src/tex2lyx/text.cpp

   1 /**
   2  * \file tex2lyx/text.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 Uwe Stöhr
   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 "FloatList.h"
  21 #include "Layout.h"
  22 #include "Length.h"
  23
  24 #include "support/lassert.h"
  25 #include "support/convert.h"
  26 #include "support/FileName.h"
  27 #include "support/filetools.h"
  28 #include "support/lstrings.h"
  29
  30 #include <algorithm>
  31 #include <iostream>
  32 #include <map>
  33 #include <sstream>
  34 #include <vector>
  35
  36 using namespace std;
  37 using namespace lyx::support;
  38
  39 namespace lyx {
  40
  41
  42 void parse_text_in_inset(Parser & p, ostream & os, unsigned flags, bool outer,
  43                 Context const & context)
  44 {
  45         Context newcontext(true, context.textclass);
  46         newcontext.font = context.font;
  47         parse_text(p, os, flags, outer, newcontext);
  48         newcontext.check_end_layout(os);
  49 }
  50
  51
  52 namespace {
  53
  54 /// parses a paragraph snippet, useful for example for \\emph{...}
  55 void parse_text_snippet(Parser & p, ostream & os, unsigned flags, bool outer,
  56                 Context & context)
  57 {
  58         Context newcontext(context);
  59         // Don't inherit the paragraph-level extra stuff
  60         newcontext.par_extra_stuff.clear();
  61         parse_text(p, os, flags, outer, newcontext);
  62         // Make sure that we don't create invalid .lyx files
  63         context.need_layout = newcontext.need_layout;
  64         context.need_end_layout = newcontext.need_end_layout;
  65 }
  66
  67
  68 /*!
  69  * Thin wrapper around parse_text_snippet() using a string.
  70  *
  71  * We completely ignore \c context.need_layout and \c context.need_end_layout,
  72  * because our return value is not used directly (otherwise the stream version
  73  * of parse_text_snippet() could be used). That means that the caller needs
  74  * to do layout management manually.
  75  * This is intended to parse text that does not create any layout changes.
  76  */
  77 string parse_text_snippet(Parser & p, unsigned flags, const bool outer,
  78                   Context & context)
  79 {
  80         Context newcontext(context);
  81         newcontext.need_layout = false;
  82         newcontext.need_end_layout = false;
  83         newcontext.new_layout_allowed = false;
  84         // Avoid warning by Context::~Context()
  85         newcontext.par_extra_stuff.clear();
  86         ostringstream os;
  87         parse_text_snippet(p, os, flags, outer, newcontext);
  88         return os.str();
  89 }
  90
  91
  92 char const * const known_latex_commands[] = { "ref", "cite", "label",
  93  "index", "printindex", "pageref", "url", "vref", "vpageref", "prettyref",
  94  "eqref", 0 };
  95
  96 /*!
  97  * natbib commands.
  98  * We can't put these into known_latex_commands because the argument order
  99  * is reversed in lyx if there are 2 arguments.
 100  * The starred forms are also known.
 101  */
 102 char const * const known_natbib_commands[] = { "cite", "citet", "citep",
 103 "citealt", "citealp", "citeauthor", "citeyear", "citeyearpar",
 104 "citefullauthor", "Citet", "Citep", "Citealt", "Citealp", "Citeauthor", 0 };
 105
 106 /*!
 107  * jurabib commands.
 108  * We can't put these into known_latex_commands because the argument order
 109  * is reversed in lyx if there are 2 arguments.
 110  * No starred form other than "cite*" known.
 111  */
 112 char const * const known_jurabib_commands[] = { "cite", "citet", "citep",
 113 "citealt", "citealp", "citeauthor", "citeyear", "citeyearpar",
 114 // jurabib commands not (yet) supported by LyX:
 115 // "fullcite",
 116 // "footcite", "footcitet", "footcitep", "footcitealt", "footcitealp",
 117 // "footciteauthor", "footciteyear", "footciteyearpar",
 118 "citefield", "citetitle", "cite*", 0 };
 119
 120 /// LaTeX names for quotes
 121 char const * const known_quotes[] = { "dq", "guillemotleft", "flqq", "og",
 122 "guillemotright", "frqq", "fg", "glq", "glqq", "textquoteleft", "grq", "grqq",
 123 "quotedblbase", "textquotedblleft", "quotesinglbase", "textquoteright", "flq",
 124 "guilsinglleft", "frq", "guilsinglright", 0};
 125
 126 /// the same as known_quotes with .lyx names
 127 char const * const known_coded_quotes[] = { "prd", "ard", "ard", "ard",
 128 "ald", "ald", "ald", "gls", "gld", "els", "els", "grd",
 129 "gld", "grd", "gls", "ers", "fls",
 130 "fls", "frs", "frs", 0};
 131
 132 /// LaTeX names for font sizes
 133 char const * const known_sizes[] = { "tiny", "scriptsize", "footnotesize",
 134 "small", "normalsize", "large", "Large", "LARGE", "huge", "Huge", 0};
 135
 136 /// the same as known_sizes with .lyx names plus a default entry
 137 char const * const known_coded_sizes[] = { "default", "tiny", "scriptsize", "footnotesize",
 138 "small", "normal", "large", "larger", "largest", "huge", "giant", 0};
 139
 140 /// LaTeX 2.09 names for font families
 141 char const * const known_old_font_families[] = { "rm", "sf", "tt", 0};
 142
 143 /// LaTeX names for font families
 144 char const * const known_font_families[] = { "rmfamily", "sffamily",
 145 "ttfamily", 0};
 146
 147 /// the same as known_old_font_families and known_font_families with .lyx names
 148 char const * const known_coded_font_families[] = { "roman", "sans",
 149 "typewriter", 0};
 150
 151 /// LaTeX 2.09 names for font series
 152 char const * const known_old_font_series[] = { "bf", 0};
 153
 154 /// LaTeX names for font series
 155 char const * const known_font_series[] = { "bfseries", "mdseries", 0};
 156
 157 /// the same as known_old_font_series and known_font_series with .lyx names
 158 char const * const known_coded_font_series[] = { "bold", "medium", 0};
 159
 160 /// LaTeX 2.09 names for font shapes
 161 char const * const known_old_font_shapes[] = { "it", "sl", "sc", 0};
 162
 163 /// LaTeX names for font shapes
 164 char const * const known_font_shapes[] = { "itshape", "slshape", "scshape",
 165 "upshape", 0};
 166
 167 /// the same as known_old_font_shapes and known_font_shapes with .lyx names
 168 char const * const known_coded_font_shapes[] = { "italic", "slanted",
 169 "smallcaps", "up", 0};
 170
 171 /*!
 172  * Graphics file extensions known by the dvips driver of the graphics package.
 173  * These extensions are used to complete the filename of an included
 174  * graphics file if it does not contain an extension.
 175  * The order must be the same that latex uses to find a file, because we
 176  * will use the first extension that matches.
 177  * This is only an approximation for the common cases. If we would want to
 178  * do it right in all cases, we would need to know which graphics driver is
 179  * used and know the extensions of every driver of the graphics package.
 180  */
 181 char const * const known_dvips_graphics_formats[] = {"eps", "ps", "eps.gz",
 182 "ps.gz", "eps.Z", "ps.Z", 0};
 183
 184 /*!
 185  * Graphics file extensions known by the pdftex driver of the graphics package.
 186  * \sa known_dvips_graphics_formats
 187  */
 188 char const * const known_pdftex_graphics_formats[] = {"png", "pdf", "jpg",
 189 "mps", "tif", 0};
 190
 191 /*!
 192  * Known file extensions for TeX files as used by \\include.
 193  */
 194 char const * const known_tex_extensions[] = {"tex", 0};
 195
 196 /// spaces known by InsetSpace
 197 char const * const known_spaces[] = { " ", "space", ",", "thinspace", "quad",
 198 "qquad", "enspace", "enskip", "negthinspace", 0};
 199
 200 /// the same as known_spaces with .lyx names
 201 char const * const known_coded_spaces[] = { "space{}", "space{}",
 202 "thinspace{}", "thinspace{}", "quad{}", "qquad{}", "enspace{}", "enskip{}",
 203 "negthinspace{}", 0};
 204
 205
 206 /// splits "x=z, y=b" into a map
 207 map<string, string> split_map(string const & s)
 208 {
 209         map<string, string> res;
 210         vector<string> v;
 211         split(s, v);
 212         for (size_t i = 0; i < v.size(); ++i) {
 213                 size_t const pos   = v[i].find('=');
 214                 string const index = v[i].substr(0, pos);
 215                 string const value = v[i].substr(pos + 1, string::npos);
 216                 res[trim(index)] = trim(value);
 217         }
 218         return res;
 219 }
 220
 221
 222 /*!
 223  * Split a LaTeX length into value and unit.
 224  * The latter can be a real unit like "pt", or a latex length variable
 225  * like "\textwidth". The unit may contain additional stuff like glue
 226  * lengths, but we don't care, because such lengths are ERT anyway.
 227  * \returns true if \p value and \p unit are valid.
 228  */
 229 bool splitLatexLength(string const & len, string & value, string & unit)
 230 {
 231         if (len.empty())
 232                 return false;
 233         const string::size_type i = len.find_first_not_of(" -+0123456789.,");
 234         //'4,5' is a valid LaTeX length number. Change it to '4.5'
 235         string const length = subst(len, ',', '.');
 236         if (i == string::npos)
 237                 return false;
 238         if (i == 0) {
 239                 if (len[0] == '\\') {
 240                         // We had something like \textwidth without a factor
 241                         value = "1.0";
 242                 } else {
 243                         return false;
 244                 }
 245         } else {
 246                 value = trim(string(length, 0, i));
 247         }
 248         if (value == "-")
 249                 value = "-1.0";
 250         // 'cM' is a valid LaTeX length unit. Change it to 'cm'
 251         if (contains(len, '\\'))
 252                 unit = trim(string(len, i));
 253         else
 254                 unit = ascii_lowercase(trim(string(len, i)));
 255         return true;
 256 }
 257
 258
 259 /// A simple function to translate a latex length to something lyx can
 260 /// understand. Not perfect, but rather best-effort.
 261 bool translate_len(string const & length, string & valstring, string & unit)
 262 {
 263         if (!splitLatexLength(length, valstring, unit))
 264                 return false;
 265         // LyX uses percent values
 266         double value;
 267         istringstream iss(valstring);
 268         iss >> value;
 269         value *= 100;
 270         ostringstream oss;
 271         oss << value;
 272         string const percentval = oss.str();
 273         // a normal length
 274         if (unit.empty() || unit[0] != '\\')
 275                 return true;
 276         string::size_type const i = unit.find(' ');
 277         string const endlen = (i == string::npos) ? string() : string(unit, i);
 278         if (unit == "\\textwidth") {
 279                 valstring = percentval;
 280                 unit = "text%" + endlen;
 281         } else if (unit == "\\columnwidth") {
 282                 valstring = percentval;
 283                 unit = "col%" + endlen;
 284         } else if (unit == "\\paperwidth") {
 285                 valstring = percentval;
 286                 unit = "page%" + endlen;
 287         } else if (unit == "\\linewidth") {
 288                 valstring = percentval;
 289                 unit = "line%" + endlen;
 290         } else if (unit == "\\paperheight") {
 291                 valstring = percentval;
 292                 unit = "pheight%" + endlen;
 293         } else if (unit == "\\textheight") {
 294                 valstring = percentval;
 295                 unit = "theight%" + endlen;
 296         }
 297         return true;
 298 }
 299
 300 }
 301
 302
 303 string translate_len(string const & length)
 304 {
 305         string unit;
 306         string value;
 307         if (translate_len(length, value, unit))
 308                 return value + unit;
 309         // If the input is invalid, return what we have.
 310         return length;
 311 }
 312
 313
 314 namespace {
 315
 316 /*!
 317  * Translates a LaTeX length into \p value, \p unit and
 318  * \p special parts suitable for a box inset.
 319  * The difference from translate_len() is that a box inset knows about
 320  * some special "units" that are stored in \p special.
 321  */
 322 void translate_box_len(string const & length, string & value, string & unit, string & special)
 323 {
 324         if (translate_len(length, value, unit)) {
 325                 if (unit == "\\height" || unit == "\\depth" ||
 326                     unit == "\\totalheight" || unit == "\\width") {
 327                         special = unit.substr(1);
 328                         // The unit is not used, but LyX requires a dummy setting
 329                         unit = "in";
 330                 } else
 331                         special = "none";
 332         } else {
 333                 value.clear();
 334                 unit = length;
 335                 special = "none";
 336         }
 337 }
 338
 339
 340 /*!
