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  *
   9  * Full author contact details are available in file CREDITS.
  10  */
  11
  12 // {[(
  13
  14 #include <config.h>
  15
  16 #include "tex2lyx.h"
  17
  18 #include "Context.h"
  19 #include "FloatList.h"
  20 #include "Layout.h"
  21 #include "Length.h"
  22
  23 #include "support/lstrings.h"
  24 #include "support/convert.h"
  25 #include "support/filetools.h"
  26
  27 #include <boost/tuple/tuple.hpp>
  28
  29 #include <iostream>
  30 #include <map>
  31 #include <sstream>
  32 #include <vector>
  33
  34 using std::cerr;
  35 using std::endl;
  36 using std::find;
  37 using std::map;
  38 using std::ostream;
  39 using std::ostringstream;
  40 using std::istringstream;
  41 using std::string;
  42 using std::vector;
  43
  44 namespace lyx {
  45
  46 using support::addExtension;
  47 using support::changeExtension;
  48 using support::FileName;
  49 using support::makeAbsPath;
  50 using support::makeRelPath;
  51 using support::rtrim;
  52 using support::suffixIs;
  53 using support::contains;
  54 using support::subst;
  55
  56
  57 void parse_text_in_inset(Parser & p, ostream & os, unsigned flags, bool outer,
  58                 Context const & context)
  59 {
  60         Context newcontext(true, context.textclass);
  61         newcontext.font = context.font;
  62         parse_text(p, os, flags, outer, newcontext);
  63         newcontext.check_end_layout(os);
  64 }
  65
  66
  67 namespace {
  68
  69 /// parses a paragraph snippet, useful for example for \\emph{...}
  70 void parse_text_snippet(Parser & p, ostream & os, unsigned flags, bool outer,
  71                 Context & context)
  72 {
  73         Context newcontext(context);
  74         // Don't inherit the extra stuff
  75         newcontext.extra_stuff.clear();
  76         parse_text(p, os, flags, outer, newcontext);
  77         // Make sure that we don't create invalid .lyx files
  78         context.need_layout = newcontext.need_layout;
  79         context.need_end_layout = newcontext.need_end_layout;
  80 }
  81
  82
  83 /*!
  84  * Thin wrapper around parse_text_snippet() using a string.
  85  *
  86  * We completely ignore \c context.need_layout and \c context.need_end_layout,
  87  * because our return value is not used directly (otherwise the stream version
  88  * of parse_text_snippet() could be used). That means that the caller needs
  89  * to do layout management manually.
  90  * This is intended to parse text that does not create any layout changes.
  91  */
  92 string parse_text_snippet(Parser & p, unsigned flags, const bool outer,
  93                   Context & context)
  94 {
  95         Context newcontext(context);
  96         newcontext.need_layout = false;
  97         newcontext.need_end_layout = false;
  98         newcontext.new_layout_allowed = false;
  99         // Avoid warning by Context::~Context()
 100         newcontext.extra_stuff.clear();
 101         ostringstream os;
 102         parse_text_snippet(p, os, flags, outer, newcontext);
 103         return os.str();
 104 }
 105
 106
 107 char const * const known_latex_commands[] = { "ref", "cite", "label",
 108  "index", "printindex", "pageref", "url", "vref", "vpageref", "prettyref",
 109  "eqref", 0 };
 110
 111 /*!
 112  * natbib commands.
 113  * We can't put these into known_latex_commands because the argument order
 114  * is reversed in lyx if there are 2 arguments.
 115  * The starred forms are also known.
 116  */
 117 char const * const known_natbib_commands[] = { "cite", "citet", "citep",
 118 "citealt", "citealp", "citeauthor", "citeyear", "citeyearpar",
 119 "citefullauthor", "Citet", "Citep", "Citealt", "Citealp", "Citeauthor", 0 };
 120
 121 /*!
 122  * jurabib commands.
 123  * We can't put these into known_latex_commands because the argument order
 124  * is reversed in lyx if there are 2 arguments.
 125  * No starred form other than "cite*" known.
 126  */
 127 char const * const known_jurabib_commands[] = { "cite", "citet", "citep",
 128 "citealt", "citealp", "citeauthor", "citeyear", "citeyearpar",
 129 // jurabib commands not (yet) supported by LyX:
 130 // "fullcite",
 131 // "footcite", "footcitet", "footcitep", "footcitealt", "footcitealp",
 132 // "footciteauthor", "footciteyear", "footciteyearpar",
 133 "citefield", "citetitle", "cite*", 0 };
 134
 135 /// LaTeX names for quotes
 136 char const * const known_quotes[] = { "glqq", "grqq", "quotedblbase",
 137 "textquotedblleft", "quotesinglbase", "guilsinglleft", "guilsinglright", 0};
 138
 139 /// the same as known_quotes with .lyx names
 140 char const * const known_coded_quotes[] = { "gld", "grd", "gld",
 141 "grd", "gls", "fls", "frs", 0};
 142
 143 /// LaTeX names for font sizes
 144 char const * const known_sizes[] = { "tiny", "scriptsize", "footnotesize",
 145 "small", "normalsize", "large", "Large", "LARGE", "huge", "Huge", 0};
 146
 147 /// the same as known_sizes with .lyx names
 148 char const * const known_coded_sizes[] = { "tiny", "scriptsize", "footnotesize",
 149 "small", "normal", "large", "larger", "largest",  "huge", "giant", 0};
 150
 151 /// LaTeX 2.09 names for font families
 152 char const * const known_old_font_families[] = { "rm", "sf", "tt", 0};
 153
 154 /// LaTeX names for font families
 155 char const * const known_font_families[] = { "rmfamily", "sffamily",
 156 "ttfamily", 0};
 157
 158 /// the same as known_old_font_families and known_font_families with .lyx names
 159 char const * const known_coded_font_families[] = { "roman", "sans",
 160 "typewriter", 0};
 161
 162 /// LaTeX 2.09 names for font series
 163 char const * const known_old_font_series[] = { "bf", 0};
 164
 165 /// LaTeX names for font series
 166 char const * const known_font_series[] = { "bfseries", "mdseries", 0};
 167
 168 /// the same as known_old_font_series and known_font_series with .lyx names
 169 char const * const known_coded_font_series[] = { "bold", "medium", 0};
 170
 171 /// LaTeX 2.09 names for font shapes
 172 char const * const known_old_font_shapes[] = { "it", "sl", "sc", 0};
 173
 174 /// LaTeX names for font shapes
 175 char const * const known_font_shapes[] = { "itshape", "slshape", "scshape",
 176 "upshape", 0};
 177
 178 /// the same as known_old_font_shapes and known_font_shapes with .lyx names
 179 char const * const known_coded_font_shapes[] = { "italic", "slanted",
 180 "smallcaps", "up", 0};
 181
 182 /*!
 183  * Graphics file extensions known by the dvips driver of the graphics package.
 184  * These extensions are used to complete the filename of an included
 185  * graphics file if it does not contain an extension.
 186  * The order must be the same that latex uses to find a file, because we
 187  * will use the first extension that matches.
 188  * This is only an approximation for the common cases. If we would want to
 189  * do it right in all cases, we would need to know which graphics driver is
 190  * used and know the extensions of every driver of the graphics package.
 191  */
 192 char const * const known_dvips_graphics_formats[] = {"eps", "ps", "eps.gz",
 193 "ps.gz", "eps.Z", "ps.Z", 0};
 194
 195 /*!
 196  * Graphics file extensions known by the pdftex driver of the graphics package.
 197  * \sa known_dvips_graphics_formats
 198  */
 199 char const * const known_pdftex_graphics_formats[] = {"png", "pdf", "jpg",
 200 "mps", "tif", 0};
 201
 202 /*!
 203  * Known file extensions for TeX files as used by \\include.
 204  */
 205 char const * const known_tex_extensions[] = {"tex", 0};
 206
 207 /// spaces known by InsetSpace
 208 char const * const known_spaces[] = { " ", "space", ",", "thinspace", "quad",
 209 "qquad", "enspace", "enskip", "negthinspace", 0};
 210
 211 /// the same as known_spaces with .lyx names
 212 char const * const known_coded_spaces[] = { "space{}", "space{}",
 213 "thinspace{}", "thinspace{}", "quad{}", "qquad{}", "enspace{}", "enskip{}",
 214 "negthinspace{}", 0};
 215
 216
 217 /// splits "x=z, y=b" into a map
 218 map<string, string> split_map(string const & s)
 219 {
 220         map<string, string> res;
 221         vector<string> v;
 222         split(s, v);
 223         for (size_t i = 0; i < v.size(); ++i) {
 224                 size_t const pos   = v[i].find('=');
 225                 string const index = v[i].substr(0, pos);
 226                 string const value = v[i].substr(pos + 1, string::npos);
 227                 res[trim(index)] = trim(value);
 228         }
 229         return res;
 230 }
 231
 232
 233 /*!
 234  * Split a LaTeX length into value and unit.
 235  * The latter can be a real unit like "pt", or a latex length variable
 236  * like "\textwidth". The unit may contain additional stuff like glue
 237  * lengths, but we don't care, because such lengths are ERT anyway.
 238  * \returns true if \p value and \p unit are valid.
 239  */
 240 bool splitLatexLength(string const & len, string & value, string & unit)
 241 {
 242         if (len.empty())
 243                 return false;
 244         const string::size_type i = len.find_first_not_of(" -+0123456789.,");
 245         //'4,5' is a valid LaTeX length number. Change it to '4.5'
 246         string const length = subst(len, ',', '.');
 247         if (i == string::npos)
 248                 return false;
 249         if (i == 0) {
 250                 if (len[0] == '\\') {
 251                         // We had something like \textwidth without a factor
 252                         value = "1.0";
 253                 } else {
 254                         return false;
 255                 }
 256         } else {
 257                 value = trim(string(length, 0, i));
 258         }
 259         if (value == "-")
 260                 value = "-1.0";
 261         // 'cM' is a valid LaTeX length unit. Change it to 'cm'
 262         if (contains(len, '\\'))
 263                 unit = trim(string(len, i));
 264         else
 265                 unit = support::ascii_lowercase(trim(string(len, i)));
 266         return true;
 267 }
 268
 269
 270 /// A simple function to translate a latex length to something lyx can
 271 /// understand. Not perfect, but rather best-effort.
 272 bool translate_len(string const & length, string & valstring, string & unit)
 273 {
 274         if (!splitLatexLength(length, valstring, unit))
 275                 return false;
 276         // LyX uses percent values
 277         double value;
 278         istringstream iss(valstring);
 279         iss >> value;
 280         value *= 100;
 281         ostringstream oss;
 282         oss << value;
 283         string const percentval = oss.str();
 284         // a normal length
 285         if (unit.empty() || unit[0] != '\\')
 286                 return true;
 287         string::size_type const i = unit.find(' ');
 288         string const endlen = (i == string::npos) ? string() : string(unit, i);
 289         if (unit == "\\textwidth") {
 290                 valstring = percentval;
 291                 unit = "text%" + endlen;
 292         } else if (unit == "\\columnwidth") {
 293                 valstring = percentval;
 294                 unit = "col%" + endlen;
 295         } else if (unit == "\\paperwidth") {
 296                 valstring = percentval;
 297                 unit = "page%" + endlen;
 298         } else if (unit == "\\linewidth") {
 299                 valstring = percentval;
 300                 unit = "line%" + endlen;
 301         } else if (unit == "\\paperheight") {
 302                 valstring = percentval;
 303                 unit = "pheight%" + endlen;
 304         } else if (unit == "\\textheight") {
 305                 valstring = percentval;
 306                 unit = "theight%" + endlen;
 307         }
 308         return true;
 309 }
 310
 311 }
 312
 313
 314 string translate_len(string const & length)
 315 {
 316         string unit;
 317         string value;
 318         if (translate_len(length, value, unit))
 319                 return value + unit;
 320         // If the input is invalid, return what we have.
 321         return length;
 322 }
 323
 324
 325 namespace {
 326
 327 /*!
 328  * Translates a LaTeX length into \p value, \p unit and
 329  * \p special parts suitable for a box inset.
