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 "Context.h"
  20 #include "Preamble.h"
  21
  22 #include "support/lassert.h"
  23 #include "support/convert.h"
  24 #include "support/lstrings.h"
  25
  26 #include <algorithm>
  27 #include <iostream>
  28 #include <sstream>
  29 #include <vector>
  30 #include <map>
  31
  32 using namespace std;
  33
  34 namespace lyx {
  35
  36
  37 namespace {
  38
  39 class ColInfo {
  40 public:
  41         ColInfo() : align('n'), valign('n'), rightlines(0), leftlines(0),
  42                 varwidth(false), decimal_point('\0'), vcolumn(false) {}
  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         /// varwidth column
  56         bool varwidth;
  57         /// decimal separator
  58         char decimal_point;
  59         /// V column type
  60         bool vcolumn;
  61 };
  62
  63
  64 /// row type for longtables
  65 enum LTRowType
  66 {
  67         /// normal row
  68         LT_NORMAL,
  69         /// part of head
  70         LT_HEAD,
  71         /// part of head on first page
  72         LT_FIRSTHEAD,
  73         /// part of foot
  74         LT_FOOT,
  75         /// part of foot on last page
  76         LT_LASTFOOT
  77 };
  78
  79
  80 class RowInfo {
  81 public:
  82         RowInfo() : topline(false), bottomline(false), type(LT_NORMAL),
  83                     caption(false), newpage(false) {}
  84         /// Does this row have any special setting?
  85         bool special() const
  86         {
  87                 return topline || bottomline || !top_space.empty() ||
  88                         !bottom_space.empty() || !interline_space.empty() ||
  89                         type != LT_NORMAL || caption || newpage;
  90         }
  91         /// horizontal line above
  92         bool topline;
  93         /// horizontal line below
  94         bool bottomline;
  95         /// Extra space between the top line and this row
  96         string top_space;
  97         /// Extra space between this row and the bottom line
  98         string bottom_space;
  99         /// Extra space between the bottom line and the next top line
 100         string interline_space;
 101         /// These are for longtabulars only
 102         /// row type (head, foot, firsthead etc.)
 103         LTRowType type;
 104         /// row for a caption
 105         bool caption;
 106         /// row for a newpage
 107         bool newpage;
 108 };
 109
 110
 111 /// the numeric values are part of the file format!
 112 enum Multi {
 113         /// A normal cell
 114         CELL_NORMAL = 0,
 115         /// A multicolumn cell. The number of columns is <tt>1 + number
 116         /// of CELL_PART_OF_MULTICOLUMN cells</tt> that follow directly
 117         CELL_BEGIN_OF_MULTICOLUMN,
 118         /// This is a dummy cell (part of a multicolumn cell)
 119         CELL_PART_OF_MULTICOLUMN,
 120         ///
 121         CELL_BEGIN_OF_MULTIROW,
 122         ///
 123         CELL_PART_OF_MULTIROW
 124 };
 125
 126
 127 class CellInfo {
 128 public:
 129         CellInfo() : multi(CELL_NORMAL), align('n'), valign('n'),
 130                      leftlines(0), rightlines(0), topline(false),
 131                      bottomline(false), topline_ltrim(false),
 132                      topline_rtrim(false), bottomline_ltrim(false),
 133                      bottomline_rtrim(false), rotate(0), mrxnum(0) {}
 134         /// cell content
 135         string content;
 136         /// multicolumn flag
 137         Multi multi;
 138         /// cell alignment
 139         char align;
 140         /// vertical cell alignment
 141         char valign;
 142         /// number of lines on the left
 143         int leftlines;
 144         /// number of lines on the right
 145         int rightlines;
 146         /// do we have a line above?
 147         bool topline;
 148         /// do we have a line below?
 149         bool bottomline;
 150         /// Left trimming of top line
 151         bool topline_ltrim;
 152         /// Right trimming of top line
 153         bool topline_rtrim;
 154         /// Left trimming of bottom line
 155         bool bottomline_ltrim;
 156         /// Right trimming of top line
 157         bool bottomline_rtrim;
 158         /// how is the cell rotated?
 159         int rotate;
 160         /// width for multicolumn cells
 161         string width;
 162         /// special formatting for multicolumn cells
 163         string special;
 164         /// multirow offset
 165         string mroffset;
 166         /// number of further multirows
 167         int mrxnum;
 168 };
 169
 170
 171 class ltType {
 172 public:
 173         // constructor
 174         ltType() : set(false), topDL(false), bottomDL(false), empty(false) {}
 175         // we have this header type (is set in the getLT... functions)
 176         bool set;
 177         // double borders on top
 178         bool topDL;
 179         // double borders on bottom
 180         bool bottomDL;
 181         // used for FirstHeader & LastFooter and if this is true
 182         // all the rows marked as FirstHeader or LastFooter are
 183         // ignored in the output and it is set to be empty!
 184         bool empty;
 185 };
 186
 187
 188 /// translate a horizontal alignment (as stored in ColInfo and CellInfo) to LyX
 189 inline char const * verbose_align(char c)
 190 {
 191         switch (c) {
 192         case 'c':
 193                 return "center";
 194         case 'r':
 195                 return "right";
 196         case 'l':
 197                 return "left";
 198         case 'd':
 199                 return "decimal";
 200         default:
 201                 return "none";
 202         }
 203 }
 204
 205
 206 /// translate a vertical alignment (as stored in ColInfo and CellInfo) to LyX
 207 inline char const * verbose_valign(char c)
 208 {
 209         // The default value for no special alignment is "top".
 210         switch (c) {
 211         case 'm':
 212                 return "middle";
 213         case 'b':
 214                 return "bottom";
 215         case 'p':
 216         default:
 217                 return "top";
 218         }
 219 }
 220
 221
 222 // stripped down from tabluar.C. We use it currently only for bools and
 223 // strings
 224 string const write_attribute(string const & name, bool const & b)
 225 {
 226         // we write only true attribute values so we remove a bit of the
 227         // file format bloat for tabulars.
 228         return b ? ' ' + name + "=\"true\"" : string();
 229 }
 230
 231
 232 string const write_attribute(string const & name, string const & s)
 233 {
 234         return s.empty() ? string() : ' ' + name + "=\"" + s + '"';
 235 }
 236
 237
 238 string const write_attribute(string const & name, int const & i)
 239 {
 240         // we write only true attribute values so we remove a bit of the
 241         // file format bloat for tabulars.
 242         return i ? write_attribute(name, convert<string>(i)) : string();
 243 }
 244
 245
 246 /*! rather brutish way to code table structure in a string:
 247
 248 \verbatim
 249   \begin{tabular}{ccc}
 250     1 & 2 & 3\\ \hline
 251     \multicolumn{2}{c}{4} & 5 //
 252     6 & 7 \\
 253     8 \endhead
 254   \end{tabular}
 255 \endverbatim
 256
 257  gets "translated" to:
 258
 259 \verbatim
 260          HLINE 1                     TAB 2 TAB 3 HLINE          HLINE LINE
 261   \hline HLINE \multicolumn{2}{c}{4} TAB 5       HLINE          HLINE LINE
 262          HLINE 6                     TAB 7       HLINE          HLINE LINE
 263          HLINE 8                                 HLINE \endhead HLINE LINE
 264 \endverbatim
 265  */
 266
 267 char const TAB   = '\001';
 268 char const LINE  = '\002';
 269 char const HLINE = '\004';
 270
 271
 272 /*!
 273  * Move the information in leftlines, rightlines, align and valign to the
 274  * special field. This is necessary if the special field is not empty,
 275  * because LyX ignores leftlines > 1, rightlines > 1, align and valign in
 276  * this case.
 277  */
 278 void ci2special(ColInfo & ci)
 279 {
 280         if (ci.width.empty() && ci.align == 'n')
 281                 // The alignment setting is already in special, since
 282                 // handle_colalign() never stores ci with these settings
 283                 // and ensures that leftlines == 0 and rightlines == 0 in
 284                 // this case.