 341  * Find a file with basename \p name in path \p path and an extension
 342  * in \p extensions.
 343  */
 344 string find_file(string const & name, string const & path,
 345                  char const * const * extensions)
 346 {
 347         // FIXME UNICODE encoding of name and path may be wrong (makeAbsPath
 348         // expects utf8)
 349         for (char const * const * what = extensions; *what; ++what) {
 350                 string const trial = addExtension(name, *what);
 351                 if (makeAbsPath(trial, path).exists())
 352                         return trial;
 353         }
 354         return string();
 355 }
 356
 357
 358 void begin_inset(ostream & os, string const & name)
 359 {
 360         os << "\n\\begin_inset " << name;
 361 }
 362
 363
 364 void end_inset(ostream & os)
 365 {
 366         os << "\n\\end_inset\n\n";
 367 }
 368
 369
 370 void skip_braces(Parser & p)
 371 {
 372         if (p.next_token().cat() != catBegin)
 373                 return;
 374         p.get_token();
 375         if (p.next_token().cat() == catEnd) {
 376                 p.get_token();
 377                 return;
 378         }
 379         p.putback();
 380 }
 381
 382
 383 void handle_ert(ostream & os, string const & s, Context & context)
 384 {
 385         // We must have a valid layout before outputting the ERT inset.
 386         context.check_layout(os);
 387         Context newcontext(true, context.textclass);
 388         begin_inset(os, "ERT");
 389         os << "\nstatus collapsed\n";
 390         newcontext.check_layout(os);
 391         for (string::const_iterator it = s.begin(), et = s.end(); it != et; ++it) {
 392                 if (*it == '\\')
 393                         os << "\n\\backslash\n";
 394                 else if (*it == '\n') {
 395                         newcontext.new_paragraph(os);
 396                         newcontext.check_layout(os);
 397                 } else
 398                         os << *it;
 399         }
 400         newcontext.check_end_layout(os);
 401         end_inset(os);
 402 }
 403
 404
 405 void handle_comment(ostream & os, string const & s, Context & context)
 406 {
 407         // TODO: Handle this better
 408         Context newcontext(true, context.textclass);
 409         begin_inset(os, "ERT");
 410         os << "\nstatus collapsed\n";
 411         newcontext.check_layout(os);
 412         for (string::const_iterator it = s.begin(), et = s.end(); it != et; ++it) {
 413                 if (*it == '\\')
 414                         os << "\n\\backslash\n";
 415                 else
 416                         os << *it;
 417         }
 418         // make sure that our comment is the last thing on the line
 419         newcontext.new_paragraph(os);
 420         newcontext.check_layout(os);
 421         newcontext.check_end_layout(os);
 422         end_inset(os);
 423 }
 424
 425
 426 Layout const * findLayout(TextClass const & textclass, string const & name)
 427 {
 428         DocumentClass::const_iterator lit = textclass.begin();
 429         DocumentClass::const_iterator len = textclass.end();
 430         for (; lit != len; ++lit)
 431                 if (lit->latexname() == name)
 432                         return &*lit;
 433         return 0;
 434 }
 435
 436
 437 void eat_whitespace(Parser &, ostream &, Context &, bool);
 438
 439
 440 Layout * captionlayout()
 441 {
 442         static Layout * lay = 0;
 443         if (!lay) {
 444                 lay = new Layout;
 445                 lay->name_ = from_ascii("Caption");
 446                 lay->latexname_ = "caption";
 447                 lay->latextype = LATEX_COMMAND;
 448                 lay->optionalargs = 1;
 449         }
 450         return lay;
 451 }
 452
 453
 454 void output_command_layout(ostream & os, Parser & p, bool outer,
 455                            Context & parent_context,
 456                            Layout const * newlayout)
 457 {
 458         parent_context.check_end_layout(os);
 459         Context context(true, parent_context.textclass, newlayout,
 460                         parent_context.layout, parent_context.font);
 461         if (parent_context.deeper_paragraph) {
 462                 // We are beginning a nested environment after a
 463                 // deeper paragraph inside the outer list environment.
 464                 // Therefore we don't need to output a "begin deeper".
 465                 context.need_end_deeper = true;
 466         }
 467         context.check_deeper(os);
 468         context.check_layout(os);
 469         if (context.layout->optionalargs > 0) {
 470                 eat_whitespace(p, os, context, false);
 471                 if (p.next_token().character() == '[') {
 472                         p.get_token(); // eat '['
 473                         begin_inset(os, "OptArg\n");
 474                         os << "status collapsed\n\n";
 475                         parse_text_in_inset(p, os, FLAG_BRACK_LAST, outer, context);
 476                         end_inset(os);
 477                         eat_whitespace(p, os, context, false);
 478                 }
 479         }
 480         parse_text(p, os, FLAG_ITEM, outer, context);
 481         context.check_end_layout(os);
 482         if (parent_context.deeper_paragraph) {
 483                 // We must suppress the "end deeper" because we
 484                 // suppressed the "begin deeper" above.
 485                 context.need_end_deeper = false;
 486         }
 487         context.check_end_deeper(os);
 488         // We don't need really a new paragraph, but
 489         // we must make sure that the next item gets a \begin_layout.
 490         parent_context.new_paragraph(os);
 491 }
 492
 493
 494 /*!
 495  * Output a space if necessary.
 496  * This function gets called for every whitespace token.
 497  *
 498  * We have three cases here:
 499  * 1. A space must be suppressed. Example: The lyxcode case below
 500  * 2. A space may be suppressed. Example: Spaces before "\par"
 501  * 3. A space must not be suppressed. Example: A space between two words
 502  *
 503  * We currently handle only 1. and 3 and from 2. only the case of
 504  * spaces before newlines as a side effect.
 505  *
 506  * 2. could be used to suppress as many spaces as possible. This has two effects:
 507  * - Reimporting LyX generated LaTeX files changes almost no whitespace
 508  * - Superflous whitespace from non LyX generated LaTeX files is removed.
 509  * The drawback is that the logic inside the function becomes
 510  * complicated, and that is the reason why it is not implemented.
 511  */
 512 void check_space(Parser const & p, ostream & os, Context & context)
 513 {
 514         Token const next = p.next_token();
 515         Token const curr = p.curr_token();
 516         // A space before a single newline and vice versa must be ignored
 517         // LyX emits a newline before \end{lyxcode}.
 518         // This newline must be ignored,
 519         // otherwise LyX will add an additional protected space.
 520         if (next.cat() == catSpace ||
 521             next.cat() == catNewline ||
 522             (next.cs() == "end" && context.layout->free_spacing && curr.cat() == catNewline)) {
 523                 return;
 524         }
 525         context.check_layout(os);
 526         os << ' ';
 527 }
 528
 529
 530 /*!
 531  * Parse all arguments of \p command
 532  */
 533 void parse_arguments(string const & command,
 534                      vector<ArgumentType> const & template_arguments,
 535                      Parser & p, ostream & os, bool outer, Context & context)
 536 {
 537         string ert = command;
 538         size_t no_arguments = template_arguments.size();
 539         for (size_t i = 0; i < no_arguments; ++i) {
 540                 switch (template_arguments[i]) {
 541                 case required:
 542                         // This argument contains regular LaTeX
 543                         handle_ert(os, ert + '{', context);
 544                         eat_whitespace(p, os, context, false);
 545                         parse_text(p, os, FLAG_ITEM, outer, context);
 546                         ert = "}";
 547                         break;
 548                 case verbatim:
 549                         // This argument may contain special characters
 550                         ert += '{' + p.verbatim_item() + '}';
 551                         break;
 552                 case optional:
 553                         ert += p.getOpt();
 554                         break;
 555                 }
 556         }
 557         handle_ert(os, ert, context);
 558 }
 559
 560
 561 /*!
 562  * Check whether \p command is a known command. If yes,
 563  * handle the command with all arguments.
 564  * \return true if the command was parsed, false otherwise.
 565  */
 566 bool parse_command(string const & command, Parser & p, ostream & os,
 567                    bool outer, Context & context)
 568 {
 569         if (known_commands.find(command) != known_commands.end()) {
 570                 parse_arguments(command, known_commands[command], p, os,
 571                                 outer, context);
 572                 return true;
 573         }
 574         return false;
 575 }
 576
 577
 578 /// Parses a minipage or parbox
 579 void parse_box(Parser & p, ostream & os, unsigned flags, bool outer,
 580                Context & parent_context, bool use_parbox)
 581 {
 582         string position;
 583         string inner_pos;
 584         // We need to set the height to the LaTeX default of 1\\totalheight
 585         // for the case when no height argument is given
 586         string height_value = "1";
 587         string height_unit = "in";
 588         string height_special = "totalheight";
 589         string latex_height;
 590         if (p.next_token().asInput() == "[") {
 591                 position = p.getArg('[', ']');
 592                 if (position != "t" && position != "c" && position != "b") {
 593                         position = "c";
 594                         cerr << "invalid position for minipage/parbox" << endl;
 595                 }
 596                 if (p.next_token().asInput() == "[") {
 597                         latex_height = p.getArg('[', ']');
 598                         translate_box_len(latex_height, height_value, height_unit, height_special);
 599
 600                         if (p.next_token().asInput() == "[") {
 601                                 inner_pos = p.getArg('[', ']');
 602                                 if (inner_pos != "c" && inner_pos != "t" &&
 603                                     inner_pos != "b" && inner_pos != "s") {
 604                                         inner_pos = position;
 605                                         cerr << "invalid inner_pos for minipage/parbox"
 606                                              << endl;
 607                                 }
 608                         }
 609                 }
 610         }
 611         string width_value;
 612         string width_unit;
 613         string const latex_width = p.verbatim_item();
 614         translate_len(latex_width, width_value, width_unit);
 615         if (contains(width_unit, '\\') || contains(height_unit, '\\')) {
 616                 // LyX can't handle length variables
 617                 ostringstream ss;
 618                 if (use_parbox)
 619                         ss << "\\parbox";
 620                 else
 621                         ss << "\\begin{minipage}";
 622                 if (!position.empty())
 623                         ss << '[' << position << ']';
 624                 if (!latex_height.empty())
 625                         ss << '[' << latex_height << ']';
 626                 if (!inner_pos.empty())
 627                         ss << '[' << inner_pos << ']';
 628                 ss << "{" << latex_width << "}";
 629                 if (use_parbox)
 630                         ss << '{';
 631                 handle_ert(os, ss.str(), parent_context);
 632                 parent_context.new_paragraph(os);
 633                 parse_text_in_inset(p, os, flags, outer, parent_context);
 634                 if (use_parbox)
 635                         handle_ert(os, "}", parent_context);
 636                 else
 637                         handle_ert(os, "\\end{minipage}", parent_context);
 638         } else {
 639                 // LyX does not like empty positions, so we have
 640                 // to set them to the LaTeX default values here.
 641                 if (position.empty())
 642                         position = "c";
 643                 if (inner_pos.empty())
 644                         inner_pos = position;
 645                 parent_context.check_layout(os);
 646                 begin_inset(os, "Box Frameless\n");
 647                 os << "position \"" << position << "\"\n";
 648                 os << "hor_pos \"c\"\n";
 649                 os << "has_inner_box 1\n";
 650                 os << "inner_pos \"" << inner_pos << "\"\n";
 651                 os << "use_parbox " << use_parbox << "\n";
 652                 os << "width \"" << width_value << width_unit << "\"\n";
 653                 os << "special \"none\"\n";
 654                 os << "height \"" << height_value << height_unit << "\"\n";
 655                 os << "height_special \"" << height_special << "\"\n";
 656                 os << "status open\n\n";
 657                 parse_text_in_inset(p, os, flags, outer, parent_context);
 658                 end_inset(os);
 659 #ifdef PRESERVE_LAYOUT
 660                 // lyx puts a % after the end of the minipage
 661                 if (p.next_token().cat() == catNewline && p.next_token().cs().size() > 1) {
 662                         // new paragraph
 663                         //handle_comment(os, "%dummy", parent_context);
 664                         p.get_token();
 665                         p.skip_spaces();
 666                         parent_context.new_paragraph(os);
 667                 }
 668                 else if (p.next_token().cat() == catSpace || p.next_token().cat() == catNewline) {
 669                         //handle_comment(os, "%dummy", parent_context);
 670                         p.get_token();
 671                         p.skip_spaces();
 672                         // We add a protected space if something real follows
 673                         if (p.good() && p.next_token().cat() != catComment) {
 674                                 os << "\\InsetSpace ~\n";
 675                         }
 676                 }
 677 #endif
 678         }
 679 }
 680
 681
 682 /// parse an unknown environment
 683 void parse_unknown_environment(Parser & p, string const & name, ostream & os,
 684                                unsigned flags, bool outer,
 685                                Context & parent_context)
 686 {
 687         if (name == "tabbing")
 688                 // We need to remember that we have to handle '\=' specially
 689                 flags |= FLAG_TABBING;
 690
 691         // We need to translate font changes and paragraphs inside the
 692         // environment to ERT if we have a non standard font.
