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