 330  * The difference from translate_len() is that a box inset knows about
 331  * some special "units" that are stored in \p special.
 332  */
 333 void translate_box_len(string const & length, string & value, string & unit, string & special)
 334 {
 335         if (translate_len(length, value, unit)) {
 336                 if (unit == "\\height" || unit == "\\depth" ||
 337                     unit == "\\totalheight" || unit == "\\width") {
 338                         special = unit.substr(1);
 339                         // The unit is not used, but LyX requires a dummy setting
 340                         unit = "in";
 341                 } else
 342                         special = "none";
 343         } else {
 344                 value.clear();
 345                 unit = length;
 346                 special = "none";
 347         }
 348 }
 349
 350
 351 /*!
 352  * Find a file with basename \p name in path \p path and an extension
 353  * in \p extensions.
 354  */
 355 string find_file(string const & name, string const & path,
 356                  char const * const * extensions)
 357 {
 358         // FIXME UNICODE encoding of name and path may be wrong (makeAbsPath
 359         // expects utf8)
 360         for (char const * const * what = extensions; *what; ++what) {
 361                 string const trial = addExtension(name, *what);
 362                 if (makeAbsPath(trial, path).exists())
 363                         return trial;
 364         }
 365         return string();
 366 }
 367
 368
 369 void begin_inset(ostream & os, string const & name)
 370 {
 371         os << "\n\\begin_inset " << name;
 372 }
 373
 374
 375 void end_inset(ostream & os)
 376 {
 377         os << "\n\\end_inset\n\n";
 378 }
 379
 380
 381 void skip_braces(Parser & p)
 382 {
 383         if (p.next_token().cat() != catBegin)
 384                 return;
 385         p.get_token();
 386         if (p.next_token().cat() == catEnd) {
 387                 p.get_token();
 388                 return;
 389         }
 390         p.putback();
 391 }
 392
 393
 394 void handle_ert(ostream & os, string const & s, Context & context)
 395 {
 396         // We must have a valid layout before outputting the ERT inset.
 397         context.check_layout(os);
 398         Context newcontext(true, context.textclass);
 399         begin_inset(os, "ERT");
 400         os << "\nstatus collapsed\n";
 401         newcontext.check_layout(os);
 402         for (string::const_iterator it = s.begin(), et = s.end(); it != et; ++it) {
 403                 if (*it == '\\')
 404                         os << "\n\\backslash\n";
 405                 else if (*it == '\n') {
 406                         newcontext.new_paragraph(os);
 407                         newcontext.check_layout(os);
 408                 } else
 409                         os << *it;
 410         }
 411         newcontext.check_end_layout(os);
 412         end_inset(os);
 413 }
 414
 415
 416 void handle_comment(ostream & os, string const & s, Context & context)
 417 {
 418         // TODO: Handle this better
 419         Context newcontext(true, context.textclass);
 420         begin_inset(os, "ERT");
 421         os << "\nstatus collapsed\n";
 422         newcontext.check_layout(os);
 423         for (string::const_iterator it = s.begin(), et = s.end(); it != et; ++it) {
 424                 if (*it == '\\')
 425                         os << "\n\\backslash\n";
 426                 else
 427                         os << *it;
 428         }
 429         // make sure that our comment is the last thing on the line
 430         newcontext.new_paragraph(os);
 431         newcontext.check_layout(os);
 432         newcontext.check_end_layout(os);
 433         end_inset(os);
 434 }
 435
 436
 437 class isLayout : public std::unary_function<LayoutPtr, bool> {
 438 public:
 439         isLayout(string const name) : name_(name) {}
 440         bool operator()(LayoutPtr const & ptr) const {
 441                 return ptr->latexname() == name_;
 442         }
 443 private:
 444         string const name_;
 445 };
 446
 447
 448 LayoutPtr findLayout(TextClass const & textclass,
 449                          string const & name)
 450 {
 451         TextClass::const_iterator beg = textclass.begin();
 452         TextClass::const_iterator end = textclass.end();
 453
 454         TextClass::const_iterator
 455                 it = std::find_if(beg, end, isLayout(name));
 456
 457         return (it == end) ? LayoutPtr() : *it;
 458 }
 459
 460
 461 void eat_whitespace(Parser &, ostream &, Context &, bool);
 462
 463
 464 void output_command_layout(ostream & os, Parser & p, bool outer,
 465                            Context & parent_context,
 466                            LayoutPtr newlayout)
 467 {
 468         parent_context.check_end_layout(os);
 469         Context context(true, parent_context.textclass, newlayout,
 470                         parent_context.layout, parent_context.font);
 471         if (parent_context.deeper_paragraph) {
 472                 // We are beginning a nested environment after a
 473                 // deeper paragraph inside the outer list environment.
 474                 // Therefore we don't need to output a "begin deeper".
 475                 context.need_end_deeper = true;
 476         }
 477         context.check_deeper(os);
 478         context.check_layout(os);
 479         if (context.layout->optionalargs > 0) {
 480                 eat_whitespace(p, os, context, false);
 481                 if (p.next_token().character() == '[') {
 482                         p.get_token(); // eat '['
 483                         begin_inset(os, "OptArg\n");
 484                         os << "status collapsed\n\n";
 485                         parse_text_in_inset(p, os, FLAG_BRACK_LAST, outer, context);
 486                         end_inset(os);
 487                         eat_whitespace(p, os, context, false);
 488                 }
 489         }
 490         parse_text(p, os, FLAG_ITEM, outer, context);
 491         context.check_end_layout(os);
 492         if (parent_context.deeper_paragraph) {
 493                 // We must suppress the "end deeper" because we
 494                 // suppressed the "begin deeper" above.
 495                 context.need_end_deeper = false;
 496         }
 497         context.check_end_deeper(os);
 498         // We don't need really a new paragraph, but
 499         // we must make sure that the next item gets a \begin_layout.
 500         parent_context.new_paragraph(os);
 501 }
 502
 503
 504 /*!
 505  * Output a space if necessary.
 506  * This function gets called for every whitespace token.
 507  *
 508  * We have three cases here:
 509  * 1. A space must be suppressed. Example: The lyxcode case below
 510  * 2. A space may be suppressed. Example: Spaces before "\par"
 511  * 3. A space must not be suppressed. Example: A space between two words
 512  *
 513  * We currently handle only 1. and 3 and from 2. only the case of
 514  * spaces before newlines as a side effect.
 515  *
 516  * 2. could be used to suppress as many spaces as possible. This has two effects:
 517  * - Reimporting LyX generated LaTeX files changes almost no whitespace
 518  * - Superflous whitespace from non LyX generated LaTeX files is removed.
 519  * The drawback is that the logic inside the function becomes
 520  * complicated, and that is the reason why it is not implemented.
 521  */
 522 void check_space(Parser const & p, ostream & os, Context & context)
 523 {
 524         Token const next = p.next_token();
 525         Token const curr = p.curr_token();
 526         // A space before a single newline and vice versa must be ignored
 527         // LyX emits a newline before \end{lyxcode}.
 528         // This newline must be ignored,
 529         // otherwise LyX will add an additional protected space.
 530         if (next.cat() == catSpace ||
 531             next.cat() == catNewline ||
 532             (next.cs() == "end" && context.layout->free_spacing && curr.cat() == catNewline)) {
 533                 return;
 534         }
 535         context.check_layout(os);
 536         os << ' ';
 537 }
 538
 539
 540 /*!
 541  * Parse all arguments of \p command
 542  */
 543 void parse_arguments(string const & command,
 544                      vector<ArgumentType> const & template_arguments,
 545                      Parser & p, ostream & os, bool outer, Context & context)
 546 {
 547         string ert = command;
 548         size_t no_arguments = template_arguments.size();
 549         for (size_t i = 0; i < no_arguments; ++i) {
 550                 switch (template_arguments[i]) {
 551                 case required:
 552                         // This argument contains regular LaTeX
 553                         handle_ert(os, ert + '{', context);
 554                         eat_whitespace(p, os, context, false);
 555                         parse_text(p, os, FLAG_ITEM, outer, context);
 556                         ert = "}";
 557                         break;
 558                 case verbatim:
 559                         // This argument may contain special characters
 560                         ert += '{' + p.verbatim_item() + '}';
 561                         break;
 562                 case optional:
 563                         ert += p.getOpt();
 564                         break;
 565                 }
 566         }
 567         handle_ert(os, ert, context);
 568 }
 569
 570
 571 /*!
 572  * Check whether \p command is a known command. If yes,
 573  * handle the command with all arguments.
 574  * \return true if the command was parsed, false otherwise.
 575  */
 576 bool parse_command(string const & command, Parser & p, ostream & os,
 577                    bool outer, Context & context)
 578 {
 579         if (known_commands.find(command) != known_commands.end()) {
 580                 parse_arguments(command, known_commands[command], p, os,
 581                                 outer, context);
 582                 return true;
 583         }
 584         return false;
 585 }
 586
 587
 588 /// Parses a minipage or parbox
 589 void parse_box(Parser & p, ostream & os, unsigned flags, bool outer,
 590                Context & parent_context, bool use_parbox)
 591 {
 592         string position;
 593         string inner_pos;
 594         // We need to set the height to the LaTeX default of 1\\totalheight
 595         // for the case when no height argument is given
 596         string height_value = "1";
 597         string height_unit = "in";
 598         string height_special = "totalheight";
 599         string latex_height;
 600         if (p.next_token().asInput() == "[") {
 601                 position = p.getArg('[', ']');
 602                 if (position != "t" && position != "c" && position != "b") {
 603                         position = "c";
 604                         cerr << "invalid position for minipage/parbox" << endl;
 605                 }
 606                 if (p.next_token().asInput() == "[") {
 607                         latex_height = p.getArg('[', ']');
 608                         translate_box_len(latex_height, height_value, height_unit, height_special);
 609
 610                         if (p.next_token().asInput() == "[") {
 611                                 inner_pos = p.getArg('[', ']');
 612                                 if (inner_pos != "c" && inner_pos != "t" &&
 613                                     inner_pos != "b" && inner_pos != "s") {
 614                                         inner_pos = position;
 615                                         cerr << "invalid inner_pos for minipage/parbox"
 616                                              << endl;
 617                                 }
 618                         }
 619                 }
 620         }
 621         string width_value;
 622         string width_unit;
 623         string const latex_width = p.verbatim_item();
 624         translate_len(latex_width, width_value, width_unit);
 625         if (contains(width_unit, '\\') || contains(height_unit, '\\')) {
 626                 // LyX can't handle length variables
 627                 ostringstream ss;
 628                 if (use_parbox)
 629                         ss << "\\parbox";
 630                 else
 631                         ss << "\\begin{minipage}";
 632                 if (!position.empty())
 633                         ss << '[' << position << ']';
 634                 if (!latex_height.empty())
 635                         ss << '[' << latex_height << ']';
 636                 if (!inner_pos.empty())
 637                         ss << '[' << inner_pos << ']';
 638                 ss << "{" << latex_width << "}";
 639                 if (use_parbox)
 640                         ss << '{';
 641                 handle_ert(os, ss.str(), parent_context);
 642                 parent_context.new_paragraph(os);
 643                 parse_text_in_inset(p, os, flags, outer, parent_context);
 644                 if (use_parbox)
 645                         handle_ert(os, "}", parent_context);
 646                 else
 647                         handle_ert(os, "\\end{minipage}", parent_context);
 648         } else {
 649                 // LyX does not like empty positions, so we have
 650                 // to set them to the LaTeX default values here.
