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.
7 * \author Jean-Marc Lasgouttes
10 * Full author contact details are available in file CREDITS.
21 #include "FloatList.h"
25 #include "support/lassert.h"
26 #include "support/convert.h"
27 #include "support/FileName.h"
28 #include "support/filetools.h"
29 #include "support/lstrings.h"
38 using namespace lyx::support;
43 void parse_text_in_inset(Parser & p, ostream & os, unsigned flags, bool outer,
44 Context const & context, InsetLayout const * layout)
46 bool const forcePlainLayout =
47 layout ? layout->forcePlainLayout() : false;
48 Context newcontext(true, context.textclass);
50 newcontext.layout = &context.textclass.plainLayout();
52 newcontext.font = context.font;
53 parse_text(p, os, flags, outer, newcontext);
54 newcontext.check_end_layout(os);
60 void parse_text_in_inset(Parser & p, ostream & os, unsigned flags, bool outer,
61 Context const & context, string const & name)
63 InsetLayout const * layout = 0;
64 DocumentClass::InsetLayouts::const_iterator it =
65 context.textclass.insetLayouts().find(from_ascii(name));
66 if (it != context.textclass.insetLayouts().end())
67 layout = &(it->second);
68 parse_text_in_inset(p, os, flags, outer, context, layout);
71 /// parses a paragraph snippet, useful for example for \\emph{...}
72 void parse_text_snippet(Parser & p, ostream & os, unsigned flags, bool outer,
75 Context newcontext(context);
76 // Don't inherit the paragraph-level extra stuff
77 newcontext.par_extra_stuff.clear();
78 parse_text(p, os, flags, outer, newcontext);
79 // Make sure that we don't create invalid .lyx files
80 context.need_layout = newcontext.need_layout;
81 context.need_end_layout = newcontext.need_end_layout;
86 * Thin wrapper around parse_text_snippet() using a string.
88 * We completely ignore \c context.need_layout and \c context.need_end_layout,
89 * because our return value is not used directly (otherwise the stream version
90 * of parse_text_snippet() could be used). That means that the caller needs
91 * to do layout management manually.
92 * This is intended to parse text that does not create any layout changes.
94 string parse_text_snippet(Parser & p, unsigned flags, const bool outer,
97 Context newcontext(context);
98 newcontext.need_layout = false;
99 newcontext.need_end_layout = false;
100 newcontext.new_layout_allowed = false;
101 // Avoid warning by Context::~Context()
102 newcontext.par_extra_stuff.clear();
104 parse_text_snippet(p, os, flags, outer, newcontext);
109 char const * const known_ref_commands[] = { "ref", "pageref", "vref",
110 "vpageref", "prettyref", "eqref", 0 };
114 * The starred forms are also known except for "citefullauthor",
115 * "citeyear" and "citeyearpar".
117 char const * const known_natbib_commands[] = { "cite", "citet", "citep",
118 "citealt", "citealp", "citeauthor", "citeyear", "citeyearpar",
119 "citefullauthor", "Citet", "Citep", "Citealt", "Citealp", "Citeauthor", 0 };
123 * No starred form other than "cite*" known.
125 char const * const known_jurabib_commands[] = { "cite", "citet", "citep",
126 "citealt", "citealp", "citeauthor", "citeyear", "citeyearpar",
127 // jurabib commands not (yet) supported by LyX:
129 // "footcite", "footcitet", "footcitep", "footcitealt", "footcitealp",
130 // "footciteauthor", "footciteyear", "footciteyearpar",
131 "citefield", "citetitle", 0 };
133 /// LaTeX names for quotes
134 char const * const known_quotes[] = { "dq", "guillemotleft", "flqq", "og",
135 "guillemotright", "frqq", "fg", "glq", "glqq", "textquoteleft", "grq", "grqq",
136 "quotedblbase", "textquotedblleft", "quotesinglbase", "textquoteright", "flq",
137 "guilsinglleft", "frq", "guilsinglright", 0};
139 /// the same as known_quotes with .lyx names
140 char const * const known_coded_quotes[] = { "prd", "ard", "ard", "ard",
141 "ald", "ald", "ald", "gls", "gld", "els", "els", "grd",
142 "gld", "grd", "gls", "ers", "fls",
143 "fls", "frs", "frs", 0};
145 /// LaTeX names for font sizes
146 char const * const known_sizes[] = { "tiny", "scriptsize", "footnotesize",
147 "small", "normalsize", "large", "Large", "LARGE", "huge", "Huge", 0};
149 /// the same as known_sizes with .lyx names
150 char const * const known_coded_sizes[] = { "tiny", "scriptsize", "footnotesize",
151 "small", "normal", "large", "larger", "largest", "huge", "giant", 0};
153 /// LaTeX 2.09 names for font families
154 char const * const known_old_font_families[] = { "rm", "sf", "tt", 0};
156 /// LaTeX names for font families
157 char const * const known_font_families[] = { "rmfamily", "sffamily",
160 /// the same as known_old_font_families and known_font_families with .lyx names
161 char const * const known_coded_font_families[] = { "roman", "sans",
164 /// LaTeX 2.09 names for font series
165 char const * const known_old_font_series[] = { "bf", 0};
167 /// LaTeX names for font series
168 char const * const known_font_series[] = { "bfseries", "mdseries", 0};
170 /// the same as known_old_font_series and known_font_series with .lyx names
171 char const * const known_coded_font_series[] = { "bold", "medium", 0};
173 /// LaTeX 2.09 names for font shapes
174 char const * const known_old_font_shapes[] = { "it", "sl", "sc", 0};
176 /// LaTeX names for font shapes
177 char const * const known_font_shapes[] = { "itshape", "slshape", "scshape",
180 /// the same as known_old_font_shapes and known_font_shapes with .lyx names
181 char const * const known_coded_font_shapes[] = { "italic", "slanted",
182 "smallcaps", "up", 0};
185 * Graphics file extensions known by the dvips driver of the graphics package.
186 * These extensions are used to complete the filename of an included
187 * graphics file if it does not contain an extension.
188 * The order must be the same that latex uses to find a file, because we
189 * will use the first extension that matches.
190 * This is only an approximation for the common cases. If we would want to
191 * do it right in all cases, we would need to know which graphics driver is
192 * used and know the extensions of every driver of the graphics package.
194 char const * const known_dvips_graphics_formats[] = {"eps", "ps", "eps.gz",
195 "ps.gz", "eps.Z", "ps.Z", 0};
198 * Graphics file extensions known by the pdftex driver of the graphics package.
199 * \sa known_dvips_graphics_formats
201 char const * const known_pdftex_graphics_formats[] = {"png", "pdf", "jpg",
205 * Known file extensions for TeX files as used by \\include.
207 char const * const known_tex_extensions[] = {"tex", 0};
209 /// spaces known by InsetSpace
210 char const * const known_spaces[] = { " ", "space", ",", "thinspace", "quad",
211 "qquad", "enspace", "enskip", "negthinspace", "hfill", "dotfill", "hrulefill",
212 "leftarrowfill", "rightarrowfill", "upbracefill", "downbracefill", 0};
214 /// the same as known_spaces with .lyx names
215 char const * const known_coded_spaces[] = { "space{}", "space{}",
216 "thinspace{}", "thinspace{}", "quad{}", "qquad{}", "enspace{}", "enskip{}",
217 "negthinspace{}", "hfill{}", "dotfill{}", "hrulefill{}", "leftarrowfill{}",
218 "rightarrowfill{}", "upbracefill{}", "downbracefill{}", 0};
220 /// These are translated by LyX to commands like "\\LyX{}", so we have to put
221 /// them in ERT. "LaTeXe" must come before "LaTeX"!
222 char const * const known_phrases[] = {"LyX", "TeX", "LaTeXe", "LaTeX", 0};
223 char const * const known_coded_phrases[] = {"LyX", "TeX", "LaTeX2e", "LaTeX", 0};
224 int const known_phrase_lengths[] = {3, 5, 7, 0};
227 /// splits "x=z, y=b" into a map and an ordered keyword vector
228 void split_map(string const & s, map<string, string> & res, vector<string> & keys)
233 keys.resize(v.size());
234 for (size_t i = 0; i < v.size(); ++i) {
235 size_t const pos = v[i].find('=');
236 string const index = trim(v[i].substr(0, pos));
237 string const value = trim(v[i].substr(pos + 1, string::npos));
245 * Split a LaTeX length into value and unit.
246 * The latter can be a real unit like "pt", or a latex length variable
247 * like "\textwidth". The unit may contain additional stuff like glue
248 * lengths, but we don't care, because such lengths are ERT anyway.
249 * \returns true if \p value and \p unit are valid.
251 bool splitLatexLength(string const & len, string & value, string & unit)
255 const string::size_type i = len.find_first_not_of(" -+0123456789.,");
256 //'4,5' is a valid LaTeX length number. Change it to '4.5'
257 string const length = subst(len, ',', '.');
258 if (i == string::npos)
261 if (len[0] == '\\') {
262 // We had something like \textwidth without a factor
268 value = trim(string(length, 0, i));
272 // 'cM' is a valid LaTeX length unit. Change it to 'cm'
273 if (contains(len, '\\'))
274 unit = trim(string(len, i));
276 unit = ascii_lowercase(trim(string(len, i)));
281 /// A simple function to translate a latex length to something LyX can
282 /// understand. Not perfect, but rather best-effort.
283 bool translate_len(string const & length, string & valstring, string & unit)
285 if (!splitLatexLength(length, valstring, unit))
287 // LyX uses percent values
289 istringstream iss(valstring);
294 string const percentval = oss.str();
296 if (unit.empty() || unit[0] != '\\')
298 string::size_type const i = unit.find(' ');
299 string const endlen = (i == string::npos) ? string() : string(unit, i);
300 if (unit == "\\textwidth") {
301 valstring = percentval;
302 unit = "text%" + endlen;
303 } else if (unit == "\\columnwidth") {
304 valstring = percentval;
305 unit = "col%" + endlen;
306 } else if (unit == "\\paperwidth") {
307 valstring = percentval;
308 unit = "page%" + endlen;
309 } else if (unit == "\\linewidth") {
310 valstring = percentval;
311 unit = "line%" + endlen;
312 } else if (unit == "\\paperheight") {
313 valstring = percentval;
314 unit = "pheight%" + endlen;
315 } else if (unit == "\\textheight") {
316 valstring = percentval;
317 unit = "theight%" + endlen;
325 string translate_len(string const & length)
329 if (translate_len(length, value, unit))
331 // If the input is invalid, return what we have.
339 * Translates a LaTeX length into \p value, \p unit and
340 * \p special parts suitable for a box inset.
341 * The difference from translate_len() is that a box inset knows about
342 * some special "units" that are stored in \p special.
344 void translate_box_len(string const & length, string & value, string & unit, string & special)
346 if (translate_len(length, value, unit)) {
347 if (unit == "\\height" || unit == "\\depth" ||
348 unit == "\\totalheight" || unit == "\\width") {
349 special = unit.substr(1);
350 // The unit is not used, but LyX requires a dummy setting
363 * Find a file with basename \p name in path \p path and an extension
366 string find_file(string const & name, string const & path,
367 char const * const * extensions)
369 for (char const * const * what = extensions; *what; ++what) {
370 string const trial = addExtension(name, *what);
371 if (makeAbsPath(trial, path).exists())
378 void begin_inset(ostream & os, string const & name)
380 os << "\n\\begin_inset " << name;
384 void begin_command_inset(ostream & os, string const & name,
385 string const & latexname)
387 begin_inset(os, "CommandInset ");
388 os << name << "\nLatexCommand " << latexname << '\n';
392 void end_inset(ostream & os)
394 os << "\n\\end_inset\n\n";
398 bool skip_braces(Parser & p)
400 if (p.next_token().cat() != catBegin)
403 if (p.next_token().cat() == catEnd) {
412 void handle_ert(ostream & os, string const & s, Context & context)
414 // We must have a valid layout before outputting the ERT inset.