 693         // Otherwise things like
 694         // \large\begin{foo}\huge bar\end{foo}
 695         // will not work.
 696         bool const specialfont =
 697                 (parent_context.font != parent_context.normalfont);
 698         bool const new_layout_allowed = parent_context.new_layout_allowed;
 699         if (specialfont)
 700                 parent_context.new_layout_allowed = false;
 701         handle_ert(os, "\\begin{" + name + "}", parent_context);
 702         parse_text_snippet(p, os, flags, outer, parent_context);
 703         handle_ert(os, "\\end{" + name + "}", parent_context);
 704         if (specialfont)
 705                 parent_context.new_layout_allowed = new_layout_allowed;
 706 }
 707
 708
 709 void parse_environment(Parser & p, ostream & os, bool outer,
 710                        Context & parent_context)
 711 {
 712         Layout const * newlayout;
 713         string const name = p.getArg('{', '}');
 714         const bool is_starred = suffixIs(name, '*');
 715         string const unstarred_name = rtrim(name, "*");
 716         active_environments.push_back(name);
 717
 718         if (is_math_env(name)) {
 719                 parent_context.check_layout(os);
 720                 begin_inset(os, "Formula ");
 721                 os << "\\begin{" << name << "}";
 722                 parse_math(p, os, FLAG_END, MATH_MODE);
 723                 os << "\\end{" << name << "}";
 724                 end_inset(os);
 725         }
 726
 727         else if (name == "tabular" || name == "longtable") {
 728                 eat_whitespace(p, os, parent_context, false);
 729                 parent_context.check_layout(os);
 730                 begin_inset(os, "Tabular ");
 731                 handle_tabular(p, os, name == "longtable", parent_context);
 732                 end_inset(os);
 733                 p.skip_spaces();
 734         }
 735
 736         else if (parent_context.textclass.floats().typeExist(unstarred_name)) {
 737                 eat_whitespace(p, os, parent_context, false);
 738                 parent_context.check_layout(os);
 739                 begin_inset(os, "Float " + unstarred_name + "\n");
 740                 if (p.next_token().asInput() == "[") {
 741                         os << "placement " << p.getArg('[', ']') << '\n';
 742                 }
 743                 os << "wide " << convert<string>(is_starred)
 744                    << "\nsideways false"
 745                    << "\nstatus open\n\n";
 746                 parse_text_in_inset(p, os, FLAG_END, outer, parent_context);
 747                 end_inset(os);
 748                 // We don't need really a new paragraph, but
 749                 // we must make sure that the next item gets a \begin_layout.
 750                 parent_context.new_paragraph(os);
 751                 p.skip_spaces();
 752         }
 753
 754         else if (name == "minipage") {
 755                 eat_whitespace(p, os, parent_context, false);
 756                 parse_box(p, os, FLAG_END, outer, parent_context, false);
 757                 p.skip_spaces();
 758         }
 759
 760         else if (name == "comment") {
 761                 eat_whitespace(p, os, parent_context, false);
 762                 parent_context.check_layout(os);
 763                 begin_inset(os, "Note Comment\n");
 764                 os << "status open\n";
 765                 parse_text_in_inset(p, os, FLAG_END, outer, parent_context);
 766                 end_inset(os);
 767                 p.skip_spaces();
 768                 skip_braces(p); // eat {} that might by set by LyX behind comments
 769         }
 770
 771         else if (name == "lyxgreyedout") {
 772                 eat_whitespace(p, os, parent_context, false);
 773                 parent_context.check_layout(os);
 774                 begin_inset(os, "Note Greyedout\n");
 775                 os << "status open\n";
 776                 parse_text_in_inset(p, os, FLAG_END, outer, parent_context);
 777                 end_inset(os);
 778                 p.skip_spaces();
 779         }
 780
 781         else if (name == "framed") {
 782                 eat_whitespace(p, os, parent_context, false);
 783                 parent_context.check_layout(os);
 784                 begin_inset(os, "Note Framed\n");
 785                 os << "status open\n";
 786                 parse_text_in_inset(p, os, FLAG_END, outer, parent_context);
 787                 end_inset(os);
 788                 p.skip_spaces();
 789         }
 790
 791         else if (name == "shaded") {
 792                 eat_whitespace(p, os, parent_context, false);
 793                 parent_context.check_layout(os);
 794                 begin_inset(os, "Note Shaded\n");
 795                 os << "status open\n";
 796                 parse_text_in_inset(p, os, FLAG_END, outer, parent_context);
 797                 end_inset(os);
 798                 p.skip_spaces();
 799         }
 800
 801         else if (!parent_context.new_layout_allowed)
 802                 parse_unknown_environment(p, name, os, FLAG_END, outer,
 803                                           parent_context);
 804
 805         // Alignment and spacing settings
 806         // FIXME (bug xxxx): These settings can span multiple paragraphs and
 807         //                                       therefore are totally broken!
 808         // Note that \centering, raggedright, and raggedleft cannot be handled, as
 809         // they are commands not environments. They are furthermore switches that
 810         // can be ended by another switches, but also by commands like \footnote or
 811         // \parbox. So the only safe way is to leave them untouched.
 812         else if (name == "center" || name == "flushleft" || name == "flushright") {
 813                 eat_whitespace(p, os, parent_context, false);
 814                 // We must begin a new paragraph if not already done
 815                 if (! parent_context.atParagraphStart()) {
 816                         parent_context.check_end_layout(os);
 817                         parent_context.new_paragraph(os);
 818                 }
 819                 if (name == "flushleft")
 820                         parent_context.add_extra_stuff("\\align left\n");
 821                 else if (name == "flushright")
 822                         parent_context.add_extra_stuff("\\align right\n");
 823                 else if (name == "center")
 824                         parent_context.add_extra_stuff("\\align center\n");
 825                 parse_text(p, os, FLAG_END, outer, parent_context);
 826                 // Just in case the environment is empty
 827                 parent_context.extra_stuff.erase();
 828                 // We must begin a new paragraph to reset the alignment
 829                 parent_context.new_paragraph(os);
 830                 p.skip_spaces();
 831         }
 832
 833         // The single '=' is meant here.
 834         else if ((newlayout = findLayout(parent_context.textclass, name)) &&
 835                   newlayout->isEnvironment()) {
 836                 eat_whitespace(p, os, parent_context, false);
 837                 Context context(true, parent_context.textclass, newlayout,
 838                                 parent_context.layout, parent_context.font);
 839                 if (parent_context.deeper_paragraph) {
 840                         // We are beginning a nested environment after a
 841                         // deeper paragraph inside the outer list environment.
 842                         // Therefore we don't need to output a "begin deeper".
 843                         context.need_end_deeper = true;
 844                 }
 845                 parent_context.check_end_layout(os);
 846                 switch (context.layout->latextype) {
 847                 case  LATEX_LIST_ENVIRONMENT:
 848                         context.add_par_extra_stuff("\\labelwidthstring "
 849                                                     + p.verbatim_item() + '\n');
 850                         p.skip_spaces();
 851                         break;
 852                 case  LATEX_BIB_ENVIRONMENT:
 853                         p.verbatim_item(); // swallow next arg
 854                         p.skip_spaces();
 855                         break;
 856                 default:
 857                         break;
 858                 }
 859                 context.check_deeper(os);
 860                 parse_text(p, os, FLAG_END, outer, context);
 861                 context.check_end_layout(os);
 862                 if (parent_context.deeper_paragraph) {
 863                         // We must suppress the "end deeper" because we
 864                         // suppressed the "begin deeper" above.
 865                         context.need_end_deeper = false;
 866                 }
 867                 context.check_end_deeper(os);
 868                 parent_context.new_paragraph(os);
 869                 p.skip_spaces();
 870         }
 871
 872         else if (name == "appendix") {
 873                 // This is no good latex style, but it works and is used in some documents...
 874                 eat_whitespace(p, os, parent_context, false);
 875                 parent_context.check_end_layout(os);
 876                 Context context(true, parent_context.textclass, parent_context.layout,
 877                                 parent_context.layout, parent_context.font);
 878                 context.check_layout(os);
 879                 os << "\\start_of_appendix\n";
 880                 parse_text(p, os, FLAG_END, outer, context);
 881                 context.check_end_layout(os);
 882                 p.skip_spaces();
 883         }
 884
 885         else if (known_environments.find(name) != known_environments.end()) {
 886                 vector<ArgumentType> arguments = known_environments[name];
 887                 // The last "argument" denotes wether we may translate the
 888                 // environment contents to LyX
 889                 // The default required if no argument is given makes us
 890                 // compatible with the reLyXre environment.
 891                 ArgumentType contents = arguments.empty() ?
 892                         required :
 893                         arguments.back();
 894                 if (!arguments.empty())
 895                         arguments.pop_back();
 896                 // See comment in parse_unknown_environment()
 897                 bool const specialfont =
 898                         (parent_context.font != parent_context.normalfont);
 899                 bool const new_layout_allowed =
 900                         parent_context.new_layout_allowed;
 901                 if (specialfont)
 902                         parent_context.new_layout_allowed = false;
 903                 parse_arguments("\\begin{" + name + "}", arguments, p, os,
 904                                 outer, parent_context);
 905                 if (contents == verbatim)
 906                         handle_ert(os, p.verbatimEnvironment(name),
 907                                    parent_context);
 908                 else
 909                         parse_text_snippet(p, os, FLAG_END, outer,
 910                                            parent_context);
 911                 handle_ert(os, "\\end{" + name + "}", parent_context);
 912                 if (specialfont)
 913                         parent_context.new_layout_allowed = new_layout_allowed;
 914         }
 915
 916         else
 917                 parse_unknown_environment(p, name, os, FLAG_END, outer,
 918                                           parent_context);
 919
 920         active_environments.pop_back();
 921 }
 922
 923
 924 /// parses a comment and outputs it to \p os.
 925 void parse_comment(Parser & p, ostream & os, Token const & t, Context & context)
 926 {
 927         LASSERT(t.cat() == catComment, return);
 928         if (!t.cs().empty()) {
 929                 context.check_layout(os);
 930                 handle_comment(os, '%' + t.cs(), context);
 931                 if (p.next_token().cat() == catNewline) {
 932                         // A newline after a comment line starts a new
 933                         // paragraph
 934                         if (context.new_layout_allowed) {
 935                                 if(!context.atParagraphStart())
 936                                         // Only start a new paragraph if not already
 937                                         // done (we might get called recursively)
 938                                         context.new_paragraph(os);
 939                         } else
 940                                 handle_ert(os, "\n", context);
 941                         eat_whitespace(p, os, context, true);
 942                 }
 943         } else {
 944                 // "%\n" combination
 945                 p.skip_spaces();
 946         }
 947 }
 948
 949
 950 /*!
 951  * Reads spaces and comments until the first non-space, non-comment token.
 952  * New paragraphs (double newlines or \\par) are handled like simple spaces
 953  * if \p eatParagraph is true.
 954  * Spaces are skipped, but comments are written to \p os.
 955  */
 956 void eat_whitespace(Parser & p, ostream & os, Context & context,
 957                     bool eatParagraph)
 958 {
 959         while (p.good()) {
 960                 Token const & t = p.get_token();
 961                 if (t.cat() == catComment)
 962                         parse_comment(p, os, t, context);
 963                 else if ((! eatParagraph && p.isParagraph()) ||
 964                          (t.cat() != catSpace && t.cat() != catNewline)) {
 965                         p.putback();
 966                         return;
 967                 }
 968         }
 969 }
 970
 971
 972 /*!
 973  * Set a font attribute, parse text and reset the font attribute.
 974  * \param attribute Attribute name (e.g. \\family, \\shape etc.)
 975  * \param currentvalue Current value of the attribute. Is set to the new
 976  * value during parsing.