 285                 return;
 286
 287         if (ci.decimal_point != '\0') {
 288                 // we only support decimal point natively
 289                 // with 'l' alignment in or 'n' alignment
 290                 // with width in second row
 291                 if (ci.align != 'l' && ci.align != 'n') {
 292                         ci.decimal_point = '\0';
 293                         return;
 294                 } else
 295                         ci.special.clear();
 296         }
 297
 298         if (!ci.width.empty()) {
 299                 string arraybackslash;
 300                 if (ci.varwidth)
 301                         arraybackslash = "\\arraybackslash";
 302                 switch (ci.align) {
 303                 case 'l':
 304                         ci.special += ">{\\raggedright" + arraybackslash + "}";
 305                         break;
 306                 case 'r':
 307                         ci.special += ">{\\raggedleft" + arraybackslash + "}";
 308                         break;
 309                 case 'c':
 310                         ci.special += ">{\\centering" + arraybackslash + "}";
 311                         break;
 312                 }
 313                 if (ci.vcolumn)
 314                         ci.special += 'V';
 315                 else if (ci.varwidth)
 316                         ci.special += 'X';
 317                 else if (ci.valign == 'n')
 318                         ci.special += 'p';
 319                 else
 320                         ci.special += ci.valign;
 321                 ci.special += '{' + ci.width + '}';
 322                 ci.width.erase();
 323         } else
 324                 ci.special += ci.align;
 325
 326         // LyX can only have one left and one right line.
 327         for (int i = 1; i < ci.leftlines; ++i)
 328                 ci.special.insert(0, "|");
 329         for (int i = 1; i < ci.rightlines; ++i)
 330                 ci.special += '|';
 331         ci.leftlines = min(ci.leftlines, 1);
 332         ci.rightlines = min(ci.rightlines, 1);
 333         ci.align = 'n';
 334         ci.valign = 'n';
 335 }
 336
 337
 338 /*!
 339  * Handle column specifications for tabulars and multicolumns.
 340  * The next token of the parser \p p must be an opening brace, and we read
 341  * everything until the matching closing brace.
 342  * The resulting column specifications are filled into \p colinfo. This is
 343  * in an intermediate form. fix_colalign() makes it suitable for LyX output.
 344  */
 345 void handle_colalign(Parser & p, vector<ColInfo> & colinfo,
 346                      ColInfo const & start)
 347 {
 348         if (p.get_token().cat() != catBegin)
 349                 cerr << "Wrong syntax for table column alignment.\n"
 350                         "Expected '{', got '" << p.curr_token().asInput()
 351                      << "'.\n";
 352
 353         ColInfo next = start;
 354         for (Token t = p.get_token(); p.good() && t.cat() != catEnd;
 355              t = p.get_token()) {
 356 #ifdef FILEDEBUG
 357                 cerr << "t: " << t << "  c: '" << t.character() << "'\n";
 358 #endif
 359
 360                 // We cannot handle comments here
 361                 if (t.cat() == catComment) {
 362                         if (t.cs().empty()) {
 363                                 // "%\n" combination
 364                                 p.skip_spaces();
 365                         } else
 366                                 cerr << "Ignoring comment: " << t.asInput();
 367                         continue;
 368                 }
 369
 370                 switch (t.character()) {
 371                         case ' ':
 372                                 // whitespace, ignore.
 373                                 break;
 374                         case 'c':
 375                         case 'l':
 376                         case 'r':
 377                                 // new column, horizontal aligned
 378                                 next.align = t.character();
 379                                 if (!next.special.empty()) {
 380                                         ci2special(next);
 381                                         // handle decimal separator
 382                                         if (next.decimal_point != '\0') {
 383                                                 if (!colinfo.empty() && colinfo.back().align == 'r') {
 384                                                         colinfo.back().align = 'd';
 385                                                         colinfo.back().decimal_point = next.decimal_point;
 386                                                 } else
 387                                                         next.decimal_point = '\0';
 388                                         }
 389                                 }
 390                                 colinfo.push_back(next);
 391                                 next = ColInfo();
 392                                 break;
 393                         case 'X':
 394                                 // varwidth column
 395                                 next.varwidth = true;
 396                                 if (!next.special.empty())
 397                                         ci2special(next);
 398                                 colinfo.push_back(next);
 399                                 next = ColInfo();
 400                                 break;
 401                         case 'V': {
 402                                 // V column type (varwidth package)
 403                                 string const s = trimSpaceAndEol(p.verbatim_item());
 404                                 // V{\linewidth} is treated as a normal column
 405                                 // (which allows for line breaks). The V type is
 406                                 // automatically set by LyX in this case
 407                                 if (s != "\\linewidth" || !next.special.empty()) {
 408                                         next.vcolumn = true;
 409                                         next.width = s;
 410                                         ci2special(next);
 411                                 }
 412                                 colinfo.push_back(next);
 413                                 next = ColInfo();
 414                                 break;
 415                         }
 416                         case 'p':
 417                         case 'b':
 418                         case 'm':
 419                                 // new column, vertical aligned box
 420                                 next.valign = t.character();
 421                                 next.width = p.verbatim_item();
 422                                 if (!next.special.empty()) {
 423                                         ci2special(next);
 424                                         // handle decimal separator
 425                                         if (next.decimal_point != '\0') {
 426                                                 if (!colinfo.empty() && colinfo.back().align == 'r') {
 427                                                         colinfo.back().align = 'd';
 428                                                         colinfo.back().decimal_point = next.decimal_point;
 429                                                 } else
 430                                                         next.decimal_point = '\0';
 431                                         }
 432                                 }
 433                                 colinfo.push_back(next);
 434                                 next = ColInfo();
 435                                 break;
 436                         case '|':
 437                                 // vertical rule
 438                                 if (colinfo.empty()) {
 439                                         if (next.special.empty())
 440                                                 ++next.leftlines;
 441                                         else
 442                                                 next.special += '|';
 443                                 } else if (colinfo.back().special.empty())
 444                                         ++colinfo.back().rightlines;
 445                                 else if (next.special.empty() && p.next_token().cat() != catEnd)
 446                                         ++next.leftlines;
 447                                 else
 448                                         colinfo.back().special += '|';
 449                                 break;
 450                         case '>': {
 451                                 // text before the next column
 452                                 string const s = trimSpaceAndEol(p.verbatim_item());
 453                                 if (next.special.empty() &&
 454                                     next.align == 'n') {
 455                                         // Maybe this can be converted to a
 456                                         // horizontal alignment setting for
 457                                         // fixed width columns
 458                                         if (s == "\\raggedleft" || s == "\\raggedleft\\arraybackslash")
 459                                                 next.align = 'r';
 460                                         else if (s == "\\raggedright" || s == "\\raggedright\\arraybackslash")
 461                                                 next.align = 'l';
 462                                         else if (s == "\\centering" || s == "\\centering\\arraybackslash")
 463                                                 next.align = 'c';
 464                                         else
 465                                                 next.special = ">{" + s + '}';
 466                                 } else
 467                                         next.special += ">{" + s + '}';
 468                                 break;
 469                         }
 470                         case '<': {
 471                                 // text after the last column
 472                                 string const s = trimSpaceAndEol(p.verbatim_item());
 473                                 if (colinfo.empty())
 474                                         // This is not possible in LaTeX.
 475                                         cerr << "Ignoring separator '<{"
 476                                              << s << "}'." << endl;
 477                                 else {
 478                                         ColInfo & ci = colinfo.back();
 479                                         ci2special(ci);
 480                                         ci.special += "<{" + s + '}';
 481                                 }
 482                                 break;
 483                         }
 484                         case '*': {
 485                                 if (p.next_token().character() != '{')
 486                                         continue;
 487                                 // *{n}{arg} means 'n' columns of type 'arg'
 488                                 string const num = p.verbatim_item();
 489                                 string const arg = p.verbatim_item();
 490                                 size_t const n = convert<unsigned int>(num);
 491                                 if (!arg.empty() && n > 0) {
 492                                         string s("{");
 493                                         for (size_t i = 0; i < n; ++i)
 494                                                 s += arg;
 495                                         s += '}';
 496                                         Parser p2(s);
 497                                         handle_colalign(p2, colinfo, next);
 498                                         next = ColInfo();
 499                                 } else {
 500                                         cerr << "Ignoring column specification"
 501                                                 " '*{" << num << "}{"
 502                                              << arg << "}'." << endl;
 503                                 }
 504                                 break;
 505                         }
 506                         case '@':
 507                                 // text instead of the column spacing
 508                         case '!': {
 509                                 // text in addition to the column spacing
 510                                 string const arg =  p.verbatim_item();
 511                                 next.special += t.character();
 512                                 next.special += '{' + arg + '}';
 513                                 string const sarg = arg.size() > 2 ? arg.substr(0, arg.size() - 1) : string();
 514                                 if (t.character() == '@' && sarg == "\\extracolsep{0pt}")
 515                                         next.decimal_point = arg.back();
 516                                 break;
 517                         }
 518                         default: {
 519                                 // try user defined column types
 520                                 // unknown column types (nargs == -1) are
 521                                 // assumed to consume no arguments
 522                                 ci2special(next);
 523                                 next.special += t.character();
 524                                 int const nargs =
 525                                         preamble.getSpecialTableColumnArguments(t.character());
 526                                 for (int i = 0; i < nargs; ++i)
 527                                         next.special += '{' +
 528                                                 p.verbatim_item() + '}';
 529                                 colinfo.push_back(next);
 530                                 next = ColInfo();
 531                                 break;
 532                         }
 533                         }
 534         }
 535
 536         // Maybe we have some column separators that need to be added to the
 537         // last column?
