outputs TeX
 660                                 parse_math(p, os, FLAG_SIMPLE, MATH_MODE);
 661                                 os << "\\]";
 662                                 p.get_token(); // skip the second '$' token
 663                         } else {
 664                                 // simple $...$  stuff
 665                                 p.putback();
 666                                 os << '$';
 667                                 // This does only work because parse_math outputs TeX
 668                                 parse_math(p, os, FLAG_SIMPLE, MATH_MODE);
 669                                 os << '$';
 670                         }
 671                 }
 672
 673                 else if (t.cat() == catSpace || t.cat() == catNewline)
 674                                 os << t.cs();
 675
 676                 else if (t.cat() == catLetter ||
 677                                t.cat() == catSuper ||
 678                                t.cat() == catSub ||
 679                                t.cat() == catOther ||
 680                                t.cat() == catActive ||
 681                                t.cat() == catParameter)
 682                         os << t.character();
 683
 684                 else if (t.cat() == catBegin) {
 685                         os << '{';
 686                         parse_table(p, os, is_long_tabular, pos,
 687                                     FLAG_BRACE_LAST);
 688                         os << '}';
 689                 }
 690
 691                 else if (t.cat() == catEnd) {
 692                         if (flags & FLAG_BRACE_LAST)
 693                                 return;
 694                         cerr << "unexpected '}'\n";
 695                 }
 696
 697                 else if (t.cat() == catAlign) {
 698                         os << TAB;
 699                         p.skip_spaces();
 700                 }
 701
 702                 else if (t.cat() == catComment)
 703                         os << t.asInput();
 704
 705                 else if (t.cs() == "(") {
 706                         os << "\\(";
 707                         // This does only work because parse_math outputs TeX
 708                         parse_math(p, os, FLAG_SIMPLE2, MATH_MODE);
 709                         os << "\\)";
 710                 }
 711
 712                 else if (t.cs() == "[") {
 713                         os << "\\[";
 714                         // This does only work because parse_math outputs TeX
 715                         parse_math(p, os, FLAG_EQUATION, MATH_MODE);
 716                         os << "\\]";
 717                 }
 718
 719                 else if (t.cs() == "begin") {
 720                         string const name = p.getArg('{', '}');
 721                         active_environments.push_back(name);
 722                         os << "\\begin{" << name << '}';
 723                         // treat the nested environment as a block, don't
 724                         // parse &, \\ etc, because they don't belong to our
 725                         // table if they appear.
 726                         os << p.verbatimEnvironment(name);
 727                         os << "\\end{" << name << '}';
 728                         active_environments.pop_back();
 729                 }
 730
 731                 else if (t.cs() == "end") {
 732                         if (flags & FLAG_END) {
 733                                 // eat environment name
 734                                 string const name = p.getArg('{', '}');
 735                                 if (name != active_environment())
 736                                         p.error("\\end{" + name + "} does not match \\begin{"
 737                                                 + active_environment() + "}");
 738                                 return;
 739                         }
 740                         p.error("found 'end' unexpectedly");
 741                 }
 742
 743                 else
 744                         os << t.asInput();
 745         }
 746
 747         // We can have comments if the last line is incomplete
 748         os << comments;
 749
 750         // We can have hline stuff if the last line is incomplete
 751         if (!hlines.empty()) {
 752                 // this does not work in LaTeX, so we ignore it
 753                 cerr << "Ignoring '" << hlines << "' at end of tabular"
 754                      << endl;
 755         }
 756 }
 757
 758
 759 void handle_hline_above(RowInfo & ri, vector<CellInfo> & ci)
 760 {
 761         ri.topline = true;
 762         for (size_t col = 0; col < ci.size(); ++col)
 763                 ci[col].topline = true;
 764 }
 765
 766
 767 void handle_hline_below(RowInfo & ri, vector<CellInfo> & ci)
 768 {
 769         ri.bottomline = true;
 770         for (size_t col = 0; col < ci.size(); ++col)
 771                 ci[col].bottomline = true;
 772 }
 773
 774
 775 } // anonymous namespace
 776
 777
 778 void handle_tabular(Parser & p, ostream & os, bool is_long_tabular,
 779                     Context & context)
 780 {
 781         string posopts = p.getOpt();
 782         if (!posopts.empty()) {
 783                 // FIXME: Convert this to ERT
 784                 if (is_long_tabular)
 785                         cerr << "horizontal longtable";
 786                 else
 787                         cerr << "vertical tabular";
 788                 cerr << " positioning '" << posopts << "' ignored\n";
 789         }
 790
 791         vector<ColInfo> colinfo;
 792
 793         // handle column formatting
 794         handle_colalign(p, colinfo, ColInfo());
 795         fix_colalign(colinfo);
 796
 797         // first scan of cells
 798         // use table mode to keep it minimal-invasive
 799         // not exactly what's TeX doing...