415 context.check_layout(os);
416 Context newcontext(true, context.textclass);
417 begin_inset(os, "ERT");
418 os << "\nstatus collapsed\n";
419 newcontext.check_layout(os);
420 for (string::const_iterator it = s.begin(), et = s.end(); it != et; ++it) {
422 os << "\n\\backslash\n";
423 else if (*it == '\n') {
424 newcontext.new_paragraph(os);
425 newcontext.check_layout(os);
429 newcontext.check_end_layout(os);
434 void handle_comment(ostream & os, string const & s, Context & context)
436 // TODO: Handle this better
437 Context newcontext(true, context.textclass);
438 begin_inset(os, "ERT");
439 os << "\nstatus collapsed\n";
440 newcontext.check_layout(os);
441 for (string::const_iterator it = s.begin(), et = s.end(); it != et; ++it) {
443 os << "\n\\backslash\n";
447 // make sure that our comment is the last thing on the line
448 newcontext.new_paragraph(os);
449 newcontext.check_layout(os);
450 newcontext.check_end_layout(os);
455 Layout const * findLayout(TextClass const & textclass, string const & name, bool command)
457 Layout const * layout = findLayoutWithoutModule(textclass, name, command);
460 if (checkModule(name, command))
461 return findLayoutWithoutModule(textclass, name, command);
466 InsetLayout const * findInsetLayout(TextClass const & textclass, string const & name, bool command)
468 InsetLayout const * insetlayout = findInsetLayoutWithoutModule(textclass, name, command);
471 if (checkModule(name, command))
472 return findInsetLayoutWithoutModule(textclass, name, command);
477 void eat_whitespace(Parser &, ostream &, Context &, bool);
481 * Skips whitespace and braces.
482 * This should be called after a command has been parsed that is not put into
483 * ERT, and where LyX adds "{}" if needed.
485 void skip_spaces_braces(Parser & p, bool keepws = false)
487 /* The following four examples produce the same typeset output and
488 should be handled by this function:
496 // Unfortunately we need to skip comments, too.
497 // We can't use eat_whitespace since writing them after the {}
498 // results in different output in some cases.
499 bool const skipped_spaces = p.skip_spaces(true);
500 bool const skipped_braces = skip_braces(p);
501 if (keepws && skipped_spaces && !skipped_braces)
502 // put back the space (it is better handled by check_space)
503 p.unskip_spaces(true);
507 void output_command_layout(ostream & os, Parser & p, bool outer,
508 Context & parent_context,
509 Layout const * newlayout)
511 TeXFont const oldFont = parent_context.font;
512 // save the current font size
513 string const size = oldFont.size;
514 // reset the font size to default, because the font size switches
515 // don't affect section headings and the like
516 parent_context.font.size = Context::normalfont.size;
517 // we only need to write the font change if we have an open layout
518 if (!parent_context.atParagraphStart())
519 output_font_change(os, oldFont, parent_context.font);
520 parent_context.check_end_layout(os);
521 Context context(true, parent_context.textclass, newlayout,
522 parent_context.layout, parent_context.font);
523 if (parent_context.deeper_paragraph) {
524 // We are beginning a nested environment after a
525 // deeper paragraph inside the outer list environment.
526 // Therefore we don't need to output a "begin deeper".
527 context.need_end_deeper = true;
529 context.check_deeper(os);
530 context.check_layout(os);
531 unsigned int optargs = 0;
532 while (optargs < context.layout->optargs) {
533 eat_whitespace(p, os, context, false);
534 if (p.next_token().cat() == catEscape ||
535 p.next_token().character() != '[')
537 p.get_token(); // eat '['
538 begin_inset(os, "OptArg\n");
539 os << "status collapsed\n\n";
540 parse_text_in_inset(p, os, FLAG_BRACK_LAST, outer, context);
542 eat_whitespace(p, os, context, false);
545 unsigned int reqargs = 0;
546 while (LYX_FORMAT >= 392 && reqargs < context.layout->reqargs) {
547 eat_whitespace(p, os, context, false);
548 if (p.next_token().cat() != catBegin)
550 p.get_token(); // eat '{'
551 begin_inset(os, "OptArg\n");
552 os << "status collapsed\n\n";
553 parse_text_in_inset(p, os, FLAG_BRACE_LAST, outer, context);
555 eat_whitespace(p, os, context, false);
558 parse_text(p, os, FLAG_ITEM, outer, context);
559 context.check_end_layout(os);
560 if (parent_context.deeper_paragraph) {
561 // We must suppress the "end deeper" because we
562 // suppressed the "begin deeper" above.
563 context.need_end_deeper = false;
565 context.check_end_deeper(os);
566 // We don't need really a new paragraph, but
567 // we must make sure that the next item gets a \begin_layout.
568 parent_context.new_paragraph(os);
569 // Set the font size to the original value. No need to output it here
570 // (Context::begin_layout() will do that if needed)
571 parent_context.font.size = size;
576 * Output a space if necessary.
577 * This function gets called for every whitespace token.
579 * We have three cases here:
580 * 1. A space must be suppressed. Example: The lyxcode case below
581 * 2. A space may be suppressed. Example: Spaces before "\par"
582 * 3. A space must not be suppressed. Example: A space between two words
584 * We currently handle only 1. and 3 and from 2. only the case of
585 * spaces before newlines as a side effect.
587 * 2. could be used to suppress as many spaces as possible. This has two effects:
588 * - Reimporting LyX generated LaTeX files changes almost no whitespace
589 * - Superflous whitespace from non LyX generated LaTeX files is removed.
590 * The drawback is that the logic inside the function becomes
591 * complicated, and that is the reason why it is not implemented.
593 void check_space(Parser & p, ostream & os, Context & context)
595 Token const next = p.next_token();
596 Token const curr = p.curr_token();
597 // A space before a single newline and vice versa must be ignored
598 // LyX emits a newline before \end{lyxcode}.
599 // This newline must be ignored,
600 // otherwise LyX will add an additional protected space.
601 if (next.cat() == catSpace ||
602 next.cat() == catNewline ||
603 (next.cs() == "end" && context.layout->free_spacing && curr.cat() == catNewline)) {
606 context.check_layout(os);
612 * Parse all arguments of \p command
614 void parse_arguments(string const & command,
615 vector<ArgumentType> const & template_arguments,
616 Parser & p, ostream & os, bool outer, Context & context)
618 string ert = command;
619 size_t no_arguments = template_arguments.size();
620 for (size_t i = 0; i < no_arguments; ++i) {
621 switch (template_arguments[i]) {
623 // This argument contains regular LaTeX
624 handle_ert(os, ert + '{', context);
625 eat_whitespace(p, os, context, false);
626 parse_text(p, os, FLAG_ITEM, outer, context);
630 // This argument consists only of a single item.
631 // The presence of '{' or not must be preserved.
633 if (p.next_token().cat() == catBegin)
634 ert += '{' + p.verbatim_item() + '}';
636 ert += p.verbatim_item();
639 // This argument may contain special characters
640 ert += '{' + p.verbatim_item() + '}';
643 // true because we must not eat whitespace
644 // if an optional arg follows me must not strip the
645 // brackets from this one
646 if (i < no_arguments - 1 &&
647 template_arguments[i+1] == optional)
648 ert += p.getFullOpt(true);
650 ert += p.getOpt(true);
654 handle_ert(os, ert, context);
659 * Check whether \p command is a known command. If yes,
660 * handle the command with all arguments.
661 * \return true if the command was parsed, false otherwise.
663 bool parse_command(string const & command, Parser & p, ostream & os,
664 bool outer, Context & context)
666 if (known_commands.find(command) != known_commands.end()) {
667 parse_arguments(command, known_commands[command], p, os,
675 /// Parses a minipage or parbox
676 void parse_box(Parser & p, ostream & os, unsigned outer_flags,
677 unsigned inner_flags, bool outer, Context & parent_context,
678 string const & outer_type, string const & special,
679 string const & inner_type)
683 string hor_pos = "c";
684 // We need to set the height to the LaTeX default of 1\\totalheight
685 // for the case when no height argument is given
686 string height_value = "1";
687 string height_unit = "in";
688 string height_special = "totalheight";
690 if (!inner_type.empty() && p.hasOpt()) {
691 position = p.getArg('[', ']');
692 if (position != "t" && position != "c" && position != "b") {
693 cerr << "invalid position " << position << " for "
694 << inner_type << endl;
698 latex_height = p.getArg('[', ']');
699 translate_box_len(latex_height, height_value, height_unit, height_special);
702 inner_pos = p.getArg('[', ']');
703 if (inner_pos != "c" && inner_pos != "t" &&
704 inner_pos != "b" && inner_pos != "s") {
705 cerr << "invalid inner_pos "
706 << inner_pos << " for "
707 << inner_type << endl;
708 inner_pos = position;
716 if (inner_type.empty()) {
718 latex_width = "\\columnwidth";
721 latex_width = p2.getArg('[', ']');
722 string const opt = p2.getArg('[', ']');
725 if (hor_pos != "l" && hor_pos != "c" &&
727 cerr << "invalid hor_pos " << hor_pos
728 << " for " << outer_type << endl;
734 latex_width = p.verbatim_item();
735 translate_len(latex_width, width_value, width_unit);
736 // LyX can't handle length variables
737 bool use_ert = contains(width_unit, '\\') || contains(height_unit, '\\');
738 if (!use_ert && !outer_type.empty() && !inner_type.empty()) {
739 // Look whether there is some content after the end of the
740 // inner box, but before the end of the outer box.
741 // If yes, we need to output ERT.
743 if (inner_flags & FLAG_END)
744 p.verbatimEnvironment(inner_type);
748 if ((outer_type == "framed" && p.next_token().asInput() != "\\end") ||
749 (outer_type != "framed" && p.next_token().cat() != catEnd)) {
750 // something is between the end of the inner box and
751 // the end of the outer box, so we need to use ERT.
758 if (!outer_type.empty()) {
759 if (outer_flags & FLAG_END)
760 ss << "\\begin{" << outer_type << '}';
762 ss << '\\' << outer_type << '{';
763 if (!special.empty())
767 if (!inner_type.empty()) {
768 if (inner_flags & FLAG_END)
769 ss << "\\begin{" << inner_type << '}';
771 ss << '\\' << inner_type;
772 if (!position.empty())
773 ss << '[' << position << ']';
774 if (!latex_height.empty())
775 ss << '[' << latex_height << ']';
776 if (!inner_pos.empty())
777 ss << '[' << inner_pos << ']';
778 ss << '{' << latex_width << '}';
779 if (!(inner_flags & FLAG_END))
782 handle_ert(os, ss.str(), parent_context);
783 if (!inner_type.empty()) {
784 parse_text(p, os, inner_flags, outer, parent_context);
785 if (inner_flags & FLAG_END)
786 handle_ert(os, "\\end{" + inner_type + '}',
789 handle_ert(os, "}", parent_context);
791 if (!outer_type.empty()) {
792 parse_text(p, os, outer_flags, outer, parent_context);
793 if (outer_flags & FLAG_END)
794 handle_ert(os, "\\end{" + outer_type + '}',
797 handle_ert(os, "}", parent_context);
800 // LyX does not like empty positions, so we have
801 // to set them to the LaTeX default values here.