 538         ci2special(next);
 539         if (!next.special.empty()) {
 540                 ColInfo & ci = colinfo.back();
 541                 ci2special(ci);
 542                 ci.special += next.special;
 543                 next.special.erase();
 544         }
 545 }
 546
 547
 548 /*!
 549  * Move the left and right lines and alignment settings of the column \p ci
 550  * to the special field if necessary.
 551  */
 552 void fix_colalign(ColInfo & ci)
 553 {
 554         if (ci.leftlines > 1 || ci.rightlines > 1)
 555                 ci2special(ci);
 556 }
 557
 558
 559 /*!
 560  * LyX can't handle more than one vertical line at the left or right side
 561  * of a column.
 562  * This function moves the left and right lines and alignment settings of all
 563  * columns in \p colinfo to the special field if necessary.
 564  */
 565 void fix_colalign(vector<ColInfo> & colinfo)
 566 {
 567         // Try to move extra leftlines to the previous column.
 568         // We do this only if both special fields are empty, otherwise we
 569         // can't tell whether the result will be the same.
 570         for (size_t col = 0; col < colinfo.size(); ++col) {
 571                 if (colinfo[col].leftlines > 1 &&
 572                     colinfo[col].special.empty() && col > 0 &&
 573                     colinfo[col - 1].rightlines == 0 &&
 574                     colinfo[col - 1].special.empty()) {
 575                         ++colinfo[col - 1].rightlines;
 576                         --colinfo[col].leftlines;
 577                 }
 578         }
 579         // Try to move extra rightlines to the next column
 580         for (size_t col = 0; col < colinfo.size(); ++col) {
 581                 if (colinfo[col].rightlines > 1 &&
 582                     colinfo[col].special.empty() &&
 583                     col < colinfo.size() - 1 &&
 584                     colinfo[col + 1].leftlines == 0 &&
 585                     colinfo[col + 1].special.empty()) {
 586                         ++colinfo[col + 1].leftlines;
 587                         --colinfo[col].rightlines;
 588                 }
 589         }
 590         // Move the lines and alignment settings to the special field if
 591         // necessary
 592         for (auto & col : colinfo)
 593                 fix_colalign(col);
 594 }
 595
 596
 597 /*!
 598  * Parse hlines and similar stuff.
 599  * \returns whether the token \p t was parsed
 600  */
 601 bool parse_hlines(Parser & p, Token const & t, string & hlines,
 602                   bool is_long_tabular)
 603 {
 604         LASSERT(t.cat() == catEscape, return false);
 605
 606         if (t.cs() == "hline" || t.cs() == "toprule" || t.cs() == "midrule" ||
 607             t.cs() == "bottomrule")
 608                 hlines += '\\' + t.cs();
 609
 610         else if (t.cs() == "cline")
 611                 hlines += "\\cline{" + p.verbatim_item() + '}';
 612
 613         else if (t.cs() == "cmidrule") {
 614                 p.pushPosition();
 615                 p.skip_spaces(true);
 616                 // We do not support the optional height argument
 617                 if (p.hasOpt())
 618                         return false;
 619                 // We support the \cmidrule(l){3-4} form but
 620                 // not the trim length parameters (l{<with>}r{<width>})
 621                 string const trim = p.getFullParentheseArg();
 622                 string const range = p.verbatim_item();
 623                 if (!trim.empty()) {
 624                         if (support::contains(trim, "{"))
 625                                 return false;
 626                         hlines += "\\cmidrule" + trim + "{" + range + "}";
 627                 } else
 628                         hlines += "\\cmidrule{" + range + '}';
 629         }
 630
 631         else if (t.cs() == "addlinespace") {
 632                 p.pushPosition();
 633                 p.skip_spaces(true);
 634                 bool const hasArgument(p.getFullArg('{', '}').first);
 635                 p.popPosition();
 636                 if (hasArgument)
 637                         hlines += "\\addlinespace{" + p.verbatim_item() + '}';
 638                 else
 639                         hlines += "\\addlinespace";
 640         }
 641
 642         else if (is_long_tabular && t.cs() == "newpage")
 643                 hlines += "\\newpage";
 644
 645         else
 646                 return false;
 647
 648         return true;
 649 }
 650
 651
 652 /// Position in a row
 653 enum RowPosition {
 654         /// At the very beginning, before the first token
 655         ROW_START,
 656         /// After the first token and before any column token
 657         IN_HLINES_START,
 658         /// After the first column token. Comments and whitespace are only
 659         /// treated as tokens in this position
 660         IN_COLUMNS,
 661         /// After the first non-column token at the end
 662         IN_HLINES_END
 663 };
 664
 665
 666 /*!
 667  * Parse table structure.
 668  * We parse tables in a two-pass process: This function extracts the table
 669  * structure (rows, columns, hlines etc.), but does not change the cell
 670  * content. The cell content is parsed in a second step in handle_tabular().
 671  */
 672 void parse_table(Parser & p, ostream & os, bool is_long_tabular,
 673                  RowPosition & pos, unsigned flags)
 674 {
 675         // table structure commands such as \hline
 676         string hlines;
 677
 678         // comments that occur at places where we can't handle them
 679         string comments;
 680
 681         while (p.good()) {
 682                 Token const & t = p.get_token();
 683
 684 #ifdef FILEDEBUG
 685                 debugToken(cerr, t, flags);
 686 #endif
 687
 688                 // comments and whitespace in hlines
 689                 switch (pos) {
 690                 case ROW_START:
 691                 case IN_HLINES_START:
 692                 case IN_HLINES_END:
 693                         if (t.cat() == catComment) {
 694                                 if (t.cs().empty())
 695                                         // line continuation
 696                                         p.skip_spaces();
 697                                 else
 698                                         // We can't handle comments here,
 699                                         // store them for later use
 700                                         comments += t.asInput();
 701                                 continue;
 702                         } else if (t.cat() == catSpace ||
 703                                    t.cat() == catNewline) {
 704                                 // whitespace is irrelevant here, we
 705                                 // need to recognize hline stuff
 706                                 p.skip_spaces();
 707                                 continue;
 708                         }
 709                         break;
 710                 case IN_COLUMNS:
 711                         break;
 712                 }
 713
 714                 // We need to handle structure stuff first in order to
 715                 // determine whether we need to output a HLINE separator
 716                 // before the row or not.
 717                 if (t.cat() == catEscape) {
 718                         if (parse_hlines(p, t, hlines, is_long_tabular)) {
 719                                 switch (pos) {
 720                                 case ROW_START:
 721                                         pos = IN_HLINES_START;
 722                                         break;
 723                                 case IN_COLUMNS:
 724                                         pos = IN_HLINES_END;
 725                                         break;
 726                                 case IN_HLINES_START:
 727                                 case IN_HLINES_END:
 728                                         break;
 729                                 }
 730                                 continue;
 731                         }
 732
 733                         else if (t.cs() == "tabularnewline" ||
 734                                  t.cs() == "\\" ||
 735                                  t.cs() == "cr") {
 736                                 if (t.cs() == "cr")
 737                                         cerr << "Warning: Converting TeX "
 738                                                 "'\\cr' to LaTeX '\\\\'."