 800         vector<string> lines;
 801         ostringstream ss;
 802         RowPosition rowpos = ROW_START;
 803         parse_table(p, ss, is_long_tabular, rowpos, FLAG_END);
 804         split(ss.str(), lines, LINE);
 805
 806         vector< vector<CellInfo> > cellinfo(lines.size());
 807         vector<RowInfo> rowinfo(lines.size());
 808
 809         // split into rows
 810         //cerr << "// split into rows\n";
 811         for (size_t row = 0; row < rowinfo.size(); ++row) {
 812
 813                 // init row
 814                 cellinfo[row].resize(colinfo.size());
 815
 816                 // split row
 817                 vector<string> dummy;
 818                 //cerr << "\n########### LINE: " << lines[row] << "########\n";
 819                 split(lines[row], dummy, HLINE);
 820
 821                 // handle horizontal line fragments
 822                 // we do only expect this for a last line without '\\'
 823                 if (dummy.size() != 3) {
 824                         if ((dummy.size() != 1 && dummy.size() != 2) ||
 825                             row != rowinfo.size() - 1)
 826                                 cerr << "unexpected dummy size: " << dummy.size()
 827                                         << " content: " << lines[row] << "\n";
 828                         dummy.resize(3);
 829                 }
 830                 lines[row] = dummy[1];
 831
 832                 //cerr << "line: " << row << " above 0: " << dummy[0] << "\n";
 833                 //cerr << "line: " << row << " below 2: " << dummy[2] <<  "\n";
 834                 //cerr << "line: " << row << " cells 1: " << dummy[1] <<  "\n";
 835
 836                 for (int i = 0; i <= 2; i += 2) {
 837                         //cerr << "   reading from line string '" << dummy[i] << "'\n";
 838                         Parser p1(dummy[i]);
 839                         while (p1.good()) {
 840                                 Token t = p1.get_token();
 841                                 //cerr << "read token: " << t << "\n";
 842                                 if (t.cs() == "hline") {
 843                                         if (i == 0) {
 844                                                 if (rowinfo[row].topline) {
 845                                                         if (row > 0) // extra bottomline above
 846                                                                 handle_hline_below(rowinfo[row - 1], cellinfo[row - 1]);
 847                                                         else
 848                                                                 cerr << "dropping extra hline\n";
 849                                                         //cerr << "below row: " << row-1 << endl;
 850                                                 } else {
 851                                                         handle_hline_above(rowinfo[row], cellinfo[row]);
 852                                                         //cerr << "above row: " << row << endl;
 853                                                 }
 854                                         } else {
 855                                                 //cerr << "below row: " << row << endl;
 856                                                 handle_hline_below(rowinfo[row], cellinfo[row]);
 857                                         }
 858                                 } else if (t.cs() == "cline") {
 859                                         string arg = p1.verbatim_item();
 860                                         //cerr << "read cline arg: '" << arg << "'\n";
 861                                         vector<string> t;
 862                                         split(arg, t, '-');
 863                                         t.resize(2);
 864                                         size_t from = convert<unsigned int>(t[0]);
 865                                         if (from == 0)
 866                                                 cerr << "Could not parse "
 867                                                         "cline start column."