802 if (position.empty())
804 if (inner_pos.empty())
805 inner_pos = position;
806 parent_context.check_layout(os);
807 begin_inset(os, "Box ");
808 if (outer_type == "framed")
810 else if (outer_type == "framebox")
812 else if (outer_type == "shadowbox")
814 else if (outer_type == "shaded")
816 else if (outer_type == "doublebox")
818 else if (outer_type.empty())
821 os << outer_type << '\n';
822 os << "position \"" << position << "\"\n";
823 os << "hor_pos \"" << hor_pos << "\"\n";
824 os << "has_inner_box " << !inner_type.empty() << "\n";
825 os << "inner_pos \"" << inner_pos << "\"\n";
826 os << "use_parbox " << (inner_type == "parbox") << '\n';
827 os << "width \"" << width_value << width_unit << "\"\n";
828 os << "special \"none\"\n";
829 os << "height \"" << height_value << height_unit << "\"\n";
830 os << "height_special \"" << height_special << "\"\n";
831 os << "status open\n\n";
832 Context context(true, parent_context.textclass);
833 context.font = parent_context.font;
835 // If we have no inner box the contens will be read with the outer box
836 if (!inner_type.empty())
837 parse_text(p, os, inner_flags, outer, context);
839 // Ensure that the end of the outer box is parsed correctly:
840 // The opening brace has been eaten by parse_outer_box()
841 if (!outer_type.empty() && (outer_flags & FLAG_ITEM)) {
842 outer_flags &= ~FLAG_ITEM;
843 outer_flags |= FLAG_BRACE_LAST;
846 // Find end of outer box, output contents if inner_type is
847 // empty and output possible comments
848 if (!outer_type.empty()) {
849 // This does not output anything but comments if
850 // inner_type is not empty (see use_ert)
851 parse_text(p, os, outer_flags, outer, context);
854 context.check_end_layout(os);
856 #ifdef PRESERVE_LAYOUT
857 // LyX puts a % after the end of the minipage
858 if (p.next_token().cat() == catNewline && p.next_token().cs().size() > 1) {
860 //handle_comment(os, "%dummy", parent_context);
863 parent_context.new_paragraph(os);
865 else if (p.next_token().cat() == catSpace || p.next_token().cat() == catNewline) {
866 //handle_comment(os, "%dummy", parent_context);
869 // We add a protected space if something real follows
870 if (p.good() && p.next_token().cat() != catComment) {
871 begin_inset(os, "space ~\n");
880 void parse_outer_box(Parser & p, ostream & os, unsigned flags, bool outer,
881 Context & parent_context, string const & outer_type,
882 string const & special)
884 eat_whitespace(p, os, parent_context, false);
885 if (flags & FLAG_ITEM) {
887 if (p.next_token().cat() == catBegin)
890 cerr << "Warning: Ignoring missing '{' after \\"
891 << outer_type << '.' << endl;
892 eat_whitespace(p, os, parent_context, false);
895 unsigned int inner_flags = 0;
896 if (outer_type == "shaded") {
897 // These boxes never have an inner box
899 } else if (p.next_token().asInput() == "\\parbox") {
900 inner = p.get_token().cs();
901 inner_flags = FLAG_ITEM;
902 } else if (p.next_token().asInput() == "\\begin") {
903 // Is this a minipage?
906 inner = p.getArg('{', '}');
908 if (inner == "minipage") {
911 eat_whitespace(p, os, parent_context, false);
912 inner_flags = FLAG_END;
916 if (inner_flags == FLAG_END) {
917 active_environments.push_back(inner);
918 parse_box(p, os, flags, FLAG_END, outer, parent_context,
919 outer_type, special, inner);
920 active_environments.pop_back();
922 parse_box(p, os, flags, inner_flags, outer, parent_context,
923 outer_type, special, inner);
928 void parse_listings(Parser & p, ostream & os, Context & parent_context)
930 parent_context.check_layout(os);
931 begin_inset(os, "listings\n");
932 os << "inline false\n"
933 << "status collapsed\n";
934 Context context(true, parent_context.textclass);
935 context.layout = &parent_context.textclass.plainLayout();
936 context.check_layout(os);
937 string const s = p.verbatimEnvironment("lstlisting");
938 for (string::const_iterator it = s.begin(), et = s.end(); it != et; ++it) {
940 os << "\n\\backslash\n";
941 else if (*it == '\n') {
942 // avoid adding an empty paragraph at the end
944 context.new_paragraph(os);
945 context.check_layout(os);
950 context.check_end_layout(os);
955 /// parse an unknown environment
956 void parse_unknown_environment(Parser & p, string const & name, ostream & os,
957 unsigned flags, bool outer,
958 Context & parent_context)
960 if (name == "tabbing")
961 // We need to remember that we have to handle '\=' specially
962 flags |= FLAG_TABBING;
964 // We need to translate font changes and paragraphs inside the
965 // environment to ERT if we have a non standard font.
966 // Otherwise things like
967 // \large\begin{foo}\huge bar\end{foo}
969 bool const specialfont =
970 (parent_context.font != parent_context.normalfont);
971 bool const new_layout_allowed = parent_context.new_layout_allowed;
973 parent_context.new_layout_allowed = false;
974 handle_ert(os, "\\begin{" + name + "}", parent_context);
975 parse_text_snippet(p, os, flags, outer, parent_context);
976 handle_ert(os, "\\end{" + name + "}", parent_context);
978 parent_context.new_layout_allowed = new_layout_allowed;
982 void parse_environment(Parser & p, ostream & os, bool outer,
983 string & last_env, Context & parent_context)
985 Layout const * newlayout;
986 InsetLayout const * newinsetlayout = 0;
987 string const name = p.getArg('{', '}');
988 const bool is_starred = suffixIs(name, '*');
989 string const unstarred_name = rtrim(name, "*");
990 active_environments.push_back(name);
992 if (is_math_env(name)) {
993 parent_context.check_layout(os);
994 begin_inset(os, "Formula ");
995 os << "\\begin{" << name << "}";
996 parse_math(p, os, FLAG_END, MATH_MODE);
997 os << "\\end{" << name << "}";
1001 else if (name == "tabular" || name == "longtable") {
1002 eat_whitespace(p, os, parent_context, false);
1003 parent_context.check_layout(os);
1004 begin_inset(os, "Tabular ");
1005 handle_tabular(p, os, name == "longtable", parent_context);
1010 else if (parent_context.textclass.floats().typeExist(unstarred_name)) {
1011 eat_whitespace(p, os, parent_context, false);
1012 parent_context.check_layout(os);
1013 begin_inset(os, "Float " + unstarred_name + "\n");
1015 os << "placement " << p.getArg('[', ']') << '\n';
1016 os << "wide " << convert<string>(is_starred)
1017 << "\nsideways false"
1018 << "\nstatus open\n\n";
1019 parse_text_in_inset(p, os, FLAG_END, outer, parent_context);
1021 // We don't need really a new paragraph, but
1022 // we must make sure that the next item gets a \begin_layout.
1023 parent_context.new_paragraph(os);
1027 else if (name == "minipage") {
1028 eat_whitespace(p, os, parent_context, false);
1029 parse_box(p, os, 0, FLAG_END, outer, parent_context, "", "", name);
1033 else if (name == "comment") {
1034 eat_whitespace(p, os, parent_context, false);
1035 parent_context.check_layout(os);
1036 begin_inset(os, "Note Comment\n");
1037 os << "status open\n";
1038 parse_text_in_inset(p, os, FLAG_END, outer, parent_context);
1041 skip_braces(p); // eat {} that might by set by LyX behind comments
1044 else if (name == "lyxgreyedout") {
1045 eat_whitespace(p, os, parent_context, false);
1046 parent_context.check_layout(os);
1047 begin_inset(os, "Note Greyedout\n");
1048 os << "status open\n";
1049 parse_text_in_inset(p, os, FLAG_END, outer, parent_context);
1054 else if (name == "framed" || name == "shaded") {
1055 eat_whitespace(p, os, parent_context, false);
1056 parse_outer_box(p, os, FLAG_END, outer, parent_context, name, "");
1060 else if (name == "lstlisting") {
1061 eat_whitespace(p, os, parent_context, false);
1062 // FIXME handle listings with parameters
1064 parse_unknown_environment(p, name, os, FLAG_END,
1065 outer, parent_context);
1067 parse_listings(p, os, parent_context);
1071 else if (!parent_context.new_layout_allowed)
1072 parse_unknown_environment(p, name, os, FLAG_END, outer,
1075 // Alignment and spacing settings
1076 // FIXME (bug xxxx): These settings can span multiple paragraphs and
1077 // therefore are totally broken!
1078 // Note that \centering, raggedright, and raggedleft cannot be handled, as
1079 // they are commands not environments. They are furthermore switches that
1080 // can be ended by another switches, but also by commands like \footnote or
1081 // \parbox. So the only safe way is to leave them untouched.
1082 else if (name == "center" || name == "centering" ||
1083 name == "flushleft" || name == "flushright" ||
1084 name == "singlespace" || name == "onehalfspace" ||
1085 name == "doublespace" || name == "spacing") {
1086 eat_whitespace(p, os, parent_context, false);
1087 // We must begin a new paragraph if not already done
1088 if (! parent_context.atParagraphStart()) {
1089 parent_context.check_end_layout(os);
1090 parent_context.new_paragraph(os);
1092 if (name == "flushleft")
1093 parent_context.add_extra_stuff("\\align left\n");
1094 else if (name == "flushright")
1095 parent_context.add_extra_stuff("\\align right\n");
1096 else if (name == "center" || name == "centering")
1097 parent_context.add_extra_stuff("\\align center\n");
1098 else if (name == "singlespace")
1099 parent_context.add_extra_stuff("\\paragraph_spacing single\n");
1100 else if (name == "onehalfspace")
1101 parent_context.add_extra_stuff("\\paragraph_spacing onehalf\n");
1102 else if (name == "doublespace")
1103 parent_context.add_extra_stuff("\\paragraph_spacing double\n");
1104 else if (name == "spacing")
1105 parent_context.add_extra_stuff("\\paragraph_spacing other " + p.verbatim_item() + "\n");
1106 parse_text(p, os, FLAG_END, outer, parent_context);
1107 // Just in case the environment is empty
1108 parent_context.extra_stuff.erase();
1109 // We must begin a new paragraph to reset the alignment
1110 parent_context.new_paragraph(os);
1114 // The single '=' is meant here.
1115 else if ((newlayout = findLayout(parent_context.textclass, name, false))) {
1116 eat_whitespace(p, os, parent_context, false);
1117 Context context(true, parent_context.textclass, newlayout,
1118 parent_context.layout, parent_context.font);
1119 if (parent_context.deeper_paragraph) {
1120 // We are beginning a nested environment after a
1121 // deeper paragraph inside the outer list environment.
1122 // Therefore we don't need to output a "begin deeper".
