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 "Preamble.h"
  20
  21 #include "support/lassert.h"
  22 #include "support/convert.h"
  23 #include "support/lstrings.h"
  24
  25 #include <iostream>
  26 #include <sstream>
  27 #include <vector>
  28 #include <map>
  29
  30 using namespace std;
  31
  32 namespace lyx {
  33
  34 // filled in preamble.cpp
  35 map<char, int> special_columns;
  36
  37
  38 namespace {
  39
  40 class ColInfo {
  41 public:
  42         ColInfo() : align('n'), valign('n'), rightlines(0), leftlines(0) {}
  43         /// column alignment
  44         char align;
  45         /// vertical alignment
  46         char valign;
  47         /// column width
  48         string width;
  49         /// special column alignment
  50         string special;
  51         /// number of lines on the right
  52         int rightlines;
  53         /// number of lines on the left
  54         int leftlines;
  55 };
  56
  57
  58 /// row type for longtables
  59 enum LTRowType
  60 {
  61         /// normal row
  62         LT_NORMAL,
  63         /// part of head
  64         LT_HEAD,
  65         /// part of head on first page
  66         LT_FIRSTHEAD,
  67         /// part of foot
  68         LT_FOOT,
  69         /// part of foot on last page
  70         LT_LASTFOOT
  71 };
  72
  73
  74 class RowInfo {
  75 public:
  76         RowInfo() : topline(false), bottomline(false), type(LT_NORMAL),
  77                     caption(false), newpage(false) {}
  78         /// horizontal line above
  79         bool topline;
  80         /// horizontal line below
  81         bool bottomline;
  82         /// Extra space between the top line and this row
  83         string top_space;
  84         /// Extra space between this row and the bottom line
  85         string bottom_space;
  86         /// Extra space between the bottom line and the next top line
  87         string interline_space;
  88         /// These are for longtabulars only
  89         /// row type (head, foot, firsthead etc.)
  90         LTRowType type;
  91         /// row for a caption
  92         bool caption;
  93         /// row for a newpage
  94         bool newpage;
  95 };
  96
  97
  98 /// the numeric values are part of the file format!
  99 enum Multicolumn {
 100         /// A normal cell
 101         CELL_NORMAL = 0,
 102         /// A multicolumn cell. The number of columns is <tt>1 + number
 103         /// of CELL_PART_OF_MULTICOLUMN cells</tt> that follow directly
 104         CELL_BEGIN_OF_MULTICOLUMN = 1,
 105         /// This is a dummy cell (part of a multicolumn cell)
 106         CELL_PART_OF_MULTICOLUMN = 2
 107 };
 108
 109
 110 class CellInfo {
 111 public:
 112         CellInfo() : multi(CELL_NORMAL), align('n'), valign('n'),
 113                      leftlines(0), rightlines(0), topline(false),
 114                      bottomline(false), rotate(false) {}
 115         /// cell content
 116         string content;
 117         /// multicolumn flag
 118         Multicolumn multi;
 119         /// cell alignment
 120         char align;
 121         /// vertical cell alignment
 122         char valign;
 123         /// number of lines on the left
 124         int leftlines;
 125         /// number of lines on the right
 126         int rightlines;
 127         /// do we have a line above?
 128         bool topline;
 129         /// do we have a line below?
 130         bool bottomline;
 131         /// is the cell rotated?
 132         bool rotate;
 133         /// width for multicolumn cells
 134         string width;
 135         /// special formatting for multicolumn cells
 136         string special;
 137 };
 138
 139
 140 /// translate a horizontal alignment (as stored in ColInfo and CellInfo) to LyX
 141 inline char const * verbose_align(char c)
 142 {
 143         switch (c) {
 144         case 'c':
 145                 return "center";
 146         case 'r':
 147                 return "right";
 148         case 'l':
 149                 return "left";
 150         default:
 151                 return "none";
 152         }
 153 }
 154
 155
 156 /// translate a vertical alignment (as stored in ColInfo and CellInfo) to LyX
 157 inline char const * verbose_valign(char c)
 158 {
 159         // The default value for no special alignment is "top".
 160         switch (c) {
 161         case 'm':
 162                 return "middle";
 163         case 'b':
 164                 return "bottom";
 165         case 'p':
 166         default:
 167                 return "top";
 168         }
 169 }
 170
 171
 172 // stripped down from tabluar.C. We use it currently only for bools and
 173 // strings
 174 string const write_attribute(string const & name, bool const & b)
 175 {
 176         // we write only true attribute values so we remove a bit of the
 177         // file format bloat for tabulars.
 178         return b ? ' ' + name + "=\"true\"" : string();
 179 }
 180
 181
 182 string const write_attribute(string const & name, string const & s)
 183 {
 184         return s.empty() ? string() : ' ' + name + "=\"" + s + '"';
 185 }
 186
 187
 188 /*! rather brutish way to code table structure in a string:
 189
 190 \verbatim
 191   \begin{tabular}{ccc}
 192     1 & 2 & 3\\ \hline
 193     \multicolumn{2}{c}{4} & 5 //
 194     6 & 7 \\
 195     8 \endhead
 196   \end{tabular}
 197 \endverbatim
 198
 199  gets "translated" to:
 200
 201 \verbatim
 202          HLINE 1                     TAB 2 TAB 3 HLINE          HLINE LINE
 203   \hline HLINE \multicolumn{2}{c}{4} TAB 5       HLINE          HLINE LINE
 204          HLINE 6                     TAB 7       HLINE          HLINE LINE
 205          HLINE 8                                 HLINE \endhead HLINE LINE
 206 \endverbatim
 207  */
 208
 209 char const TAB   = '\001';
 210 char const LINE  = '\002';
 211 char const HLINE = '\004';
 212
 213
 214 /*!
 215  * Move the information in leftlines, rightlines, align and valign to the
 216  * special field. This is necessary if the special field is not empty,
 217  * because LyX ignores leftlines > 1, rightlines > 1, align and valign in
 218  * this case.
