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