1123 context.need_end_deeper = true;
1125 parent_context.check_end_layout(os);
1126 if (last_env == name) {
1127 // we need to output a separator since LyX would export
1128 // the two environments as one otherwise (bug 5716)
1129 docstring const sep = from_ascii("--Separator--");
1130 TeX2LyXDocClass const & textclass(parent_context.textclass);
1131 if (textclass.hasLayout(sep)) {
1132 Context newcontext(parent_context);
1133 newcontext.layout = &(textclass[sep]);
1134 newcontext.check_layout(os);
1135 newcontext.check_end_layout(os);
1137 parent_context.check_layout(os);
1138 begin_inset(os, "Note Note\n");
1139 os << "status closed\n";
1140 Context newcontext(true, textclass,
1141 &(textclass.defaultLayout()));
1142 newcontext.check_layout(os);
1143 newcontext.check_end_layout(os);
1145 parent_context.check_end_layout(os);
1148 switch (context.layout->latextype) {
1149 case LATEX_LIST_ENVIRONMENT:
1150 context.add_par_extra_stuff("\\labelwidthstring "
1151 + p.verbatim_item() + '\n');
1154 case LATEX_BIB_ENVIRONMENT:
1155 p.verbatim_item(); // swallow next arg
1161 context.check_deeper(os);
1162 parse_text(p, os, FLAG_END, outer, context);
1163 context.check_end_layout(os);
1164 if (parent_context.deeper_paragraph) {
1165 // We must suppress the "end deeper" because we
1166 // suppressed the "begin deeper" above.
1167 context.need_end_deeper = false;
1169 context.check_end_deeper(os);
1170 parent_context.new_paragraph(os);
1174 // The single '=' is meant here.
1175 else if ((newinsetlayout = findInsetLayout(parent_context.textclass, name, false))) {
1176 eat_whitespace(p, os, parent_context, false);
1177 parent_context.check_layout(os);
1178 begin_inset(os, "Flex ");
1179 os << to_utf8(newinsetlayout->name()) << '\n'
1180 << "status collapsed\n";
1181 parse_text_in_inset(p, os, FLAG_END, false, parent_context, newinsetlayout);
1185 else if (name == "appendix") {
1186 // This is no good latex style, but it works and is used in some documents...
1187 eat_whitespace(p, os, parent_context, false);
1188 parent_context.check_end_layout(os);
1189 Context context(true, parent_context.textclass, parent_context.layout,
1190 parent_context.layout, parent_context.font);
1191 context.check_layout(os);
1192 os << "\\start_of_appendix\n";
1193 parse_text(p, os, FLAG_END, outer, context);
1194 context.check_end_layout(os);
1198 else if (known_environments.find(name) != known_environments.end()) {
1199 vector<ArgumentType> arguments = known_environments[name];
1200 // The last "argument" denotes wether we may translate the
1201 // environment contents to LyX
1202 // The default required if no argument is given makes us
1203 // compatible with the reLyXre environment.
1204 ArgumentType contents = arguments.empty() ?
1207 if (!arguments.empty())
1208 arguments.pop_back();
1209 // See comment in parse_unknown_environment()
1210 bool const specialfont =
1211 (parent_context.font != parent_context.normalfont);
1212 bool const new_layout_allowed =
1213 parent_context.new_layout_allowed;
1215 parent_context.new_layout_allowed = false;
1216 parse_arguments("\\begin{" + name + "}", arguments, p, os,
1217 outer, parent_context);
1218 if (contents == verbatim)
1219 handle_ert(os, p.verbatimEnvironment(name),
1222 parse_text_snippet(p, os, FLAG_END, outer,
1224 handle_ert(os, "\\end{" + name + "}", parent_context);
1226 parent_context.new_layout_allowed = new_layout_allowed;
1230 parse_unknown_environment(p, name, os, FLAG_END, outer,
1234 active_environments.pop_back();
1238 /// parses a comment and outputs it to \p os.
1239 void parse_comment(Parser & p, ostream & os, Token const & t, Context & context)
1241 LASSERT(t.cat() == catComment, return);
1242 if (!t.cs().empty()) {
1243 context.check_layout(os);
1244 handle_comment(os, '%' + t.cs(), context);
1245 if (p.next_token().cat() == catNewline) {
1246 // A newline after a comment line starts a new
1248 if (context.new_layout_allowed) {
1249 if(!context.atParagraphStart())
1250 // Only start a new paragraph if not already
1251 // done (we might get called recursively)
1252 context.new_paragraph(os);
1254 handle_ert(os, "\n", context);
1255 eat_whitespace(p, os, context, true);
1258 // "%\n" combination
1265 * Reads spaces and comments until the first non-space, non-comment token.
1266 * New paragraphs (double newlines or \\par) are handled like simple spaces
1267 * if \p eatParagraph is true.
1268 * Spaces are skipped, but comments are written to \p os.
1270 void eat_whitespace(Parser & p, ostream & os, Context & context,
1274 Token const & t = p.get_token();
1275 if (t.cat() == catComment)
1276 parse_comment(p, os, t, context);
1277 else if ((! eatParagraph && p.isParagraph()) ||
1278 (t.cat() != catSpace && t.cat() != catNewline)) {
1287 * Set a font attribute, parse text and reset the font attribute.
1288 * \param attribute Attribute name (e.g. \\family, \\shape etc.)
1289 * \param currentvalue Current value of the attribute. Is set to the new
1290 * value during parsing.
1291 * \param newvalue New value of the attribute
1293 void parse_text_attributes(Parser & p, ostream & os, unsigned flags, bool outer,
1294 Context & context, string const & attribute,
1295 string & currentvalue, string const & newvalue)
1297 context.check_layout(os);
1298 string const oldvalue = currentvalue;
1299 currentvalue = newvalue;
1300 os << '\n' << attribute << ' ' << newvalue << "\n";
1301 parse_text_snippet(p, os, flags, outer, context);
1302 context.check_layout(os);
1303 os << '\n' << attribute << ' ' << oldvalue << "\n";
1304 currentvalue = oldvalue;
1308 /// get the arguments of a natbib or jurabib citation command
1309 void get_cite_arguments(Parser & p, bool natbibOrder,
1310 string & before, string & after)
1312 // We need to distinguish "" and "[]", so we can't use p.getOpt().
1314 // text before the citation
1316 // text after the citation
1317 after = p.getFullOpt();
1319 if (!after.empty()) {
1320 before = p.getFullOpt();
1321 if (natbibOrder && !before.empty())
1322 swap(before, after);
1327 /// Convert filenames with TeX macros and/or quotes to something LyX
1329 string const normalize_filename(string const & name)
1331 Parser p(trim(name, "\""));
1334 Token const & t = p.get_token();
1335 if (t.cat() != catEscape)
1337 else if (t.cs() == "lyxdot") {
1338 // This is used by LyX for simple dots in relative
1342 } else if (t.cs() == "space") {
1352 /// Convert \p name from TeX convention (relative to master file) to LyX
1353 /// convention (relative to .lyx file) if it is relative
1354 void fix_relative_filename(string & name)
1356 if (FileName::isAbsolute(name))
1359 name = to_utf8(makeRelPath(from_utf8(makeAbsPath(name, getMasterFilePath()).absFileName()),
1360 from_utf8(getParentFilePath())));
1364 /// Parse a NoWeb Scrap section. The initial "<<" is already parsed.
1365 void parse_noweb(Parser & p, ostream & os, Context & context)
1367 // assemble the rest of the keyword
1371 Token const & t = p.get_token();
1372 if (t.asInput() == ">" && p.next_token().asInput() == ">") {
1375 scrap = (p.good() && p.next_token().asInput() == "=");
1377 name += p.get_token().asInput();
1380 name += t.asInput();
1383 if (!scrap || !context.new_layout_allowed ||
1384 !context.textclass.hasLayout(from_ascii("Scrap"))) {
1385 cerr << "Warning: Could not interpret '" << name
1386 << "'. Ignoring it." << endl;
1390 // We use new_paragraph instead of check_end_layout because the stuff
1391 // following the noweb chunk needs to start with a \begin_layout.
1392 // This may create a new paragraph even if there was none in the
1393 // noweb file, but the alternative is an invalid LyX file. Since
1394 // noweb code chunks are implemented with a layout style in LyX they
1395 // always must be in an own paragraph.
1396 context.new_paragraph(os);
1397 Context newcontext(true, context.textclass,
1398 &context.textclass[from_ascii("Scrap")]);
1399 newcontext.check_layout(os);
1402 Token const & t = p.get_token();
1403 // We abuse the parser a bit, because this is no TeX syntax
1405 if (t.cat() == catEscape)
1406 os << subst(t.asInput(), "\\", "\n\\backslash\n");
1409 begin_inset(oss, "Newline newline");
1411 os << subst(t.asInput(), "\n", oss.str());
1413 // The scrap chunk is ended by an @ at the beginning of a line.
1414 // After the @ the line may contain a comment and/or
1415 // whitespace, but nothing else.
1416 if (t.asInput() == "@" && p.prev_token().cat() == catNewline &&
1417 (p.next_token().cat() == catSpace ||
1418 p.next_token().cat() == catNewline ||
1419 p.next_token().cat() == catComment)) {
1420 while (p.good() && p.next_token().cat() == catSpace)
1421 os << p.get_token().asInput();
1422 if (p.next_token().cat() == catComment)
1423 // The comment includes a final '\n'
1424 os << p.get_token().asInput();
1426 if (p.next_token().cat() == catNewline)
1433 newcontext.check_end_layout(os);
1437 /// detects \\def, \\long\\def and \\global\\long\\def with ws and comments
1438 bool is_macro(Parser & p)
1440 Token first = p.curr_token();
1441 if (first.cat() != catEscape || !p.good())
1443 if (first.cs() == "def")
1445 if (first.cs() != "global" && first.cs() != "long")
1447 Token second = p.get_token();
1449 while (p.good() && !p.isParagraph() && (second.cat() == catSpace ||
1450 second.cat() == catNewline || second.cat() == catComment)) {
1451 second = p.get_token();
1454 bool secondvalid = second.cat() == catEscape;
1456 bool thirdvalid = false;
1457 if (p.good() && first.cs() == "global" && secondvalid &&
1458 second.cs() == "long") {
1459 third = p.get_token();
1461 while (p.good() && !p.isParagraph() &&
1462 (third.cat() == catSpace ||
1463 third.cat() == catNewline ||
1464 third.cat() == catComment)) {
1465 third = p.get_token();
1468 thirdvalid = third.cat() == catEscape;
1470 for (int i = 0; i < pos; ++i)
1475 return (first.cs() == "global" || first.cs() == "long") &&
1476 second.cs() == "def";
1477 return first.cs() == "global" && second.cs() == "long" &&
1478 third.cs() == "def";
1482 /// Parse a macro definition (assumes that is_macro() returned true)
1483 void parse_macro(Parser & p, ostream & os, Context & context)
1485 context.check_layout(os);
1486 Token first = p.curr_token();
1489 string command = first.asInput();
1490 if (first.cs() != "def") {
1492 eat_whitespace(p, os, context, false);
1493 second = p.curr_token();
1494 command += second.asInput();
1495 if (second.cs() != "def") {
1497 eat_whitespace(p, os, context, false);
1498 third = p.curr_token();
1499 command += third.asInput();
1502 eat_whitespace(p, os, context, false);
1503 string const name = p.get_token().cs();
1504 eat_whitespace(p, os, context, false);
1510 while (p.next_token().cat() != catBegin) {
1511 if (p.next_token().cat() == catParameter) {
1516 // followed by number?
1517 if (p.next_token().cat() == catOther) {
1518 char c = p.getChar();
1520 // number = current arity + 1?