 651                 if (position.empty())
 652                         position = "c";
 653                 if (inner_pos.empty())
 654                         inner_pos = position;
 655                 parent_context.check_layout(os);
 656                 begin_inset(os, "Box Frameless\n");
 657                 os << "position \"" << position << "\"\n";
 658                 os << "hor_pos \"c\"\n";
 659                 os << "has_inner_box 1\n";
 660                 os << "inner_pos \"" << inner_pos << "\"\n";
 661                 os << "use_parbox " << use_parbox << "\n";
 662                 os << "width \"" << width_value << width_unit << "\"\n";
 663                 os << "special \"none\"\n";
 664                 os << "height \"" << height_value << height_unit << "\"\n";
 665                 os << "height_special \"" << height_special << "\"\n";
 666                 os << "status open\n\n";
 667                 parse_text_in_inset(p, os, flags, outer, parent_context);
 668                 end_inset(os);
 669 #ifdef PRESERVE_LAYOUT
 670                 // lyx puts a % after the end of the minipage
 671                 if (p.next_token().cat() == catNewline && p.next_token().cs().size() > 1) {
 672                         // new paragraph
 673                         //handle_comment(os, "%dummy", parent_context);
 674                         p.get_token();
 675                         p.skip_spaces();
 676                         parent_context.new_paragraph(os);
 677                 }
 678                 else if (p.next_token().cat() == catSpace || p.next_token().cat() == catNewline) {
 679                         //handle_comment(os, "%dummy", parent_context);
 680                         p.get_token();
 681                         p.skip_spaces();
 682                         // We add a protected space if something real follows
 683                         if (p.good() && p.next_token().cat() != catComment) {
 684                                 os << "\\InsetSpace ~\n";
 685                         }
 686                 }
 687 #endif
 688         }
 689 }
 690
 691
 692 /// parse an unknown environment
 693 void parse_unknown_environment(Parser & p, string const & name, ostream & os,
 694                                unsigned flags, bool outer,
 695                                Context & parent_context)
 696 {
 697         if (name == "tabbing")
 698                 // We need to remember that we have to handle '\=' specially
 699                 flags |= FLAG_TABBING;
 700
 701         // We need to translate font changes and paragraphs inside the
 702         // environment to ERT if we have a non standard font.
 703         // Otherwise things like
 704         // \large\begin{foo}\huge bar\end{foo}
 705         // will not work.
 706         bool const specialfont =
 707                 (parent_context.font != parent_context.normalfont);
 708         bool const new_layout_allowed = parent_context.new_layout_allowed;
 709         if (specialfont)
 710                 parent_context.new_layout_allowed = false;
 711         handle_ert(os, "\\begin{" + name + "}", parent_context);
 712         parse_text_snippet(p, os, flags, outer, parent_context);
 713         handle_ert(os, "\\end{" + name + "}", parent_context);
 714         if (specialfont)
 715                 parent_context.new_layout_allowed = new_layout_allowed;
 716 }
 717
 718
 719 void parse_environment(Parser & p, ostream & os, bool outer,
 720                        Context & parent_context)
 721 {
 722         LayoutPtr newlayout;
 723         string const name = p.getArg('{', '}');
 724         const bool is_starred = suffixIs(name, '*');
 725         string const unstarred_name = rtrim(name, "*");
 726         active_environments.push_back(name);
 727
 728         if (is_math_env(name)) {
 729                 parent_context.check_layout(os);
 730                 begin_inset(os, "Formula ");
 731                 os << "\\begin{" << name << "}";
 732                 parse_math(p, os, FLAG_END, MATH_MODE);
 733                 os << "\\end{" << name << "}";
 734                 end_inset(os);
 735         }
 736
 737         else if (name == "tabular" || name == "longtable") {
 738                 eat_whitespace(p, os, parent_context, false);
 739                 parent_context.check_layout(os);
 740                 begin_inset(os, "Tabular ");
 741                 handle_tabular(p, os, name == "longtable", parent_context);
 742                 end_inset(os);
 743                 p.skip_spaces();
 744         }
 745
 746         else if (parent_context.textclass.floats().typeExist(unstarred_name)) {
 747                 eat_whitespace(p, os, parent_context, false);
 748                 parent_context.check_layout(os);
 749                 begin_inset(os, "Float " + unstarred_name + "\n");
 750                 if (p.next_token().asInput() == "[") {
 751                         os << "placement " << p.getArg('[', ']') << '\n';
 752                 }
 753                 os << "wide " << convert<string>(is_starred)
 754                    << "\nsideways false"
 755                    << "\nstatus open\n\n";
 756                 parse_text_in_inset(p, os, FLAG_END, outer, parent_context);
 757                 end_inset(os);
 758                 // We don't need really a new paragraph, but
 759                 // we must make sure that the next item gets a \begin_layout.
 760                 parent_context.new_paragraph(os);
 761                 p.skip_spaces();
 762         }
 763
 764         else if (name == "minipage") {
 765                 eat_whitespace(p, os, parent_context, false);
 766                 parse_box(p, os, FLAG_END, outer, parent_context, false);
 767                 p.skip_spaces();
 768         }
 769
 770         else if (name == "comment") {
 771                 eat_whitespace(p, os, parent_context, false);
 772                 parent_context.check_layout(os);
 773                 begin_inset(os, "Note Comment\n");
 774                 os << "status open\n";
 775                 parse_text_in_inset(p, os, FLAG_END, outer, parent_context);
 776                 end_inset(os);
 777                 p.skip_spaces();
 778         }
 779
 780         else if (name == "lyxgreyedout") {
 781                 eat_whitespace(p, os, parent_context, false);
 782                 parent_context.check_layout(os);
 783                 begin_inset(os, "Note Greyedout\n");
 784                 os << "status open\n";
 785                 parse_text_in_inset(p, os, FLAG_END, outer, parent_context);
 786                 end_inset(os);
 787                 p.skip_spaces();
 788         }
 789
 790         else if (name == "framed") {
 791                 eat_whitespace(p, os, parent_context, false);
 792                 parent_context.check_layout(os);
 793                 begin_inset(os, "Note Framed\n");
 794                 os << "status open\n";
 795                 parse_text_in_inset(p, os, FLAG_END, outer, parent_context);
 796                 end_inset(os);
 797                 p.skip_spaces();
 798         }
 799
 800         else if (name == "shaded") {
 801                 eat_whitespace(p, os, parent_context, false);
 802                 parent_context.check_layout(os);
 803                 begin_inset(os, "Note Shaded\n");
 804                 os << "status open\n";
 805                 parse_text_in_inset(p, os, FLAG_END, outer, parent_context);
 806                 end_inset(os);
 807                 p.skip_spaces();
 808         }
 809
 810         else if (!parent_context.new_layout_allowed)
 811                 parse_unknown_environment(p, name, os, FLAG_END, outer,
 812                                           parent_context);
 813
 814         // Alignment settings
 815         else if (name == "center" || name == "flushleft" || name == "flushright" ||
 816                  name == "centering" || name == "raggedright" || name == "raggedleft") {
 817                 eat_whitespace(p, os, parent_context, false);
 818                 // We must begin a new paragraph if not already done
 819                 if (! parent_context.atParagraphStart()) {
 820                         parent_context.check_end_layout(os);
 821                         parent_context.new_paragraph(os);
 822                 }
 823                 if (name == "flushleft" || name == "raggedright")
 824                         parent_context.add_extra_stuff("\\align left\n");
 825                 else if (name == "flushright" || name == "raggedleft")
 826                         parent_context.add_extra_stuff("\\align right\n");
 827                 else
 828                         parent_context.add_extra_stuff("\\align center\n");
 829                 parse_text(p, os, FLAG_END, outer, parent_context);
 830                 // Just in case the environment is empty ..
 831                 parent_context.extra_stuff.erase();
 832                 // We must begin a new paragraph to reset the alignment
 833                 parent_context.new_paragraph(os);
 834                 p.skip_spaces();
 835         }
 836
 837         // The single '=' is meant here.
 838         else if ((newlayout = findLayout(parent_context.textclass, name)).get() &&
 839                   newlayout->isEnvironment()) {
 840                 eat_whitespace(p, os, parent_context, false);
 841                 Context context(true, parent_context.textclass, newlayout,
 842                                 parent_context.layout, parent_context.font);
 843                 if (parent_context.deeper_paragraph) {
 844                         // We are beginning a nested environment after a
 845                         // deeper paragraph inside the outer list environment.
 846                         // Therefore we don't need to output a "begin deeper".
 847                         context.need_end_deeper = true;
 848                 }
 849                 parent_context.check_end_layout(os);
 850                 switch (context.layout->latextype) {
 851                 case  LATEX_LIST_ENVIRONMENT:
 852                         context.extra_stuff = "\\labelwidthstring "
 853                                 + p.verbatim_item() + '\n';
 854                         p.skip_spaces();
 855                         break;
 856                 case  LATEX_BIB_ENVIRONMENT:
 857                         p.verbatim_item(); // swallow next arg
 858                         p.skip_spaces();
 859                         break;
 860                 default:
 861                         break;
 862                 }
 863                 context.check_deeper(os);
 864                 parse_text(p, os, FLAG_END, outer, context);
 865                 context.check_end_layout(os);
 866                 if (parent_context.deeper_paragraph) {
 867                         // We must suppress the "end deeper" because we
 868                         // suppressed the "begin deeper" above.
 869                         context.need_end_deeper = false;
 870                 }
 871                 context.check_end_deeper(os);
 872                 parent_context.new_paragraph(os);
 873                 p.skip_spaces();
 874         }
 875
 876         else if (name == "appendix") {
 877                 // This is no good latex style, but it works and is used in some documents...
 878                 eat_whitespace(p, os, parent_context, false);
 879                 parent_context.check_end_layout(os);
 880                 Context context(true, parent_context.textclass, parent_context.layout,
 881                                 parent_context.layout, parent_context.font);
 882                 context.check_layout(os);
 883                 os << "\\start_of_appendix\n";
 884                 parse_text(p, os, FLAG_END, outer, context);
 885                 context.check_end_layout(os);
 886                 p.skip_spaces();
 887         }
 888
 889         else if (known_environments.find(name) != known_environments.end()) {
 890                 vector<ArgumentType> arguments = known_environments[name];
 891                 // The last "argument" denotes wether we may translate the
 892                 // environment contents to LyX
 893                 // The default required if no argument is given makes us
 894                 // compatible with the reLyXre environment.
 895                 ArgumentType contents = arguments.empty() ?
 896                         required :
 897                         arguments.back();
 898                 if (!arguments.empty())
 899                         arguments.pop_back();
 900                 // See comment in parse_unknown_environment()
 901                 bool const specialfont =
 902                         (parent_context.font != parent_context.normalfont);
 903                 bool const new_layout_allowed =
 904                         parent_context.new_layout_allowed;
 905                 if (specialfont)
 906                         parent_context.new_layout_allowed = false;
 907                 parse_arguments("\\begin{" + name + "}", arguments, p, os,
 908                                 outer, parent_context);
 909                 if (contents == verbatim)
 910                         handle_ert(os, p.verbatimEnvironment(name),
 911                                    parent_context);
 912                 else
 913                         parse_text_snippet(p, os, FLAG_END, outer,
 914                                            parent_context);
 915                 handle_ert(os, "\\end{" + name + "}", parent_context);
 916                 if (specialfont)
 917                         parent_context.new_layout_allowed = new_layout_allowed;
 918         }
 919
 920         else
 921                 parse_unknown_environment(p, name, os, FLAG_END, outer,
 922                                           parent_context);
 923
 924         active_environments.pop_back();
 925 }
 926
 927
 928 /// parses a comment and outputs it to \p os.
 929 void parse_comment(Parser & p, ostream & os, Token const & t, Context & context)
 930 {
 931         BOOST_ASSERT(t.cat() == catComment);
 932         if (!t.cs().empty()) {
 933                 context.check_layout(os);
 934                 handle_comment(os, '%' + t.cs(), context);
 935                 if (p.next_token().cat() == catNewline) {
 936                         // A newline after a comment line starts a new
 937                         // paragraph
 938                         if (context.new_layout_allowed) {
 939                                 if(!context.atParagraphStart())
 940                                         // Only start a new paragraph if not already
 941                                         // done (we might get called recursively)
 942                                         context.new_paragraph(os);
 943                         } else
 944                                 handle_ert(os, "\n", context);
 945                         eat_whitespace(p, os, context, true);
 946                 }
 947         } else {
 948                 // "%\n" combination
 949                 p.skip_spaces();
 950         }
 951 }
 952
 953
 954 /*!
 955  * Reads spaces and comments until the first non-space, non-comment token.
