src/tex2lyx/table.cpp

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