1521 if (c == arity + '0' + 1)
1526 paramtext += p.get_token().cs();
1528 paramtext += p.get_token().cs();
1533 // only output simple (i.e. compatible) macro as FormulaMacros
1534 string ert = '\\' + name + ' ' + paramtext + '{' + p.verbatim_item() + '}';
1536 context.check_layout(os);
1537 begin_inset(os, "FormulaMacro");
1538 os << "\n\\def" << ert;
1541 handle_ert(os, command + ert, context);
1544 } // anonymous namespace
1547 void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
1550 Layout const * newlayout = 0;
1551 InsetLayout const * newinsetlayout = 0;
1552 // Store the latest bibliographystyle and nocite{*} option
1553 // (needed for bibtex inset)
1555 string bibliographystyle;
1556 bool const use_natbib = used_packages.find("natbib") != used_packages.end();
1557 bool const use_jurabib = used_packages.find("jurabib") != used_packages.end();
1560 Token const & t = p.get_token();
1563 debugToken(cerr, t, flags);
1566 if (flags & FLAG_ITEM) {
1567 if (t.cat() == catSpace)
1570 flags &= ~FLAG_ITEM;
1571 if (t.cat() == catBegin) {
1572 // skip the brace and collect everything to the next matching
1574 flags |= FLAG_BRACE_LAST;
1578 // handle only this single token, leave the loop if done
1579 flags |= FLAG_LEAVE;
1582 if (t.cat() != catEscape && t.character() == ']' &&
1583 (flags & FLAG_BRACK_LAST))
1585 if (t.cat() == catEnd && (flags & FLAG_BRACE_LAST))
1588 // If there is anything between \end{env} and \begin{env} we
1589 // don't need to output a separator.
1590 if (t.cat() != catSpace && t.cat() != catNewline &&
1591 t.asInput() != "\\begin")
1597 if (t.cat() == catMath) {
1598 // we are inside some text mode thingy, so opening new math is allowed
1599 context.check_layout(os);
1600 begin_inset(os, "Formula ");
1601 Token const & n = p.get_token();
1602 if (n.cat() == catMath && outer) {
1603 // TeX's $$...$$ syntax for displayed math
1605 parse_math(p, os, FLAG_SIMPLE, MATH_MODE);
1607 p.get_token(); // skip the second '$' token
1609 // simple $...$ stuff
1612 parse_math(p, os, FLAG_SIMPLE, MATH_MODE);
1618 else if (t.cat() == catSuper || t.cat() == catSub)
1619 cerr << "catcode " << t << " illegal in text mode\n";
1621 // Basic support for english quotes. This should be
1622 // extended to other quotes, but is not so easy (a
1623 // left english quote is the same as a right german
1625 else if (t.asInput() == "`" && p.next_token().asInput() == "`") {
1626 context.check_layout(os);
1627 begin_inset(os, "Quotes ");
1633 else if (t.asInput() == "'" && p.next_token().asInput() == "'") {
1634 context.check_layout(os);
1635 begin_inset(os, "Quotes ");
1642 else if (t.asInput() == ">" && p.next_token().asInput() == ">") {
1643 context.check_layout(os);
1644 begin_inset(os, "Quotes ");
1651 else if (t.asInput() == "<" && p.next_token().asInput() == "<") {
1652 context.check_layout(os);
1653 begin_inset(os, "Quotes ");
1660 else if (t.asInput() == "<"
1661 && p.next_token().asInput() == "<" && noweb_mode) {
1663 parse_noweb(p, os, context);
1666 else if (t.cat() == catSpace || (t.cat() == catNewline && ! p.isParagraph()))
1667 check_space(p, os, context);
1669 else if (t.character() == '[' && noweb_mode &&
1670 p.next_token().character() == '[') {
1671 // These can contain underscores
1673 string const s = p.getFullOpt() + ']';
1674 if (p.next_token().character() == ']')
1677 cerr << "Warning: Inserting missing ']' in '"
1678 << s << "'." << endl;
1679 handle_ert(os, s, context);
1682 else if (t.cat() == catLetter) {
1683 context.check_layout(os);
1684 // Workaround for bug 4752.
1685 // FIXME: This whole code block needs to be removed
1686 // when the bug is fixed and tex2lyx produces
1687 // the updated file format.
1688 // The replacement algorithm in LyX is so stupid that
1689 // it even translates a phrase if it is part of a word.
1690 bool handled = false;
1691 for (int const * l = known_phrase_lengths; *l; ++l) {
1692 string phrase = t.cs();
1693 for (int i = 1; i < *l && p.next_token().isAlnumASCII(); ++i)
1694 phrase += p.get_token().cs();
1695 if (is_known(phrase, known_coded_phrases)) {
1696 handle_ert(os, phrase, context);
1700 for (size_t i = 1; i < phrase.length(); ++i)
1708 else if (t.cat() == catOther ||
1709 t.cat() == catAlign ||
1710 t.cat() == catParameter) {
1711 // This translates "&" to "\\&" which may be wrong...
1712 context.check_layout(os);
1716 else if (p.isParagraph()) {
1717 if (context.new_layout_allowed)
1718 context.new_paragraph(os);
1720 handle_ert(os, "\\par ", context);
1721 eat_whitespace(p, os, context, true);
1724 else if (t.cat() == catActive) {
1725 context.check_layout(os);
1726 if (t.character() == '~') {
1727 if (context.layout->free_spacing)
1730 begin_inset(os, "space ~\n");
1737 else if (t.cat() == catBegin &&
1738 p.next_token().cat() == catEnd) {
1740 Token const prev = p.prev_token();
1742 if (p.next_token().character() == '`' ||
1743 (prev.character() == '-' &&
1744 p.next_token().character() == '-'))
1745 ; // ignore it in {}`` or -{}-
1747 handle_ert(os, "{}", context);
1751 else if (t.cat() == catBegin) {
1752 context.check_layout(os);
1753 // special handling of font attribute changes
1754 Token const prev = p.prev_token();
1755 Token const next = p.next_token();
1756 TeXFont const oldFont = context.font;
1757 if (next.character() == '[' ||
1758 next.character() == ']' ||
1759 next.character() == '*') {
1761 if (p.next_token().cat() == catEnd) {
1766 handle_ert(os, "{", context);
1767 parse_text_snippet(p, os,
1770 handle_ert(os, "}", context);
1772 } else if (! context.new_layout_allowed) {
1773 handle_ert(os, "{", context);
1774 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1776 handle_ert(os, "}", context);
1777 } else if (is_known(next.cs(), known_sizes)) {
1778 // next will change the size, so we must
1780 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1782 if (!context.atParagraphStart())
1784 << context.font.size << "\n";
1785 } else if (is_known(next.cs(), known_font_families)) {
1786 // next will change the font family, so we
1787 // must reset it here
1788 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1790 if (!context.atParagraphStart())
1792 << context.font.family << "\n";
1793 } else if (is_known(next.cs(), known_font_series)) {
1794 // next will change the font series, so we
1795 // must reset it here
1796 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1798 if (!context.atParagraphStart())
1800 << context.font.series << "\n";
1801 } else if (is_known(next.cs(), known_font_shapes)) {
1802 // next will change the font shape, so we
1803 // must reset it here
1804 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1806 if (!context.atParagraphStart())
1808 << context.font.shape << "\n";
1809 } else if (is_known(next.cs(), known_old_font_families) ||
1810 is_known(next.cs(), known_old_font_series) ||
1811 is_known(next.cs(), known_old_font_shapes)) {
1812 // next will change the font family, series
1813 // and shape, so we must reset it here
1814 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1816 if (!context.atParagraphStart())
1818 << context.font.family
1820 << context.font.series
1822 << context.font.shape << "\n";
1824 handle_ert(os, "{", context);
1825 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1827 handle_ert(os, "}", context);
1831 else if (t.cat() == catEnd) {
1832 if (flags & FLAG_BRACE_LAST) {
1835 cerr << "stray '}' in text\n";
1836 handle_ert(os, "}", context);
1839 else if (t.cat() == catComment)
1840 parse_comment(p, os, t, context);
1843 // control sequences
1846 else if (t.cs() == "(") {
1847 context.check_layout(os);
1848 begin_inset(os, "Formula");
1850 parse_math(p, os, FLAG_SIMPLE2, MATH_MODE);
1855 else if (t.cs() == "[") {
1856 context.check_layout(os);
1857 begin_inset(os, "Formula");
1859 parse_math(p, os, FLAG_EQUATION, MATH_MODE);
1864 else if (t.cs() == "begin")
1865 parse_environment(p, os, outer, last_env, context);
1867 else if (t.cs() == "end") {
1868 if (flags & FLAG_END) {
1869 // eat environment name
1870 string const name = p.getArg('{', '}');
1871 if (name != active_environment())
1872 cerr << "\\end{" + name + "} does not match \\begin{"
1873 + active_environment() + "}\n";
1876 p.error("found 'end' unexpectedly");
1879 else if (t.cs() == "item") {
1882 bool optarg = false;
1883 if (p.next_token().cat() != catEscape &&
1884 p.next_token().character() == '[') {
1885 p.get_token(); // eat '['
1886 s = parse_text_snippet(p, FLAG_BRACK_LAST,
1891 context.check_layout(os);
1892 if (context.has_item) {
1893 // An item in an unknown list-like environment
1894 // FIXME: Do this in check_layout()!
1895 context.has_item = false;
1897 handle_ert(os, "\\item", context);
1899 handle_ert(os, "\\item ", context);
1902 if (context.layout->labeltype != LABEL_MANUAL) {
1903 // LyX does not support \item[\mybullet]
1904 // in itemize environments
1905 handle_ert(os, "[", context);
1907 handle_ert(os, "]", context);
1908 } else if (!s.empty()) {
1909 // The space is needed to separate the
1910 // item from the rest of the sentence.
1912 eat_whitespace(p, os, context, false);
1917 else if (t.cs() == "bibitem") {
1919 context.check_layout(os);
1920 begin_command_inset(os, "bibitem", "bibitem");
1921 os << "label \"" << p.getArg('[', ']') << "\"\n";
1922 os << "key \"" << p.verbatim_item() << "\"\n";
1926 else if (is_macro(p))
1927 parse_macro(p, os, context);
1929 else if (t.cs() == "noindent") {
1931 context.add_par_extra_stuff("\\noindent\n");
1934 else if (t.cs() == "appendix") {
1935 context.add_par_extra_stuff("\\start_of_appendix\n");
1936 // We need to start a new paragraph. Otherwise the
1937 // appendix in 'bla\appendix\chapter{' would start
1939 context.new_paragraph(os);
1940 // We need to make sure that the paragraph is
1941 // generated even if it is empty. Otherwise the
1942 // appendix in '\par\appendix\par\chapter{' would
1944 context.check_layout(os);
1945 // FIXME: This is a hack to prevent paragraph
1946 // deletion if it is empty. Handle this better!
1948 "%dummy comment inserted by tex2lyx to "
1949 "ensure that this paragraph is not empty",
1951 // Both measures above may generate an additional
1952 // empty paragraph, but that does not hurt, because
1953 // whitespace does not matter here.
1954 eat_whitespace(p, os, context, true);
1957 // Starred section headings
1958 // Must attempt to parse "Section*" before "Section".