 868                                                      << endl;
 869                                         else
 870                                                 // 1 based index -> 0 based
 871                                                 --from;
 872                                         if (from >= colinfo.size()) {
 873                                                 cerr << "cline starts at non "
 874                                                         "existing column "
 875                                                      << (from + 1) << endl;
 876                                                 from = colinfo.size() - 1;
 877                                         }
 878                                         size_t to = convert<unsigned int>(t[1]);
 879                                         if (to == 0)
 880                                                 cerr << "Could not parse "
 881                                                         "cline end column."
 882                                                      << endl;
 883                                         else
 884                                                 // 1 based index -> 0 based
 885                                                 --to;
 886                                         if (to >= colinfo.size()) {
 887                                                 cerr << "cline ends at non "
 888                                                         "existing column "
 889                                                      << (to + 1) << endl;
 890                                                 to = colinfo.size() - 1;
 891                                         }
 892                                         for (size_t col = from; col <= to; ++col) {
 893                                                 //cerr << "row: " << row << " col: " << col << " i: " << i << endl;
 894                                                 if (i == 0) {
 895                                                         rowinfo[row].topline = true;
 896                                                         cellinfo[row][col].topline = true;
 897                                                 } else {
 898                                                         rowinfo[row].bottomline = true;
 899                                                         cellinfo[row][col].bottomline = true;
 900                                                 }
 901                                         }
 902                                 } else if (t.cs() == "endhead") {
 903                                         if (i > 0)
 904                                                 rowinfo[row].type = LT_HEAD;
 905                                         for (int r = row - 1; r >= 0; --r) {
 906                                                 if (rowinfo[r].type != LT_NORMAL)
 907                                                         break;
 908                                                 rowinfo[r].type = LT_HEAD;
 909                                         }
 910                                 } else if (t.cs() == "endfirsthead") {
 911                                         if (i > 0)
 912                                                 rowinfo[row].type = LT_FIRSTHEAD;
 913                                         for (int r = row - 1; r >= 0; --r) {
 914                                                 if (rowinfo[r].type != LT_NORMAL)
 915                                                         break;
 916                                                 rowinfo[r].type = LT_FIRSTHEAD;
 917                                         }
 918                                 } else if (t.cs() == "endfoot") {
 919                                         if (i > 0)
 920                                                 rowinfo[row].type = LT_FOOT;
 921                                         for (int r = row - 1; r >= 0; --r) {
 922                                                 if (rowinfo[r].type != LT_NORMAL)
 923                                                         break;
 924                                                 rowinfo[r].type = LT_FOOT;
 925                                         }
 926                                 } else if (t.cs() == "endlastfoot") {
 927                                         if (i > 0)
 928                                                 rowinfo[row].type = LT_LASTFOOT;
 929                                         for (int r = row - 1; r >= 0; --r) {
 930                                                 if (rowinfo[r].type != LT_NORMAL)
 931                                                         break;
 932                                                 rowinfo[r].type = LT_LASTFOOT;
 933                                         }
 934                                 } else if (t.cs() == "newpage") {
 935                                         if (i == 0) {
 936                                                 if (row > 0)
 937                                                         rowinfo[row - 1].newpage = true;
 938                                                 else
 939                                                         // This does not work in LaTeX
 940                                                         cerr << "Ignoring "
 941                                                                 "'\\newpage' "
 942                                                                 "before rows."
