src/tex2lyx/table.cpp

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