 219  */
 220 void ci2special(ColInfo & ci)
 221 {
 222         if (ci.width.empty() && ci.align == 'n')
 223                 // The alignment setting is already in special, since
 224                 // handle_colalign() never stores ci with these settings
 225                 // and ensures that leftlines == 0 and rightlines == 0 in
 226                 // this case.
 227                 return;
 228
 229         if (!ci.width.empty()) {
 230                 switch (ci.align) {
 231                 case 'l':
 232                         ci.special += ">{\\raggedright}";
 233                         break;
 234                 case 'r':
 235                         ci.special += ">{\\raggedleft}";
 236                         break;
 237                 case 'c':
 238                         ci.special += ">{\\centering}";
 239                         break;
 240                 }
 241                 if (ci.valign == 'n')
 242                         ci.special += 'p';
 243                 else
 244                         ci.special += ci.valign;
 245                 ci.special += '{' + ci.width + '}';
 246                 ci.width.erase();
 247         } else
 248                 ci.special += ci.align;
 249
 250         // LyX can only have one left and one right line.
 251         for (int i = 1; i < ci.leftlines; ++i)
 252                 ci.special.insert(0, "|");
 253         for (int i = 1; i < ci.rightlines; ++i)
 254                 ci.special += '|';
 255         ci.leftlines = min(ci.leftlines, 1);
 256         ci.rightlines = min(ci.rightlines, 1);
 257         ci.align = 'n';
 258         ci.valign = 'n';
 259 }
 260
 261
 262 /*!
 263  * Handle column specifications for tabulars and multicolumns.
 264  * The next token of the parser \p p must be an opening brace, and we read
 265  * everything until the matching closing brace.
 266  * The resulting column specifications are filled into \p colinfo. This is
 267  * in an intermediate form. fix_colalign() makes it suitable for LyX output.
 268  */
 269 void handle_colalign(Parser & p, vector<ColInfo> & colinfo,
 270                      ColInfo const & start)
 271 {
 272         if (p.get_token().cat() != catBegin)
 273                 cerr << "Wrong syntax for table column alignment.\n"
 274                         "Expected '{', got '" << p.curr_token().asInput()
 275                      << "'.\n";
 276
 277         ColInfo next = start;
 278         for (Token t = p.get_token(); p.good() && t.cat() != catEnd;
 279              t = p.get_token()) {
 280 #ifdef FILEDEBUG
 281                 cerr << "t: " << t << "  c: '" << t.character() << "'\n";
 282 #endif
 283
 284                 // We cannot handle comments here
 285                 if (t.cat() == catComment) {
 286                         if (t.cs().empty()) {
 287                                 // "%\n" combination
 288                                 p.skip_spaces();
 289                         } else
 290                                 cerr << "Ignoring comment: " << t.asInput();
 291                         continue;
 292                 }
 293
 294                 switch (t.character()) {
 295                         case 'c':
 296                         case 'l':
 297                         case 'r':
 298                                 // new column, horizontal aligned
 299                                 next.align = t.character();
 300                                 if (!next.special.empty())
 301                                         ci2special(next);
 302                                 colinfo.push_back(next);
 303                                 next = ColInfo();
 304                                 break;
 305                         case 'p':
 306                         case 'b':
 307                         case 'm':
 308                                 // new column, vertical aligned box
 309                                 next.valign = t.character();
 310                                 next.width = p.verbatim_item();
 311                                 if (!next.special.empty())
 312                                         ci2special(next);
 313                                 colinfo.push_back(next);
 314                                 next = ColInfo();
 315                                 break;
 316                         case '|':
 317                                 // vertical rule
 318                                 if (colinfo.empty()) {
 319                                         if (next.special.empty())
 320                                                 ++next.leftlines;
 321                                         else
 322                                                 next.special += '|';
 323                                 } else if (colinfo.back().special.empty())
 324                                         ++colinfo.back().rightlines;
 325                                 else if (next.special.empty())
 326                                         ++next.leftlines;
 327                                 else
 328                                         colinfo.back().special += '|';
 329                                 break;
 330                         case '>': {
 331                                 // text before the next column
 332                                 string const s = trimSpaceAndEol(p.verbatim_item());
 333                                 if (next.special.empty() &&
 334                                     next.align == 'n') {
 335                                         // Maybe this can be converted to a
 336                                         // horizontal alignment setting for
 337                                         // fixed width columns
 338                                         if (s == "\\raggedleft")
 339                                                 next.align = 'r';
 340                                         else if (s == "\\raggedright")
 341                                                 next.align = 'l';
 342                                         else if (s == "\\centering")
 343                                                 next.align = 'c';
 344                                         else
 345                                                 next.special = ">{" + s + '}';
 346                                 } else
 347                                         next.special += ">{" + s + '}';
 348                                 break;
 349                         }
 350                         case '<': {
 351                                 // text after the last column
 352                                 string const s = trimSpaceAndEol(p.verbatim_item());
 353                                 if (colinfo.empty())
 354                                         // This is not possible in LaTeX.