 956  * New paragraphs (double newlines or \\par) are handled like simple spaces
 957  * if \p eatParagraph is true.
 958  * Spaces are skipped, but comments are written to \p os.
 959  */
 960 void eat_whitespace(Parser & p, ostream & os, Context & context,
 961                     bool eatParagraph)
 962 {
 963         while (p.good()) {
 964                 Token const & t = p.get_token();
 965                 if (t.cat() == catComment)
 966                         parse_comment(p, os, t, context);
 967                 else if ((! eatParagraph && p.isParagraph()) ||
 968                          (t.cat() != catSpace && t.cat() != catNewline)) {
 969                         p.putback();
 970                         return;
 971                 }
 972         }
 973 }
 974
 975
 976 /*!
 977  * Set a font attribute, parse text and reset the font attribute.
 978  * \param attribute Attribute name (e.g. \\family, \\shape etc.)
 979  * \param currentvalue Current value of the attribute. Is set to the new
 980  * value during parsing.
 981  * \param newvalue New value of the attribute
 982  */
 983 void parse_text_attributes(Parser & p, ostream & os, unsigned flags, bool outer,
 984                            Context & context, string const & attribute,
 985                            string & currentvalue, string const & newvalue)
 986 {
 987         context.check_layout(os);
 988         string const oldvalue = currentvalue;
 989         currentvalue = newvalue;
 990         os << '\n' << attribute << ' ' << newvalue << "\n";
 991         parse_text_snippet(p, os, flags, outer, context);
 992         context.check_layout(os);
 993         os << '\n' << attribute << ' ' << oldvalue << "\n";
 994         currentvalue = oldvalue;
 995 }
 996
 997
 998 /// get the arguments of a natbib or jurabib citation command
 999 std::pair<string, string> getCiteArguments(Parser & p, bool natbibOrder)
1000 {
1001         // We need to distinguish "" and "[]", so we can't use p.getOpt().
1002
1003         // text before the citation
1004         string before;
1005         // text after the citation
1006         string after = p.getFullOpt();
1007
1008         if (!after.empty()) {
1009                 before = p.getFullOpt();
1010                 if (natbibOrder && !before.empty())
1011                         std::swap(before, after);
1012         }
1013         return std::make_pair(before, after);
1014 }
1015
1016
1017 /// Convert filenames with TeX macros and/or quotes to something LyX can understand
1018 string const normalize_filename(string const & name)
1019 {
1020         Parser p(trim(name, "\""));
1021         ostringstream os;
1022         while (p.good()) {
1023                 Token const & t = p.get_token();
1024                 if (t.cat() != catEscape)
1025                         os << t.asInput();
1026                 else if (t.cs() == "lyxdot") {
1027                         // This is used by LyX for simple dots in relative
1028                         // names
1029                         os << '.';
1030                         p.skip_spaces();
1031                 } else if (t.cs() == "space") {
1032                         os << ' ';
1033                         p.skip_spaces();
1034                 } else
1035                         os << t.asInput();
1036         }
1037         return os.str();
1038 }
1039
1040
1041 /// Convert \p name from TeX convention (relative to master file) to LyX
1042 /// convention (relative to .lyx file) if it is relative
1043 void fix_relative_filename(string & name)
1044 {
1045         if (lyx::support::absolutePath(name))
1046                 return;
1047         // FIXME UNICODE encoding of name may be wrong (makeAbsPath expects
1048         // utf8)
1049         name = to_utf8(makeRelPath(from_utf8(makeAbsPath(name, getMasterFilePath()).absFilename()),
1050                                    from_utf8(getParentFilePath())));
1051 }
1052
1053
1054 /// Parse a NoWeb Scrap section. The initial "<<" is already parsed.
1055 void parse_noweb(Parser & p, ostream & os, Context & context)
1056 {
1057         // assemble the rest of the keyword
1058         string name("<<");
1059         bool scrap = false;
1060         while (p.good()) {
1061                 Token const & t = p.get_token();
1062                 if (t.asInput() == ">" && p.next_token().asInput() == ">") {
1063                         name += ">>";
1064                         p.get_token();
1065                         scrap = (p.good() && p.next_token().asInput() == "=");
1066                         if (scrap)
1067                                 name += p.get_token().asInput();
1068                         break;
1069                 }
1070                 name += t.asInput();
1071         }
1072
1073         if (!scrap || !context.new_layout_allowed ||
1074             !context.textclass.hasLayout(from_ascii("Scrap"))) {
1075                 cerr << "Warning: Could not interpret '" << name
1076                      << "'. Ignoring it." << endl;
1077                 return;
1078         }
1079
1080         // We use new_paragraph instead of check_end_layout because the stuff
1081         // following the noweb chunk needs to start with a \begin_layout.
1082         // This may create a new paragraph even if there was none in the
1083         // noweb file, but the alternative is an invalid LyX file. Since
1084         // noweb code chunks are implemented with a layout style in LyX they
1085         // always must be in an own paragraph.
1086         context.new_paragraph(os);
1087         Context newcontext(true, context.textclass,
1088                 context.textclass[from_ascii("Scrap")]);
1089         newcontext.check_layout(os);
1090         os << name;
1091         while (p.good()) {
1092                 Token const & t = p.get_token();
1093                 // We abuse the parser a bit, because this is no TeX syntax
1094                 // at all.
1095                 if (t.cat() == catEscape)
1096                         os << subst(t.asInput(), "\\", "\n\\backslash\n");
1097                 else
1098                         os << subst(t.asInput(), "\n", "\n\\newline\n");
1099                 // The scrap chunk is ended by an @ at the beginning of a line.
1100                 // After the @ the line may contain a comment and/or
1101                 // whitespace, but nothing else.
1102                 if (t.asInput() == "@" && p.prev_token().cat() == catNewline &&
1103                     (p.next_token().cat() == catSpace ||
1104                      p.next_token().cat() == catNewline ||
1105                      p.next_token().cat() == catComment)) {
1106                         while (p.good() && p.next_token().cat() == catSpace)
1107                                 os << p.get_token().asInput();
1108                         if (p.next_token().cat() == catComment)
1109                                 // The comment includes a final '\n'
1110                                 os << p.get_token().asInput();
1111                         else {
1112                                 if (p.next_token().cat() == catNewline)
1113                                         p.get_token();
1114                                 os << '\n';
1115                         }
1116                         break;
1117                 }
1118         }
1119         newcontext.check_end_layout(os);
1120 }
1121
1122 } // anonymous namespace
1123
1124
1125 void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
1126                 Context & context)
1127 {
1128         LayoutPtr newlayout;
1129         // Store the latest bibliographystyle (needed for bibtex inset)
1130         string bibliographystyle;
1131         bool const use_natbib = used_packages.find("natbib") != used_packages.end();
1132         bool const use_jurabib = used_packages.find("jurabib") != used_packages.end();
1133         while (p.good()) {
1134                 Token const & t = p.get_token();
1135
1136 #ifdef FILEDEBUG
1137                 cerr << "t: " << t << " flags: " << flags << "\n";
1138 #endif
1139
1140                 if (flags & FLAG_ITEM) {
1141                         if (t.cat() == catSpace)
1142                                 continue;
1143
1144                         flags &= ~FLAG_ITEM;
1145                         if (t.cat() == catBegin) {
1146                                 // skip the brace and collect everything to the next matching
1147                                 // closing brace
1148                                 flags |= FLAG_BRACE_LAST;
1149                                 continue;
1150                         }
1151
1152                         // handle only this single token, leave the loop if done
1153                         flags |= FLAG_LEAVE;
1154                 }
1155
1156                 if (t.character() == ']' && (flags & FLAG_BRACK_LAST))
1157                         return;
1158
1159                 //
1160                 // cat codes
1161                 //
1162                 if (t.cat() == catMath) {
1163                         // we are inside some text mode thingy, so opening new math is allowed
1164                         context.check_layout(os);
1165                         begin_inset(os, "Formula ");
1166                         Token const & n = p.get_token();
1167                         if (n.cat() == catMath && outer) {
1168                                 // TeX's $$...$$ syntax for displayed math
1169                                 os << "\\[";
1170                                 parse_math(p, os, FLAG_SIMPLE, MATH_MODE);
1171                                 os << "\\]";
1172                                 p.get_token(); // skip the second '$' token
1173                         } else {
1174                                 // simple $...$  stuff
1175                                 p.putback();
1176                                 os << '$';
1177                                 parse_math(p, os, FLAG_SIMPLE, MATH_MODE);
1178                                 os << '$';
1179                         }
1180                         end_inset(os);
1181                 }
1182
1183                 else if (t.cat() == catSuper || t.cat() == catSub)
1184                         cerr << "catcode " << t << " illegal in text mode\n";
1185
1186                 // Basic support for english quotes. This should be
1187                 // extended to other quotes, but is not so easy (a
1188                 // left english quote is the same as a right german
1189                 // quote...)
1190                 else if (t.asInput() == "`"
1191                          && p.next_token().asInput() == "`") {
1192                         context.check_layout(os);
1193                         begin_inset(os, "Quotes ");
1194                         os << "eld";
1195                         end_inset(os);
1196                         p.get_token();
1197                         skip_braces(p);
1198                 }
1199                 else if (t.asInput() == "'"
1200                          && p.next_token().asInput() == "'") {
1201                         context.check_layout(os);
1202                         begin_inset(os, "Quotes ");
1203                         os << "erd";
1204                         end_inset(os);
1205                         p.get_token();
1206                         skip_braces(p);
1207                 }
1208
1209                 else if (t.asInput() == "<"
1210                          && p.next_token().asInput() == "<" && noweb_mode) {
1211                         p.get_token();
1212                         parse_noweb(p, os, context);
1213                 }
1214
1215                 else if (t.cat() == catSpace || (t.cat() == catNewline && ! p.isParagraph()))
1216                         check_space(p, os, context);
1217
1218                 else if (t.character() == '[' && noweb_mode &&
1219                          p.next_token().character() == '[') {
1220                         // These can contain underscores
1221                         p.putback();
1222                         string const s = p.getFullOpt() + ']';
1223                         if (p.next_token().character() == ']')
1224                                 p.get_token();
1225                         else
1226                                 cerr << "Warning: Inserting missing ']' in '"
1227                                      << s << "'." << endl;
1228                         handle_ert(os, s, context);
1229                 }
1230
1231                 else if (t.cat() == catLetter ||
1232                                t.cat() == catOther ||
1233                                t.cat() == catAlign ||
1234                                t.cat() == catParameter) {
1235                         // This translates "&" to "\\&" which may be wrong...