 739                                              << endl;
 740                                 // stuff before the line break
 741                                 os << comments << HLINE << hlines << HLINE
 742                                    << LINE;
 743                                 //cerr << "hlines: " << hlines << endl;
 744                                 hlines.erase();
 745                                 comments.erase();
 746                                 pos = ROW_START;
 747                                 continue;
 748                         }
 749
 750                         else if (is_long_tabular &&
 751                                  (t.cs() == "endhead" ||
 752                                   t.cs() == "endfirsthead" ||
 753                                   t.cs() == "endfoot" ||
 754                                   t.cs() == "endlastfoot")) {
 755                                 hlines += t.asInput();
 756                                 switch (pos) {
 757                                 case IN_COLUMNS:
 758                                 case IN_HLINES_END:
 759                                         // these commands are implicit line
 760                                         // breaks
 761                                         os << comments << HLINE << hlines
 762                                            << HLINE << LINE;
 763                                         hlines.erase();
 764                                         comments.erase();
 765                                         pos = ROW_START;
 766                                         break;
 767                                 case ROW_START:
 768                                         pos = IN_HLINES_START;
 769                                         break;
 770                                 case IN_HLINES_START:
 771                                         break;
 772                                 }
 773                                 continue;
 774                         }
 775                 }
 776
 777                 // We need a HLINE separator if we either have no hline
 778                 // stuff at all and are just starting a row or if we just
 779                 // got the first non-hline token.
 780                 switch (pos) {
 781                 case ROW_START:
 782                         // no hline tokens exist, first token at row start
 783                 case IN_HLINES_START:
 784                         // hline tokens exist, first non-hline token at row
 785                         // start
 786                         os << hlines << HLINE << comments;
 787                         hlines.erase();
 788                         comments.erase();
 789                         pos = IN_COLUMNS;
 790                         break;
 791                 case IN_HLINES_END:
 792                         // Oops, there is still cell content or unsupported
 793                         // booktabs commands after hline stuff. The latter are
 794                         // moved to the cell, and the first does not work in
 795                         // LaTeX, so we ignore the hlines.
 796                         os << comments;
 797                         comments.erase();
 798                         if (support::contains(hlines, "\\hline") ||
 799                             support::contains(hlines, "\\cline") ||
 800                             support::contains(hlines, "\\newpage"))
 801                                 cerr << "Ignoring '" << hlines
 802                                      << "' in a cell" << endl;
 803                         else
 804                                 os << hlines;
 805                         hlines.erase();
 806                         pos = IN_COLUMNS;
 807                         break;
 808                 case IN_COLUMNS:
 809                         break;
 810                 }
 811
 812                 // If we come here we have normal cell content
 813                 //
 814                 // cat codes
 815                 //
 816                 if (t.cat() == catMath) {
 817                         // we are inside some text mode thingy, so opening new math is allowed
 818                         Token const & n = p.get_token();
 819                         if (n.cat() == catMath) {
 820                                 // TeX's $$...$$ syntax for displayed math
 821                                 os << "\\[";
 822                                 // This does only work because parse_math outputs TeX
 823                                 parse_math(p, os, FLAG_SIMPLE, MATH_MODE);
 824                                 os << "\\]";
 825                                 p.get_token(); // skip the second '$' token
 826                         } else {
 827                                 // simple $...$  stuff
 828                                 p.putback();
 829                                 os << '$';
 830                                 // This does only work because parse_math outputs TeX
 831                                 parse_math(p, os, FLAG_SIMPLE, MATH_MODE);
 832                                 os << '$';
 833                         }
 834                 }
 835
 836                 else if (t.cat() == catSpace
 837                          || t.cat() == catNewline
 838                          || t.cat() == catLetter
 839                          || t.cat() == catSuper
 840                          || t.cat() == catSub
 841                          || t.cat() == catOther
 842                          || t.cat() == catActive
 843                          || t.cat() == catParameter)
 844                         os << t.cs();
 845
 846                 else if (t.cat() == catBegin) {
 847                         os << '{';
 848                         parse_table(p, os, is_long_tabular, pos,
 849                                     FLAG_BRACE_LAST);
 850                         os << '}';
 851                 }
 852
 853                 else if (t.cat() == catEnd) {
 854                         if (flags & FLAG_BRACE_LAST)
 855                                 return;
 856                         cerr << "unexpected '}'\n";
 857                 }
 858
 859                 else if (t.cat() == catAlign) {
 860                         os << TAB;
 861                         p.skip_spaces();
 862                 }
 863
 864                 else if (t.cat() == catComment)
 865                         os << t.asInput();
 866
 867                 else if (t.cs() == "(") {
 868                         os << "\\(";
 869                         // This does only work because parse_math outputs TeX
 870                         parse_math(p, os, FLAG_SIMPLE2, MATH_MODE);
 871                         os << "\\)";
 872                 }
 873
 874                 else if (t.cs() == "[") {
 875                         os << "\\[";
 876                         // This does only work because parse_math outputs TeX
 877                         parse_math(p, os, FLAG_EQUATION, MATH_MODE);
 878                         os << "\\]";
 879                 }
 880
 881                 else if (t.cs() == "begin") {
 882                         string const name = p.getArg('{', '}');
 883                         active_environments.push_back(name);
 884                         os << "\\begin{" << name << '}';
 885                         // treat the nested environment as a block, don't
 886                         // parse &, \\ etc, because they don't belong to our
 887                         // table if they appear.
 888                         os << p.ertEnvironment(name);
 889                         os << "\\end{" << name << '}';
 890                         active_environments.pop_back();
 891                 }
 892
 893                 else if (t.cs() == "end") {
 894                         if (flags & FLAG_END) {
 895                                 // eat environment name
 896                                 string const name = p.getArg('{', '}');
 897                                 if (name != active_environment())
 898                                         p.error("\\end{" + name + "} does not match \\begin{"
 899                                                 + active_environment() + "}");
 900                                 return;
 901                         }
 902                         p.error("found 'end' unexpectedly");
 903                 }
 904
 905                 else
 906                         os << t.asInput();
 907         }
 908
 909         // We can have comments if the last line is incomplete
 910         os << comments;
 911
 912         // We can have hline stuff if the last line is incomplete
 913         if (!hlines.empty()) {
 914                 // this does not work in LaTeX, so we ignore it
 915                 cerr << "Ignoring '" << hlines << "' at end of tabular"
 916                      << endl;
 917         }
 918 }
 919
 920
 921 void handle_hline_above(RowInfo & ri, vector<CellInfo> & ci)
 922 {
 923         ri.topline = true;
 924         for (auto & col : ci)
 925                 col.topline = true;
 926 }
 927
 928
 929 void handle_hline_below(RowInfo & ri, vector<CellInfo> & ci)
 930 {
 931         ri.bottomline = true;
 932         for (auto & col : ci)
 933                 col.bottomline = true;
 934 }
 935
 936
 937 void parse_cell_content(ostringstream & os2, Parser & parse, unsigned int flags, Context & newcontext,
 938                         vector< vector<CellInfo> > & cellinfo, vector<ColInfo> & colinfo,
 939                         size_t const row, size_t const col)
 940 {
 941         bool turn = false;
 942         int rotate = 0;
 943         bool varwidth = false;
 944         if (parse.next_token().cs() == "begin") {
 945                 parse.pushPosition();
 946                 parse.get_token();
 947                 string const env = parse.getArg('{', '}');
 948                 if (env == "sideways" || env == "turn") {
 949                         string angle = "90";
 950                         if (env == "turn") {
 951                                 turn = true;
 952                                 angle = parse.getArg('{', '}');
 953                         }
 954                         active_environments.push_back(env);
 955                         parse.ertEnvironment(env);
 956                         active_environments.pop_back();
 957                         parse.skip_spaces();
 958                         if (!parse.good() && support::isStrInt(angle))
 959                                 rotate = convert<int>(angle);
 960                 } else if (env == "cellvarwidth") {
 961                         active_environments.push_back(env);
 962                         parse.ertEnvironment(env);
 963                         active_environments.pop_back();
 964                         parse.skip_spaces();
 965                         varwidth = true;
 966                 }
 967                 parse.popPosition();
 968         }
 969         if (rotate != 0) {
 970                 cellinfo[row][col].rotate = rotate;
 971                 parse.get_token();
 972                 active_environments.push_back(parse.getArg('{', '}'));
 973                 if (turn)
 974                         parse.getArg('{', '}');
 975                 parse_text_in_inset(parse, os2, FLAG_END, false, newcontext);
 976                 active_environments.pop_back();
 977                 preamble.registerAutomaticallyLoadedPackage("rotating");
 978         } else if (varwidth) {
 979                 parse.get_token();
 980                 active_environments.push_back(parse.getArg('{', '}'));
 981                 // valign arg
 982                 if (parse.hasOpt())
 983                         cellinfo[row][col].valign = parse.getArg('[', ']')[1];
 984                 newcontext.in_table_cell = true;
 985                 parse_text_in_inset(parse, os2, FLAG_END, false, newcontext);
 986                 if (cellinfo[row][col].multi == CELL_NORMAL)
 987                         colinfo[col].align = newcontext.cell_align;
 988                 else
 989                         cellinfo[row][col].align = newcontext.cell_align;
 990                 active_environments.pop_back();
 991                 preamble.registerAutomaticallyLoadedPackage("varwidth");
 992         } else {
 993                 parse_text_in_inset(parse, os2, flags, false, newcontext);
 994         }
 995 }
 996
 997
 998 } // anonymous namespace
 999
1000
1001 void handle_tabular(Parser & p, ostream & os, string const & name,
1002                     string const & tabularwidth, string const & halign,
1003                     Context const & context)
1004 {
1005         Context newcontext = context;
1006         bool const is_long_tabular(name == "longtable" || name == "xltabular");
1007         bool booktabs = false;
1008         string tabularvalignment("middle");
1009         string posopts = p.getOpt();
1010         if (!posopts.empty()) {
1011                 if (posopts == "[t]")
1012                         tabularvalignment = "top";
1013                 else if (posopts == "[b]")
1014                         tabularvalignment = "bottom";
1015                 else
1016                         cerr << "vertical tabular positioning '"
1017                              << posopts << "' ignored\n";
1018         }
1019
1020         vector<ColInfo> colinfo;
1021
1022         // handle column formatting
1023         handle_colalign(p, colinfo, ColInfo());
1024         fix_colalign(colinfo);
1025
1026         // first scan of cells
1027         // use table mode to keep it minimal-invasive
1028         // not exactly what's TeX doing...