 355                                         cerr << "Ignoring separator '<{"
 356                                              << s << "}'." << endl;
 357                                 else {
 358                                         ColInfo & ci = colinfo.back();
 359                                         ci2special(ci);
 360                                         ci.special += "<{" + s + '}';
 361                                 }
 362                                 break;
 363                         }
 364                         case '*': {
 365                                 // *{n}{arg} means 'n' columns of type 'arg'
 366                                 string const num = p.verbatim_item();
 367                                 string const arg = p.verbatim_item();
 368                                 size_t const n = convert<unsigned int>(num);
 369                                 if (!arg.empty() && n > 0) {
 370                                         string s("{");
 371                                         for (size_t i = 0; i < n; ++i)
 372                                                 s += arg;
 373                                         s += '}';
 374                                         Parser p2(s);
 375                                         handle_colalign(p2, colinfo, next);
 376                                         next = ColInfo();
 377                                 } else {
 378                                         cerr << "Ignoring column specification"
 379                                                 " '*{" << num << "}{"
 380                                              << arg << "}'." << endl;
 381                                 }
 382                                 break;
 383                         }
 384                         case '@':
 385                                 // text instead of the column spacing
 386                         case '!':
 387                                 // text in addition to the column spacing
 388                                 next.special += t.character();
 389                                 next.special += '{' + p.verbatim_item() + '}';
 390                                 break;
 391                         default:
 392                                 // try user defined column types
 393                                 if (special_columns.find(t.character()) !=
 394                                     special_columns.end()) {
 395                                         ci2special(next);
 396                                         next.special += t.character();
 397                                         int const nargs =
 398                                                 special_columns[t.character()];
 399                                         for (int i = 0; i < nargs; ++i)
 400                                                 next.special += '{' +
 401                                                         p.verbatim_item() +
 402                                                         '}';
 403                                         colinfo.push_back(next);
 404                                         next = ColInfo();
 405                                 } else
 406                                         cerr << "Ignoring column specification"
 407                                                 " '" << t << "'." << endl;
 408                                 break;
 409                         }
 410         }
 411
 412         // Maybe we have some column separators that need to be added to the
 413         // last column?
 414         ci2special(next);
 415         if (!next.special.empty()) {
 416                 ColInfo & ci = colinfo.back();
 417                 ci2special(ci);
 418                 ci.special += next.special;
 419                 next.special.erase();
 420         }
 421 }
 422
 423
 424 /*!
 425  * Move the left and right lines and alignment settings of the column \p ci
 426  * to the special field if necessary.
 427  */
 428 void fix_colalign(ColInfo & ci)
 429 {
 430         if (ci.leftlines > 1 || ci.rightlines > 1)
 431                 ci2special(ci);
 432 }
 433
 434
 435 /*!
 436  * LyX can't handle more than one vertical line at the left or right side
 437  * of a column.
 438  * This function moves the left and right lines and alignment settings of all
 439  * columns in \p colinfo to the special field if necessary.
 440  */
 441 void fix_colalign(vector<ColInfo> & colinfo)
 442 {
 443         // Try to move extra leftlines to the previous column.
 444         // We do this only if both special fields are empty, otherwise we
 445         // can't tell wether the result will be the same.
 446         for (size_t col = 0; col < colinfo.size(); ++col) {
 447                 if (colinfo[col].leftlines > 1 &&
 448                     colinfo[col].special.empty() && col > 0 &&
 449                     colinfo[col - 1].rightlines == 0 &&
 450                     colinfo[col - 1].special.empty()) {
 451                         ++colinfo[col - 1].rightlines;
 452                         --colinfo[col].leftlines;
 453                 }
 454         }
 455         // Try to move extra rightlines to the next column
 456         for (size_t col = 0; col < colinfo.size(); ++col) {
 457                 if (colinfo[col].rightlines > 1 &&
 458                     colinfo[col].special.empty() &&
 459                     col < colinfo.size() - 1 &&
 460                     colinfo[col + 1].leftlines == 0 &&
 461                     colinfo[col + 1].special.empty()) {
 462                         ++colinfo[col + 1].leftlines;
 463                         --colinfo[col].rightlines;
 464                 }
 465         }
 466         // Move the lines and alignment settings to the special field if
 467         // necessary
 468         for (size_t col = 0; col < colinfo.size(); ++col)
 469                 fix_colalign(colinfo[col]);
 470 }
 471
 472
 473 /*!
 474  * Parse hlines and similar stuff.
 475  * \returns wether the token \p t was parsed
 476  */
 477 bool parse_hlines(Parser & p, Token const & t, string & hlines,
 478                   bool is_long_tabular)
 479 {
 480         LASSERT(t.cat() == catEscape, return false);
 481
 482         if (t.cs() == "hline" || t.cs() == "toprule" || t.cs() == "midrule" ||
 483             t.cs() == "bottomrule")
 484                 hlines += '\\' + t.cs();
 485
 486         else if (t.cs() == "cline")
 487                 hlines += "\\cline{" + p.verbatim_item() + '}';
 488
 489         else if (t.cs() == "cmidrule") {
 490                 // We cannot handle the \cmidrule(l){3-4} form
 491                 p.pushPosition();
 492                 p.skip_spaces(true);
 493                 bool const hasParentheses(p.getFullArg('(', ')').first);
 494                 p.popPosition();
 495                 if (hasParentheses)
 496                         return false;
 497                 hlines += "\\cmidrule{" + p.verbatim_item() + '}';
 498         }
 499
 500         else if (t.cs() == "addlinespace") {
 501                 p.pushPosition();
 502                 p.skip_spaces(true);
 503                 bool const hasArgument(p.getFullArg('{', '}').first);
 504                 p.popPosition();
 505                 if (hasArgument)
 506                         hlines += "\\addlinespace{" + p.verbatim_item() + '}';
 507                 else
 508                         hlines += "\\addlinespace";
 509         }
 510
 511         else if (is_long_tabular && t.cs() == "newpage")
 512                 hlines += "\\newpage";
 513
 514         else
 515                 return false;
 516
 517         return true;
 518 }
 519
 520
 521 /// Position in a row
 522 enum RowPosition {
 523         /// At the very beginning, before the first token
 524         ROW_START,
 525         /// After the first token and before any column token
 526         IN_HLINES_START,
 527         /// After the first column token. Comments and whitespace are only
 528         /// treated as tokens in this position
 529         IN_COLUMNS,
 530         /// After the first non-column token at the end
 531         IN_HLINES_END
 532 };
 533
 534
 535 /*!
 536  * Parse table structure.
 537  * We parse tables in a two-pass process: This function extracts the table
 538  * structure (rows, columns, hlines etc.), but does not change the cell
 539  * content. The cell content is parsed in a second step in handle_tabular().