 943                                                              << endl;
 944                                         } else
 945                                                 rowinfo[row].newpage = true;
 946                                 } else {
 947                                         cerr << "unexpected line token: " << t << endl;
 948                                 }
 949                         }
 950                 }
 951
 952                 // split into cells
 953                 vector<string> cells;
 954                 split(lines[row], cells, TAB);
 955                 for (size_t col = 0, cell = 0; cell < cells.size();
 956                      ++col, ++cell) {
 957                         //cerr << "cell content: '" << cells[cell] << "'\n";
 958                         if (col >= colinfo.size()) {
 959                                 // This does not work in LaTeX
 960                                 cerr << "Ignoring extra cell '"
 961                                      << cells[cell] << "'." << endl;
 962                                 continue;
 963                         }
 964                         Parser p(cells[cell]);
 965                         p.skip_spaces();
 966                         //cells[cell] << "'\n";
 967                         if (p.next_token().cs() == "multicolumn") {
 968                                 // how many cells?
 969                                 p.get_token();
 970                                 size_t const ncells =
 971                                         convert<unsigned int>(p.verbatim_item());
 972
 973                                 // special cell properties alignment
 974                                 vector<ColInfo> t;
 975                                 handle_colalign(p, t, ColInfo());
 976                                 ColInfo & ci = t.front();
 977
 978                                 // The logic of LyX for multicolumn vertical
 979                                 // lines is too complicated to reproduce it
 980                                 // here (see LyXTabular::TeXCellPreamble()).
 981                                 // Therefore we simply put everything in the
 982                                 // special field.
 983                                 ci2special(ci);
 984
 985                                 cellinfo[row][col].multi      = CELL_BEGIN_OF_MULTICOLUMN;
 986                                 cellinfo[row][col].align      = ci.align;
 987                                 cellinfo[row][col].special    = ci.special;
 988                                 cellinfo[row][col].leftlines  = ci.leftlines;
 989                                 cellinfo[row][col].rightlines = ci.rightlines;
 990                                 ostringstream os;
 991                                 parse_text_in_inset(p, os, FLAG_ITEM, false, context);
 992                                 if (!cellinfo[row][col].content.empty()) {
 993                                         // This may or may not work in LaTeX,
 994                                         // but it does not work in LyX.
 995                                         // FIXME: Handle it correctly!
 996                                         cerr << "Moving cell content '"
 997                                              << cells[cell]
 998                                              << "' into a multicolumn cell. "
 999                                                 "This will probably not work."
1000                                              << endl;
1001                                 }
1002                                 cellinfo[row][col].content += os.str();
1003
1004                                 // add dummy cells for multicol
1005                                 for (size_t i = 0; i < ncells - 1 && col < colinfo.size(); ++i) {
1006                                         ++col;
1007                                         cellinfo[row][col].multi = CELL_PART_OF_MULTICOLUMN;
1008                                         cellinfo[row][col].align = 'c';
1009                                 }
1010
1011                         } else {
1012                                 cellinfo[row][col].leftlines  = colinfo[col].leftlines;
1013                                 cellinfo[row][col].rightlines = colinfo[col].rightlines;
1014                                 cellinfo[row][col].align      = colinfo[col].align;
1015                                 ostringstream os;
1016                                 parse_text_in_inset(p, os, FLAG_CELL, false, context);
1017                                 cellinfo[row][col].content += os.str();
1018                         }
1019                 }
1020
1021                 //cerr << "//  handle almost empty last row what we have\n";
1022                 // handle almost empty last row
1023                 if (row && lines[row].empty() && row + 1 == rowinfo.size()) {
1024                         //cerr << "remove empty last line\n";
1025                         if (rowinfo[row].topline)
1026                                 rowinfo[row - 1].bottomline = true;
1027                         for (size_t col = 0; col < colinfo.size(); ++col)
1028                                 if (cellinfo[row][col].topline)
1029                                         cellinfo[row - 1][col].bottomline = true;
1030                         rowinfo.pop_back();
1031                 }
1032         }
1033
1034         // Now we have the table structure and content in rowinfo, colinfo
1035         // and cellinfo.
1036         // Unfortunately LyX has some limitations that we need to work around.
1037
1038         // Convert cells with special content to multicolumn cells
1039         // (LyX ignores the special field for non-multicolumn cells).