 977  * \param newvalue New value of the attribute
 978  */
 979 void parse_text_attributes(Parser & p, ostream & os, unsigned flags, bool outer,
 980                            Context & context, string const & attribute,
 981                            string & currentvalue, string const & newvalue)
 982 {
 983         context.check_layout(os);
 984         string const oldvalue = currentvalue;
 985         currentvalue = newvalue;
 986         os << '\n' << attribute << ' ' << newvalue << "\n";
 987         parse_text_snippet(p, os, flags, outer, context);
 988         context.check_layout(os);
 989         os << '\n' << attribute << ' ' << oldvalue << "\n";
 990         currentvalue = oldvalue;
 991 }
 992
 993
 994 /// get the arguments of a natbib or jurabib citation command
 995 void get_cite_arguments(Parser & p, bool natbibOrder,
 996         string & before, string & after)
 997 {
 998         // We need to distinguish "" and "[]", so we can't use p.getOpt().
 999
1000         // text before the citation
1001         before.clear();
1002         // text after the citation
1003         after = p.getFullOpt();
1004
1005         if (!after.empty()) {
1006                 before = p.getFullOpt();
1007                 if (natbibOrder && !before.empty())
1008                         swap(before, after);
1009         }
1010 }
1011
1012
1013 /// Convert filenames with TeX macros and/or quotes to something LyX
1014 /// can understand
1015 string const normalize_filename(string const & name)
1016 {
1017         Parser p(trim(name, "\""));
1018         ostringstream os;
1019         while (p.good()) {
1020                 Token const & t = p.get_token();
1021                 if (t.cat() != catEscape)
1022                         os << t.asInput();
1023                 else if (t.cs() == "lyxdot") {
1024                         // This is used by LyX for simple dots in relative
1025                         // names
1026                         os << '.';
1027                         p.skip_spaces();
1028                 } else if (t.cs() == "space") {
1029                         os << ' ';
1030                         p.skip_spaces();
1031                 } else
1032                         os << t.asInput();
1033         }
1034         return os.str();
1035 }
1036
1037
1038 /// Convert \p name from TeX convention (relative to master file) to LyX
1039 /// convention (relative to .lyx file) if it is relative
1040 void fix_relative_filename(string & name)
1041 {
1042         FileName fname(name);
1043         if (fname.isAbsolute())
1044                 return;
1045
1046         // FIXME UNICODE encoding of name may be wrong (makeAbsPath expects
1047         // utf8)
1048         name = to_utf8(makeRelPath(from_utf8(makeAbsPath(name, getMasterFilePath()).absFilename()),
1049                                    from_utf8(getParentFilePath())));
1050 }
1051
1052
1053 /// Parse a NoWeb Scrap section. The initial "<<" is already parsed.
1054 void parse_noweb(Parser & p, ostream & os, Context & context)
1055 {
1056         // assemble the rest of the keyword
1057         string name("<<");
1058         bool scrap = false;
1059         while (p.good()) {
1060                 Token const & t = p.get_token();
1061                 if (t.asInput() == ">" && p.next_token().asInput() == ">") {
1062                         name += ">>";
1063                         p.get_token();
1064                         scrap = (p.good() && p.next_token().asInput() == "=");
1065                         if (scrap)
1066                                 name += p.get_token().asInput();
1067                         break;
1068                 }
1069                 name += t.asInput();
1070         }
1071
1072         if (!scrap || !context.new_layout_allowed ||
1073             !context.textclass.hasLayout(from_ascii("Scrap"))) {
1074                 cerr << "Warning: Could not interpret '" << name
1075                      << "'. Ignoring it." << endl;
1076                 return;
1077         }
1078
1079         // We use new_paragraph instead of check_end_layout because the stuff
1080         // following the noweb chunk needs to start with a \begin_layout.
1081         // This may create a new paragraph even if there was none in the
1082         // noweb file, but the alternative is an invalid LyX file. Since
1083         // noweb code chunks are implemented with a layout style in LyX they
1084         // always must be in an own paragraph.
1085         context.new_paragraph(os);
1086         Context newcontext(true, context.textclass,
1087                 &context.textclass[from_ascii("Scrap")]);
1088         newcontext.check_layout(os);
1089         os << name;
1090         while (p.good()) {
1091                 Token const & t = p.get_token();
1092                 // We abuse the parser a bit, because this is no TeX syntax
1093                 // at all.
1094                 if (t.cat() == catEscape)
1095                         os << subst(t.asInput(), "\\", "\n\\backslash\n");
1096                 else
1097                         os << subst(t.asInput(), "\n", "\n\\newline\n");
1098                 // The scrap chunk is ended by an @ at the beginning of a line.
1099                 // After the @ the line may contain a comment and/or
1100                 // whitespace, but nothing else.
1101                 if (t.asInput() == "@" && p.prev_token().cat() == catNewline &&
1102                     (p.next_token().cat() == catSpace ||
1103                      p.next_token().cat() == catNewline ||
1104                      p.next_token().cat() == catComment)) {
1105                         while (p.good() && p.next_token().cat() == catSpace)
1106                                 os << p.get_token().asInput();
1107                         if (p.next_token().cat() == catComment)
1108                                 // The comment includes a final '\n'
1109                                 os << p.get_token().asInput();
1110                         else {
1111                                 if (p.next_token().cat() == catNewline)
1112                                         p.get_token();
1113                                 os << '\n';
1114                         }
1115                         break;
1116                 }
1117         }
1118         newcontext.check_end_layout(os);
1119 }
1120
1121 } // anonymous namespace
1122
1123
1124 void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
1125                 Context & context)
1126 {
1127         Layout const * newlayout = 0;
1128         // store the current selectlanguage to be used after \foreignlanguage
1129         string selectlang;
1130         // Store the latest bibliographystyle (needed for bibtex inset)
1131         string bibliographystyle;
1132         bool const use_natbib = used_packages.find("natbib") != used_packages.end();
1133         bool const use_jurabib = used_packages.find("jurabib") != used_packages.end();
1134         while (p.good()) {
1135                 Token const & t = p.get_token();
1136
1137 #ifdef FILEDEBUG
1138                 cerr << "t: " << t << " flags: " << flags << "\n";
1139 #endif
1140
1141                 if (flags & FLAG_ITEM) {
1142                         if (t.cat() == catSpace)
1143                                 continue;
1144
1145                         flags &= ~FLAG_ITEM;
1146                         if (t.cat() == catBegin) {
1147                                 // skip the brace and collect everything to the next matching
1148                                 // closing brace
1149                                 flags |= FLAG_BRACE_LAST;
1150                                 continue;
1151                         }
1152
1153                         // handle only this single token, leave the loop if done
1154                         flags |= FLAG_LEAVE;
1155                 }
1156
1157                 if (t.character() == ']' && (flags & FLAG_BRACK_LAST))
1158                         return;
1159
1160                 //
1161                 // cat codes
1162                 //
1163                 if (t.cat() == catMath) {
1164                         // we are inside some text mode thingy, so opening new math is allowed
1165                         context.check_layout(os);
1166                         begin_inset(os, "Formula ");
1167                         Token const & n = p.get_token();
1168                         if (n.cat() == catMath && outer) {
1169                                 // TeX's $$...$$ syntax for displayed math
1170                                 os << "\\[";
1171                                 parse_math(p, os, FLAG_SIMPLE, MATH_MODE);
1172                                 os << "\\]";
1173                                 p.get_token(); // skip the second '$' token
1174                         } else {
1175                                 // simple $...$  stuff
1176                                 p.putback();
1177                                 os << '$';
1178                                 parse_math(p, os, FLAG_SIMPLE, MATH_MODE);
1179                                 os << '$';
1180                         }
1181                         end_inset(os);
1182                 }
1183
1184                 else if (t.cat() == catSuper || t.cat() == catSub)
1185                         cerr << "catcode " << t << " illegal in text mode\n";
1186
1187                 // Basic support for english quotes. This should be
1188                 // extended to other quotes, but is not so easy (a
1189                 // left english quote is the same as a right german
1190                 // quote...)
1191                 else if (t.asInput() == "`" && p.next_token().asInput() == "`") {
1192                         context.check_layout(os);
1193                         begin_inset(os, "Quotes ");
1194                         os << "eld";
1195                         end_inset(os);
1196                         p.get_token();
1197                         skip_braces(p);
1198                 }
1199                 else if (t.asInput() == "'" && p.next_token().asInput() == "'") {
1200                         context.check_layout(os);
1201                         begin_inset(os, "Quotes ");
1202                         os << "erd";
1203                         end_inset(os);
1204                         p.get_token();
1205                         skip_braces(p);
1206                 }
1207
1208                 else if (t.asInput() == ">" && p.next_token().asInput() == ">") {
1209                         context.check_layout(os);
1210                         begin_inset(os, "Quotes ");
1211                         os << "ald";
1212                         end_inset(os);
1213                         p.get_token();
1214                         skip_braces(p);
1215                 }
1216
1217                 else if (t.asInput() == "<" && p.next_token().asInput() == "<") {
1218                         context.check_layout(os);
1219                         begin_inset(os, "Quotes ");
1220                         os << "ard";
1221                         end_inset(os);
1222                         p.get_token();
1223                         skip_braces(p);
1224                 }
1225
1226                 else if (t.asInput() == "<"
1227                          && p.next_token().asInput() == "<" && noweb_mode) {
1228                         p.get_token();
1229                         parse_noweb(p, os, context);
1230                 }
1231
1232                 else if (t.cat() == catSpace || (t.cat() == catNewline && ! p.isParagraph()))
1233                         check_space(p, os, context);
1234
1235                 else if (t.character() == '[' && noweb_mode &&
1236                          p.next_token().character() == '[') {
1237                         // These can contain underscores
1238                         p.putback();
1239                         string const s = p.getFullOpt() + ']';
1240                         if (p.next_token().character() == ']')
1241                                 p.get_token();
1242                         else
1243                                 cerr << "Warning: Inserting missing ']' in '"
1244                                      << s << "'." << endl;
1245                         handle_ert(os, s, context);
1246                 }
1247
1248                 else if (t.cat() == catLetter ||
1249                                t.cat() == catOther ||
1250                                t.cat() == catAlign ||
1251                                t.cat() == catParameter) {
1252                         // This translates "&" to "\\&" which may be wrong...