 540  */
 541 void parse_table(Parser & p, ostream & os, bool is_long_tabular,
 542                  RowPosition & pos, unsigned flags)
 543 {
 544         // table structure commands such as \hline
 545         string hlines;
 546
 547         // comments that occur at places where we can't handle them
 548         string comments;
 549
 550         while (p.good()) {
 551                 Token const & t = p.get_token();
 552
 553 #ifdef FILEDEBUG
 554                 debugToken(cerr, t, flags);
 555 #endif
 556
 557                 // comments and whitespace in hlines
 558                 switch (pos) {
 559                 case ROW_START:
 560                 case IN_HLINES_START:
 561                 case IN_HLINES_END:
 562                         if (t.cat() == catComment) {
 563                                 if (t.cs().empty())
 564                                         // line continuation
 565                                         p.skip_spaces();
 566                                 else
 567                                         // We can't handle comments here,
 568                                         // store them for later use
 569                                         comments += t.asInput();
 570                                 continue;
 571                         } else if (t.cat() == catSpace ||
 572                                    t.cat() == catNewline) {
 573                                 // whitespace is irrelevant here, we
 574                                 // need to recognize hline stuff
 575                                 p.skip_spaces();
 576                                 continue;
 577                         }
 578                         break;
 579                 case IN_COLUMNS:
 580                         break;
 581                 }
 582
 583                 // We need to handle structure stuff first in order to
 584                 // determine wether we need to output a HLINE separator
 585                 // before the row or not.
 586                 if (t.cat() == catEscape) {
 587                         if (parse_hlines(p, t, hlines, is_long_tabular)) {
 588                                 switch (pos) {
 589                                 case ROW_START:
 590                                         pos = IN_HLINES_START;
 591                                         break;
 592                                 case IN_COLUMNS:
 593                                         pos = IN_HLINES_END;
 594                                         break;
 595                                 case IN_HLINES_START:
 596                                 case IN_HLINES_END:
 597                                         break;
 598                                 }
 599                                 continue;
 600                         }
 601
 602                         else if (t.cs() == "tabularnewline" ||
 603                                  t.cs() == "\\" ||
 604                                  t.cs() == "cr") {
 605                                 if (t.cs() == "cr")
 606                                         cerr << "Warning: Converting TeX "
 607                                                 "'\\cr' to LaTeX '\\\\'."
 608                                              << endl;
 609                                 // stuff before the line break
 610                                 os << comments << HLINE << hlines << HLINE
 611                                    << LINE;
 612                                 //cerr << "hlines: " << hlines << endl;
 613                                 hlines.erase();
 614                                 comments.erase();
 615                                 pos = ROW_START;
 616                                 continue;
 617                         }
 618
 619                         else if (is_long_tabular &&
 620                                  (t.cs() == "endhead" ||
 621                                   t.cs() == "endfirsthead" ||
 622                                   t.cs() == "endfoot" ||
 623                                   t.cs() == "endlastfoot")) {
 624                                 hlines += t.asInput();
 625                                 switch (pos) {
 626                                 case IN_COLUMNS:
 627                                 case IN_HLINES_END:
 628                                         // these commands are implicit line
 629                                         // breaks
 630                                         os << comments << HLINE << hlines
 631                                            << HLINE << LINE;
 632                                         hlines.erase();
 633                                         comments.erase();
 634                                         pos = ROW_START;
 635                                         break;
 636                                 case ROW_START:
 637                                         pos = IN_HLINES_START;
 638                                         break;
 639                                 case IN_HLINES_START:
 640                                         break;
 641                                 }
 642                                 continue;
 643                         }
 644
 645                 }
 646
 647                 // We need a HLINE separator if we either have no hline
 648                 // stuff at all and are just starting a row or if we just
 649                 // got the first non-hline token.
 650                 switch (pos) {
 651                 case ROW_START:
 652                         // no hline tokens exist, first token at row start
 653                 case IN_HLINES_START:
 654                         // hline tokens exist, first non-hline token at row
 655                         // start
 656                         os << hlines << HLINE << comments;
 657                         hlines.erase();
 658                         comments.erase();
 659                         pos = IN_COLUMNS;
 660                         break;
 661                 case IN_HLINES_END:
 662                         // Oops, there is still cell content or unsupported
 663                         // booktabs commands after hline stuff. The latter are
 664                         // moved to the cell, and the first does not work in
 665                         // LaTeX, so we ignore the hlines.
 666                         os << comments;
 667                         comments.erase();
 668                         if (support::contains(hlines, "\\hline") ||
 669                             support::contains(hlines, "\\cline") ||
 670                             support::contains(hlines, "\\newpage"))
 671                                 cerr << "Ignoring '" << hlines
 672                                      << "' in a cell" << endl;
 673                         else
 674                                 os << hlines;
 675                         hlines.erase();
 676                         pos = IN_COLUMNS;
 677                         break;
 678                 case IN_COLUMNS:
 679                         break;
 680                 }
 681
 682                 // If we come here we have normal cell content
 683                 //
 684                 // cat codes
 685                 //
 686                 if (t.cat() == catMath) {
 687                         // we are inside some text mode thingy, so opening new math is allowed
 688                         Token const & n = p.get_token();
 689                         if (n.cat() == catMath) {
 690                                 // TeX's $$...$$ syntax for displayed math
 691                                 os << "\\[";
 692                                 // This does only work because parse_math outputs TeX
 693                                 parse_math(p, os, FLAG_SIMPLE, MATH_MODE);
 694                                 os << "\\]";
 695                                 p.get_token(); // skip the second '$' token
 696                         } else {
 697                                 // simple $...$  stuff
 698                                 p.putback();
 699                                 os << '$';
 700                                 // This does only work because parse_math outputs TeX
 701                                 parse_math(p, os, FLAG_SIMPLE, MATH_MODE);
 702                                 os << '$';
 703                         }
 704                 }
 705
 706                 else if (t.cat() == catSpace
 707                          || t.cat() == catNewline
 708                          || t.cat() == catLetter
 709                          || t.cat() == catSuper
 710                          || t.cat() == catSub
 711                          || t.cat() == catOther
 712                          || t.cat() == catActive
 713                          || t.cat() == catParameter)
 714                         os << t.cs();
 715
 716                 else if (t.cat() == catBegin) {
 717                         os << '{';
 718                         parse_table(p, os, is_long_tabular, pos,
 719                                     FLAG_BRACE_LAST);
 720                         os << '}';
 721                 }
 722
 723                 else if (t.cat() == catEnd) {
 724                         if (flags & FLAG_BRACE_LAST)
 725                                 return;
 726                         cerr << "unexpected '}'\n";
 727                 }
 728
 729                 else if (t.cat() == catAlign) {
 730                         os << TAB;
 731                         p.skip_spaces();
 732                 }
 733
 734                 else if (t.cat() == catComment)
 735                         os << t.asInput();
 736
 737                 else if (t.cs() == "(") {
 738                         os << "\\(";
 739                         // This does only work because parse_math outputs TeX
 740                         parse_math(p, os, FLAG_SIMPLE2, MATH_MODE);
 741                         os << "\\)";
 742                 }
 743
 744                 else if (t.cs() == "[") {
 745                         os << "\\[";
 746                         // This does only work because parse_math outputs TeX
 747                         parse_math(p, os, FLAG_EQUATION, MATH_MODE);
 748                         os << "\\]";
 749                 }
 750
 751                 else if (t.cs() == "begin") {
 752                         string const name = p.getArg('{', '}');
 753                         active_environments.push_back(name);
 754                         os << "\\begin{" << name << '}';
 755                         // treat the nested environment as a block, don't
 756                         // parse &, \\ etc, because they don't belong to our
 757                         // table if they appear.