1029         vector<string> lines;
1030         ostringstream ss;
1031         RowPosition rowpos = ROW_START;
1032         parse_table(p, ss, is_long_tabular, rowpos, FLAG_END);
1033         split(ss.str(), lines, LINE);
1034
1035         vector< vector<CellInfo> > cellinfo(lines.size());
1036         vector<RowInfo> rowinfo(lines.size());
1037         ltType endfirsthead;
1038         ltType endhead;
1039         ltType endfoot;
1040         ltType endlastfoot;
1041
1042         // split into rows
1043         //cerr << "// split into rows\n";
1044         for (size_t row = 0; row < rowinfo.size();) {
1045
1046                 // init row
1047                 cellinfo[row].resize(colinfo.size());
1048                 bool deletelastrow = false;
1049
1050                 // split row
1051                 vector<string> dummy;
1052                 //cerr << "\n########### LINE: " << lines[row] << "########\n";
1053                 split(lines[row], dummy, HLINE);
1054
1055                 // handle horizontal line fragments
1056                 // we do only expect this for a last line without '\\'
1057                 if (dummy.size() != 3) {
1058                         if ((dummy.size() != 1 && dummy.size() != 2) ||
1059                             row != rowinfo.size() - 1)
1060                                 cerr << "unexpected dummy size: " << dummy.size()
1061                                         << " content: " << lines[row] << "\n";
1062                         dummy.resize(3);
1063                 }
1064                 lines[row] = dummy[1];
1065
1066                 //cerr << "line: " << row << " above 0: " << dummy[0] << "\n";
1067                 //cerr << "line: " << row << " below 2: " << dummy[2] <<  "\n";
1068                 //cerr << "line: " << row << " cells 1: " << dummy[1] <<  "\n";
1069
1070                 for (int i = 0; i <= 2; i += 2) {
1071                         //cerr << "   reading from line string '" << dummy[i] << "'\n";
1072                         Parser p1(dummy[size_type(i)]);
1073                         while (p1.good()) {
1074                                 Token t = p1.get_token();
1075                                 //cerr << "read token: " << t << "\n";
1076                                 if (t.cs() == "hline" || t.cs() == "toprule" ||
1077                                     t.cs() == "midrule" ||
1078                                     t.cs() == "bottomrule") {
1079                                         if (t.cs() != "hline")
1080                                                 booktabs = true;
1081                                         if (i == 0) {
1082                                                 if (rowinfo[row].topline) {
1083                                                         if (row > 0) // extra bottomline above
1084                                                                 handle_hline_below(rowinfo[row - 1], cellinfo[row - 1]);
1085                                                         else
1086                                                                 cerr << "dropping extra "
1087                                                                      << t.cs() << '\n';
1088                                                         //cerr << "below row: " << row-1 << endl;
1089                                                 } else {
1090                                                         handle_hline_above(rowinfo[row], cellinfo[row]);
1091                                                         //cerr << "above row: " << row << endl;
1092                                                 }
1093                                         } else {
1094                                                 //cerr << "below row: " << row << endl;
1095                                                 handle_hline_below(rowinfo[row], cellinfo[row]);
1096                                         }
1097                                 } else if (t.cs() == "cline" || t.cs() == "cmidrule") {
1098                                         string trim;
1099                                         if (t.cs() == "cmidrule") {
1100                                                 booktabs = true;
1101                                                 trim = p1.getFullParentheseArg();
1102                                         }
1103                                         string arg = p1.verbatim_item();
1104                                         //cerr << "read " << t.cs() << " arg: '" << arg << "', trim: '" << trim << "'\n";
1105                                         vector<string> cols;
1106                                         split(arg, cols, '-');
1107                                         cols.resize(2);
1108                                         size_t from = convert<unsigned int>(cols[0]);
1109                                         if (from == 0)
1110                                                 cerr << "Could not parse "
1111                                                      << t.cs() << " start column."
1112                                                      << endl;
1113                                         else
1114                                                 // 1 based index -> 0 based
1115                                                 --from;
1116                                         if (from >= colinfo.size()) {
1117                                                 cerr << t.cs() << " starts at "
1118                                                         "non existing column "
1119                                                      << (from + 1) << endl;
1120                                                 from = colinfo.size() - 1;
1121                                         }
1122                                         size_t to = convert<unsigned int>(cols[1]);
1123                                         if (to == 0)
1124                                                 cerr << "Could not parse "
1125                                                      << t.cs() << " end column."
1126                                                      << endl;
1127                                         else
1128                                                 // 1 based index -> 0 based
1129                                                 --to;
1130                                         if (to >= colinfo.size()) {
1131                                                 cerr << t.cs() << " ends at "
1132                                                         "non existing column "
1133                                                      << (to + 1) << endl;
1134                                                 to = colinfo.size() - 1;
1135                                         }
1136                                         for (size_t col = from; col <= to; ++col) {
1137                                                 //cerr << "row: " << row << " col: " << col << " i: " << i << endl;
1138                                                 if (i == 0) {
1139                                                         rowinfo[row].topline = true;
1140                                                         cellinfo[row][col].topline = true;
1141                                                         if (support::contains(trim, 'l') && col == from) {
1142                                                                 //rowinfo[row].topline_ltrim = true;
1143                                                                 cellinfo[row][col].topline_ltrim = true;
1144                                                         }
1145                                                         else if (support::contains(trim, 'r') && col == to) {
1146                                                                 //rowinfo[row].topline_rtrim = true;
1147                                                                 cellinfo[row][col].topline_rtrim = true;
1148                                                         }
1149                                                 } else {
1150                                                         rowinfo[row].bottomline = true;
1151                                                         cellinfo[row][col].bottomline = true;
1152                                                         if (support::contains(trim, 'l') && col == from) {
1153                                                                 //rowinfo[row].bottomline_ltrim = true;
1154                                                                 cellinfo[row][col].bottomline_ltrim = true;
1155                                                         }
1156                                                         else if (support::contains(trim, 'r') && col == to) {
1157                                                                 //rowinfo[row].bottomline_rtrim = true;
1158                                                                 cellinfo[row][col].bottomline_rtrim = true;
1159                                                         }
1160                                                 }
1161                                         }
1162                                 } else if (t.cs() == "addlinespace") {
1163                                         booktabs = true;
1164                                         string const opt = p.next_token().cat() == catBegin ?