1253                         context.check_layout(os);
1254                         os << t.character();
1255                 }
1256
1257                 else if (p.isParagraph()) {
1258                         if (context.new_layout_allowed)
1259                                 context.new_paragraph(os);
1260                         else
1261                                 handle_ert(os, "\\par ", context);
1262                         eat_whitespace(p, os, context, true);
1263                 }
1264
1265                 else if (t.cat() == catActive) {
1266                         context.check_layout(os);
1267                         if (t.character() == '~') {
1268                                 if (context.layout->free_spacing)
1269                                         os << ' ';
1270                                 else
1271                                         os << "\\InsetSpace ~\n";
1272                         } else
1273                                 os << t.character();
1274                 }
1275
1276                 else if (t.cat() == catBegin &&
1277                          p.next_token().cat() == catEnd) {
1278                         // {}
1279                         Token const prev = p.prev_token();
1280                         p.get_token();
1281                         if (p.next_token().character() == '`' ||
1282                             (prev.character() == '-' &&
1283                              p.next_token().character() == '-'))
1284                                 ; // ignore it in {}`` or -{}-
1285                         else
1286                                 handle_ert(os, "{}", context);
1287
1288                 }
1289
1290                 else if (t.cat() == catBegin) {
1291                         context.check_layout(os);
1292                         // special handling of font attribute changes
1293                         Token const prev = p.prev_token();
1294                         Token const next = p.next_token();
1295                         TeXFont const oldFont = context.font;
1296                         if (next.character() == '[' ||
1297                             next.character() == ']' ||
1298                             next.character() == '*') {
1299                                 p.get_token();
1300                                 if (p.next_token().cat() == catEnd) {
1301                                         os << next.character();
1302                                         p.get_token();
1303                                 } else {
1304                                         p.putback();
1305                                         handle_ert(os, "{", context);
1306                                         parse_text_snippet(p, os,
1307                                                         FLAG_BRACE_LAST,
1308                                                         outer, context);
1309                                         handle_ert(os, "}", context);
1310                                 }
1311                         } else if (! context.new_layout_allowed) {
1312                                 handle_ert(os, "{", context);
1313                                 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1314                                                    outer, context);
1315                                 handle_ert(os, "}", context);
1316                         } else if (is_known(next.cs(), known_sizes)) {
1317                                 // next will change the size, so we must
1318                                 // reset it here
1319                                 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1320                                                    outer, context);
1321                                 if (!context.atParagraphStart())
1322                                         os << "\n\\size "
1323                                            << context.font.size << "\n";
1324                         } else if (is_known(next.cs(), known_font_families)) {
1325                                 // next will change the font family, so we
1326                                 // must reset it here
1327                                 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1328                                                    outer, context);
1329                                 if (!context.atParagraphStart())
1330                                         os << "\n\\family "
1331                                            << context.font.family << "\n";
1332                         } else if (is_known(next.cs(), known_font_series)) {
1333                                 // next will change the font series, so we
1334                                 // must reset it here
1335                                 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1336                                                    outer, context);
1337                                 if (!context.atParagraphStart())
1338                                         os << "\n\\series "
1339                                            << context.font.series << "\n";
1340                         } else if (is_known(next.cs(), known_font_shapes)) {
1341                                 // next will change the font shape, so we
1342                                 // must reset it here
1343                                 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1344                                                    outer, context);
1345                                 if (!context.atParagraphStart())
1346                                         os << "\n\\shape "
1347                                            << context.font.shape << "\n";
1348                         } else if (is_known(next.cs(), known_old_font_families) ||
1349                                    is_known(next.cs(), known_old_font_series) ||
1350                                    is_known(next.cs(), known_old_font_shapes)) {
1351                                 // next will change the font family, series
1352                                 // and shape, so we must reset it here
1353                                 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1354                                                    outer, context);
1355                                 if (!context.atParagraphStart())
1356                                         os <<  "\n\\family "
1357                                            << context.font.family
1358                                            << "\n\\series "
1359                                            << context.font.series
1360                                            << "\n\\shape "
1361                                            << context.font.shape << "\n";
1362                         } else {
1363                                 handle_ert(os, "{", context);
1364                                 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1365                                                    outer, context);
1366                                 handle_ert(os, "}", context);
1367                         }
1368                 }
1369
1370                 else if (t.cat() == catEnd) {
1371                         if (flags & FLAG_BRACE_LAST) {
1372                                 return;
1373                         }
1374                         cerr << "stray '}' in text\n";
1375                         handle_ert(os, "}", context);
1376                 }
1377
1378                 else if (t.cat() == catComment)
1379                         parse_comment(p, os, t, context);
1380
1381                 //
1382                 // control sequences
1383                 //
1384
1385                 else if (t.cs() == "(") {
1386                         context.check_layout(os);
1387                         begin_inset(os, "Formula");
1388                         os << " \\(";
1389                         parse_math(p, os, FLAG_SIMPLE2, MATH_MODE);
1390                         os << "\\)";
1391                         end_inset(os);
1392                 }
1393
1394                 else if (t.cs() == "[") {
1395                         context.check_layout(os);
1396                         begin_inset(os, "Formula");
1397                         os << " \\[";
1398                         parse_math(p, os, FLAG_EQUATION, MATH_MODE);
1399                         os << "\\]";
1400                         end_inset(os);
1401                 }
1402
1403                 else if (t.cs() == "begin")
1404                         parse_environment(p, os, outer, context);
1405
1406                 else if (t.cs() == "end") {
1407                         if (flags & FLAG_END) {
1408                                 // eat environment name
1409                                 string const name = p.getArg('{', '}');
1410                                 if (name != active_environment())
1411                                         cerr << "\\end{" + name + "} does not match \\begin{"
1412                                                 + active_environment() + "}\n";
1413                                 return;
1414                         }
1415                         p.error("found 'end' unexpectedly");
1416                 }
1417
1418                 else if (t.cs() == "item") {
1419                         p.skip_spaces();
1420                         string s;
1421                         bool optarg = false;
1422                         if (p.next_token().character() == '[') {
1423                                 p.get_token(); // eat '['
1424                                 s = parse_text_snippet(p, FLAG_BRACK_LAST,
1425                                                        outer, context);
1426                                 optarg = true;
1427                         }
1428                         context.set_item();
1429                         context.check_layout(os);
1430                         if (context.has_item) {
1431                                 // An item in an unknown list-like environment
1432                                 // FIXME: Do this in check_layout()!
1433                                 context.has_item = false;
1434                                 if (optarg)
1435                                         handle_ert(os, "\\item", context);
1436                                 else
1437                                         handle_ert(os, "\\item ", context);
1438                         }
1439                         if (optarg) {
1440                                 if (context.layout->labeltype != LABEL_MANUAL) {
1441                                         // lyx does not support \item[\mybullet]
1442                                         // in itemize environments
1443                                         handle_ert(os, "[", context);
1444                                         os << s;
1445                                         handle_ert(os, "]", context);
1446                                 } else if (!s.empty()) {
1447                                         // The space is needed to separate the
1448                                         // item from the rest of the sentence.
1449                                         os << s << ' ';
1450                                         eat_whitespace(p, os, context, false);
1451                                 }
1452                         }
1453                 }
1454
1455                 else if (t.cs() == "bibitem") {
1456                         context.set_item();
1457                         context.check_layout(os);
1458                         os << "\\bibitem ";
1459                         os << p.getOpt();
1460                         os << '{' << p.verbatim_item() << '}' << "\n";
1461                 }
1462
1463                 else if (t.cs() == "def") {
1464                         context.check_layout(os);
1465                         eat_whitespace(p, os, context, false);
1466                         string name = p.get_token().cs();
1467                         eat_whitespace(p, os, context, false);
1468
1469                         // parameter text
1470                         bool simple = true;
1471                         string paramtext;
1472                         int arity = 0;
1473                         while (p.next_token().cat() != catBegin) {
1474                                 if (p.next_token().cat() == catParameter) {
1475                                         // # found
1476                                         p.get_token();
1477                                         paramtext += "#";
1478
1479                                         // followed by number?
1480                                         if (p.next_token().cat() == catOther) {
1481                                                 char c = p.getChar();
1482                                                 paramtext += c;
1483                                                 // number = current arity + 1?
1484                                                 if (c == arity + '0' + 1)
1485                                                         ++arity;
1486                                                 else
1487                                                         simple = false;
1488                                         } else
1489                                                 paramtext += p.get_token().asString();
1490                                 } else {
1491                                         paramtext += p.get_token().asString();
1492                                         simple = false;
1493                                 }
1494                         }
1495
1496                         // only output simple (i.e. compatible) macro as FormulaMacros
1497                         string ert = "\\def\\" + name + ' ' + paramtext + '{' + p.verbatim_item() + '}';
1498                         if (simple) {
1499                                 context.check_layout(os);
1500                                 begin_inset(os, "FormulaMacro");
1501                                 os << "\n" << ert;
1502                                 end_inset(os);
1503                         } else
1504                                 handle_ert(os, ert, context);
1505                 }
1506
1507                 else if (t.cs() == "noindent") {
1508                         p.skip_spaces();
1509                         context.add_par_extra_stuff("\\noindent\n");
1510                 }
1511
1512                 else if (t.cs() == "appendix") {
1513                         context.add_par_extra_stuff("\\start_of_appendix\n");
1514                         // We need to start a new paragraph. Otherwise the
1515                         // appendix in 'bla\appendix\chapter{' would start
1516                         // too late.
1517                         context.new_paragraph(os);
1518                         // We need to make sure that the paragraph is
1519                         // generated even if it is empty. Otherwise the
1520                         // appendix in '\par\appendix\par\chapter{' would
1521                         // start too late.
1522                         context.check_layout(os);
1523                         // FIXME: This is a hack to prevent paragraph
1524                         // deletion if it is empty. Handle this better!
1525                         handle_comment(os,
1526                                 "%dummy comment inserted by tex2lyx to "
1527                                 "ensure that this paragraph is not empty",
1528                                 context);
1529                         // Both measures above may generate an additional
1530                         // empty paragraph, but that does not hurt, because
1531                         // whitespace does not matter here.
1532                         eat_whitespace(p, os, context, true);
1533                 }
1534
1535                 // Starred section headings
1536                 // Must attempt to parse "Section*" before "Section".
1537                 else if ((p.next_token().asInput() == "*") &&
1538                          context.new_layout_allowed &&
1539                          (newlayout = findLayout(context.textclass, t.cs() + '*')) &&
1540                          newlayout->isCommand()) {
1541                         TeXFont const oldFont = context.font;
1542                         // save the current font size
1543                         string const size = oldFont.size;
1544                         // reset the font size to default, because the font size switches don't
1545                         // affect section headings and the like
1546                         context.font.size = known_coded_sizes[0];
1547                         output_font_change(os, oldFont, context.font);
1548                         // write the layout
1549                         p.get_token();
1550                         output_command_layout(os, p, outer, context, newlayout);
1551                         // set the font size to the original value
1552                         context.font.size = size;
1553                         output_font_change(os, oldFont, context.font);
1554                         p.skip_spaces();
1555                 }
1556
1557                 // Section headings and the like
1558                 else if (context.new_layout_allowed &&
1559                          (newlayout = findLayout(context.textclass, t.cs())) &&
1560                          newlayout->isCommand()) {
1561                         TeXFont const oldFont = context.font;
1562                         // save the current font size
1563                         string const size = oldFont.size;
1564                         // reset the font size to default, because the font size switches don't
1565                         // affect section headings and the like
1566                         context.font.size = known_coded_sizes[0];
1567                         output_font_change(os, oldFont, context.font);
1568                         // write the layout
1569                         output_command_layout(os, p, outer, context, newlayout);
1570                         // set the font size to the original value
1571                         context.font.size = size;
1572                         output_font_change(os, oldFont, context.font);
1573                         p.skip_spaces();
1574                 }
1575
1576                 // Special handling for \caption
1577                 // FIXME: remove this when InsetCaption is supported.
1578                 else if (context.new_layout_allowed &&
1579                          t.cs() == captionlayout()->latexname()) {
1580                         output_command_layout(os, p, outer, context,
1581                                               captionlayout());
1582                         p.skip_spaces();
1583                 }
1584
1585                 else if (t.cs() == "includegraphics") {
1586                         bool const clip = p.next_token().asInput() == "*";
1587                         if (clip)
1588                                 p.get_token();
1589                         map<string, string> opts = split_map(p.getArg('[', ']'));
1590                         if (clip)
1591                                 opts["clip"] = string();
1592                         string name = normalize_filename(p.verbatim_item());
1593
1594                         string const path = getMasterFilePath();
1595                         // We want to preserve relative / absolute filenames,
1596                         // therefore path is only used for testing
1597                         // FIXME UNICODE encoding of name and path may be
1598                         // wrong (makeAbsPath expects utf8)
1599                         if (!makeAbsPath(name, path).exists()) {
1600                                 // The file extension is probably missing.
1601                                 // Now try to find it out.
1602                                 string const dvips_name =
1603                                         find_file(name, path,
1604                                                   known_dvips_graphics_formats);
1605                                 string const pdftex_name =
1606                                         find_file(name, path,
1607                                                   known_pdftex_graphics_formats);
1608                                 if (!dvips_name.empty()) {
1609                                         if (!pdftex_name.empty()) {
1610                                                 cerr << "This file contains the "
1611                                                         "latex snippet\n"
1612                                                         "\"\\includegraphics{"
1613                                                      << name << "}\".\n"
1614                                                         "However, files\n\""
1615                                                      << dvips_name << "\" and\n\""
1616                                                      << pdftex_name << "\"\n"
1617                                                         "both exist, so I had to make a "
1618                                                         "choice and took the first one.\n"
1619                                                         "Please move the unwanted one "
1620                                                         "someplace else and try again\n"
1621                                                         "if my choice was wrong."