 758                         os << p.verbatimEnvironment(name);
 759                         os << "\\end{" << name << '}';
 760                         active_environments.pop_back();
 761                 }
 762
 763                 else if (t.cs() == "end") {
 764                         if (flags & FLAG_END) {
 765                                 // eat environment name
 766                                 string const name = p.getArg('{', '}');
 767                                 if (name != active_environment())
 768                                         p.error("\\end{" + name + "} does not match \\begin{"
 769                                                 + active_environment() + "}");
 770                                 return;
 771                         }
 772                         p.error("found 'end' unexpectedly");
 773                 }
 774
 775                 else
 776                         os << t.asInput();
 777         }
 778
 779         // We can have comments if the last line is incomplete
 780         os << comments;
 781
 782         // We can have hline stuff if the last line is incomplete
 783         if (!hlines.empty()) {
 784                 // this does not work in LaTeX, so we ignore it
 785                 cerr << "Ignoring '" << hlines << "' at end of tabular"
 786                      << endl;
 787         }
 788 }
 789
 790
 791 void handle_hline_above(RowInfo & ri, vector<CellInfo> & ci)
 792 {
 793         ri.topline = true;
 794         for (size_t col = 0; col < ci.size(); ++col)
 795                 ci[col].topline = true;
 796 }
 797
 798
 799 void handle_hline_below(RowInfo & ri, vector<CellInfo> & ci)
 800 {
 801         ri.bottomline = true;
 802         for (size_t col = 0; col < ci.size(); ++col)
 803                 ci[col].bottomline = true;
 804 }
 805
 806
 807 } // anonymous namespace
 808
 809
 810 void handle_tabular(Parser & p, ostream & os, string const & name,
 811                     string const & tabularwidth, Context & context)
 812 {
 813         bool const is_long_tabular(name == "longtable");
 814         bool booktabs = false;
 815         string tabularvalignment("middle");
 816         string posopts = p.getOpt();
 817         if (!posopts.empty()) {
 818                 // FIXME: Convert this to ERT
 819                 if (is_long_tabular)
 820                         cerr << "horizontal longtable positioning '"
 821                              << posopts << "' ignored\n";
 822                 else if (posopts == "[t]")
 823                         tabularvalignment = "top";
 824                 else if (posopts == "[b]")
 825                         tabularvalignment = "bottom";
 826                 else
 827                         cerr << "vertical tabular positioning '"
 828                              << posopts << "' ignored\n";
 829         }
 830
 831         vector<ColInfo> colinfo;
 832
 833         // handle column formatting
 834         handle_colalign(p, colinfo, ColInfo());
 835         fix_colalign(colinfo);
 836
 837         // first scan of cells
 838         // use table mode to keep it minimal-invasive
 839         // not exactly what's TeX doing...
 840         vector<string> lines;
 841         ostringstream ss;
 842         RowPosition rowpos = ROW_START;
 843         parse_table(p, ss, is_long_tabular, rowpos, FLAG_END);
 844         split(ss.str(), lines, LINE);
 845
 846         vector< vector<CellInfo> > cellinfo(lines.size());
 847         vector<RowInfo> rowinfo(lines.size());
 848
 849         // split into rows
 850         //cerr << "// split into rows\n";
 851         for (size_t row = 0; row < rowinfo.size(); ++row) {
 852
 853                 // init row
 854                 cellinfo[row].resize(colinfo.size());
 855
 856                 // split row
 857                 vector<string> dummy;
 858                 //cerr << "\n########### LINE: " << lines[row] << "########\n";
 859                 split(lines[row], dummy, HLINE);
 860
 861                 // handle horizontal line fragments
 862                 // we do only expect this for a last line without '\\'
 863                 if (dummy.size() != 3) {
 864                         if ((dummy.size() != 1 && dummy.size() != 2) ||
 865                             row != rowinfo.size() - 1)
 866                                 cerr << "unexpected dummy size: " << dummy.size()
 867                                         << " content: " << lines[row] << "\n";
 868                         dummy.resize(3);
 869                 }
 870                 lines[row] = dummy[1];
 871
 872                 //cerr << "line: " << row << " above 0: " << dummy[0] << "\n";
 873                 //cerr << "line: " << row << " below 2: " << dummy[2] <<  "\n";
 874                 //cerr << "line: " << row << " cells 1: " << dummy[1] <<  "\n";
 875
 876                 for (int i = 0; i <= 2; i += 2) {
 877                         //cerr << "   reading from line string '" << dummy[i] << "'\n";
 878                         Parser p1(dummy[i]);
 879                         while (p1.good()) {
 880                                 Token t = p1.get_token();
 881                                 //cerr << "read token: " << t << "\n";
 882                                 if (t.cs() == "hline" || t.cs() == "toprule" ||
 883                                     t.cs() == "midrule" ||
 884                                     t.cs() == "bottomrule") {
 885                                         if (t.cs() != "hline")
 886                                                 booktabs = true;
 887                                         if (i == 0) {
 888                                                 if (rowinfo[row].topline) {
 889                                                         if (row > 0) // extra bottomline above
 890                                                                 handle_hline_below(rowinfo[row - 1], cellinfo[row - 1]);
 891                                                         else
 892                                                                 cerr << "dropping extra "
 893                                                                      << t.cs() << '\n';
 894                                                         //cerr << "below row: " << row-1 << endl;
 895                                                 } else {
 896                                                         handle_hline_above(rowinfo[row], cellinfo[row]);
 897                                                         //cerr << "above row: " << row << endl;
 898                                                 }
 899                                         } else {
 900                                                 //cerr << "below row: " << row << endl;
 901                                                 handle_hline_below(rowinfo[row], cellinfo[row]);
 902                                         }
 903                                 } else if (t.cs() == "cline" || t.cs() == "cmidrule") {
 904                                         if (t.cs() == "cmidrule")
 905                                                 booktabs = true;
 906                                         string arg = p1.verbatim_item();
 907                                         //cerr << "read " << t.cs() << " arg: '" << arg << "'\n";
 908                                         vector<string> cols;
 909                                         split(arg, cols, '-');
 910                                         cols.resize(2);
 911                                         size_t from = convert<unsigned int>(cols[0]);
 912                                         if (from == 0)
 913                                                 cerr << "Could not parse "
 914                                                      << t.cs() << " start column."