1959 else if ((p.next_token().asInput() == "*") &&
1960 context.new_layout_allowed &&
1961 (newlayout = findLayout(context.textclass, t.cs() + '*', true))) {
1964 output_command_layout(os, p, outer, context, newlayout);
1968 // Section headings and the like
1969 else if (context.new_layout_allowed &&
1970 (newlayout = findLayout(context.textclass, t.cs(), true))) {
1972 output_command_layout(os, p, outer, context, newlayout);
1976 else if (t.cs() == "caption") {
1978 context.check_layout(os);
1980 begin_inset(os, "Caption\n\n");
1981 Context newcontext(true, context.textclass);
1982 newcontext.font = context.font;
1983 newcontext.check_layout(os);
1984 if (p.next_token().cat() != catEscape &&
1985 p.next_token().character() == '[') {
1986 p.get_token(); // eat '['
1987 begin_inset(os, "OptArg\n");
1988 os << "status collapsed\n";
1989 parse_text_in_inset(p, os, FLAG_BRACK_LAST, outer, context);
1991 eat_whitespace(p, os, context, false);
1993 parse_text(p, os, FLAG_ITEM, outer, context);
1994 context.check_end_layout(os);
1995 // We don't need really a new paragraph, but
1996 // we must make sure that the next item gets a \begin_layout.
1997 context.new_paragraph(os);
2000 newcontext.check_end_layout(os);
2003 else if (t.cs() == "includegraphics") {
2004 bool const clip = p.next_token().asInput() == "*";
2007 string const arg = p.getArg('[', ']');
2008 map<string, string> opts;
2009 vector<string> keys;
2010 split_map(arg, opts, keys);
2012 opts["clip"] = string();
2013 string name = normalize_filename(p.verbatim_item());
2015 string const path = getMasterFilePath();
2016 // We want to preserve relative / absolute filenames,
2017 // therefore path is only used for testing
2018 if (!makeAbsPath(name, path).exists()) {
2019 // The file extension is probably missing.
2020 // Now try to find it out.
2021 string const dvips_name =
2022 find_file(name, path,
2023 known_dvips_graphics_formats);
2024 string const pdftex_name =
2025 find_file(name, path,
2026 known_pdftex_graphics_formats);
2027 if (!dvips_name.empty()) {
2028 if (!pdftex_name.empty()) {
2029 cerr << "This file contains the "
2031 "\"\\includegraphics{"
2033 "However, files\n\""
2034 << dvips_name << "\" and\n\""
2035 << pdftex_name << "\"\n"
2036 "both exist, so I had to make a "
2037 "choice and took the first one.\n"
2038 "Please move the unwanted one "
2039 "someplace else and try again\n"
2040 "if my choice was wrong."
2044 } else if (!pdftex_name.empty()) {
2050 if (makeAbsPath(name, path).exists())
2051 fix_relative_filename(name);
2053 cerr << "Warning: Could not find graphics file '"
2054 << name << "'." << endl;
2056 context.check_layout(os);
2057 begin_inset(os, "Graphics ");
2058 os << "\n\tfilename " << name << '\n';
2059 if (opts.find("width") != opts.end())
2061 << translate_len(opts["width"]) << '\n';
2062 if (opts.find("height") != opts.end())
2064 << translate_len(opts["height"]) << '\n';
2065 if (opts.find("scale") != opts.end()) {
2066 istringstream iss(opts["scale"]);
2070 os << "\tscale " << val << '\n';
2072 if (opts.find("angle") != opts.end()) {
2073 os << "\trotateAngle "
2074 << opts["angle"] << '\n';
2075 vector<string>::const_iterator a =
2076 find(keys.begin(), keys.end(), "angle");
2077 vector<string>::const_iterator s =
2078 find(keys.begin(), keys.end(), "width");
2079 if (s == keys.end())
2080 s = find(keys.begin(), keys.end(), "height");
2081 if (s == keys.end())
2082 s = find(keys.begin(), keys.end(), "scale");
2083 if (s != keys.end() && distance(s, a) > 0)
2084 os << "\tscaleBeforeRotation\n";
2086 if (opts.find("origin") != opts.end()) {
2088 string const opt = opts["origin"];
2089 if (opt.find('l') != string::npos) ss << "left";
2090 if (opt.find('r') != string::npos) ss << "right";
2091 if (opt.find('c') != string::npos) ss << "center";
2092 if (opt.find('t') != string::npos) ss << "Top";
2093 if (opt.find('b') != string::npos) ss << "Bottom";
2094 if (opt.find('B') != string::npos) ss << "Baseline";
2095 if (!ss.str().empty())
2096 os << "\trotateOrigin " << ss.str() << '\n';
2098 cerr << "Warning: Ignoring unknown includegraphics origin argument '" << opt << "'\n";
2100 if (opts.find("keepaspectratio") != opts.end())
2101 os << "\tkeepAspectRatio\n";
2102 if (opts.find("clip") != opts.end())
2104 if (opts.find("draft") != opts.end())
2106 if (opts.find("bb") != opts.end())
2107 os << "\tBoundingBox "
2108 << opts["bb"] << '\n';
2109 int numberOfbbOptions = 0;
2110 if (opts.find("bbllx") != opts.end())
2111 numberOfbbOptions++;
2112 if (opts.find("bblly") != opts.end())
2113 numberOfbbOptions++;
2114 if (opts.find("bburx") != opts.end())
2115 numberOfbbOptions++;
2116 if (opts.find("bbury") != opts.end())
2117 numberOfbbOptions++;
2118 if (numberOfbbOptions == 4)
2119 os << "\tBoundingBox "
2120 << opts["bbllx"] << " " << opts["bblly"] << " "
2121 << opts["bburx"] << " " << opts["bbury"] << '\n';
2122 else if (numberOfbbOptions > 0)
2123 cerr << "Warning: Ignoring incomplete includegraphics boundingbox arguments.\n";
2124 numberOfbbOptions = 0;
2125 if (opts.find("natwidth") != opts.end())
2126 numberOfbbOptions++;
2127 if (opts.find("natheight") != opts.end())
2128 numberOfbbOptions++;
2129 if (numberOfbbOptions == 2)
2130 os << "\tBoundingBox 0bp 0bp "
2131 << opts["natwidth"] << " " << opts["natheight"] << '\n';
2132 else if (numberOfbbOptions > 0)
2133 cerr << "Warning: Ignoring incomplete includegraphics boundingbox arguments.\n";
2134 ostringstream special;
2135 if (opts.find("hiresbb") != opts.end())
2136 special << "hiresbb,";
2137 if (opts.find("trim") != opts.end())
2139 if (opts.find("viewport") != opts.end())
2140 special << "viewport=" << opts["viewport"] << ',';
2141 if (opts.find("totalheight") != opts.end())
2142 special << "totalheight=" << opts["totalheight"] << ',';
2143 if (opts.find("type") != opts.end())
2144 special << "type=" << opts["type"] << ',';
2145 if (opts.find("ext") != opts.end())
2146 special << "ext=" << opts["ext"] << ',';
2147 if (opts.find("read") != opts.end())
2148 special << "read=" << opts["read"] << ',';
2149 if (opts.find("command") != opts.end())
2150 special << "command=" << opts["command"] << ',';
2151 string s_special = special.str();
2152 if (!s_special.empty()) {
2153 // We had special arguments. Remove the trailing ','.
2154 os << "\tspecial " << s_special.substr(0, s_special.size() - 1) << '\n';
2156 // TODO: Handle the unknown settings better.
2157 // Warn about invalid options.
2158 // Check whether some option was given twice.
2162 else if (t.cs() == "footnote" ||
2163 (t.cs() == "thanks" && context.layout->intitle)) {
2165 context.check_layout(os);
2166 begin_inset(os, "Foot\n");
2167 os << "status collapsed\n\n";
2168 parse_text_in_inset(p, os, FLAG_ITEM, false, context);
2172 else if (t.cs() == "marginpar") {
2174 context.check_layout(os);
2175 begin_inset(os, "Marginal\n");
2176 os << "status collapsed\n\n";
2177 parse_text_in_inset(p, os, FLAG_ITEM, false, context);
2181 else if (t.cs() == "ensuremath") {
2183 context.check_layout(os);
2184 string const s = p.verbatim_item();
2185 //FIXME: this never triggers in UTF8
2186 if (s == "\xb1" || s == "\xb3" || s == "\xb2" || s == "\xb5")
2189 handle_ert(os, "\\ensuremath{" + s + "}",
2193 else if (t.cs() == "makeindex" || t.cs() == "maketitle") {
2194 // FIXME: Somehow prevent title layouts if
2195 // "maketitle" was not found
2197 skip_spaces_braces(p);
2200 else if (t.cs() == "tableofcontents") {
2201 context.check_layout(os);
2202 begin_command_inset(os, "toc", "tableofcontents");
2204 skip_spaces_braces(p);
2207 else if (t.cs() == "listoffigures") {
2208 context.check_layout(os);
2209 begin_inset(os, "FloatList figure\n");
2211 skip_spaces_braces(p);
2214 else if (t.cs() == "listoftables") {
2215 context.check_layout(os);
2216 begin_inset(os, "FloatList table\n");
2218 skip_spaces_braces(p);
2221 else if (t.cs() == "listof") {
2222 p.skip_spaces(true);
2223 string const name = p.get_token().cs();
2224 if (context.textclass.floats().typeExist(name)) {
2225 context.check_layout(os);
2226 begin_inset(os, "FloatList ");
2229 p.get_token(); // swallow second arg
2231 handle_ert(os, "\\listof{" + name + "}", context);
2234 else if (t.cs() == "textrm")
2235 parse_text_attributes(p, os, FLAG_ITEM, outer,
2236 context, "\\family",
2237 context.font.family, "roman");
2239 else if (t.cs() == "textsf")
2240 parse_text_attributes(p, os, FLAG_ITEM, outer,
2241 context, "\\family",
2242 context.font.family, "sans");
2244 else if (t.cs() == "texttt")
2245 parse_text_attributes(p, os, FLAG_ITEM, outer,
2246 context, "\\family",
2247 context.font.family, "typewriter");
2249 else if (t.cs() == "textmd")
2250 parse_text_attributes(p, os, FLAG_ITEM, outer,
2251 context, "\\series",
2252 context.font.series, "medium");
2254 else if (t.cs() == "textbf")
2255 parse_text_attributes(p, os, FLAG_ITEM, outer,
2256 context, "\\series",
2257 context.font.series, "bold");
2259 else if (t.cs() == "textup")
2260 parse_text_attributes(p, os, FLAG_ITEM, outer,
2262 context.font.shape, "up");
2264 else if (t.cs() == "textit")
2265 parse_text_attributes(p, os, FLAG_ITEM, outer,
2267 context.font.shape, "italic");
2269 else if (t.cs() == "textsl")
2270 parse_text_attributes(p, os, FLAG_ITEM, outer,
2272 context.font.shape, "slanted");
2274 else if (t.cs() == "textsc")
2275 parse_text_attributes(p, os, FLAG_ITEM, outer,
2277 context.font.shape, "smallcaps");
2279 else if (t.cs() == "textnormal" || t.cs() == "normalfont") {
2280 context.check_layout(os);
2281 TeXFont oldFont = context.font;
2282 context.font.init();
2283 context.font.size = oldFont.size;
2284 os << "\n\\family " << context.font.family << "\n";
2285 os << "\n\\series " << context.font.series << "\n";
2286 os << "\n\\shape " << context.font.shape << "\n";
2287 if (t.cs() == "textnormal") {
2288 parse_text_snippet(p, os, FLAG_ITEM, outer, context);
2289 output_font_change(os, context.font, oldFont);
2290 context.font = oldFont;
2292 eat_whitespace(p, os, context, false);
2295 else if (t.cs() == "textcolor") {
2296 // scheme is \textcolor{color name}{text}
2297 string const color = p.verbatim_item();
2298 // we only support the predefined colors of the color package
2299 if (color == "black" || color == "blue" || color == "cyan"
2300 || color == "green" || color == "magenta" || color == "red"
2301 || color == "white" || color == "yellow") {
2302 context.check_layout(os);
2303 os << "\n\\color " << color << "\n";
2304 parse_text_snippet(p, os, FLAG_ITEM, outer, context);
2305 context.check_layout(os);
2306 os << "\n\\color inherit\n";
2308 // for custom defined colors
2309 handle_ert(os, t.asInput() + "{" + color + "}", context);
2312 else if (t.cs() == "underbar") {
2313 // Do NOT handle \underline.