1622                                                      << endl;
1623                                         }
1624                                         name = dvips_name;
1625                                 } else if (!pdftex_name.empty())
1626                                         name = pdftex_name;
1627                         }
1628
1629                         // FIXME UNICODE encoding of name and path may be
1630                         // wrong (makeAbsPath expects utf8)
1631                         if (makeAbsPath(name, path).exists())
1632                                 fix_relative_filename(name);
1633                         else
1634                                 cerr << "Warning: Could not find graphics file '"
1635                                      << name << "'." << endl;
1636
1637                         context.check_layout(os);
1638                         begin_inset(os, "Graphics ");
1639                         os << "\n\tfilename " << name << '\n';
1640                         if (opts.find("width") != opts.end())
1641                                 os << "\twidth "
1642                                    << translate_len(opts["width"]) << '\n';
1643                         if (opts.find("height") != opts.end())
1644                                 os << "\theight "
1645                                    << translate_len(opts["height"]) << '\n';
1646                         if (opts.find("scale") != opts.end()) {
1647                                 istringstream iss(opts["scale"]);
1648                                 double val;
1649                                 iss >> val;
1650                                 val = val*100;
1651                                 os << "\tscale " << val << '\n';
1652                         }
1653                         if (opts.find("angle") != opts.end())
1654                                 os << "\trotateAngle "
1655                                    << opts["angle"] << '\n';
1656                         if (opts.find("origin") != opts.end()) {
1657                                 ostringstream ss;
1658                                 string const opt = opts["origin"];
1659                                 if (opt.find('l') != string::npos) ss << "left";
1660                                 if (opt.find('r') != string::npos) ss << "right";
1661                                 if (opt.find('c') != string::npos) ss << "center";
1662                                 if (opt.find('t') != string::npos) ss << "Top";
1663                                 if (opt.find('b') != string::npos) ss << "Bottom";
1664                                 if (opt.find('B') != string::npos) ss << "Baseline";
1665                                 if (!ss.str().empty())
1666                                         os << "\trotateOrigin " << ss.str() << '\n';
1667                                 else
1668                                         cerr << "Warning: Ignoring unknown includegraphics origin argument '" << opt << "'\n";
1669                         }
1670                         if (opts.find("keepaspectratio") != opts.end())
1671                                 os << "\tkeepAspectRatio\n";
1672                         if (opts.find("clip") != opts.end())
1673                                 os << "\tclip\n";
1674                         if (opts.find("draft") != opts.end())
1675                                 os << "\tdraft\n";
1676                         if (opts.find("bb") != opts.end())
1677                                 os << "\tBoundingBox "
1678                                    << opts["bb"] << '\n';
1679                         int numberOfbbOptions = 0;
1680                         if (opts.find("bbllx") != opts.end())
1681                                 numberOfbbOptions++;
1682                         if (opts.find("bblly") != opts.end())
1683                                 numberOfbbOptions++;
1684                         if (opts.find("bburx") != opts.end())
1685                                 numberOfbbOptions++;
1686                         if (opts.find("bbury") != opts.end())
1687                                 numberOfbbOptions++;
1688                         if (numberOfbbOptions == 4)
1689                                 os << "\tBoundingBox "
1690                                    << opts["bbllx"] << " " << opts["bblly"] << " "
1691                                    << opts["bburx"] << " " << opts["bbury"] << '\n';
1692                         else if (numberOfbbOptions > 0)
1693                                 cerr << "Warning: Ignoring incomplete includegraphics boundingbox arguments.\n";
1694                         numberOfbbOptions = 0;
1695                         if (opts.find("natwidth") != opts.end())
1696                                 numberOfbbOptions++;
1697                         if (opts.find("natheight") != opts.end())
1698                                 numberOfbbOptions++;
1699                         if (numberOfbbOptions == 2)
1700                                 os << "\tBoundingBox 0bp 0bp "
1701                                    << opts["natwidth"] << " " << opts["natheight"] << '\n';
1702                         else if (numberOfbbOptions > 0)
1703                                 cerr << "Warning: Ignoring incomplete includegraphics boundingbox arguments.\n";
1704                         ostringstream special;
1705                         if (opts.find("hiresbb") != opts.end())
1706                                 special << "hiresbb,";
1707                         if (opts.find("trim") != opts.end())
1708                                 special << "trim,";
1709                         if (opts.find("viewport") != opts.end())
1710                                 special << "viewport=" << opts["viewport"] << ',';
1711                         if (opts.find("totalheight") != opts.end())
1712                                 special << "totalheight=" << opts["totalheight"] << ',';
1713                         if (opts.find("type") != opts.end())
1714                                 special << "type=" << opts["type"] << ',';
1715                         if (opts.find("ext") != opts.end())
1716                                 special << "ext=" << opts["ext"] << ',';
1717                         if (opts.find("read") != opts.end())
1718                                 special << "read=" << opts["read"] << ',';
1719                         if (opts.find("command") != opts.end())
1720                                 special << "command=" << opts["command"] << ',';
1721                         string s_special = special.str();
1722                         if (!s_special.empty()) {
1723                                 // We had special arguments. Remove the trailing ','.
1724                                 os << "\tspecial " << s_special.substr(0, s_special.size() - 1) << '\n';
1725                         }
1726                         // TODO: Handle the unknown settings better.
1727                         // Warn about invalid options.
1728                         // Check whether some option was given twice.
1729                         end_inset(os);
1730                 }
1731
1732                 else if (t.cs() == "footnote" ||
1733                          (t.cs() == "thanks" && context.layout->intitle)) {
1734                         p.skip_spaces();
1735                         context.check_layout(os);
1736                         begin_inset(os, "Foot\n");
1737                         os << "status collapsed\n\n";
1738                         parse_text_in_inset(p, os, FLAG_ITEM, false, context);
1739                         end_inset(os);
1740                 }
1741
1742                 else if (t.cs() == "marginpar") {
1743                         p.skip_spaces();
1744                         context.check_layout(os);
1745                         begin_inset(os, "Marginal\n");
1746                         os << "status collapsed\n\n";
1747                         parse_text_in_inset(p, os, FLAG_ITEM, false, context);
1748                         end_inset(os);
1749                 }
1750
1751                 else if (t.cs() == "ensuremath") {
1752                         p.skip_spaces();
1753                         context.check_layout(os);
1754                         string const s = p.verbatim_item();
1755                         if (s == "\xb1" || s == "\xb3" || s == "\xb2" || s == "\xb5")
1756                                 os << s;
1757                         else
1758                                 handle_ert(os, "\\ensuremath{" + s + "}",
1759                                            context);
1760                 }
1761
1762                 else if (t.cs() == "hfill") {
1763                         context.check_layout(os);
1764                         os << "\n\\hfill\n";
1765                         skip_braces(p);
1766                         p.skip_spaces();
1767                 }
1768
1769                 else if (t.cs() == "makeindex" || t.cs() == "maketitle") {
1770                         // FIXME: Somehow prevent title layouts if
1771                         // "maketitle" was not found
1772                         p.skip_spaces();
1773                         skip_braces(p); // swallow this
1774                 }
1775
1776                 else if (t.cs() == "tableofcontents") {
1777                         p.skip_spaces();
1778                         context.check_layout(os);
1779                         begin_inset(os, "LatexCommand \\tableofcontents\n");
1780                         end_inset(os);
1781                         skip_braces(p); // swallow this
1782                 }
1783
1784                 else if (t.cs() == "listoffigures") {
1785                         p.skip_spaces();
1786                         context.check_layout(os);
1787                         begin_inset(os, "FloatList figure\n");
1788                         end_inset(os);
1789                         skip_braces(p); // swallow this
1790                 }
1791
1792                 else if (t.cs() == "listoftables") {
1793                         p.skip_spaces();
1794                         context.check_layout(os);
1795                         begin_inset(os, "FloatList table\n");
1796                         end_inset(os);
1797                         skip_braces(p); // swallow this
1798                 }
1799
1800                 else if (t.cs() == "listof") {
1801                         p.skip_spaces(true);
1802                         string const name = p.get_token().asString();
1803                         if (context.textclass.floats().typeExist(name)) {
1804                                 context.check_layout(os);
1805                                 begin_inset(os, "FloatList ");
1806                                 os << name << "\n";
1807                                 end_inset(os);
1808                                 p.get_token(); // swallow second arg
1809                         } else
1810                                 handle_ert(os, "\\listof{" + name + "}", context);
1811                 }
1812
1813                 else if (t.cs() == "textrm")
1814                         parse_text_attributes(p, os, FLAG_ITEM, outer,
1815                                               context, "\\family",
1816                                               context.font.family, "roman");
1817
1818                 else if (t.cs() == "textsf")
1819                         parse_text_attributes(p, os, FLAG_ITEM, outer,
1820                                               context, "\\family",
1821                                               context.font.family, "sans");
1822
1823                 else if (t.cs() == "texttt")
1824                         parse_text_attributes(p, os, FLAG_ITEM, outer,
1825                                               context, "\\family",
1826                                               context.font.family, "typewriter");
1827
1828                 else if (t.cs() == "textmd")
1829                         parse_text_attributes(p, os, FLAG_ITEM, outer,
1830                                               context, "\\series",
1831                                               context.font.series, "medium");
1832
1833                 else if (t.cs() == "textbf")
1834                         parse_text_attributes(p, os, FLAG_ITEM, outer,
1835                                               context, "\\series",
1836                                               context.font.series, "bold");
1837
1838                 else if (t.cs() == "textup")
1839                         parse_text_attributes(p, os, FLAG_ITEM, outer,
1840                                               context, "\\shape",
1841                                               context.font.shape, "up");
1842
1843                 else if (t.cs() == "textit")
1844                         parse_text_attributes(p, os, FLAG_ITEM, outer,
1845                                               context, "\\shape",
1846                                               context.font.shape, "italic");
1847
1848                 else if (t.cs() == "textsl")
1849                         parse_text_attributes(p, os, FLAG_ITEM, outer,
1850                                               context, "\\shape",
1851                                               context.font.shape, "slanted");
1852
1853                 else if (t.cs() == "textsc")
1854                         parse_text_attributes(p, os, FLAG_ITEM, outer,
1855                                               context, "\\shape",
1856                                               context.font.shape, "smallcaps");
1857
1858                 else if (t.cs() == "textnormal" || t.cs() == "normalfont") {
1859                         context.check_layout(os);
1860                         TeXFont oldFont = context.font;
1861                         context.font.init();
1862                         context.font.size = oldFont.size;
1863                         os << "\n\\family " << context.font.family << "\n";
1864                         os << "\n\\series " << context.font.series << "\n";
1865                         os << "\n\\shape " << context.font.shape << "\n";
1866                         if (t.cs() == "textnormal") {
1867                                 parse_text_snippet(p, os, FLAG_ITEM, outer, context);
1868                                 output_font_change(os, context.font, oldFont);
1869                                 context.font = oldFont;
1870                         } else
1871                                 eat_whitespace(p, os, context, false);
1872                 }
1873
1874                 else if (t.cs() == "textcolor") {
1875                         // scheme is \textcolor{color name}{text}
1876                         string const color = p.verbatim_item();
1877                         // we only support the predefined colors of the color package
1878                         if (color == "black" || color == "blue" || color == "cyan"
1879                                 || color == "green" || color == "magenta" || color == "red"
1880                                 || color == "white" || color == "yellow") {
1881                                         context.check_layout(os);
1882                                         os << "\n\\color " << color << "\n";
1883                                         parse_text_snippet(p, os, FLAG_ITEM, outer, context);
1884                                         context.check_layout(os);
1885                                         os << "\n\\color inherit\n";
1886                         } else
1887                                 // for custom defined colors
1888                                 handle_ert(os, t.asInput() + "{" + color + "}", context);
1889                 }
1890
1891                 else if (t.cs() == "underbar") {
1892                         // Do NOT handle \underline.
1893                         // \underbar cuts through y, g, q, p etc.,
1894                         // \underline does not.
1895                         context.check_layout(os);
1896                         os << "\n\\bar under\n";
1897                         parse_text_snippet(p, os, FLAG_ITEM, outer, context);
1898                         context.check_layout(os);
1899                         os << "\n\\bar default\n";
1900                 }
1901
1902                 else if (t.cs() == "emph" || t.cs() == "noun") {
1903                         context.check_layout(os);
1904                         os << "\n\\" << t.cs() << " on\n";
1905                         parse_text_snippet(p, os, FLAG_ITEM, outer, context);
1906                         context.check_layout(os);
1907                         os << "\n\\" << t.cs() << " default\n";
1908                 }
1909
1910                 else if (t.cs() == "lyxline") {
1911                         context.check_layout(os);
1912                         os << "\\lyxline";
1913                 }
1914
1915                 else if (use_natbib &&
1916                          is_known(t.cs(), known_natbib_commands) &&
1917                          ((t.cs() != "citefullauthor" &&
1918                            t.cs() != "citeyear" &&
1919                            t.cs() != "citeyearpar") ||
1920                           p.next_token().asInput() != "*")) {
1921                         context.check_layout(os);
1922                         // tex                       lyx
1923                         // \citet[before][after]{a}  \citet[after][before]{a}
1924                         // \citet[before][]{a}       \citet[][before]{a}
1925                         // \citet[after]{a}          \citet[after]{a}
1926                         // \citet{a}                 \citet{a}
1927                         string command = '\\' + t.cs();
1928                         if (p.next_token().asInput() == "*") {
1929                                 command += '*';
1930                                 p.get_token();
1931                         }
1932                         if (command == "\\citefullauthor")
1933                                 // alternative name for "\\citeauthor*"
1934                                 command = "\\citeauthor*";
1935
1936                         // text before the citation
1937                         string before;
1938                         // text after the citation
1939                         string after;
1940                         get_cite_arguments(p, true, before, after);
1941
1942                         if (command == "\\cite") {
1943                                 // \cite without optional argument means
1944                                 // \citet, \cite with at least one optional
1945                                 // argument means \citep.