1165                                                         p.verbatim_item() : string();
1166                                         if (i == 0) {
1167                                                 if (opt.empty())
1168                                                         rowinfo[row].top_space = "default";
1169                                                 else
1170                                                         rowinfo[row].top_space = translate_len(opt);
1171                                         } else if (rowinfo[row].bottomline) {
1172                                                 if (opt.empty())
1173                                                         rowinfo[row].bottom_space = "default";
1174                                                 else
1175                                                         rowinfo[row].bottom_space = translate_len(opt);
1176                                         } else {
1177                                                 if (opt.empty())
1178                                                         rowinfo[row].interline_space = "default";
1179                                                 else
1180                                                         rowinfo[row].interline_space = translate_len(opt);
1181                                         }
1182                                 } else if (t.cs() == "endhead") {
1183                                         if (i == 0)
1184                                                 endhead.empty = true;
1185                                         else
1186                                                 rowinfo[row].type = LT_HEAD;
1187                                         for (int r = row - 1; r >= 0; --r) {
1188                                                 if (rowinfo[r].type != LT_NORMAL)
1189                                                         break;
1190                                                 rowinfo[r].type = LT_HEAD;
1191                                                 endhead.empty = false;
1192                                         }
1193                                         endhead.set = true;
1194                                 } else if (t.cs() == "endfirsthead") {
1195                                         if (i == 0)
1196                                                 endfirsthead.empty = true;
1197                                         else
1198                                                 rowinfo[row].type = LT_FIRSTHEAD;
1199                                         for (int r = row - 1; r >= 0; --r) {
1200                                                 if (rowinfo[r].type != LT_NORMAL)
1201                                                         break;
1202                                                 rowinfo[r].type = LT_FIRSTHEAD;
1203                                                 endfirsthead.empty = false;
1204                                         }
1205                                         endfirsthead.set = true;
1206                                 } else if (t.cs() == "endfoot") {
1207                                         if (i == 0)
1208                                                 endfoot.empty = true;
1209                                         else
1210                                                 rowinfo[row].type = LT_FOOT;
1211                                         for (int r = row - 1; r >= 0; --r) {
1212                                                 if (rowinfo[r].type != LT_NORMAL)
1213                                                         break;
1214                                                 rowinfo[r].type = LT_FOOT;
1215                                                 endfoot.empty = false;
1216                                         }
1217                                         endfoot.set = true;
1218                                 } else if (t.cs() == "endlastfoot") {
1219                                         if (i == 0)
1220                                                 endlastfoot.empty = true;
1221                                         else
1222                                                 rowinfo[row].type = LT_LASTFOOT;
1223                                         for (int r = row - 1; r >= 0; --r) {
1224                                                 if (rowinfo[r].type != LT_NORMAL)
1225                                                         break;
1226                                                 rowinfo[r].type = LT_LASTFOOT;
1227                                                 endlastfoot.empty = false;
1228                                         }
1229                                         endlastfoot.set = true;
1230                                 } else if (t.cs() == "newpage") {
1231                                         if (i == 0) {
1232                                                 if (row > 0)
1233                                                         rowinfo[row - 1].newpage = true;
1234                                                 else
1235                                                         // This does not work in LaTeX
1236                                                         cerr << "Ignoring "
1237                                                                 "'\\newpage' "
1238                                                                 "before rows."
1239                                                              << endl;
1240                                         } else
1241                                                 rowinfo[row].newpage = true;
1242                                 } else {
1243                                         cerr << "unexpected line token: " << t << endl;
1244                                 }
1245                         }
1246                 }
1247
1248                 // LyX ends headers and footers always with \tabularnewline.
1249                 // This causes one additional row in the output.
1250                 // If the last row of a header/footer is empty, we can work
1251                 // around that by removing it.
1252                 if (row > 1) {
1253                         RowInfo test = rowinfo[row-1];
1254                         test.type = LT_NORMAL;
1255                         if (lines[row-1].empty() && !test.special()) {
1256                                 switch (rowinfo[row-1].type) {
1257                                 case LT_FIRSTHEAD:
1258                                         if (rowinfo[row].type != LT_FIRSTHEAD &&
1259                                             rowinfo[row-2].type == LT_FIRSTHEAD)
1260                                                 deletelastrow = true;
1261                                         break;
1262                                 case LT_HEAD:
1263                                         if (rowinfo[row].type != LT_HEAD &&
1264                                             rowinfo[row-2].type == LT_HEAD)
1265                                                 deletelastrow = true;
1266                                         break;
1267                                 case LT_FOOT:
1268                                         if (rowinfo[row].type != LT_FOOT &&
1269                                             rowinfo[row-2].type == LT_FOOT)
1270                                                 deletelastrow = true;
1271                                         break;
1272                                 case LT_LASTFOOT:
1273                                         if (rowinfo[row].type != LT_LASTFOOT &&
1274                                             rowinfo[row-2].type == LT_LASTFOOT)
1275                                                 deletelastrow = true;
1276                                         break;
1277                                 case LT_NORMAL:
1278                                         break;
1279                                 }
1280                         }
1281                 }
1282
1283                 if (deletelastrow) {
1284                         lines.erase(lines.begin() + (row - 1));
1285                         rowinfo.erase(rowinfo.begin() + (row - 1));
1286                         cellinfo.erase(cellinfo.begin() + (row - 1));
1287                         continue;
1288                 }
1289
1290                 // split into cells
1291                 vector<string> cells;
1292                 split(lines[row], cells, TAB);
1293                 for (size_t col = 0, cell = 0; cell < cells.size();
1294                      ++col, ++cell) {
1295                         //cerr << "cell content: '" << cells[cell] << "'\n";
1296                         if (col >= colinfo.size()) {
1297                                 // This does not work in LaTeX
1298                                 cerr << "Ignoring extra cell '"
1299                                      << cells[cell] << "'." << endl;
1300                                 continue;
1301                         }
1302                         string cellcont = cells[cell];
1303                         // For decimal cells, ass the content of the second one to the first one
1304                         // of a pair.
1305                         if (colinfo[col].decimal_point != '\0' && colinfo[col].align == 'd' && cell < cells.size() - 1)
1306                                 cellcont += colinfo[col].decimal_point + cells[cell + 1];
1307                         Parser parse(cellcont);
1308                         parse.skip_spaces();
1309                         //cells[cell] << "'\n";
1310                         if (parse.next_token().cs() == "multirow") {
1311                                 // We do not support the vpos arg yet.
1312                                 if (parse.hasOpt()) {
1313                                         string const vpos = parse.getArg('[', ']');
1314                                         parse.skip_spaces(true);
1315                                         cerr << "Ignoring multirow's vpos arg '"
1316                                              << vpos << "'!" << endl;
1317                                 }
1318                                 // how many cells?
1319                                 parse.get_token();
1320                                 size_t const ncells =
1321                                         convert<unsigned int>(parse.verbatim_item());
1322                                 // We do not support the bigstrut arg yet.
1323                                 if (parse.hasOpt()) {
1324                                         string const bs = parse.getArg('[', ']');
1325                                         parse.skip_spaces(true);
1326                                         cerr << "Ignoring multirow's bigstrut arg '"
1327                                              << bs << "'!" << endl;
1328                                 }
1329                                 // the width argument
1330                                 string const width = parse.getArg('{', '}');
1331                                 // the vmove arg
1332                                 string vmove;
1333                                 if (parse.hasOpt()) {
1334                                         vmove = parse.getArg('[', ']');
1335                                         parse.skip_spaces(true);
1336                                 }
1337
1338                                 if (width != "*" && width != "=")
1339                                         colinfo[col].width = width;
1340                                 if (!vmove.empty())
1341                                         cellinfo[row][col].mroffset = vmove;
1342                                 cellinfo[row][col].multi = CELL_BEGIN_OF_MULTIROW;
1343                                 cellinfo[row][col].leftlines  = colinfo[col].leftlines;
1344                                 cellinfo[row][col].rightlines = colinfo[col].rightlines;
1345                                 cellinfo[row][col].mrxnum = ncells - 1;
1346
1347                                 ostringstream os2;
1348                                 parse.get_token();// skip {
1349                                 parse_cell_content(os2, parse, FLAG_BRACE_LAST, newcontext,
1350                                                         cellinfo, colinfo,
1351                                                         row, col);
1352                                 parse.get_token();// skip }
1353                                 if (!cellinfo[row][col].content.empty()) {
1354                                         // This may or may not work in LaTeX,
1355                                         // but it does not work in LyX.
1356                                         // FIXME: Handle it correctly!
1357                                         cerr << "Moving cell content '"
1358                                              << cells[cell]
1359                                              << "' into a multirow cell. "
1360                                                 "This will probably not work."
1361                                              << endl;
1362                                 }
1363                                 cellinfo[row][col].content += os2.str();
1364                         } else if (parse.next_token().cs() == "multicolumn") {
1365                                 // how many cells?
1366                                 parse.get_token();
1367                                 size_t const ncells =
1368                                         convert<unsigned int>(parse.verbatim_item());
1369
1370                                 // special cell properties alignment
1371                                 vector<ColInfo> t;
1372                                 handle_colalign(parse, t, ColInfo());
1373                                 parse.skip_spaces(true);
1374                                 ColInfo & ci = t.front();
1375
1376                                 // The logic of LyX for multicolumn vertical
1377                                 // lines is too complicated to reproduce it
1378                                 // here (see LyXTabular::TeXCellPreamble()).