1236                         context.check_layout(os);
1237                         os << t.character();
1238                 }
1239
1240                 else if (p.isParagraph()) {
1241                         if (context.new_layout_allowed)
1242                                 context.new_paragraph(os);
1243                         else
1244                                 handle_ert(os, "\\par ", context);
1245                         eat_whitespace(p, os, context, true);
1246                 }
1247
1248                 else if (t.cat() == catActive) {
1249                         context.check_layout(os);
1250                         if (t.character() == '~') {
1251                                 if (context.layout->free_spacing)
1252                                         os << ' ';
1253                                 else
1254                                         os << "\\InsetSpace ~\n";
1255                         } else
1256                                 os << t.character();
1257                 }
1258
1259                 else if (t.cat() == catBegin &&
1260                          p.next_token().cat() == catEnd) {
1261                         // {}
1262                         Token const prev = p.prev_token();
1263                         p.get_token();
1264                         if (p.next_token().character() == '`' ||
1265                             (prev.character() == '-' &&
1266                              p.next_token().character() == '-'))
1267                                 ; // ignore it in {}`` or -{}-
1268                         else
1269                                 handle_ert(os, "{}", context);
1270
1271                 }
1272
1273                 else if (t.cat() == catBegin) {
1274                         context.check_layout(os);
1275                         // special handling of font attribute changes
1276                         Token const prev = p.prev_token();
1277                         Token const next = p.next_token();
1278                         TeXFont const oldFont = context.font;
1279                         if (next.character() == '[' ||
1280                             next.character() == ']' ||
1281                             next.character() == '*') {
1282                                 p.get_token();
1283                                 if (p.next_token().cat() == catEnd) {
1284                                         os << next.character();
1285                                         p.get_token();
1286                                 } else {
1287                                         p.putback();
1288                                         handle_ert(os, "{", context);
1289                                         parse_text_snippet(p, os,
1290                                                         FLAG_BRACE_LAST,
1291                                                         outer, context);
1292                                         handle_ert(os, "}", context);
1293                                 }
1294                         } else if (! context.new_layout_allowed) {
1295                                 handle_ert(os, "{", context);
1296                                 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1297                                                    outer, context);
1298                                 handle_ert(os, "}", context);
1299                         } else if (is_known(next.cs(), known_sizes)) {
1300                                 // next will change the size, so we must
1301                                 // reset it here
1302                                 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1303                                                    outer, context);
1304                                 if (!context.atParagraphStart())
1305                                         os << "\n\\size "
1306                                            << context.font.size << "\n";
1307                         } else if (is_known(next.cs(), known_font_families)) {
1308                                 // next will change the font family, so we
1309                                 // must reset it here
1310                                 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1311                                                    outer, context);
1312                                 if (!context.atParagraphStart())
1313                                         os << "\n\\family "
1314                                            << context.font.family << "\n";
1315                         } else if (is_known(next.cs(), known_font_series)) {
1316                                 // next will change the font series, so we
1317                                 // must reset it here
1318                                 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1319                                                    outer, context);
1320                                 if (!context.atParagraphStart())
1321                                         os << "\n\\series "
1322                                            << context.font.series << "\n";
1323                         } else if (is_known(next.cs(), known_font_shapes)) {
1324                                 // next will change the font shape, so we
1325                                 // must reset it here
1326                                 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1327                                                    outer, context);
1328                                 if (!context.atParagraphStart())
1329                                         os << "\n\\shape "
1330                                            << context.font.shape << "\n";
1331                         } else if (is_known(next.cs(), known_old_font_families) ||
1332                                    is_known(next.cs(), known_old_font_series) ||
1333                                    is_known(next.cs(), known_old_font_shapes)) {
1334                                 // next will change the font family, series
1335                                 // and shape, so we must reset it here
1336                                 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1337                                                    outer, context);
1338                                 if (!context.atParagraphStart())
1339                                         os <<  "\n\\family "
1340                                            << context.font.family
1341                                            << "\n\\series "
1342                                            << context.font.series
1343                                            << "\n\\shape "
1344                                            << context.font.shape << "\n";
1345                         } else {
1346                                 handle_ert(os, "{", context);
1347                                 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1348                                                    outer, context);
1349                                 handle_ert(os, "}", context);
1350                         }
1351                 }
1352
1353                 else if (t.cat() == catEnd) {
1354                         if (flags & FLAG_BRACE_LAST) {
1355                                 return;
1356                         }
1357                         cerr << "stray '}' in text\n";
1358                         handle_ert(os, "}", context);
1359                 }
1360
1361                 else if (t.cat() == catComment)
1362                         parse_comment(p, os, t, context);
1363
1364                 //
1365                 // control sequences
1366                 //
1367
1368                 else if (t.cs() == "(") {
1369                         context.check_layout(os);
1370                         begin_inset(os, "Formula");
1371                         os << " \\(";
1372                         parse_math(p, os, FLAG_SIMPLE2, MATH_MODE);
1373                         os << "\\)";
1374                         end_inset(os);
1375                 }
1376
1377                 else if (t.cs() == "[") {
1378                         context.check_layout(os);
1379                         begin_inset(os, "Formula");
1380                         os << " \\[";
1381                         parse_math(p, os, FLAG_EQUATION, MATH_MODE);
1382                         os << "\\]";
1383                         end_inset(os);
1384                 }
1385
1386                 else if (t.cs() == "begin")
1387                         parse_environment(p, os, outer, context);
1388
1389                 else if (t.cs() == "end") {
1390                         if (flags & FLAG_END) {
1391                                 // eat environment name
1392                                 string const name = p.getArg('{', '}');
1393                                 if (name != active_environment())
1394                                         cerr << "\\end{" + name + "} does not match \\begin{"
1395                                                 + active_environment() + "}\n";
1396                                 return;
1397                         }
1398                         p.error("found 'end' unexpectedly");
1399                 }
1400
1401                 else if (t.cs() == "item") {
1402                         p.skip_spaces();
1403                         string s;
1404                         bool optarg = false;
1405                         if (p.next_token().character() == '[') {
1406                                 p.get_token(); // eat '['
1407                                 s = parse_text_snippet(p, FLAG_BRACK_LAST,
1408                                                        outer, context);
1409                                 optarg = true;
1410                         }
1411                         context.set_item();
1412                         context.check_layout(os);
1413                         if (context.has_item) {
1414                                 // An item in an unknown list-like environment
1415                                 // FIXME: Do this in check_layout()!
1416                                 context.has_item = false;
1417                                 if (optarg)
1418                                         handle_ert(os, "\\item", context);
1419                                 else
1420                                         handle_ert(os, "\\item ", context);
1421                         }
1422                         if (optarg) {
1423                                 if (context.layout->labeltype != LABEL_MANUAL) {
1424                                         // lyx does not support \item[\mybullet]
1425                                         // in itemize environments
1426                                         handle_ert(os, "[", context);
1427                                         os << s;
1428                                         handle_ert(os, "]", context);
1429                                 } else if (!s.empty()) {
1430                                         // The space is needed to separate the
1431                                         // item from the rest of the sentence.
1432                                         os << s << ' ';
1433                                         eat_whitespace(p, os, context, false);
1434                                 }
1435                         }
1436                 }
1437
1438                 else if (t.cs() == "bibitem") {
1439                         context.set_item();
1440                         context.check_layout(os);
1441                         os << "\\bibitem ";
1442                         os << p.getOpt();
1443                         os << '{' << p.verbatim_item() << '}' << "\n";
1444                 }
1445
1446                 else if (t.cs() == "def") {
1447                         context.check_layout(os);
1448                         eat_whitespace(p, os, context, false);
1449                         string name = p.get_token().cs();
1450                         eat_whitespace(p, os, context, false);
1451
1452                         // parameter text
1453                         bool simple = true;
1454                         string paramtext;
1455                         int arity = 0;
1456                         while (p.next_token().cat() != catBegin) {
1457                                 if (p.next_token().cat() == catParameter) {
1458                                         // # found
1459                                         p.get_token();
1460                                         paramtext += "#";
1461
1462                                         // followed by number?
1463                                         if (p.next_token().cat() == catOther) {
1464                                                 char c = p.getChar();
1465                                                 paramtext += c;
1466                                                 // number = current arity + 1?
1467                                                 if (c == arity + '0' + 1)
1468                                                         ++arity;
1469                                                 else
1470                                                         simple = false;
1471                                         } else
1472                                                 paramtext += p.get_token().asString();
1473                                 } else {
1474                                         paramtext += p.get_token().asString();
1475                                         simple = false;
1476                                 }
1477                         }
1478
1479                         // only output simple (i.e. compatible) macro as FormulaMacros
1480                         string ert = "\\def\\" + name + ' ' + paramtext + '{' + p.verbatim_item() + '}';
1481                         if (simple) {
1482                                 context.check_layout(os);
1483                                 begin_inset(os, "FormulaMacro");
1484                                 os << "\n" << ert;
1485                                 end_inset(os);
1486                         } else
1487                                 handle_ert(os, ert, context);
1488                 }
1489
1490                 else if (t.cs() == "noindent") {
1491                         p.skip_spaces();
1492                         context.add_extra_stuff("\\noindent\n");
1493                 }
1494
1495                 else if (t.cs() == "appendix") {
1496                         context.add_extra_stuff("\\start_of_appendix\n");
1497                         // We need to start a new paragraph. Otherwise the
1498                         // appendix in 'bla\appendix\chapter{' would start
1499                         // too late.
1500                         context.new_paragraph(os);
1501                         // We need to make sure that the paragraph is
1502                         // generated even if it is empty. Otherwise the
1503                         // appendix in '\par\appendix\par\chapter{' would
1504                         // start too late.
1505                         context.check_layout(os);
1506                         // FIXME: This is a hack to prevent paragraph
1507                         // deletion if it is empty. Handle this better!
1508                         handle_comment(os,
1509                                 "%dummy comment inserted by tex2lyx to "
1510                                 "ensure that this paragraph is not empty",
1511                                 context);
1512                         // Both measures above may generate an additional
1513                         // empty paragraph, but that does not hurt, because
1514                         // whitespace does not matter here.
1515                         eat_whitespace(p, os, context, true);
1516                 }
1517
1518                 // Must attempt to parse "Section*" before "Section".
1519                 else if ((p.next_token().asInput() == "*") &&
1520                          context.new_layout_allowed &&
1521                          // The single '=' is meant here.
1522                          (newlayout = findLayout(context.textclass,
1523                                                  t.cs() + '*')).get() &&
1524                          newlayout->isCommand()) {
1525                         p.get_token();
1526                         output_command_layout(os, p, outer, context, newlayout);
1527                         p.skip_spaces();
1528                 }
1529
1530                 // The single '=' is meant here.
1531                 else if (context.new_layout_allowed &&
1532                          (newlayout = findLayout(context.textclass, t.cs())).get() &&
1533                          newlayout->isCommand()) {
1534                         output_command_layout(os, p, outer, context, newlayout);
1535                         p.skip_spaces();
1536                 }
1537
1538                 // Special handling for \caption
1539                 // FIXME: remove this when InsetCaption is supported.
1540                 else if (context.new_layout_allowed &&
1541                          t.cs() == captionlayout->latexname()) {
1542                         output_command_layout(os, p, outer, context,
1543                                               captionlayout);
1544                         p.skip_spaces();
1545                 }
1546
1547                 else if (t.cs() == "includegraphics") {
1548                         bool const clip = p.next_token().asInput() == "*";
1549                         if (clip)
1550                                 p.get_token();
1551                         map<string, string> opts = split_map(p.getArg('[', ']'));
1552                         if (clip)
1553                                 opts["clip"] = string();
1554                         string name = normalize_filename(p.verbatim_item());
1555
1556                         string const path = getMasterFilePath();
1557                         // We want to preserve relative / absolute filenames,
1558                         // therefore path is only used for testing
1559                         // FIXME UNICODE encoding of name and path may be
1560                         // wrong (makeAbsPath expects utf8)
1561                         if (!makeAbsPath(name, path).exists()) {
1562                                 // The file extension is probably missing.
1563                                 // Now try to find it out.
1564                                 string const dvips_name =
1565                                         find_file(name, path,
1566                                                   known_dvips_graphics_formats);
1567                                 string const pdftex_name =
1568                                         find_file(name, path,
1569                                                   known_pdftex_graphics_formats);
1570                                 if (!dvips_name.empty()) {
1571                                         if (!pdftex_name.empty()) {
1572                                                 cerr << "This file contains the "
1573                                                         "latex snippet\n"
1574                                                         "\"\\includegraphics{"
1575                                                      << name << "}\".\n"
1576                                                         "However, files\n\""
1577                                                      << dvips_name << "\" and\n\""
1578                                                      << pdftex_name << "\"\n"
1579                                                         "both exist, so I had to make a "
1580                                                         "choice and took the first one.\n"
1581                                                         "Please move the unwanted one "
1582                                                         "someplace else and try again\n"
1583                                                         "if my choice was wrong."