1946                                 if (before.empty() && after.empty())
1947                                         command = "\\citet";
1948                                 else
1949                                         command = "\\citep";
1950                         }
1951                         if (before.empty() && after == "[]")
1952                                 // avoid \citet[]{a}
1953                                 after.erase();
1954                         else if (before == "[]" && after == "[]") {
1955                                 // avoid \citet[][]{a}
1956                                 before.erase();
1957                                 after.erase();
1958                         }
1959                         begin_inset(os, "LatexCommand ");
1960                         os << command << after << before
1961                            << '{' << p.verbatim_item() << "}\n";
1962                         end_inset(os);
1963                 }
1964
1965                 else if (use_jurabib &&
1966                          is_known(t.cs(), known_jurabib_commands)) {
1967                         context.check_layout(os);
1968                         string const command = '\\' + t.cs();
1969                         char argumentOrder = '\0';
1970                         vector<string> const & options = used_packages["jurabib"];
1971                         if (find(options.begin(), options.end(),
1972                                       "natbiborder") != options.end())
1973                                 argumentOrder = 'n';
1974                         else if (find(options.begin(), options.end(),
1975                                            "jurabiborder") != options.end())
1976                                 argumentOrder = 'j';
1977
1978                         // text before the citation
1979                         string before;
1980                         // text after the citation
1981                         string after;
1982                         get_cite_arguments(p, argumentOrder != 'j', before, after);
1983
1984                         string const citation = p.verbatim_item();
1985                         if (!before.empty() && argumentOrder == '\0') {
1986                                 cerr << "Warning: Assuming argument order "
1987                                         "of jurabib version 0.6 for\n'"
1988                                      << command << before << after << '{'
1989                                      << citation << "}'.\n"
1990                                         "Add 'jurabiborder' to the jurabib "
1991                                         "package options if you used an\n"
1992                                         "earlier jurabib version." << endl;
1993                         }
1994                         begin_inset(os, "LatexCommand ");
1995                         os << command << after << before
1996                            << '{' << citation << "}\n";
1997                         end_inset(os);
1998                 }
1999
2000                 else if (is_known(t.cs(), known_latex_commands)) {
2001                         // This needs to be after the check for natbib and
2002                         // jurabib commands, because "cite" has different
2003                         // arguments with natbib and jurabib.
2004                         context.check_layout(os);
2005                         begin_inset(os, "LatexCommand ");
2006                         os << '\\' << t.cs();
2007                         // lyx cannot handle newlines in a latex command
2008                         // FIXME: Move the substitution into parser::getOpt()?
2009                         os << subst(p.getOpt(), "\n", " ");
2010                         os << subst(p.getOpt(), "\n", " ");
2011                         os << '{' << subst(p.verbatim_item(), "\n", " ") << "}\n";
2012                         end_inset(os);
2013                 }
2014
2015                 else if (is_known(t.cs(), known_quotes)) {
2016                         char const * const * where = is_known(t.cs(), known_quotes);
2017                         context.check_layout(os);
2018                         begin_inset(os, "Quotes ");
2019                         os << known_coded_quotes[where - known_quotes];
2020                         end_inset(os);
2021                         // LyX adds {} after the quote, so we have to eat
2022                         // spaces here if there are any before a possible
2023                         // {} pair.
2024                         eat_whitespace(p, os, context, false);
2025                         skip_braces(p);
2026                 }
2027
2028                 else if (is_known(t.cs(), known_sizes) &&
2029                          context.new_layout_allowed) {
2030                         char const * const * where = is_known(t.cs(), known_sizes);
2031                         context.check_layout(os);
2032                         TeXFont const oldFont = context.font;
2033                         // the font size index differs by 1, because the known_coded_sizes
2034                         // has additionally a "default" entry
2035                         context.font.size = known_coded_sizes[where - known_sizes + 1];
2036                         output_font_change(os, oldFont, context.font);
2037                         eat_whitespace(p, os, context, false);
2038                 }
2039
2040                 else if (is_known(t.cs(), known_font_families) &&
2041                          context.new_layout_allowed) {
2042                         char const * const * where =
2043                                 is_known(t.cs(), known_font_families);
2044                         context.check_layout(os);
2045                         TeXFont const oldFont = context.font;
2046                         context.font.family =
2047                                 known_coded_font_families[where - known_font_families];
2048                         output_font_change(os, oldFont, context.font);
2049                         eat_whitespace(p, os, context, false);
2050                 }
2051
2052                 else if (is_known(t.cs(), known_font_series) &&
2053                          context.new_layout_allowed) {
2054                         char const * const * where =
2055                                 is_known(t.cs(), known_font_series);
2056                         context.check_layout(os);
2057                         TeXFont const oldFont = context.font;
2058                         context.font.series =
2059                                 known_coded_font_series[where - known_font_series];
2060                         output_font_change(os, oldFont, context.font);
2061                         eat_whitespace(p, os, context, false);
2062                 }
2063
2064                 else if (is_known(t.cs(), known_font_shapes) &&
2065                          context.new_layout_allowed) {
2066                         char const * const * where =
2067                                 is_known(t.cs(), known_font_shapes);
2068                         context.check_layout(os);
2069                         TeXFont const oldFont = context.font;
2070                         context.font.shape =
2071                                 known_coded_font_shapes[where - known_font_shapes];
2072                         output_font_change(os, oldFont, context.font);
2073                         eat_whitespace(p, os, context, false);
2074                 }
2075                 else if (is_known(t.cs(), known_old_font_families) &&
2076                          context.new_layout_allowed) {
2077                         char const * const * where =
2078                                 is_known(t.cs(), known_old_font_families);
2079                         context.check_layout(os);
2080                         TeXFont const oldFont = context.font;
2081                         context.font.init();
2082                         context.font.size = oldFont.size;
2083                         context.font.family =
2084                                 known_coded_font_families[where - known_old_font_families];
2085                         output_font_change(os, oldFont, context.font);
2086                         eat_whitespace(p, os, context, false);
2087                 }
2088
2089                 else if (is_known(t.cs(), known_old_font_series) &&
2090                          context.new_layout_allowed) {
2091                         char const * const * where =
2092                                 is_known(t.cs(), known_old_font_series);
2093                         context.check_layout(os);
2094                         TeXFont const oldFont = context.font;
2095                         context.font.init();
2096                         context.font.size = oldFont.size;
2097                         context.font.series =
2098                                 known_coded_font_series[where - known_old_font_series];
2099                         output_font_change(os, oldFont, context.font);
2100                         eat_whitespace(p, os, context, false);
2101                 }
2102
2103                 else if (is_known(t.cs(), known_old_font_shapes) &&
2104                          context.new_layout_allowed) {
2105                         char const * const * where =
2106                                 is_known(t.cs(), known_old_font_shapes);
2107                         context.check_layout(os);
2108                         TeXFont const oldFont = context.font;
2109                         context.font.init();
2110                         context.font.size = oldFont.size;
2111                         context.font.shape =
2112                                 known_coded_font_shapes[where - known_old_font_shapes];
2113                         output_font_change(os, oldFont, context.font);
2114                         eat_whitespace(p, os, context, false);
2115                 }
2116
2117                 else if (t.cs() == "selectlanguage") {
2118                         context.check_layout(os);
2119                         // save the language for the case that a \foreignlanguage is used
2120                         selectlang = subst(p.verbatim_item(), "\n", " ");
2121                         os << "\\lang " << selectlang << "\n";
2122
2123                 }
2124
2125                 else if (t.cs() == "foreignlanguage") {
2126                         context.check_layout(os);
2127                         os << "\n\\lang " << subst(p.verbatim_item(), "\n", " ") << "\n";
2128                         os << subst(p.verbatim_item(), "\n", " ");
2129                         // set back to last selectlanguage
2130                         os << "\n\\lang " << selectlang << "\n";
2131                 }
2132
2133                 else if (t.cs() == "inputencoding")
2134                         // write nothing because this is done by LyX using the "\lang"
2135                         // information given by selectlanguage and foreignlanguage
2136                         subst(p.verbatim_item(), "\n", " ");
2137
2138                 else if (t.cs() == "LyX" || t.cs() == "TeX"
2139                          || t.cs() == "LaTeX") {
2140                         context.check_layout(os);
2141                         os << t.cs();
2142                         skip_braces(p); // eat {}
2143                 }
2144
2145                 else if (t.cs() == "LaTeXe") {
2146                         context.check_layout(os);
2147                         os << "LaTeX2e";
2148                         skip_braces(p); // eat {}
2149                 }
2150
2151                 else if (t.cs() == "ldots") {
2152                         context.check_layout(os);
2153                         skip_braces(p);
2154                         os << "\\SpecialChar \\ldots{}\n";
2155                 }
2156
2157                 else if (t.cs() == "lyxarrow") {
2158                         context.check_layout(os);
2159                         os << "\\SpecialChar \\menuseparator\n";
2160                         skip_braces(p);
2161                 }
2162
2163                 else if (t.cs() == "textcompwordmark") {
2164                         context.check_layout(os);
2165                         os << "\\SpecialChar \\textcompwordmark{}\n";
2166                         skip_braces(p);
2167                 }
2168
2169                 else if (t.cs() == "@" && p.next_token().asInput() == ".") {
2170                         context.check_layout(os);
2171                         os << "\\SpecialChar \\@.\n";
2172                         p.get_token();
2173                 }
2174
2175                 else if (t.cs() == "-") {
2176                         context.check_layout(os);
2177                         os << "\\SpecialChar \\-\n";
2178                 }
2179
2180                 else if (t.cs() == "textasciitilde") {
2181                         context.check_layout(os);
2182                         os << '~';
2183                         skip_braces(p);
2184                 }
2185
2186                 else if (t.cs() == "textasciicircum") {
2187                         context.check_layout(os);
2188                         os << '^';
2189                         skip_braces(p);
2190                 }
2191
2192                 else if (t.cs() == "textbackslash") {
2193                         context.check_layout(os);
2194                         os << "\n\\backslash\n";
2195                         skip_braces(p);
2196                 }
2197
2198                 else if (t.cs() == "_" || t.cs() == "&" || t.cs() == "#"
2199                             || t.cs() == "$" || t.cs() == "{" || t.cs() == "}"
2200                             || t.cs() == "%") {
2201                         context.check_layout(os);
2202                         os << t.cs();
2203                 }
2204
2205                 else if (t.cs() == "char") {
2206                         context.check_layout(os);
2207                         if (p.next_token().character() == '`') {
2208                                 p.get_token();
2209                                 if (p.next_token().cs() == "\"") {
2210                                         p.get_token();
2211                                         os << '"';
2212                                         skip_braces(p);
2213                                 } else {
2214                                         handle_ert(os, "\\char`", context);
2215                                 }
2216                         } else {
2217                                 handle_ert(os, "\\char", context);
2218                         }
2219                 }
2220
2221                 else if (t.cs() == "verb") {
2222                         context.check_layout(os);
2223                         char const delimiter = p.next_token().character();
2224                         string const arg = p.getArg(delimiter, delimiter);
2225                         ostringstream oss;
2226                         oss << "\\verb" << delimiter << arg << delimiter;
2227                         handle_ert(os, oss.str(), context);
2228                 }
2229
2230                 else if (t.cs() == "\"") {
2231                         context.check_layout(os);
2232                         string const name = p.verbatim_item();
2233                              if (name == "a") os << '\xe4';
2234                         else if (name == "o") os << '\xf6';
2235                         else if (name == "u") os << '\xfc';
2236                         else if (name == "A") os << '\xc4';
2237                         else if (name == "O") os << '\xd6';
2238                         else if (name == "U") os << '\xdc';
2239                         else handle_ert(os, "\"{" + name + "}", context);
2240                 }
2241
2242                 // Problem: \= creates a tabstop inside the tabbing environment
2243                 // and else an accent. In the latter case we really would want
2244                 // \={o} instead of \= o.