 915                                                      << endl;
 916                                         else
 917                                                 // 1 based index -> 0 based
 918                                                 --from;
 919                                         if (from >= colinfo.size()) {
 920                                                 cerr << t.cs() << " starts at "
 921                                                         "non existing column "
 922                                                      << (from + 1) << endl;
 923                                                 from = colinfo.size() - 1;
 924                                         }
 925                                         size_t to = convert<unsigned int>(cols[1]);
 926                                         if (to == 0)
 927                                                 cerr << "Could not parse "
 928                                                      << t.cs() << " end column."
 929                                                      << endl;
 930                                         else
 931                                                 // 1 based index -> 0 based
 932                                                 --to;
 933                                         if (to >= colinfo.size()) {
 934                                                 cerr << t.cs() << " ends at "
 935                                                         "non existing column "
 936                                                      << (to + 1) << endl;
 937                                                 to = colinfo.size() - 1;
 938                                         }
 939                                         for (size_t col = from; col <= to; ++col) {
 940                                                 //cerr << "row: " << row << " col: " << col << " i: " << i << endl;
 941                                                 if (i == 0) {
 942                                                         rowinfo[row].topline = true;
 943                                                         cellinfo[row][col].topline = true;
 944                                                 } else {
 945                                                         rowinfo[row].bottomline = true;
 946                                                         cellinfo[row][col].bottomline = true;
 947                                                 }
 948                                         }
 949                                 } else if (t.cs() == "addlinespace") {
 950                                         booktabs = true;
 951                                         string const opt = p.next_token().cat() == catBegin ?
 952                                                         p.verbatim_item() : string();
 953                                         if (i == 0) {
 954                                                 if (opt.empty())
 955                                                         rowinfo[row].top_space = "default";
 956                                                 else
 957                                                         rowinfo[row].top_space = translate_len(opt);
 958                                         } else if (rowinfo[row].bottomline) {
 959                                                 if (opt.empty())
 960                                                         rowinfo[row].bottom_space = "default";
 961                                                 else
 962                                                         rowinfo[row].bottom_space = translate_len(opt);
 963                                         } else {
 964                                                 if (opt.empty())
 965                                                         rowinfo[row].interline_space = "default";
 966                                                 else
 967                                                         rowinfo[row].interline_space = translate_len(opt);
 968                                         }
 969                                 } else if (t.cs() == "endhead") {
 970                                         if (i > 0)
 971                                                 rowinfo[row].type = LT_HEAD;
 972                                         for (int r = row - 1; r >= 0; --r) {
 973                                                 if (rowinfo[r].type != LT_NORMAL)
 974                                                         break;
 975                                                 rowinfo[r].type = LT_HEAD;
 976                                         }
 977                                 } else if (t.cs() == "endfirsthead") {
 978                                         if (i > 0)
 979                                                 rowinfo[row].type = LT_FIRSTHEAD;
 980                                         for (int r = row - 1; r >= 0; --r) {
 981                                                 if (rowinfo[r].type != LT_NORMAL)
 982                                                         break;
 983                                                 rowinfo[r].type = LT_FIRSTHEAD;
 984                                         }
 985                                 } else if (t.cs() == "endfoot") {
 986                                         if (i > 0)
 987                                                 rowinfo[row].type = LT_FOOT;
 988                                         for (int r = row - 1; r >= 0; --r) {
 989                                                 if (rowinfo[r].type != LT_NORMAL)
 990                                                         break;
 991                                                 rowinfo[r].type = LT_FOOT;
 992                                         }
 993                                 } else if (t.cs() == "endlastfoot") {
 994                                         if (i > 0)
 995                                                 rowinfo[row].type = LT_LASTFOOT;
 996                                         for (int r = row - 1; r >= 0; --r) {
 997                                                 if (rowinfo[r].type != LT_NORMAL)
 998                                                         break;
 999                                                 rowinfo[r].type = LT_LASTFOOT;
1000                                         }
1001                                 } else if (t.cs() == "newpage") {
1002                                         if (i == 0) {
1003                                                 if (row > 0)
1004                                                         rowinfo[row - 1].newpage = true;
1005                                                 else
1006                                                         // This does not work in LaTeX
1007                                                         cerr << "Ignoring "
1008                                                                 "'\\newpage' "
1009                                                                 "before rows."