2314 // \underbar cuts through y, g, q, p etc.,
2315 // \underline does not.
2316 context.check_layout(os);
2317 os << "\n\\bar under\n";
2318 parse_text_snippet(p, os, FLAG_ITEM, outer, context);
2319 context.check_layout(os);
2320 os << "\n\\bar default\n";
2323 else if (t.cs() == "emph" || t.cs() == "noun") {
2324 context.check_layout(os);
2325 os << "\n\\" << t.cs() << " on\n";
2326 parse_text_snippet(p, os, FLAG_ITEM, outer, context);
2327 context.check_layout(os);
2328 os << "\n\\" << t.cs() << " default\n";
2331 else if (t.cs() == "lyxline") {
2332 context.check_layout(os);
2336 else if (is_known(t.cs(), known_phrases) ||
2337 (t.cs() == "protect" &&
2338 p.next_token().cat() == catEscape &&
2339 is_known(p.next_token().cs(), known_phrases))) {
2340 // LyX sometimes puts a \protect in front, so we have to ignore it
2341 // FIXME: This needs to be changed when bug 4752 is fixed.
2342 char const * const * where = is_known(
2343 t.cs() == "protect" ? p.get_token().cs() : t.cs(),
2345 context.check_layout(os);
2346 os << known_coded_phrases[where - known_phrases];
2347 skip_spaces_braces(p);
2350 else if (is_known(t.cs(), known_ref_commands)) {
2351 string const opt = p.getOpt();
2353 context.check_layout(os);
2354 begin_command_inset(os, "ref", t.cs());
2355 // LyX cannot handle newlines in a latex command
2356 // FIXME: Move the substitution into parser::getOpt()?
2357 os << "reference \""
2358 << subst(p.verbatim_item(), "\n", " ")
2362 // LyX does not support optional arguments of ref commands
2363 handle_ert(os, t.asInput() + '[' + opt + "]{" +
2364 p.verbatim_item() + "}", context);
2369 else if (use_natbib &&
2370 is_known(t.cs(), known_natbib_commands) &&
2371 ((t.cs() != "citefullauthor" &&
2372 t.cs() != "citeyear" &&
2373 t.cs() != "citeyearpar") ||
2374 p.next_token().asInput() != "*")) {
2375 context.check_layout(os);
2376 string command = t.cs();
2377 if (p.next_token().asInput() == "*") {
2381 if (command == "citefullauthor")
2382 // alternative name for "\\citeauthor*"
2383 command = "citeauthor*";
2385 // text before the citation
2387 // text after the citation
2389 get_cite_arguments(p, true, before, after);
2391 if (command == "cite") {
2392 // \cite without optional argument means
2393 // \citet, \cite with at least one optional
2394 // argument means \citep.
2395 if (before.empty() && after.empty())
2400 if (before.empty() && after == "[]")
2401 // avoid \citet[]{a}
2403 else if (before == "[]" && after == "[]") {
2404 // avoid \citet[][]{a}
2408 // remove the brackets around after and before
2409 if (!after.empty()) {
2411 after.erase(after.length() - 1, 1);
2412 // LyX cannot handle newlines in the parameter
2413 after = subst(after, "\n", " ");
2415 if (!before.empty()) {
2417 before.erase(before.length() - 1, 1);
2418 // LyX cannot handle newlines in the parameter
2419 before = subst(before, "\n", " ");
2421 begin_command_inset(os, "citation", command);
2422 os << "after " << '"' << after << '"' << "\n";
2423 os << "before " << '"' << before << '"' << "\n";
2424 os << "key " << '"' << p.verbatim_item() << '"' << "\n";
2428 else if (use_jurabib &&
2429 is_known(t.cs(), known_jurabib_commands) &&
2430 (t.cs() == "cite" || p.next_token().asInput() != "*")) {
2431 context.check_layout(os);
2432 string command = t.cs();
2433 if (p.next_token().asInput() == "*") {
2437 char argumentOrder = '\0';
2438 vector<string> const & options = used_packages["jurabib"];
2439 if (find(options.begin(), options.end(),
2440 "natbiborder") != options.end())
2441 argumentOrder = 'n';
2442 else if (find(options.begin(), options.end(),
2443 "jurabiborder") != options.end())
2444 argumentOrder = 'j';
2446 // text before the citation
2448 // text after the citation
2450 get_cite_arguments(p, argumentOrder != 'j', before, after);
2452 string const citation = p.verbatim_item();
2453 if (!before.empty() && argumentOrder == '\0') {
2454 cerr << "Warning: Assuming argument order "
2455 "of jurabib version 0.6 for\n'"
2456 << command << before << after << '{'
2457 << citation << "}'.\n"
2458 "Add 'jurabiborder' to the jurabib "
2459 "package options if you used an\n"
2460 "earlier jurabib version." << endl;
2462 if (!after.empty()) {
2464 after.erase(after.length() - 1, 1);
2466 if (!before.empty()) {
2468 before.erase(before.length() - 1, 1);
2470 begin_command_inset(os, "citation", command);
2471 os << "after " << '"' << after << '"' << "\n";
2472 os << "before " << '"' << before << '"' << "\n";
2473 os << "key " << '"' << citation << '"' << "\n";
2477 else if (t.cs() == "cite"
2478 || t.cs() == "nocite") {
2479 context.check_layout(os);
2480 // LyX cannot handle newlines in a latex command
2481 string after = subst(p.getArg('[', ']'), "\n", " ");
2482 string key = subst(p.verbatim_item(), "\n", " ");
2483 // store the case that it is "\nocite{*}" to use it later for
2486 begin_command_inset(os, "citation", t.cs());
2487 os << "after " << '"' << after << '"' << "\n";
2488 os << "key " << '"' << key << '"' << "\n";
2490 } else if (t.cs() == "nocite")
2494 else if (t.cs() == "index") {
2495 context.check_layout(os);
2496 begin_inset(os, "Index\n");
2497 os << "status collapsed\n";
2498 parse_text_in_inset(p, os, FLAG_ITEM, false, context, "Index");
2502 else if (t.cs() == "nomenclature") {
2503 context.check_layout(os);
2504 begin_command_inset(os, "nomenclature", "nomenclature");
2505 // LyX cannot handle newlines in a latex command
2506 string prefix = subst(p.getArg('[', ']'), "\n", " ");
2507 if (!prefix.empty())
2508 os << "prefix " << '"' << prefix << '"' << "\n";
2509 os << "symbol " << '"' << subst(p.verbatim_item(), "\n", " ") << '"' << "\n";
2510 os << "description " << '"' << subst(p.verbatim_item(), "\n", " ") << '"' << "\n";
2514 else if (t.cs() == "label") {
2515 context.check_layout(os);
2516 begin_command_inset(os, "label", "label");
2517 // LyX cannot handle newlines in a latex command
2518 os << "name " << '"' << subst(p.verbatim_item(), "\n", " ") << '"' << "\n";
2522 else if (t.cs() == "printindex") {
2523 context.check_layout(os);
2524 begin_command_inset(os, "index_print", "printindex");
2526 skip_spaces_braces(p);
2529 else if (t.cs() == "printnomenclature") {
2530 context.check_layout(os);
2531 begin_command_inset(os, "nomencl_print", "printnomenclature");
2533 skip_spaces_braces(p);
2536 else if (LYX_FORMAT >= 408 &&
2537 (t.cs() == "textsuperscript" || t.cs() == "textsubscript")) {
2538 context.check_layout(os);
2539 begin_inset(os, "script ");
2540 os << t.cs().substr(4) << '\n';
2541 parse_text_in_inset(p, os, FLAG_ITEM, false, context);
2545 else if (is_known(t.cs(), known_quotes)) {
2546 char const * const * where = is_known(t.cs(), known_quotes);
2547 context.check_layout(os);
2548 begin_inset(os, "Quotes ");
2549 os << known_coded_quotes[where - known_quotes];
2551 // LyX adds {} after the quote, so we have to eat
2552 // spaces here if there are any before a possible
2554 eat_whitespace(p, os, context, false);
2558 else if (is_known(t.cs(), known_sizes) &&
2559 context.new_layout_allowed) {
2560 char const * const * where = is_known(t.cs(), known_sizes);
2561 context.check_layout(os);
2562 TeXFont const oldFont = context.font;
2563 context.font.size = known_coded_sizes[where - known_sizes];
2564 output_font_change(os, oldFont, context.font);
2565 eat_whitespace(p, os, context, false);
2568 else if (is_known(t.cs(), known_font_families) &&
2569 context.new_layout_allowed) {
2570 char const * const * where =
2571 is_known(t.cs(), known_font_families);
2572 context.check_layout(os);
2573 TeXFont const oldFont = context.font;
2574 context.font.family =
2575 known_coded_font_families[where - known_font_families];
2576 output_font_change(os, oldFont, context.font);
2577 eat_whitespace(p, os, context, false);
2580 else if (is_known(t.cs(), known_font_series) &&
2581 context.new_layout_allowed) {
2582 char const * const * where =
2583 is_known(t.cs(), known_font_series);
2584 context.check_layout(os);
2585 TeXFont const oldFont = context.font;
2586 context.font.series =
2587 known_coded_font_series[where - known_font_series];
2588 output_font_change(os, oldFont, context.font);
2589 eat_whitespace(p, os, context, false);
2592 else if (is_known(t.cs(), known_font_shapes) &&
2593 context.new_layout_allowed) {
2594 char const * const * where =
2595 is_known(t.cs(), known_font_shapes);
2596 context.check_layout(os);
2597 TeXFont const oldFont = context.font;
2598 context.font.shape =
2599 known_coded_font_shapes[where - known_font_shapes];
2600 output_font_change(os, oldFont, context.font);
2601 eat_whitespace(p, os, context, false);
2603 else if (is_known(t.cs(), known_old_font_families) &&
2604 context.new_layout_allowed) {
2605 char const * const * where =
2606 is_known(t.cs(), known_old_font_families);
2607 context.check_layout(os);
2608 TeXFont const oldFont = context.font;
2609 context.font.init();
2610 context.font.size = oldFont.size;
2611 context.font.family =
2612 known_coded_font_families[where - known_old_font_families];
2613 output_font_change(os, oldFont, context.font);
2614 eat_whitespace(p, os, context, false);
2617 else if (is_known(t.cs(), known_old_font_series) &&
2618 context.new_layout_allowed) {
2619 char const * const * where =
2620 is_known(t.cs(), known_old_font_series);
2621 context.check_layout(os);
2622 TeXFont const oldFont = context.font;
2623 context.font.init();
2624 context.font.size = oldFont.size;
2625 context.font.series =
2626 known_coded_font_series[where - known_old_font_series];
2627 output_font_change(os, oldFont, context.font);
2628 eat_whitespace(p, os, context, false);
2631 else if (is_known(t.cs(), known_old_font_shapes) &&
2632 context.new_layout_allowed) {
2633 char const * const * where =
2634 is_known(t.cs(), known_old_font_shapes);
2635 context.check_layout(os);