2245                 else if (t.cs() == "=" && (flags & FLAG_TABBING))
2246                         handle_ert(os, t.asInput(), context);
2247
2248                 else if (t.cs() == "H" || t.cs() == "c" || t.cs() == "^"
2249                          || t.cs() == "'" || t.cs() == "`"
2250                          || t.cs() == "~" || t.cs() == "." || t.cs() == "=") {
2251                         // we need the trim as the LyX parser chokes on such spaces
2252                         // The argument of InsetLatexAccent is parsed as a
2253                         // subset of LaTeX, so don't parse anything here,
2254                         // but use the raw argument.
2255                         // Otherwise we would convert \~{\i} wrongly.
2256                         // This will of course not translate \~{\ss} to \~{ß},
2257                         // but that does at least compile and does only look
2258                         // strange on screen.
2259                         context.check_layout(os);
2260                         os << "\\i \\" << t.cs() << "{"
2261                            << trim(p.verbatim_item(), " ")
2262                            << "}\n";
2263                 }
2264
2265                 else if (t.cs() == "ss") {
2266                         context.check_layout(os);
2267                         os << "\xdf";
2268                         skip_braces(p); // eat {}
2269                 }
2270
2271                 else if (t.cs() == "i" || t.cs() == "j" || t.cs() == "l" ||
2272                          t.cs() == "L") {
2273                         context.check_layout(os);
2274                         os << "\\i \\" << t.cs() << "{}\n";
2275                         skip_braces(p); // eat {}
2276                 }
2277
2278                 else if (t.cs() == "\\") {
2279                         context.check_layout(os);
2280                         string const next = p.next_token().asInput();
2281                         if (next == "[")
2282                                 handle_ert(os, "\\\\" + p.getOpt(), context);
2283                         else if (next == "*") {
2284                                 p.get_token();
2285                                 handle_ert(os, "\\\\*" + p.getOpt(), context);
2286                         }
2287                         else {
2288                                 os << "\n\\newline\n";
2289                         }
2290                 }
2291
2292                 else if (t.cs() == "newline" ||
2293                         t.cs() == "linebreak") {
2294                         context.check_layout(os);
2295                         os << "\n\\" << t.cs() << "\n";
2296                         skip_braces(p); // eat {}
2297                 }
2298
2299                 else if (t.cs() == "input" || t.cs() == "include"
2300                          || t.cs() == "verbatiminput") {
2301                         string name = '\\' + t.cs();
2302                         if (t.cs() == "verbatiminput"
2303                             && p.next_token().asInput() == "*")
2304                                 name += p.get_token().asInput();
2305                         context.check_layout(os);
2306                         begin_inset(os, "Include ");
2307                         string filename(normalize_filename(p.getArg('{', '}')));
2308                         string const path = getMasterFilePath();
2309                         // We want to preserve relative / absolute filenames,
2310                         // therefore path is only used for testing
2311                         // FIXME UNICODE encoding of filename and path may be
2312                         // wrong (makeAbsPath expects utf8)
2313                         if ((t.cs() == "include" || t.cs() == "input") &&
2314                             !makeAbsPath(filename, path).exists()) {
2315                                 // The file extension is probably missing.
2316                                 // Now try to find it out.
2317                                 string const tex_name =
2318                                         find_file(filename, path,
2319                                                   known_tex_extensions);
2320                                 if (!tex_name.empty())
2321                                         filename = tex_name;
2322                         }
2323                         // FIXME UNICODE encoding of filename and path may be
2324                         // wrong (makeAbsPath expects utf8)
2325                         if (makeAbsPath(filename, path).exists()) {
2326                                 string const abstexname =
2327                                         makeAbsPath(filename, path).absFilename();
2328                                 string const abslyxname =
2329                                         changeExtension(abstexname, ".lyx");
2330                                 fix_relative_filename(filename);
2331                                 string const lyxname =
2332                                         changeExtension(filename, ".lyx");
2333                                 if (t.cs() != "verbatiminput" &&
2334                                     tex2lyx(abstexname, FileName(abslyxname))) {
2335                                         os << name << '{' << lyxname << "}\n";
2336                                 } else {
2337                                         os << name << '{' << filename << "}\n";
2338                                 }
2339                         } else {
2340                                 cerr << "Warning: Could not find included file '"
2341                                      << filename << "'." << endl;
2342                                 os << name << '{' << filename << "}\n";
2343                         }
2344                         os << "preview false\n";
2345                         end_inset(os);
2346                 }
2347
2348                 else if (t.cs() == "bibliographystyle") {
2349                         // store new bibliographystyle
2350                         bibliographystyle = p.verbatim_item();
2351                         // output new bibliographystyle.
2352                         // This is only necessary if used in some other macro than \bibliography.
2353                         handle_ert(os, "\\bibliographystyle{" + bibliographystyle + "}", context);
2354                 }
2355
2356                 else if (t.cs() == "bibliography") {
2357                         context.check_layout(os);
2358                         begin_inset(os, "LatexCommand ");
2359                         os << "\\bibtex";
2360                         // Do we have a bibliographystyle set?
2361                         if (!bibliographystyle.empty()) {
2362                                 os << '[' << bibliographystyle << ']';
2363                         }
2364                         os << '{' << p.verbatim_item() << "}\n";
2365                         end_inset(os);
2366                 }
2367
2368                 else if (t.cs() == "parbox")
2369                         parse_box(p, os, FLAG_ITEM, outer, context, true);
2370
2371                 //\makebox() is part of the picture environment and different from \makebox{}
2372                 //\makebox{} will be parsed by parse_box when bug 2956 is fixed
2373                 else if (t.cs() == "makebox") {
2374                         string arg = t.asInput();
2375                         if (p.next_token().character() == '(')
2376                                 //the syntax is: \makebox(x,y)[position]{content}
2377                                 arg += p.getFullParentheseArg();
2378                         else
2379                                 //the syntax is: \makebox[width][position]{content}
2380                                 arg += p.getFullOpt();
2381                         handle_ert(os, arg + p.getFullOpt(), context);
2382                 }
2383
2384                 else if (t.cs() == "smallskip" ||
2385                          t.cs() == "medskip" ||
2386                          t.cs() == "bigskip" ||
2387                          t.cs() == "vfill") {
2388                         context.check_layout(os);
2389                         begin_inset(os, "VSpace ");
2390                         os << t.cs();
2391                         end_inset(os);
2392                         skip_braces(p);
2393                 }
2394
2395                 else if (is_known(t.cs(), known_spaces)) {
2396                         char const * const * where = is_known(t.cs(), known_spaces);
2397                         context.check_layout(os);
2398                         os << "\\InsetSpace ";
2399                         os << '\\' << known_coded_spaces[where - known_spaces]
2400                            << '\n';
2401                         // LaTeX swallows whitespace after all spaces except
2402                         // "\\,". We have to do that here, too, because LyX
2403                         // adds "{}" which would make the spaces significant.
2404                         if (t.cs() !=  ",")
2405                                 eat_whitespace(p, os, context, false);
2406                         // LyX adds "{}" after all spaces except "\\ " and
2407                         // "\\,", so we have to remove "{}".
2408                         // "\\,{}" is equivalent to "\\," in LaTeX, so we
2409                         // remove the braces after "\\,", too.
2410                         if (t.cs() != " ")
2411                                 skip_braces(p);
2412                 }
2413
2414                 else if (t.cs() == "newpage" ||
2415                         t.cs() == "pagebreak" ||
2416                         t.cs() == "clearpage" ||
2417                         t.cs() == "cleardoublepage") {
2418                         context.check_layout(os);
2419                         os << "\n\\" << t.cs() << "\n";
2420                         skip_braces(p); // eat {}
2421                 }
2422
2423                 else if (t.cs() == "newcommand" ||
2424                          t.cs() == "providecommand" ||
2425                          t.cs() == "renewcommand") {
2426                         // these could be handled by parse_command(), but
2427                         // we need to call add_known_command() here.
2428                         string name = t.asInput();
2429                         if (p.next_token().asInput() == "*") {
2430                                 // Starred form. Eat '*'
2431                                 p.get_token();
2432                                 name += '*';
2433                         }
2434                         string const command = p.verbatim_item();
2435                         string const opt1 = p.getOpt();
2436                         string const opt2 = p.getFullOpt();
2437                         add_known_command(command, opt1, !opt2.empty());
2438                         string const ert = name + '{' + command + '}' +
2439                                            opt1 + opt2 +
2440                                            '{' + p.verbatim_item() + '}';
2441
2442                         context.check_layout(os);
2443                         begin_inset(os, "FormulaMacro");
2444                         os << "\n" << ert;
2445                         end_inset(os);
2446                 }
2447
2448                 else if (t.cs() == "vspace") {
2449                         bool starred = false;
2450                         if (p.next_token().asInput() == "*") {
2451                                 p.get_token();
2452                                 starred = true;
2453                         }
2454                         string const length = p.verbatim_item();
2455                         string unit;
2456                         string valstring;
2457                         bool valid = splitLatexLength(length, valstring, unit);
2458                         bool known_vspace = false;
2459                         bool known_unit = false;
2460                         double value;
2461                         if (valid) {
2462                                 istringstream iss(valstring);
2463                                 iss >> value;
2464                                 if (value == 1.0) {
2465                                         if (unit == "\\smallskipamount") {
2466                                                 unit = "smallskip";
2467                                                 known_vspace = true;
2468                                         } else if (unit == "\\medskipamount") {
2469                                                 unit = "medskip";
2470                                                 known_vspace = true;
2471                                         } else if (unit == "\\bigskipamount") {
2472                                                 unit = "bigskip";
2473                                                 known_vspace = true;
2474                                         } else if (unit == "\\fill") {
2475                                                 unit = "vfill";
2476                                                 known_vspace = true;
2477                                         }
2478                                 }
2479                                 if (!known_vspace) {
2480                                         switch (unitFromString(unit)) {
2481                                         case Length::SP:
2482                                         case Length::PT:
2483                                         case Length::BP:
2484                                         case Length::DD:
2485                                         case Length::MM:
2486                                         case Length::PC:
2487                                         case Length::CC:
2488                                         case Length::CM:
2489                                         case Length::IN:
2490                                         case Length::EX:
2491                                         case Length::EM:
2492                                         case Length::MU:
2493                                                 known_unit = true;
2494                                                 break;
2495                                         default:
2496                                                 break;
2497                                         }
2498                                 }
2499                         }
2500
2501                         if (known_unit || known_vspace) {
2502                                 // Literal length or known variable
2503                                 context.check_layout(os);
2504                                 begin_inset(os, "VSpace ");
2505                                 if (known_unit)
2506                                         os << value;
2507                                 os << unit;
2508                                 if (starred)
2509                                         os << '*';
2510                                 end_inset(os);
2511                         } else {
2512                                 // LyX can't handle other length variables in Inset VSpace
2513                                 string name = t.asInput();
2514                                 if (starred)
2515                                         name += '*';
2516                                 if (valid) {
2517                                         if (value == 1.0)
2518                                                 handle_ert(os, name + '{' + unit + '}', context);
2519                                         else if (value == -1.0)
2520                                                 handle_ert(os, name + "{-" + unit + '}', context);
2521                                         else
2522                                                 handle_ert(os, name + '{' + valstring + unit + '}', context);
2523                                 } else
2524                                         handle_ert(os, name + '{' + length + '}', context);
2525                         }
2526                 }
2527
2528                 else {
2529                         //cerr << "#: " << t << " mode: " << mode << endl;
2530                         // heuristic: read up to next non-nested space
2531                         /*
2532                         string s = t.asInput();
2533                         string z = p.verbatim_item();
2534                         while (p.good() && z != " " && z.size()) {
2535                                 //cerr << "read: " << z << endl;
2536                                 s += z;
2537                                 z = p.verbatim_item();
2538                         }
2539                         cerr << "found ERT: " << s << endl;
2540                         handle_ert(os, s + ' ', context);
2541                         */
2542                         string name = t.asInput();
2543                         if (p.next_token().asInput() == "*") {
2544                                 // Starred commands like \vspace*{}
2545                                 p.get_token();                          // Eat '*'
2546                                 name += '*';
2547                         }
2548                         if (! parse_command(name, p, os, outer, context))
2549                                 handle_ert(os, name, context);
2550                 }
2551
2552                 if (flags & FLAG_LEAVE) {
2553                         flags &= ~FLAG_LEAVE;
2554                         break;
2555                 }
2556         }
2557 }
2558
2559 // }])
2560
2561
2562 } // namespace lyx