1584                                                      << endl;
1585                                         }
1586                                         name = dvips_name;
1587                                 } else if (!pdftex_name.empty())
1588                                         name = pdftex_name;
1589                         }
1590
1591                         // FIXME UNICODE encoding of name and path may be
1592                         // wrong (makeAbsPath expects utf8)
1593                         if (makeAbsPath(name, path).exists())
1594                                 fix_relative_filename(name);
1595                         else
1596                                 cerr << "Warning: Could not find graphics file '"
1597                                      << name << "'." << endl;
1598
1599                         context.check_layout(os);
1600                         begin_inset(os, "Graphics ");
1601                         os << "\n\tfilename " << name << '\n';
1602                         if (opts.find("width") != opts.end())
1603                                 os << "\twidth "
1604                                    << translate_len(opts["width"]) << '\n';
1605                         if (opts.find("height") != opts.end())
1606                                 os << "\theight "
1607                                    << translate_len(opts["height"]) << '\n';
1608                         if (opts.find("scale") != opts.end()) {
1609                                 istringstream iss(opts["scale"]);
1610                                 double val;
1611                                 iss >> val;
1612                                 val = val*100;
1613                                 os << "\tscale " << val << '\n';
1614                         }
1615                         if (opts.find("angle") != opts.end())
1616                                 os << "\trotateAngle "
1617                                    << opts["angle"] << '\n';
1618                         if (opts.find("origin") != opts.end()) {
1619                                 ostringstream ss;
1620                                 string const opt = opts["origin"];
1621                                 if (opt.find('l') != string::npos) ss << "left";
1622                                 if (opt.find('r') != string::npos) ss << "right";
1623                                 if (opt.find('c') != string::npos) ss << "center";
1624                                 if (opt.find('t') != string::npos) ss << "Top";
1625                                 if (opt.find('b') != string::npos) ss << "Bottom";
1626                                 if (opt.find('B') != string::npos) ss << "Baseline";
1627                                 if (!ss.str().empty())
1628                                         os << "\trotateOrigin " << ss.str() << '\n';
1629                                 else
1630                                         cerr << "Warning: Ignoring unknown includegraphics origin argument '" << opt << "'\n";
1631                         }
1632                         if (opts.find("keepaspectratio") != opts.end())
1633                                 os << "\tkeepAspectRatio\n";
1634                         if (opts.find("clip") != opts.end())
1635                                 os << "\tclip\n";
1636                         if (opts.find("draft") != opts.end())
1637                                 os << "\tdraft\n";
1638                         if (opts.find("bb") != opts.end())
1639                                 os << "\tBoundingBox "
1640                                    << opts["bb"] << '\n';
1641                         int numberOfbbOptions = 0;
1642                         if (opts.find("bbllx") != opts.end())
1643                                 numberOfbbOptions++;
1644                         if (opts.find("bblly") != opts.end())
1645                                 numberOfbbOptions++;
1646                         if (opts.find("bburx") != opts.end())
1647                                 numberOfbbOptions++;
1648                         if (opts.find("bbury") != opts.end())
1649                                 numberOfbbOptions++;
1650                         if (numberOfbbOptions == 4)
1651                                 os << "\tBoundingBox "
1652                                    << opts["bbllx"] << opts["bblly"]
1653                                    << opts["bburx"] << opts["bbury"] << '\n';
1654                         else if (numberOfbbOptions > 0)
1655                                 cerr << "Warning: Ignoring incomplete includegraphics boundingbox arguments.\n";
1656                         numberOfbbOptions = 0;
1657                         if (opts.find("natwidth") != opts.end())
1658                                 numberOfbbOptions++;
1659                         if (opts.find("natheight") != opts.end())
1660                                 numberOfbbOptions++;
1661                         if (numberOfbbOptions == 2)
1662                                 os << "\tBoundingBox 0bp 0bp "
1663                                    << opts["natwidth"] << opts["natheight"] << '\n';
1664                         else if (numberOfbbOptions > 0)
1665                                 cerr << "Warning: Ignoring incomplete includegraphics boundingbox arguments.\n";
1666                         ostringstream special;
1667                         if (opts.find("hiresbb") != opts.end())
1668                                 special << "hiresbb,";
1669                         if (opts.find("trim") != opts.end())
1670                                 special << "trim,";
1671                         if (opts.find("viewport") != opts.end())
1672                                 special << "viewport=" << opts["viewport"] << ',';
1673                         if (opts.find("totalheight") != opts.end())
1674                                 special << "totalheight=" << opts["totalheight"] << ',';
1675                         if (opts.find("type") != opts.end())
1676                                 special << "type=" << opts["type"] << ',';
1677                         if (opts.find("ext") != opts.end())
1678                                 special << "ext=" << opts["ext"] << ',';
1679                         if (opts.find("read") != opts.end())
1680                                 special << "read=" << opts["read"] << ',';
1681                         if (opts.find("command") != opts.end())
1682                                 special << "command=" << opts["command"] << ',';
1683                         string s_special = special.str();
1684                         if (!s_special.empty()) {
1685                                 // We had special arguments. Remove the trailing ','.
1686                                 os << "\tspecial " << s_special.substr(0, s_special.size() - 1) << '\n';
1687                         }
1688                         // TODO: Handle the unknown settings better.
1689                         // Warn about invalid options.
1690                         // Check whether some option was given twice.
1691                         end_inset(os);
1692                 }
1693
1694                 else if (t.cs() == "footnote" ||
1695                          (t.cs() == "thanks" && context.layout->intitle)) {
1696                         p.skip_spaces();
1697                         context.check_layout(os);
1698                         begin_inset(os, "Foot\n");
1699                         os << "status collapsed\n\n";
1700                         parse_text_in_inset(p, os, FLAG_ITEM, false, context);
1701                         end_inset(os);
1702                 }
1703
1704                 else if (t.cs() == "marginpar") {
1705                         p.skip_spaces();
1706                         context.check_layout(os);
1707                         begin_inset(os, "Marginal\n");
1708                         os << "status collapsed\n\n";
1709                         parse_text_in_inset(p, os, FLAG_ITEM, false, context);
1710                         end_inset(os);
1711                 }
1712
1713                 else if (t.cs() == "ensuremath") {
1714                         p.skip_spaces();
1715                         context.check_layout(os);
1716                         string const s = p.verbatim_item();
1717                         if (s == "\xb1" || s == "\xb3" || s == "\xb2" || s == "\xb5")
1718                                 os << s;
1719                         else
1720                                 handle_ert(os, "\\ensuremath{" + s + "}",
1721                                            context);
1722                 }
1723
1724                 else if (t.cs() == "hfill") {
1725                         context.check_layout(os);
1726                         os << "\n\\hfill\n";
1727                         skip_braces(p);
1728                         p.skip_spaces();
1729                 }
1730
1731                 else if (t.cs() == "makeindex" || t.cs() == "maketitle") {
1732                         // FIXME: Somehow prevent title layouts if
1733                         // "maketitle" was not found
1734                         p.skip_spaces();
1735                         skip_braces(p); // swallow this
1736                 }
1737
1738                 else if (t.cs() == "tableofcontents") {
1739                         p.skip_spaces();
1740                         context.check_layout(os);
1741                         begin_inset(os, "LatexCommand \\tableofcontents\n");
1742                         end_inset(os);
1743                         skip_braces(p); // swallow this
1744                 }
1745
1746                 else if (t.cs() == "listoffigures") {
1747                         p.skip_spaces();
1748                         context.check_layout(os);
1749                         begin_inset(os, "FloatList figure\n");
1750                         end_inset(os);
1751                         skip_braces(p); // swallow this
1752                 }
1753
1754                 else if (t.cs() == "listoftables") {
1755                         p.skip_spaces();
1756                         context.check_layout(os);
1757                         begin_inset(os, "FloatList table\n");
1758                         end_inset(os);
1759                         skip_braces(p); // swallow this
1760                 }
1761
1762                 else if (t.cs() == "listof") {
1763                         p.skip_spaces(true);
1764                         string const name = p.get_token().asString();
1765                         if (context.textclass.floats().typeExist(name)) {
1766                                 context.check_layout(os);
1767                                 begin_inset(os, "FloatList ");
1768                                 os << name << "\n";
1769                                 end_inset(os);
1770                                 p.get_token(); // swallow second arg
1771                         } else
1772                                 handle_ert(os, "\\listof{" + name + "}", context);
1773                 }
1774
1775                 else if (t.cs() == "textrm")
1776                         parse_text_attributes(p, os, FLAG_ITEM, outer,
1777                                               context, "\\family",
1778                                               context.font.family, "roman");
1779
1780                 else if (t.cs() == "textsf")
1781                         parse_text_attributes(p, os, FLAG_ITEM, outer,
1782                                               context, "\\family",
1783                                               context.font.family, "sans");
1784
1785                 else if (t.cs() == "texttt")
1786                         parse_text_attributes(p, os, FLAG_ITEM, outer,
1787                                               context, "\\family",
1788                                               context.font.family, "typewriter");
1789
1790                 else if (t.cs() == "textmd")
1791                         parse_text_attributes(p, os, FLAG_ITEM, outer,
1792                                               context, "\\series",
1793                                               context.font.series, "medium");
1794
1795                 else if (t.cs() == "textbf")
1796                         parse_text_attributes(p, os, FLAG_ITEM, outer,
1797                                               context, "\\series",
1798                                               context.font.series, "bold");
1799
1800                 else if (t.cs() == "textup")
1801                         parse_text_attributes(p, os, FLAG_ITEM, outer,
1802                                               context, "\\shape",
1803                                               context.font.shape, "up");
1804
1805                 else if (t.cs() == "textit")
1806                         parse_text_attributes(p, os, FLAG_ITEM, outer,
1807                                               context, "\\shape",
1808                                               context.font.shape, "italic");
1809
1810                 else if (t.cs() == "textsl")
1811                         parse_text_attributes(p, os, FLAG_ITEM, outer,
1812                                               context, "\\shape",
1813                                               context.font.shape, "slanted");
1814
1815                 else if (t.cs() == "textsc")
1816                         parse_text_attributes(p, os, FLAG_ITEM, outer,
1817                                               context, "\\shape",
1818                                               context.font.shape, "smallcaps");
1819
1820                 else if (t.cs() == "textnormal" || t.cs() == "normalfont") {
1821                         context.check_layout(os);
1822                         TeXFont oldFont = context.font;
1823                         context.font.init();
1824                         context.font.size = oldFont.size;
1825                         os << "\n\\family " << context.font.family << "\n";
1826                         os << "\n\\series " << context.font.series << "\n";
1827                         os << "\n\\shape " << context.font.shape << "\n";
1828                         if (t.cs() == "textnormal") {
1829                                 parse_text_snippet(p, os, FLAG_ITEM, outer, context);
1830                                 output_font_change(os, context.font, oldFont);
1831                                 context.font = oldFont;
1832                         } else
1833                                 eat_whitespace(p, os, context, false);
1834                 }
1835
1836                 else if (t.cs() == "underbar") {
1837                         // Do NOT handle \underline.
1838                         // \underbar cuts through y, g, q, p etc.,
1839                         // \underline does not.
1840                         context.check_layout(os);
1841                         os << "\n\\bar under\n";
1842                         parse_text_snippet(p, os, FLAG_ITEM, outer, context);
1843                         context.check_layout(os);
1844                         os << "\n\\bar default\n";
1845                 }
1846
1847                 else if (t.cs() == "emph" || t.cs() == "noun") {
1848                         context.check_layout(os);
1849                         os << "\n\\" << t.cs() << " on\n";
1850                         parse_text_snippet(p, os, FLAG_ITEM, outer, context);
1851                         context.check_layout(os);
1852                         os << "\n\\" << t.cs() << " default\n";
1853                 }
1854
1855                 else if (use_natbib &&
1856                          is_known(t.cs(), known_natbib_commands) &&
1857                          ((t.cs() != "citefullauthor" &&
1858                            t.cs() != "citeyear" &&
1859                            t.cs() != "citeyearpar") ||
1860                           p.next_token().asInput() != "*")) {
1861                         context.check_layout(os);
1862                         // tex                       lyx
1863                         // \citet[before][after]{a}  \citet[after][before]{a}
1864                         // \citet[before][]{a}       \citet[][before]{a}
1865                         // \citet[after]{a}          \citet[after]{a}
1866                         // \citet{a}                 \citet{a}
1867                         string command = '\\' + t.cs();
1868                         if (p.next_token().asInput() == "*") {
1869                                 command += '*';
1870                                 p.get_token();
1871                         }
1872                         if (command == "\\citefullauthor")
1873                                 // alternative name for "\\citeauthor*"
1874                                 command = "\\citeauthor*";
1875
1876                         // text before the citation
1877                         string before;
1878                         // text after the citation
1879                         string after;
1880
1881                         boost::tie(before, after) = getCiteArguments(p, true);
1882                         if (command == "\\cite") {
1883                                 // \cite without optional argument means
1884                                 // \citet, \cite with at least one optional
1885                                 // argument means \citep.
