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