1010                                                              << endl;
1011                                         } else
1012                                                 rowinfo[row].newpage = true;
1013                                 } else {
1014                                         cerr << "unexpected line token: " << t << endl;
1015                                 }
1016                         }
1017                 }
1018
1019                 // split into cells
1020                 vector<string> cells;
1021                 split(lines[row], cells, TAB);
1022                 for (size_t col = 0, cell = 0; cell < cells.size();
1023                      ++col, ++cell) {
1024                         //cerr << "cell content: '" << cells[cell] << "'\n";
1025                         if (col >= colinfo.size()) {
1026                                 // This does not work in LaTeX
1027                                 cerr << "Ignoring extra cell '"
1028                                      << cells[cell] << "'." << endl;
1029                                 continue;
1030                         }
1031                         Parser p(cells[cell]);
1032                         p.skip_spaces();
1033                         //cells[cell] << "'\n";
1034                         if (p.next_token().cs() == "multicolumn") {
1035                                 // how many cells?
1036                                 p.get_token();
1037                                 size_t const ncells =
1038                                         convert<unsigned int>(p.verbatim_item());
1039
1040                                 // special cell properties alignment
1041                                 vector<ColInfo> t;
1042                                 handle_colalign(p, t, ColInfo());
1043                                 p.skip_spaces(true);
1044                                 ColInfo & ci = t.front();
1045
1046                                 // The logic of LyX for multicolumn vertical
1047                                 // lines is too complicated to reproduce it
1048                                 // here (see LyXTabular::TeXCellPreamble()).
1049                                 // Therefore we simply put everything in the
1050                                 // special field.
1051                                 ci2special(ci);
1052
1053                                 cellinfo[row][col].multi      = CELL_BEGIN_OF_MULTICOLUMN;
1054                                 cellinfo[row][col].align      = ci.align;
1055                                 cellinfo[row][col].special    = ci.special;
1056                                 cellinfo[row][col].leftlines  = ci.leftlines;
1057                                 cellinfo[row][col].rightlines = ci.rightlines;
1058                                 ostringstream os;
1059                                 parse_text_in_inset(p, os, FLAG_ITEM, false, context);
1060                                 if (!cellinfo[row][col].content.empty()) {
1061                                         // This may or may not work in LaTeX,
1062                                         // but it does not work in LyX.
1063                                         // FIXME: Handle it correctly!
1064                                         cerr << "Moving cell content '"
1065                                              << cells[cell]
1066                                              << "' into a multicolumn cell. "
1067                                                 "This will probably not work."
1068                                              << endl;
1069                                 }
1070                                 cellinfo[row][col].content += os.str();
1071
1072                                 // add dummy cells for multicol
1073                                 for (size_t i = 0; i < ncells - 1 && col < colinfo.size(); ++i) {
1074                                         ++col;
1075                                         cellinfo[row][col].multi = CELL_PART_OF_MULTICOLUMN;
1076                                         cellinfo[row][col].align = 'c';
1077                                 }
1078
1079                         } else if (col == 0 && is_long_tabular &&
1080                                    p.next_token().cs() == "caption") {
1081                                 // longtable caption support in LyX is a hack:
1082                                 // Captions require a row of their own with
1083                                 // the caption flag set to true, having only
1084                                 // one multicolumn cell. The contents of that
1085                                 // cell must contain exactly one caption inset
1086                                 // and nothing else.
1087                                 rowinfo[row].caption = true;
1088                                 for (size_t c = 1; c < cells.size(); ++c) {
1089                                         if (!cells[c].empty()) {
1090                                                 cerr << "Moving cell content '"
1091                                                      << cells[c]
1092                                                      << "' into the caption cell. "
1093                                                         "This will probably not work."
1094                                                      << endl;
1095                                                 cells[0] += cells[c];
1096                                         }
1097                                 }
1098                                 cells.resize(1);
1099                                 cellinfo[row][col].align      = colinfo[col].align;
1100                                 cellinfo[row][col].multi      = CELL_BEGIN_OF_MULTICOLUMN;
1101                                 ostringstream os;
1102                                 parse_text_in_inset(p, os, FLAG_CELL, false, context);
1103                                 cellinfo[row][col].content += os.str();
1104                                 // add dummy multicolumn cells
1105                                 for (size_t c = 1; c < colinfo.size(); ++c)
1106                                         cellinfo[row][c].multi = CELL_PART_OF_MULTICOLUMN;
1107
1108                         } else {
1109                                 cellinfo[row][col].leftlines  = colinfo[col].leftlines;
1110                                 cellinfo[row][col].rightlines = colinfo[col].rightlines;
1111                                 cellinfo[row][col].align      = colinfo[col].align;
1112                                 ostringstream os;
1113                                 parse_text_in_inset(p, os, FLAG_CELL, false, context);
1114                                 cellinfo[row][col].content += os.str();
1115                         }
1116                 }
1117
1118                 //cerr << "//  handle almost empty last row what we have\n";
1119                 // handle almost empty last row
1120                 if (row && lines[row].empty() && row + 1 == rowinfo.size()) {
1121                         //cerr << "remove empty last line\n";
1122                         if (rowinfo[row].topline)
1123                                 rowinfo[row - 1].bottomline = true;
1124                         for (size_t col = 0; col < colinfo.size(); ++col)
1125                                 if (cellinfo[row][col].topline)
1126                                         cellinfo[row - 1][col].bottomline = true;
1127                         rowinfo.pop_back();
1128                 }
1129         }
1130
1131         // Now we have the table structure and content in rowinfo, colinfo
1132         // and cellinfo.
1133         // Unfortunately LyX has some limitations that we need to work around.
1134
1135         // Convert cells with special content to multicolumn cells
1136         // (LyX ignores the special field for non-multicolumn cells).