1886                                 if (before.empty() && after.empty())
1887                                         command = "\\citet";
1888                                 else
1889                                         command = "\\citep";
1890                         }
1891                         if (before.empty() && after == "[]")
1892                                 // avoid \citet[]{a}
1893                                 after.erase();
1894                         else if (before == "[]" && after == "[]") {
1895                                 // avoid \citet[][]{a}
1896                                 before.erase();
1897                                 after.erase();
1898                         }
1899                         begin_inset(os, "LatexCommand ");
1900                         os << command << after << before
1901                            << '{' << p.verbatim_item() << "}\n";
1902                         end_inset(os);
1903                 }
1904
1905                 else if (use_jurabib &&
1906                          is_known(t.cs(), known_jurabib_commands)) {
1907                         context.check_layout(os);
1908                         string const command = '\\' + t.cs();
1909                         char argumentOrder = '\0';
1910                         vector<string> const & options = used_packages["jurabib"];
1911                         if (std::find(options.begin(), options.end(),
1912                                       "natbiborder") != options.end())
1913                                 argumentOrder = 'n';
1914                         else if (std::find(options.begin(), options.end(),
1915                                            "jurabiborder") != options.end())
1916                                 argumentOrder = 'j';
1917
1918                         // text before the citation
1919                         string before;
1920                         // text after the citation
1921                         string after;
1922
1923                         boost::tie(before, after) =
1924                                 getCiteArguments(p, argumentOrder != 'j');
1925                         string const citation = p.verbatim_item();
1926                         if (!before.empty() && argumentOrder == '\0') {
1927                                 cerr << "Warning: Assuming argument order "
1928                                         "of jurabib version 0.6 for\n'"
1929                                      << command << before << after << '{'
1930                                      << citation << "}'.\n"
1931                                         "Add 'jurabiborder' to the jurabib "
1932                                         "package options if you used an\n"
1933                                         "earlier jurabib version." << endl;
1934                         }
1935                         begin_inset(os, "LatexCommand ");
1936                         os << command << after << before
1937                            << '{' << citation << "}\n";
1938                         end_inset(os);
1939                 }
1940
1941                 else if (is_known(t.cs(), known_latex_commands)) {
1942                         // This needs to be after the check for natbib and
1943                         // jurabib commands, because "cite" has different
1944                         // arguments with natbib and jurabib.
1945                         context.check_layout(os);
1946                         begin_inset(os, "LatexCommand ");
1947                         os << '\\' << t.cs();
1948                         // lyx cannot handle newlines in a latex command
1949                         // FIXME: Move the substitution into parser::getOpt()?
1950                         os << subst(p.getOpt(), "\n", " ");
1951                         os << subst(p.getOpt(), "\n", " ");
1952                         os << '{' << subst(p.verbatim_item(), "\n", " ") << "}\n";
1953                         end_inset(os);
1954                 }
1955
1956                 else if (is_known(t.cs(), known_quotes)) {
1957                         char const * const * where = is_known(t.cs(), known_quotes);
1958                         context.check_layout(os);
1959                         begin_inset(os, "Quotes ");
1960                         os << known_coded_quotes[where - known_quotes];
1961                         end_inset(os);
1962                         // LyX adds {} after the quote, so we have to eat
1963                         // spaces here if there are any before a possible
1964                         // {} pair.
1965                         eat_whitespace(p, os, context, false);
1966                         skip_braces(p);
1967                 }
1968
1969                 else if (is_known(t.cs(), known_sizes) &&
1970                          context.new_layout_allowed) {
1971                         char const * const * where = is_known(t.cs(), known_sizes);
1972                         context.check_layout(os);
1973                         TeXFont const oldFont = context.font;
1974                         context.font.size = known_coded_sizes[where - known_sizes];
1975                         output_font_change(os, oldFont, context.font);
1976                         eat_whitespace(p, os, context, false);
1977                 }
1978
1979                 else if (is_known(t.cs(), known_font_families) &&
1980                          context.new_layout_allowed) {
1981                         char const * const * where =
1982                                 is_known(t.cs(), known_font_families);
1983                         context.check_layout(os);
1984                         TeXFont const oldFont = context.font;
1985                         context.font.family =
1986                                 known_coded_font_families[where - known_font_families];
1987                         output_font_change(os, oldFont, context.font);
1988                         eat_whitespace(p, os, context, false);
1989                 }
1990
1991                 else if (is_known(t.cs(), known_font_series) &&
1992                          context.new_layout_allowed) {
1993                         char const * const * where =
1994                                 is_known(t.cs(), known_font_series);
1995                         context.check_layout(os);
1996                         TeXFont const oldFont = context.font;
1997                         context.font.series =
1998                                 known_coded_font_series[where - known_font_series];
1999                         output_font_change(os, oldFont, context.font);
2000                         eat_whitespace(p, os, context, false);
2001                 }
2002
2003                 else if (is_known(t.cs(), known_font_shapes) &&
2004                          context.new_layout_allowed) {
2005                         char const * const * where =
2006                                 is_known(t.cs(), known_font_shapes);
2007                         context.check_layout(os);
2008                         TeXFont const oldFont = context.font;
2009                         context.font.shape =
2010                                 known_coded_font_shapes[where - known_font_shapes];
2011                         output_font_change(os, oldFont, context.font);
2012                         eat_whitespace(p, os, context, false);
2013                 }
2014                 else if (is_known(t.cs(), known_old_font_families) &&
2015                          context.new_layout_allowed) {
2016                         char const * const * where =
2017                                 is_known(t.cs(), known_old_font_families);
2018                         context.check_layout(os);
2019                         TeXFont const oldFont = context.font;
2020                         context.font.init();
2021                         context.font.size = oldFont.size;
2022                         context.font.family =
2023                                 known_coded_font_families[where - known_old_font_families];
2024                         output_font_change(os, oldFont, context.font);
2025                         eat_whitespace(p, os, context, false);
2026                 }
2027
2028                 else if (is_known(t.cs(), known_old_font_series) &&
2029                          context.new_layout_allowed) {
2030                         char const * const * where =
2031                                 is_known(t.cs(), known_old_font_series);
2032                         context.check_layout(os);
2033                         TeXFont const oldFont = context.font;
2034                         context.font.init();
2035                         context.font.size = oldFont.size;
2036                         context.font.series =
2037                                 known_coded_font_series[where - known_old_font_series];
2038                         output_font_change(os, oldFont, context.font);
2039                         eat_whitespace(p, os, context, false);
2040                 }
2041
2042                 else if (is_known(t.cs(), known_old_font_shapes) &&
2043                          context.new_layout_allowed) {
2044                         char const * const * where =
2045                                 is_known(t.cs(), known_old_font_shapes);
2046                         context.check_layout(os);
2047                         TeXFont const oldFont = context.font;
2048                         context.font.init();
2049                         context.font.size = oldFont.size;
2050                         context.font.shape =
2051                                 known_coded_font_shapes[where - known_old_font_shapes];
2052                         output_font_change(os, oldFont, context.font);
2053                         eat_whitespace(p, os, context, false);
2054                 }
2055
2056                 else if (t.cs() == "LyX" || t.cs() == "TeX"
2057                          || t.cs() == "LaTeX") {
2058                         context.check_layout(os);
2059                         os << t.cs();
2060                         skip_braces(p); // eat {}
2061                 }
2062
2063                 else if (t.cs() == "LaTeXe") {
2064                         context.check_layout(os);
2065                         os << "LaTeX2e";
2066                         skip_braces(p); // eat {}
2067                 }
2068
2069                 else if (t.cs() == "ldots") {
2070                         context.check_layout(os);
2071                         skip_braces(p);
2072                         os << "\\SpecialChar \\ldots{}\n";
2073                 }
2074
2075                 else if (t.cs() == "lyxarrow") {
2076                         context.check_layout(os);
2077                         os << "\\SpecialChar \\menuseparator\n";
2078                         skip_braces(p);
2079                 }
2080
2081                 else if (t.cs() == "textcompwordmark") {
2082                         context.check_layout(os);
2083                         os << "\\SpecialChar \\textcompwordmark{}\n";
2084                         skip_braces(p);
2085                 }
2086
2087                 else if (t.cs() == "@" && p.next_token().asInput() == ".") {
2088                         context.check_layout(os);
2089                         os << "\\SpecialChar \\@.\n";
2090                         p.get_token();
2091                 }
2092
2093                 else if (t.cs() == "-") {
2094                         context.check_layout(os);
2095                         os << "\\SpecialChar \\-\n";
2096                 }
2097
2098                 else if (t.cs() == "textasciitilde") {
2099                         context.check_layout(os);
2100                         os << '~';
2101                         skip_braces(p);
2102                 }
2103
2104                 else if (t.cs() == "textasciicircum") {
2105                         context.check_layout(os);
2106                         os << '^';
2107                         skip_braces(p);
2108                 }
2109
2110                 else if (t.cs() == "textbackslash") {
2111                         context.check_layout(os);
2112                         os << "\n\\backslash\n";
2113                         skip_braces(p);
2114                 }
2115
2116                 else if (t.cs() == "_" || t.cs() == "&" || t.cs() == "#"
2117                             || t.cs() == "$" || t.cs() == "{" || t.cs() == "}"
2118                             || t.cs() == "%") {
2119                         context.check_layout(os);
2120                         os << t.cs();
2121                 }
2122
2123                 else if (t.cs() == "char") {
2124                         context.check_layout(os);
2125                         if (p.next_token().character() == '`') {
2126                                 p.get_token();
2127                                 if (p.next_token().cs() == "\"") {
2128                                         p.get_token();
2129                                         os << '"';
2130                                         skip_braces(p);
2131                                 } else {
2132                                         handle_ert(os, "\\char`", context);
2133                                 }
2134                         } else {
2135                                 handle_ert(os, "\\char", context);
2136                         }
2137                 }
2138
2139                 else if (t.cs() == "verb") {
2140                         context.check_layout(os);
2141                         char const delimiter = p.next_token().character();
2142                         string const arg = p.getArg(delimiter, delimiter);
2143                         ostringstream oss;
2144                         oss << "\\verb" << delimiter << arg << delimiter;
2145                         handle_ert(os, oss.str(), context);
2146                 }
2147
2148                 else if (t.cs() == "\"") {
2149                         context.check_layout(os);
2150                         string const name = p.verbatim_item();
2151                              if (name == "a") os << '\xe4';
2152                         else if (name == "o") os << '\xf6';
2153                         else if (name == "u") os << '\xfc';
2154                         else if (name == "A") os << '\xc4';
2155                         else if (name == "O") os << '\xd6';
2156                         else if (name == "U") os << '\xdc';
2157                         else handle_ert(os, "\"{" + name + "}", context);
2158                 }
2159
2160                 // Problem: \= creates a tabstop inside the tabbing environment
2161                 // and else an accent. In the latter case we really would want
2162                 // \={o} instead of \= o.
2163                 else if (t.cs() == "=" && (flags & FLAG_TABBING))
2164                         handle_ert(os, t.asInput(), context);
2165
2166                 else if (t.cs() == "H" || t.cs() == "c" || t.cs() == "^"
2167                          || t.cs() == "'" || t.cs() == "`"
2168                          || t.cs() == "~" || t.cs() == "." || t.cs() == "=") {
2169                         // we need the trim as the LyX parser chokes on such spaces
2170                         // The argument of InsetLatexAccent is parsed as a
2171                         // subset of LaTeX, so don't parse anything here,
2172                         // but use the raw argument.
2173                         // Otherwise we would convert \~{\i} wrongly.
2174                         // This will of course not translate \~{\ss} to \~{ß},
2175                         // but that does at least compile and does only look
2176                         // strange on screen.