1379                                 // Therefore we simply put everything in the
1380                                 // special field.
1381                                 ci2special(ci);
1382
1383                                 cellinfo[row][col].multi      = CELL_BEGIN_OF_MULTICOLUMN;
1384                                 cellinfo[row][col].align      = ci.align;
1385                                 cellinfo[row][col].special    = ci.special;
1386                                 cellinfo[row][col].leftlines  = ci.leftlines;
1387                                 cellinfo[row][col].rightlines = ci.rightlines;
1388                                 ostringstream os2;
1389                                 parse.get_token();// skip {
1390                                 parse_cell_content(os2, parse, FLAG_BRACE_LAST, newcontext,
1391                                                         cellinfo, colinfo,
1392                                                         row, col);
1393                                 parse.get_token();// skip }
1394                                 if (!cellinfo[row][col].content.empty()) {
1395                                         // This may or may not work in LaTeX,
1396                                         // but it does not work in LyX.
1397                                         // FIXME: Handle it correctly!
1398                                         cerr << "Moving cell content '"
1399                                              << cells[cell]
1400                                              << "' into a multicolumn cell. "
1401                                                 "This will probably not work."
1402                                              << endl;
1403                                 }
1404                                 cellinfo[row][col].content += os2.str();
1405
1406                                 // add dummy cells for multicol
1407                                 for (size_t i = 0; i < ncells - 1; ++i) {
1408                                         ++col;
1409                                         if (col >= colinfo.size()) {
1410                                                 cerr << "The cell '"
1411                                                         << cells[cell]
1412                                                         << "' specifies "
1413                                                         << convert<string>(ncells)
1414                                                         << " columns while the table has only "
1415                                                         << convert<string>(colinfo.size())
1416                                                         << " columns!"
1417                                                         << " Therefore the surplus columns will be ignored."
1418                                                         << endl;
1419                                                 break;
1420                                         }
1421                                         cellinfo[row][col].multi = CELL_PART_OF_MULTICOLUMN;
1422                                         cellinfo[row][col].align = 'c';
1423                                 }
1424
1425                         } else if (col == 0 && colinfo.size() > 1 &&
1426                                    is_long_tabular &&
1427                                    parse.next_token().cs() == "caption") {
1428                                 // longtable caption support in LyX is a hack:
1429                                 // Captions require a row of their own with
1430                                 // the caption flag set to true, having only
1431                                 // one multicolumn cell. The contents of that
1432                                 // cell must contain exactly one caption inset
1433                                 // and nothing else.
1434                                 // Fortunately, the caption flag is only needed
1435                                 // for tables with more than one column.
1436                                 rowinfo[row].caption = true;
1437                                 for (size_t c = 1; c < cells.size(); ++c) {
1438                                         if (!cells[c].empty()) {
1439                                                 cerr << "Moving cell content '"
1440                                                      << cells[c]
1441                                                      << "' into the caption cell. "
1442                                                         "This will probably not work."
1443                                                      << endl;
1444                                                 cells[0] += cells[c];
1445                                         }
1446                                 }
1447                                 cells.resize(1);
1448                                 cellinfo[row][col].align      = colinfo[col].align;
1449                                 cellinfo[row][col].multi      = CELL_BEGIN_OF_MULTICOLUMN;
1450                                 ostringstream os2;
1451                                 parse_text_in_inset(parse, os2, FLAG_CELL, false, newcontext);
1452                                 cellinfo[row][col].content += os2.str();
1453                                 // add dummy multicolumn cells
1454                                 for (size_t c = 1; c < colinfo.size(); ++c)
1455                                         cellinfo[row][c].multi = CELL_PART_OF_MULTICOLUMN;
1456                         } else if (col == 0 && colinfo.size() > 1 &&
1457                                    is_long_tabular &&
1458                                    parse.next_token().cs() == "caption") {
1459                                 // longtable caption support in LyX is a hack:
1460                                 // Captions require a row of their own with
1461                                 // the caption flag set to true, having only
1462                                 // one multicolumn cell. The contents of that
1463                                 // cell must contain exactly one caption inset
1464                                 // and nothing else.
1465                                 // Fortunately, the caption flag is only needed
1466                                 // for tables with more than one column.
1467                                 rowinfo[row].caption = true;
1468                                 for (size_t c = 1; c < cells.size(); ++c) {
1469                                         if (!cells[c].empty()) {
1470                                                 cerr << "Moving cell content '"
1471                                                      << cells[c]
1472                                                      << "' into the caption cell. "
1473                                                         "This will probably not work."
1474                                                      << endl;
1475                                                 cells[0] += cells[c];
1476                                         }
1477                                 }
1478                                 cells.resize(1);
1479                                 cellinfo[row][col].align      = colinfo[col].align;
1480                                 cellinfo[row][col].multi      = CELL_BEGIN_OF_MULTICOLUMN;
1481                                 ostringstream os2;
1482                                 parse_text_in_inset(parse, os2, FLAG_CELL, false, newcontext);
1483                                 cellinfo[row][col].content += os2.str();
1484                                 // add dummy multicolumn cells
1485                                 for (size_t c = 1; c < colinfo.size(); ++c)
1486                                         cellinfo[row][c].multi = CELL_PART_OF_MULTICOLUMN;
1487                         } else {
1488                                 ostringstream os2;
1489                                 parse_cell_content(os2, parse, FLAG_CELL, newcontext,
1490                                                         cellinfo, colinfo,
1491                                                         row, col);
1492                                 cellinfo[row][col].leftlines  = colinfo[col].leftlines;
1493                                 cellinfo[row][col].rightlines = colinfo[col].rightlines;
1494                                 cellinfo[row][col].align      = colinfo[col].align;
1495                                 cellinfo[row][col].content += os2.str();
1496                         }
1497                 }
1498
1499                 //cerr << "//  handle almost empty last row what we have\n";
1500                 // handle almost empty last row
1501                 if (row && lines[row].empty() && row + 1 == rowinfo.size()) {
1502                         //cerr << "remove empty last line\n";
1503                         if (rowinfo[row].topline)
1504                                 rowinfo[row - 1].bottomline = true;
1505                         for (size_t col = 0; col < colinfo.size(); ++col)
1506                                 if (cellinfo[row][col].topline)
1507                                         cellinfo[row - 1][col].bottomline = true;
1508                         rowinfo.pop_back();
1509                 }
1510
1511                 ++row;
1512         }
1513
1514         // Now we have the table structure and content in rowinfo, colinfo
1515         // and cellinfo.
1516         // Unfortunately LyX has some limitations that we need to work around.
1517
1518         // Some post work
1519         for (size_t row = 0; row < rowinfo.size(); ++row) {
1520                 for (size_t col = 0; col < cellinfo[row].size(); ++col) {
1521                         // Convert cells with special content to multicolumn cells
1522                         // (LyX ignores the special field for non-multicolumn cells).
1523                         if (cellinfo[row][col].multi == CELL_NORMAL &&
1524                             !cellinfo[row][col].special.empty())
1525                                 cellinfo[row][col].multi = CELL_BEGIN_OF_MULTICOLUMN;
1526                         // Add multirow dummy cells
1527                         if (row > 1 && (cellinfo[row - 1][col].multi == CELL_PART_OF_MULTIROW
1528                                         || cellinfo[row - 1][col].multi == CELL_BEGIN_OF_MULTIROW)
1529                                     && cellinfo[row - 1][col].mrxnum > 0) {
1530                                 // add dummy cells for multirow
1531                                 cellinfo[row][col].multi = CELL_PART_OF_MULTIROW;
1532                                 cellinfo[row][col].align = 'c';
1533                                 cellinfo[row][col].mrxnum = cellinfo[row - 1][col].mrxnum - 1;
1534                         }
1535                 }
1536         }
1537
1538         // Distribute lines from rows/columns to cells
1539         // The code was stolen from convert_tablines() in lyx2lyx/lyx_1_6.py.
1540         // Each standard cell inherits the settings of the corresponding
1541         // rowinfo/colinfo. This works because all cells with individual
1542         // settings were converted to multicolumn cells above.
1543         // Each multicolumn cell inherits the settings of the rowinfo/colinfo
1544         // corresponding to the first column of the multicolumn cell (default
1545         // of the multicol package). This works because the special field
1546         // overrides the line fields.