1137         for (size_t row = 0; row < rowinfo.size(); ++row) {
1138                 for (size_t col = 0; col < cellinfo[row].size(); ++col) {
1139                         if (cellinfo[row][col].multi == CELL_NORMAL &&
1140                             !cellinfo[row][col].special.empty())
1141                                 cellinfo[row][col].multi = CELL_BEGIN_OF_MULTICOLUMN;
1142                 }
1143         }
1144
1145         // Distribute lines from rows/columns to cells
1146         // The code was stolen from convert_tablines() in lyx2lyx/lyx_1_6.py.
1147         // Each standard cell inherits the settings of the corresponding
1148         // rowinfo/colinfo. This works because all cells with individual
1149         // settings were converted to multicolumn cells above.
1150         // Each multicolumn cell inherits the settings of the rowinfo/colinfo
1151         // corresponding to the first column of the multicolumn cell (default
1152         // of the multicol package). This works because the special field
1153         // overrides the line fields.
1154         for (size_t row = 0; row < rowinfo.size(); ++row) {
1155                 for (size_t col = 0; col < cellinfo[row].size(); ++col) {
1156                         if (cellinfo[row][col].multi == CELL_NORMAL) {
1157                                 cellinfo[row][col].topline = rowinfo[row].topline;
1158                                 cellinfo[row][col].bottomline = rowinfo[row].bottomline;
1159                                 cellinfo[row][col].leftlines = colinfo[col].leftlines;
1160                                 cellinfo[row][col].rightlines = colinfo[col].rightlines;
1161                         } else if (cellinfo[row][col].multi == CELL_BEGIN_OF_MULTICOLUMN) {
1162                                 size_t s = col + 1;
1163                                 while (s < cellinfo[row].size() &&
1164                                        cellinfo[row][s].multi == CELL_PART_OF_MULTICOLUMN)
1165                                         s++;
1166                                 if (s < cellinfo[row].size() &&
1167                                     cellinfo[row][s].multi != CELL_BEGIN_OF_MULTICOLUMN)
1168                                         cellinfo[row][col].rightlines = colinfo[col].rightlines;
1169                                 if (col > 0 && cellinfo[row][col-1].multi == CELL_NORMAL)
1170                                         cellinfo[row][col].leftlines = colinfo[col].leftlines;
1171                         }
1172                 }
1173         }
1174
1175         if (booktabs)
1176                 preamble.registerAutomaticallyLoadedPackage("booktabs");
1177         if (is_long_tabular)
1178                 preamble.registerAutomaticallyLoadedPackage("longtable");
1179
1180         //cerr << "// output what we have\n";
1181         // output what we have
1182         os << "\n<lyxtabular version=\"3\" rows=\"" << rowinfo.size()
1183            << "\" columns=\"" << colinfo.size() << "\">\n";
1184         os << "<features"
1185            << write_attribute("rotate", false)
1186            << write_attribute("booktabs", booktabs)
1187            << write_attribute("islongtable", is_long_tabular);
1188         if (!is_long_tabular)
1189                 os << write_attribute("tabularvalignment", tabularvalignment)
1190                    << write_attribute("tabularwidth", tabularwidth);
1191         os << ">\n";
1192
1193         //cerr << "// after header\n";
1194         for (size_t col = 0; col < colinfo.size(); ++col) {
1195                 os << "<column alignment=\""
1196                    << verbose_align(colinfo[col].align) << "\""
1197                    << " valignment=\""
1198                    << verbose_valign(colinfo[col].valign) << "\""
1199                    << write_attribute("width", translate_len(colinfo[col].width))
1200                    << write_attribute("special", colinfo[col].special)
1201                    << ">\n";
1202         }
1203         //cerr << "// after cols\n";
1204
1205         for (size_t row = 0; row < rowinfo.size(); ++row) {
1206                 os << "<row"
1207                    << write_attribute("topspace", rowinfo[row].top_space)
1208                    << write_attribute("bottomspace", rowinfo[row].bottom_space)
1209                    << write_attribute("interlinespace", rowinfo[row].interline_space)
1210                    << write_attribute("endhead",
1211                                       rowinfo[row].type == LT_HEAD)
1212                    << write_attribute("endfirsthead",
1213                                       rowinfo[row].type == LT_FIRSTHEAD)
1214                    << write_attribute("endfoot",
1215                                       rowinfo[row].type == LT_FOOT)
1216                    << write_attribute("endlastfoot",
1217                                       rowinfo[row].type == LT_LASTFOOT)
1218                    << write_attribute("newpage", rowinfo[row].newpage)
1219                    << write_attribute("caption", rowinfo[row].caption)
1220                    << ">\n";
1221                 for (size_t col = 0; col < colinfo.size(); ++col) {
1222                         CellInfo const & cell = cellinfo[row][col];
1223                         os << "<cell";
1224                         if (cell.multi != CELL_NORMAL)
1225                                 os << " multicolumn=\"" << cell.multi << "\"";
1226                         os << " alignment=\"" << verbose_align(cell.align)
1227                            << "\""
1228                            << " valignment=\"" << verbose_valign(cell.valign)
1229                            << "\""
1230                            << write_attribute("topline", cell.topline)
1231                            << write_attribute("bottomline", cell.bottomline)
1232                            << write_attribute("leftline", cell.leftlines > 0)
1233                            << write_attribute("rightline", cell.rightlines > 0)
1234                            << write_attribute("rotate", cell.rotate);
1235                         //cerr << "\nrow: " << row << " col: " << col;
1236                         //if (cell.topline)
1237                         //      cerr << " topline=\"true\"";
1238                         //if (cell.bottomline)
1239                         //      cerr << " bottomline=\"true\"";
1240                         os << " usebox=\"none\""
1241                            << write_attribute("width", translate_len(cell.width));
1242                         if (cell.multi != CELL_NORMAL)
1243                                 os << write_attribute("special", cell.special);
1244                         os << ">"
1245                            << "\n\\begin_inset Text\n"
1246                            << cell.content
1247                            << "\n\\end_inset\n"
1248                            << "</cell>\n";
1249                 }
1250                 os << "</row>\n";
1251         }
1252
1253         os << "</lyxtabular>\n";
1254 }
1255
1256
1257
1258
1259 // }])
1260
1261
1262 } // namespace lyx