1040         for (size_t row = 0; row < rowinfo.size(); ++row) {
1041                 for (size_t col = 0; col < cellinfo[row].size(); ++col) {
1042                         if (cellinfo[row][col].multi == CELL_NORMAL &&
1043                             !cellinfo[row][col].special.empty())
1044                                 cellinfo[row][col].multi = CELL_BEGIN_OF_MULTICOLUMN;
1045                 }
1046         }
1047
1048         //cerr << "// output what we have\n";
1049         // output what we have
1050         os << "\n<lyxtabular version=\"3\" rows=\"" << rowinfo.size()
1051            << "\" columns=\"" << colinfo.size() << "\">\n";
1052         os << "<features"
1053            << write_attribute("rotate", false)
1054            << write_attribute("islongtable", is_long_tabular)
1055            << ">\n";
1056
1057         //cerr << "// after header\n";
1058         for (size_t col = 0; col < colinfo.size(); ++col) {
1059                 os << "<column alignment=\""
1060                    << verbose_align(colinfo[col].align) << "\""
1061                    << " valignment=\""
1062                    << verbose_valign(colinfo[col].valign) << "\""
1063                    << write_attribute("leftline", colinfo[col].leftlines > 0)
1064                    << write_attribute("rightline", colinfo[col].rightlines > 0)
1065                    << write_attribute("width", translate_len(colinfo[col].width))
1066                    << write_attribute("special", colinfo[col].special)
1067                    << ">\n";
1068         }
1069         //cerr << "// after cols\n";
1070
1071         for (size_t row = 0; row < rowinfo.size(); ++row) {
1072                 os << "<row"
1073                    << write_attribute("topline", rowinfo[row].topline)
1074                    << write_attribute("bottomline", rowinfo[row].bottomline)
1075                    << write_attribute("endhead",
1076                                       rowinfo[row].type == LT_HEAD)
1077                    << write_attribute("endfirsthead",
1078                                       rowinfo[row].type == LT_FIRSTHEAD)
1079                    << write_attribute("endfoot",
1080                                       rowinfo[row].type == LT_FOOT)
1081                    << write_attribute("endlastfoot",
1082                                       rowinfo[row].type == LT_LASTFOOT)
1083                    << write_attribute("newpage", rowinfo[row].newpage)
1084                    << ">\n";
1085                 for (size_t col = 0; col < colinfo.size(); ++col) {
1086                         CellInfo const & cell = cellinfo[row][col];
1087                         os << "<cell";
1088                         if (cell.multi != CELL_NORMAL)
1089                                 os << " multicolumn=\"" << cell.multi << "\"";
1090                         os << " alignment=\"" << verbose_align(cell.align)
1091                            << "\""
1092                            << " valignment=\"" << verbose_valign(cell.valign)
1093                            << "\""
1094                            << write_attribute("topline", cell.topline)
1095                            << write_attribute("bottomline", cell.bottomline)
1096                            << write_attribute("leftline", cell.leftlines > 0)
1097                            << write_attribute("rightline", cell.rightlines > 0)
1098                            << write_attribute("rotate", cell.rotate);
1099                         //cerr << "\nrow: " << row << " col: " << col;
1100                         //if (cell.topline)
1101                         //      cerr << " topline=\"true\"";
1102                         //if (cell.bottomline)
1103                         //      cerr << " bottomline=\"true\"";
1104                         os << " usebox=\"none\""
1105                            << write_attribute("width", translate_len(cell.width));
1106                         if (cell.multi != CELL_NORMAL)
1107                                 os << write_attribute("special", cell.special);
1108                         os << ">"
1109                            << "\n\\begin_inset Text\n"
1110                            << cell.content
1111                            << "\n\\end_inset\n"
1112                            << "</cell>\n";
1113                 }
1114                 os << "</row>\n";
1115         }
1116
1117         os << "</lyxtabular>\n";
1118 }
1119
1120
1121
1122
1123 // }])
1124
1125
1126 } // namespace lyx