2177                         context.check_layout(os);
2178                         os << "\\i \\" << t.cs() << "{"
2179                            << trim(p.verbatim_item(), " ")
2180                            << "}\n";
2181                 }
2182
2183                 else if (t.cs() == "ss") {
2184                         context.check_layout(os);
2185                         os << "\xdf";
2186                         skip_braces(p); // eat {}
2187                 }
2188
2189                 else if (t.cs() == "i" || t.cs() == "j" || t.cs() == "l" ||
2190                          t.cs() == "L") {
2191                         context.check_layout(os);
2192                         os << "\\i \\" << t.cs() << "{}\n";
2193                         skip_braces(p); // eat {}
2194                 }
2195
2196                 else if (t.cs() == "\\") {
2197                         context.check_layout(os);
2198                         string const next = p.next_token().asInput();
2199                         if (next == "[")
2200                                 handle_ert(os, "\\\\" + p.getOpt(), context);
2201                         else if (next == "*") {
2202                                 p.get_token();
2203                                 handle_ert(os, "\\\\*" + p.getOpt(), context);
2204                         }
2205                         else {
2206                                 os << "\n\\newline\n";
2207                         }
2208                 }
2209
2210                 else if (t.cs() == "newline") {
2211                         context.check_layout(os);
2212                         os << "\n\\" << t.cs() << "\n";
2213                         skip_braces(p); // eat {}
2214                 }
2215
2216                 else if (t.cs() == "href") {
2217                         context.check_layout(os);
2218                         begin_inset(os, "CommandInset ");
2219                         os << t.cs() << "\n";
2220                         os << "LatexCommand " << t.cs() << "\n";
2221                         bool erase = false;
2222                         size_t pos;
2223                         // the first argument is "type:target", "type:" is optional
2224                         // the second argument the name
2225                         string href_target = subst(p.verbatim_item(), "\n", " ");
2226                         string href_name = subst(p.verbatim_item(), "\n", " ");
2227                         string href_type;
2228                         // serach for the ":" to divide type from target
2229                         if ((pos = href_target.find(":", 0)) != string::npos){
2230                                 href_type = href_target;
2231                                 href_type.erase(pos + 1, href_type.length());
2232                                 href_target.erase(0, pos + 1);
2233                             erase = true;
2234                         }
2235                         os << "name " << '"' << href_name << '"' << "\n";
2236                         os << "target " << '"' << href_target << '"' << "\n";
2237                         if(erase)
2238                                 os << "type " << '"' << href_type << '"' << "\n";
2239                         end_inset(os);
2240                 }
2241
2242                 else if (t.cs() == "input" || t.cs() == "include"
2243                          || t.cs() == "verbatiminput") {
2244                         string name = '\\' + t.cs();
2245                         if (t.cs() == "verbatiminput"
2246                             && p.next_token().asInput() == "*")
2247                                 name += p.get_token().asInput();
2248                         context.check_layout(os);
2249                         begin_inset(os, "Include ");
2250                         string filename(normalize_filename(p.getArg('{', '}')));
2251                         string const path = getMasterFilePath();
2252                         // We want to preserve relative / absolute filenames,
2253                         // therefore path is only used for testing
2254                         // FIXME UNICODE encoding of filename and path may be
2255                         // wrong (makeAbsPath expects utf8)
2256                         if ((t.cs() == "include" || t.cs() == "input") &&
2257                             !makeAbsPath(filename, path).exists()) {
2258                                 // The file extension is probably missing.
2259                                 // Now try to find it out.
2260                                 string const tex_name =
2261                                         find_file(filename, path,
2262                                                   known_tex_extensions);
2263                                 if (!tex_name.empty())
2264                                         filename = tex_name;
2265                         }
2266                         // FIXME UNICODE encoding of filename and path may be
2267                         // wrong (makeAbsPath expects utf8)
2268                         if (makeAbsPath(filename, path).exists()) {
2269                                 string const abstexname =
2270                                         makeAbsPath(filename, path).absFilename();
2271                                 string const abslyxname =
2272                                         changeExtension(abstexname, ".lyx");
2273                                 fix_relative_filename(filename);
2274                                 string const lyxname =
2275                                         changeExtension(filename, ".lyx");
2276                                 if (t.cs() != "verbatiminput" &&
2277                                     tex2lyx(abstexname, FileName(abslyxname))) {
2278                                         os << name << '{' << lyxname << "}\n";
2279                                 } else {
2280                                         os << name << '{' << filename << "}\n";
2281                                 }
2282                         } else {
2283                                 cerr << "Warning: Could not find included file '"
2284                                      << filename << "'." << endl;
2285                                 os << name << '{' << filename << "}\n";
2286                         }
2287                         os << "preview false\n";
2288                         end_inset(os);
2289                 }
2290
2291                 else if (t.cs() == "bibliographystyle") {
2292                         // store new bibliographystyle
2293                         bibliographystyle = p.verbatim_item();
2294                         // output new bibliographystyle.
2295                         // This is only necessary if used in some other macro than \bibliography.
2296                         handle_ert(os, "\\bibliographystyle{" + bibliographystyle + "}", context);
2297                 }
2298
2299                 else if (t.cs() == "bibliography") {
2300                         context.check_layout(os);
2301                         begin_inset(os, "LatexCommand ");
2302                         os << "\\bibtex";
2303                         // Do we have a bibliographystyle set?
2304                         if (!bibliographystyle.empty()) {
2305                                 os << '[' << bibliographystyle << ']';
2306                         }
2307                         os << '{' << p.verbatim_item() << "}\n";
2308                         end_inset(os);
2309                 }
2310
2311                 else if (t.cs() == "parbox")
2312                         parse_box(p, os, FLAG_ITEM, outer, context, true);
2313
2314                 else if (t.cs() == "smallskip" ||
2315                          t.cs() == "medskip" ||
2316                          t.cs() == "bigskip" ||
2317                          t.cs() == "vfill") {
2318                         context.check_layout(os);
2319                         begin_inset(os, "VSpace ");
2320                         os << t.cs();
2321                         end_inset(os);
2322                         skip_braces(p);
2323                 }
2324
2325                 else if (is_known(t.cs(), known_spaces)) {
2326                         char const * const * where = is_known(t.cs(), known_spaces);
2327                         context.check_layout(os);
2328                         begin_inset(os, "InsetSpace ");
2329                         os << '\\' << known_coded_spaces[where - known_spaces]
2330                            << '\n';
2331                         // LaTeX swallows whitespace after all spaces except
2332                         // "\\,". We have to do that here, too, because LyX
2333                         // adds "{}" which would make the spaces significant.
2334                         if (t.cs() !=  ",")
2335                                 eat_whitespace(p, os, context, false);
2336                         // LyX adds "{}" after all spaces except "\\ " and
2337                         // "\\,", so we have to remove "{}".
2338                         // "\\,{}" is equivalent to "\\," in LaTeX, so we
2339                         // remove the braces after "\\,", too.
2340                         if (t.cs() != " ")
2341                                 skip_braces(p);
2342                 }
2343
2344                 else if (t.cs() == "newpage" ||
2345                         t.cs() == "pagebreak" ||
2346                         t.cs() == "clearpage" ||
2347                         t.cs() == "cleardoublepage") {
2348                         context.check_layout(os);
2349                         os << "\n\\" << t.cs() << "\n";
2350                         skip_braces(p); // eat {}
2351                 }
2352
2353                 else if (t.cs() == "newcommand" ||
2354                          t.cs() == "providecommand" ||
2355                          t.cs() == "renewcommand") {
2356                         // these could be handled by parse_command(), but
2357                         // we need to call add_known_command() here.
2358                         string name = t.asInput();
2359                         if (p.next_token().asInput() == "*") {
2360                                 // Starred form. Eat '*'
2361                                 p.get_token();
2362                                 name += '*';
2363                         }
2364                         string const command = p.verbatim_item();
2365                         string const opt1 = p.getOpt();
2366                         string const opt2 = p.getFullOpt();
2367                         add_known_command(command, opt1, !opt2.empty());
2368                         string const ert = name + '{' + command + '}' +
2369                                            opt1 + opt2 +
2370                                            '{' + p.verbatim_item() + '}';
2371
2372                         context.check_layout(os);
2373                         begin_inset(os, "FormulaMacro");
2374                         os << "\n" << ert;
2375                         end_inset(os);
2376                 }
2377
2378                 else if (t.cs() == "vspace") {
2379                         bool starred = false;
2380                         if (p.next_token().asInput() == "*") {
2381                                 p.get_token();
2382                                 starred = true;
2383                         }
2384                         string const length = p.verbatim_item();
2385                         string unit;
2386                         string valstring;
2387                         bool valid = splitLatexLength(length, valstring, unit);
2388                         bool known_vspace = false;
2389                         bool known_unit = false;
2390                         double value;
2391                         if (valid) {
2392                                 istringstream iss(valstring);
2393                                 iss >> value;
2394                                 if (value == 1.0) {
2395                                         if (unit == "\\smallskipamount") {
2396                                                 unit = "smallskip";
2397                                                 known_vspace = true;
2398                                         } else if (unit == "\\medskipamount") {
2399                                                 unit = "medskip";
2400                                                 known_vspace = true;
2401                                         } else if (unit == "\\bigskipamount") {
2402                                                 unit = "bigskip";
2403                                                 known_vspace = true;
2404                                         } else if (unit == "\\fill") {
2405                                                 unit = "vfill";
2406                                                 known_vspace = true;
2407                                         }
2408                                 }
2409                                 if (!known_vspace) {
2410                                         switch (unitFromString(unit)) {
2411                                         case Length::SP:
2412                                         case Length::PT:
2413                                         case Length::BP:
2414                                         case Length::DD:
2415                                         case Length::MM:
2416                                         case Length::PC:
2417                                         case Length::CC:
2418                                         case Length::CM:
2419                                         case Length::IN:
2420                                         case Length::EX:
2421                                         case Length::EM:
2422                                         case Length::MU:
2423                                                 known_unit = true;
2424                                                 break;
2425                                         default:
2426                                                 break;
2427                                         }
2428                                 }
2429                         }
2430
2431                         if (known_unit || known_vspace) {
2432                                 // Literal length or known variable
2433                                 context.check_layout(os);
2434                                 begin_inset(os, "VSpace ");
2435                                 if (known_unit)
2436                                         os << value;
2437                                 os << unit;
2438                                 if (starred)
2439                                         os << '*';
2440                                 end_inset(os);
2441                         } else {
2442                                 // LyX can't handle other length variables in Inset VSpace
2443                                 string name = t.asInput();
2444                                 if (starred)
2445                                         name += '*';
2446                                 if (valid) {
2447                                         if (value == 1.0)
2448                                                 handle_ert(os, name + '{' + unit + '}', context);
2449                                         else if (value == -1.0)
2450                                                 handle_ert(os, name + "{-" + unit + '}', context);
2451                                         else
2452                                                 handle_ert(os, name + '{' + valstring + unit + '}', context);
2453                                 } else
2454                                         handle_ert(os, name + '{' + length + '}', context);
2455                         }
2456                 }
2457
2458                 else {
2459                         //cerr << "#: " << t << " mode: " << mode << endl;
2460                         // heuristic: read up to next non-nested space
2461                         /*
2462                         string s = t.asInput();
2463                         string z = p.verbatim_item();
2464                         while (p.good() && z != " " && z.size()) {
2465                                 //cerr << "read: " << z << endl;
2466                                 s += z;
2467                                 z = p.verbatim_item();
2468                         }
2469                         cerr << "found ERT: " << s << endl;
2470                         handle_ert(os, s + ' ', context);
2471                         */
2472                         string name = t.asInput();
2473                         if (p.next_token().asInput() == "*") {
2474                                 // Starred commands like \vspace*{}
2475                                 p.get_token();                          // Eat '*'
2476                                 name += '*';
2477                         }
2478                         if (! parse_command(name, p, os, outer, context))
2479                                 handle_ert(os, name, context);
2480                 }
2481
2482                 if (flags & FLAG_LEAVE) {
2483                         flags &= ~FLAG_LEAVE;
2484                         break;
2485                 }
2486         }
2487 }
2488
2489 // }])
2490
2491
2492 } // namespace lyx