1547         for (size_t row = 0; row < rowinfo.size(); ++row) {
1548                 for (size_t col = 0; col < cellinfo[row].size(); ++col) {
1549                         if (cellinfo[row][col].multi == CELL_NORMAL) {
1550                                 cellinfo[row][col].topline = rowinfo[row].topline;
1551                                 cellinfo[row][col].bottomline = rowinfo[row].bottomline;
1552                                 cellinfo[row][col].leftlines = colinfo[col].leftlines;
1553                                 cellinfo[row][col].rightlines = colinfo[col].rightlines;
1554                         } else if (cellinfo[row][col].multi == CELL_BEGIN_OF_MULTICOLUMN) {
1555                                 size_t s = col + 1;
1556                                 while (s < cellinfo[row].size() &&
1557                                        cellinfo[row][s].multi == CELL_PART_OF_MULTICOLUMN)
1558                                         s++;
1559                                 if (s < cellinfo[row].size() &&
1560                                     cellinfo[row][s].multi != CELL_BEGIN_OF_MULTICOLUMN)
1561                                         cellinfo[row][col].rightlines = colinfo[col].rightlines;
1562                                 if (col > 0 && cellinfo[row][col-1].multi == CELL_NORMAL)
1563                                         cellinfo[row][col].leftlines = colinfo[col].leftlines;
1564                         } else if (cellinfo[row][col].multi == CELL_BEGIN_OF_MULTIROW) {
1565                                 size_t s = row + 1;
1566                                 while (s < rowinfo.size() &&
1567                                        cellinfo[s][col].multi == CELL_PART_OF_MULTIROW)
1568                                         s++;
1569                                 if (s < cellinfo[row].size() &&
1570                                     cellinfo[s][col].multi != CELL_BEGIN_OF_MULTIROW)
1571                                         cellinfo[row][col].bottomline = rowinfo[row].bottomline;
1572                                 if (row > 0 && cellinfo[row - 1][col].multi == CELL_NORMAL)
1573                                         cellinfo[row][col].topline = rowinfo[row].topline;
1574                         }
1575                 }
1576         }
1577
1578         if (booktabs)
1579                 preamble.registerAutomaticallyLoadedPackage("booktabs");
1580         if (name == "longtable")
1581                 preamble.registerAutomaticallyLoadedPackage("longtable");
1582         else if (name == "xltabular")
1583                 preamble.registerAutomaticallyLoadedPackage("xltabular");
1584         else if (name == "tabularx")
1585                 preamble.registerAutomaticallyLoadedPackage("tabularx");
1586
1587         //cerr << "// output what we have\n";
1588         // output what we have
1589         size_type cols = colinfo.size();
1590         for (auto const & col : colinfo) {
1591                 if (col.decimal_point != '\0' && col.align != 'd')
1592                         --cols;
1593         }
1594         os << "\n<lyxtabular version=\"3\" rows=\"" << rowinfo.size()
1595            << "\" columns=\"" << cols << "\">\n";
1596         os << "<features"
1597            << write_attribute("rotate", context.tablerotation)
1598            << write_attribute("booktabs", booktabs)
1599            << write_attribute("islongtable", is_long_tabular);
1600         if (is_long_tabular) {
1601                 os << write_attribute("firstHeadTopDL", endfirsthead.topDL)
1602                    << write_attribute("firstHeadBottomDL", endfirsthead.bottomDL)
1603                    << write_attribute("firstHeadEmpty", endfirsthead.empty)
1604                    << write_attribute("headTopDL", endhead.topDL)
1605                    << write_attribute("headBottomDL", endhead.bottomDL)
1606                    << write_attribute("footTopDL", endfoot.topDL)
1607                    << write_attribute("footBottomDL", endfoot.bottomDL)
1608                    << write_attribute("lastFootTopDL", endlastfoot.topDL)
1609                    << write_attribute("lastFootBottomDL", endlastfoot.bottomDL)
1610                    << write_attribute("lastFootEmpty", endlastfoot.empty);
1611                 if (!halign.empty())
1612                         os << write_attribute("longtabularalignment", halign);
1613         } else
1614                 os << write_attribute("tabularvalignment", tabularvalignment);
1615
1616         os << write_attribute("tabularwidth", tabularwidth) << ">\n";
1617
1618         //cerr << "// after header\n";
1619         for (auto const & col : colinfo) {
1620                 if (col.decimal_point != '\0' && col.align != 'd')
1621                         continue;
1622                 os << "<column alignment=\""
1623                            << verbose_align(col.align) << "\"";
1624                 if (col.decimal_point != '\0')
1625                         os << " decimal_point=\"" << col.decimal_point << "\"";
1626                 os << " valignment=\""
1627                    << verbose_valign(col.valign) << "\""
1628                    << write_attribute("width", translate_len(col.width))
1629                    << write_attribute("special", col.special)
1630                    << write_attribute("varwidth", col.varwidth)
1631                    << ">\n";
1632         }
1633         //cerr << "// after cols\n";
1634
1635         for (size_t row = 0; row < rowinfo.size(); ++row) {
1636                 os << "<row"
1637                    << write_attribute("topspace", rowinfo[row].top_space)
1638                    << write_attribute("bottomspace", rowinfo[row].bottom_space)
1639                    << write_attribute("interlinespace", rowinfo[row].interline_space)
1640                    << write_attribute("endhead",
1641                                       rowinfo[row].type == LT_HEAD)
1642                    << write_attribute("endfirsthead",
1643                                       rowinfo[row].type == LT_FIRSTHEAD)
1644                    << write_attribute("endfoot",
1645                                       rowinfo[row].type == LT_FOOT)
1646                    << write_attribute("endlastfoot",
1647                                       rowinfo[row].type == LT_LASTFOOT)
1648                    << write_attribute("newpage", rowinfo[row].newpage)
1649                    << write_attribute("caption", rowinfo[row].caption)
1650                    << ">\n";
1651                 for (size_t col = 0; col < colinfo.size(); ++col) {
1652                         CellInfo const & cell = cellinfo[row][col];
1653                         if (colinfo[col].decimal_point != '\0' && colinfo[col].align != 'd')
1654                                 // These are the second columns in a salign pair. Skip.
1655                                 continue;
1656                         os << "<cell";
1657                         if ((cell.multi == CELL_BEGIN_OF_MULTICOLUMN
1658                                     || cell.multi == CELL_PART_OF_MULTICOLUMN)
1659                                    && colinfo[col].align != 'd')
1660                                 os << " multicolumn=\"" << cell.multi << "\"";
1661                         if (cell.multi == CELL_BEGIN_OF_MULTIROW
1662                             || cell.multi == CELL_PART_OF_MULTIROW)
1663                                 os << " multirow=\"" << cell.multi << "\"";
1664                         os << " alignment=\"" << verbose_align(cell.align)
1665                            << "\""
1666                            << " valignment=\"" << verbose_valign(cell.valign)
1667                            << "\""
1668                            << write_attribute("topline", cell.topline)
1669                            << write_attribute("toplineltrim", cell.topline_ltrim)
1670                            << write_attribute("toplinertrim", cell.topline_rtrim)
1671                            << write_attribute("bottomline", cell.bottomline)
1672                            << write_attribute("bottomlineltrim", cell.bottomline_ltrim)
1673                            << write_attribute("bottomlinertrim", cell.bottomline_rtrim)
1674                            << write_attribute("leftline", cell.leftlines > 0)
1675                            << write_attribute("rightline", cell.rightlines > 0)
1676                            << write_attribute("rotate", cell.rotate)
1677                            << write_attribute("mroffset", cell.mroffset);
1678                         //cerr << "\nrow: " << row << " col: " << col;
1679                         //if (cell.topline)
1680                         //      cerr << " topline=\"true\"";
1681                         //if (cell.bottomline)
1682                         //      cerr << " bottomline=\"true\"";
1683                         os << " usebox=\"none\""
1684                            << write_attribute("width", translate_len(cell.width));
1685                         if (cell.multi != CELL_NORMAL)
1686                                 os << write_attribute("special", cell.special);
1687                         os << ">"
1688                            << "\n\\begin_inset Text\n"
1689                            << cell.content
1690                            << "\n\\end_inset\n"
1691                            << "</cell>\n";
1692                 }
1693                 os << "</row>\n";
1694         }
1695
1696         os << "</lyxtabular>\n";
1697 }
1698
1699
1700
1701
1702 // }])
1703
1704
1705 } // namespace lyx