2636 TeXFont const oldFont = context.font;
2637 context.font.init();
2638 context.font.size = oldFont.size;
2639 context.font.shape =
2640 known_coded_font_shapes[where - known_old_font_shapes];
2641 output_font_change(os, oldFont, context.font);
2642 eat_whitespace(p, os, context, false);
2645 else if (t.cs() == "selectlanguage") {
2646 context.check_layout(os);
2647 // save the language for the case that a
2648 // \foreignlanguage is used
2650 context.font.language = babel2lyx(p.verbatim_item());
2651 os << "\n\\lang " << context.font.language << "\n";
2654 else if (t.cs() == "foreignlanguage") {
2655 string const lang = babel2lyx(p.verbatim_item());
2656 parse_text_attributes(p, os, FLAG_ITEM, outer,
2658 context.font.language, lang);
2661 else if (t.cs() == "inputencoding") {
2662 // nothing to write here
2663 string const enc = subst(p.verbatim_item(), "\n", " ");
2667 else if (t.cs() == "ldots") {
2668 context.check_layout(os);
2669 os << "\\SpecialChar \\ldots{}\n";
2670 skip_spaces_braces(p);
2673 else if (t.cs() == "lyxarrow") {
2674 context.check_layout(os);
2675 os << "\\SpecialChar \\menuseparator\n";
2676 skip_spaces_braces(p);
2679 else if (t.cs() == "textcompwordmark") {
2680 context.check_layout(os);
2681 os << "\\SpecialChar \\textcompwordmark{}\n";
2682 skip_spaces_braces(p);
2685 else if (t.cs() == "slash") {
2686 context.check_layout(os);
2687 os << "\\SpecialChar \\slash{}\n";
2688 skip_spaces_braces(p);
2691 else if (t.cs() == "nobreakdash" && p.next_token().asInput() == "-") {
2692 context.check_layout(os);
2693 os << "\\SpecialChar \\nobreakdash-\n";
2697 else if (t.cs() == "textquotedbl") {
2698 context.check_layout(os);
2703 else if (t.cs() == "@" && p.next_token().asInput() == ".") {
2704 context.check_layout(os);
2705 os << "\\SpecialChar \\@.\n";
2709 else if (t.cs() == "-") {
2710 context.check_layout(os);
2711 os << "\\SpecialChar \\-\n";
2714 else if (t.cs() == "textasciitilde") {
2715 context.check_layout(os);
2717 skip_spaces_braces(p);
2720 else if (t.cs() == "textasciicircum") {
2721 context.check_layout(os);
2723 skip_spaces_braces(p);
2726 else if (t.cs() == "textbackslash") {
2727 context.check_layout(os);
2728 os << "\n\\backslash\n";
2729 skip_spaces_braces(p);
2732 else if (t.cs() == "_" || t.cs() == "&" || t.cs() == "#"
2733 || t.cs() == "$" || t.cs() == "{" || t.cs() == "}"
2735 context.check_layout(os);
2739 else if (t.cs() == "char") {
2740 context.check_layout(os);
2741 if (p.next_token().character() == '`') {
2743 if (p.next_token().cs() == "\"") {
2748 handle_ert(os, "\\char`", context);
2751 handle_ert(os, "\\char", context);
2755 else if (t.cs() == "verb") {
2756 context.check_layout(os);
2757 char const delimiter = p.next_token().character();
2758 string const arg = p.getArg(delimiter, delimiter);
2760 oss << "\\verb" << delimiter << arg << delimiter;
2761 handle_ert(os, oss.str(), context);
2764 // Problem: \= creates a tabstop inside the tabbing environment
2765 // and else an accent. In the latter case we really would want
2766 // \={o} instead of \= o.
2767 else if (t.cs() == "=" && (flags & FLAG_TABBING))
2768 handle_ert(os, t.asInput(), context);
2770 // accents (see Table 6 in Comprehensive LaTeX Symbol List)
2771 else if (t.cs().size() == 1
2772 && contains("\"'.=^`bcdHkrtuv~", t.cs())) {
2773 context.check_layout(os);
2774 // try to see whether the string is in unicodesymbols
2776 string command = t.asInput() + "{"
2777 + trim(p.verbatim_item())
2779 docstring s = encodings.fromLaTeXCommand(from_utf8(command), rem);
2782 cerr << "When parsing " << command
2783 << ", result is " << to_utf8(s)
2784 << "+" << to_utf8(rem) << endl;
2787 // we did not find a non-ert version
2788 handle_ert(os, command, context);
2791 else if (t.cs() == "\\") {
2792 context.check_layout(os);
2794 handle_ert(os, "\\\\" + p.getOpt(), context);
2795 else if (p.next_token().asInput() == "*") {
2797 // getOpt() eats the following space if there
2798 // is no optional argument, but that is OK
2799 // here since it has no effect in the output.
2800 handle_ert(os, "\\\\*" + p.getOpt(), context);
2803 begin_inset(os, "Newline newline");
2808 else if (t.cs() == "newline" ||
2809 (t.cs() == "linebreak" && !p.hasOpt())) {
2810 context.check_layout(os);
2811 begin_inset(os, "Newline ");
2814 skip_spaces_braces(p);
2817 else if (t.cs() == "input" || t.cs() == "include"
2818 || t.cs() == "verbatiminput") {
2819 string name = t.cs();
2820 if (t.cs() == "verbatiminput"
2821 && p.next_token().asInput() == "*")
2822 name += p.get_token().asInput();
2823 context.check_layout(os);
2824 string filename(normalize_filename(p.getArg('{', '}')));
2825 string const path = getMasterFilePath();
2826 // We want to preserve relative / absolute filenames,
2827 // therefore path is only used for testing
2828 if ((t.cs() == "include" || t.cs() == "input") &&
2829 !makeAbsPath(filename, path).exists()) {
2830 // The file extension is probably missing.
2831 // Now try to find it out.
2832 string const tex_name =
2833 find_file(filename, path,
2834 known_tex_extensions);
2835 if (!tex_name.empty())
2836 filename = tex_name;
2838 bool external = false;
2840 if (makeAbsPath(filename, path).exists()) {
2841 string const abstexname =
2842 makeAbsPath(filename, path).absFileName();
2843 string const abslyxname =
2844 changeExtension(abstexname, ".lyx");
2845 string const absfigname =
2846 changeExtension(abstexname, ".fig");
2847 fix_relative_filename(filename);
2848 string const lyxname =
2849 changeExtension(filename, ".lyx");
2851 external = FileName(absfigname).exists();
2852 if (t.cs() == "input") {
2853 string const ext = getExtension(abstexname);
2855 // Combined PS/LaTeX:
2856 // x.eps, x.pstex_t (old xfig)
2857 // x.pstex, x.pstex_t (new xfig, e.g. 3.2.5)
2858 FileName const absepsname(
2859 changeExtension(abstexname, ".eps"));
2860 FileName const abspstexname(
2861 changeExtension(abstexname, ".pstex"));
2862 bool const xfigeps =
2863 (absepsname.exists() ||
2864 abspstexname.exists()) &&
2867 // Combined PDF/LaTeX:
2868 // x.pdf, x.pdftex_t (old xfig)
2869 // x.pdf, x.pdf_t (new xfig, e.g. 3.2.5)
2870 FileName const abspdfname(
2871 changeExtension(abstexname, ".pdf"));
2872 bool const xfigpdf =
2873 abspdfname.exists() &&
2874 (ext == "pdftex_t" || ext == "pdf_t");
2878 // Combined PS/PDF/LaTeX:
2879 // x_pspdftex.eps, x_pspdftex.pdf, x.pspdftex
2880 string const absbase2(
2881 removeExtension(abstexname) + "_pspdftex");
2882 FileName const abseps2name(
2883 addExtension(absbase2, ".eps"));
2884 FileName const abspdf2name(
2885 addExtension(absbase2, ".pdf"));
2886 bool const xfigboth =
2887 abspdf2name.exists() &&
2888 abseps2name.exists() && ext == "pspdftex";
2890 xfig = xfigpdf || xfigeps || xfigboth;
2891 external = external && xfig;
2894 outname = changeExtension(filename, ".fig");
2896 // Don't try to convert, the result
2897 // would be full of ERT.
2899 } else if (t.cs() != "verbatiminput" &&
2900 tex2lyx(abstexname, FileName(abslyxname),
2907 cerr << "Warning: Could not find included file '"
2908 << filename << "'." << endl;
2912 begin_inset(os, "External\n");
2913 os << "\ttemplate XFig\n"
2914 << "\tfilename " << outname << '\n';
2916 begin_command_inset(os, "include", name);
2917 os << "preview false\n"
2918 "filename \"" << outname << "\"\n";
2923 else if (t.cs() == "bibliographystyle") {
2924 // store new bibliographystyle
2925 bibliographystyle = p.verbatim_item();
2926 // If any other command than \bibliography and
2927 // \nocite{*} follows, we need to output the style
2928 // (because it might be used by that command).
2929 // Otherwise, it will automatically be output by LyX.
2932 for (Token t2 = p.get_token(); p.good(); t2 = p.get_token()) {
2933 if (t2.cat() == catBegin)
2935 if (t2.cat() != catEscape)
2937 if (t2.cs() == "nocite") {
2938 if (p.getArg('{', '}') == "*")
2940 } else if (t2.cs() == "bibliography")
2947 "\\bibliographystyle{" + bibliographystyle + '}',
2952 else if (t.cs() == "bibliography") {
2953 context.check_layout(os);
2954 begin_command_inset(os, "bibtex", "bibtex");
2955 if (!btprint.empty()) {
2956 os << "btprint " << '"' << "btPrintAll" << '"' << "\n";
2957 // clear the string because the next BibTeX inset can be without the
2958 // \nocite{*} option
2961 os << "bibfiles " << '"' << p.verbatim_item() << '"' << "\n";
2962 // Do we have a bibliographystyle set?
2963 if (!bibliographystyle.empty())
2964 os << "options " << '"' << bibliographystyle << '"' << "\n";
2968 else if (t.cs() == "parbox")
2969 parse_box(p, os, 0, FLAG_ITEM, outer, context, "", "", t.cs());
2971 else if (t.cs() == "ovalbox" || t.cs() == "Ovalbox" ||
2972 t.cs() == "shadowbox" || t.cs() == "doublebox")
2973 parse_outer_box(p, os, FLAG_ITEM, outer, context, t.cs(), "");
2975 else if (t.cs() == "framebox") {
2976 string special = p.getFullOpt();
2977 special += p.getOpt();
2978 parse_outer_box(p, os, FLAG_ITEM, outer, context, t.cs(), special);
2981 //\makebox() is part of the picture environment and different from \makebox{}
2982 //\makebox{} will be parsed by parse_box when bug 2956 is fixed
2983 else if (t.cs() == "makebox") {
2984 string arg = t.asInput();
2985 if (p.next_token().character() == '(')
2986 //the syntax is: \makebox(x,y)[position]{content}
2987 arg += p.getFullParentheseArg();
2989 //the syntax is: \makebox[width][position]{content}
2990 arg += p.getFullOpt();
2991 handle_ert(os, arg + p.getFullOpt(), context);
2994 else if (t.cs() == "smallskip" ||
2995 t.cs() == "medskip" ||
2996 t.cs() == "bigskip" ||
2997 t.cs() == "vfill") {
2998 context.check_layout(os);
2999 begin_inset(os, "VSpace ");
3002 skip_spaces_braces(p);
3005 else if (is_known(t.cs(), known_spaces)) {
3006 char const * const * where = is_known(t.cs(), known_spaces);
3007 context.check_layout(os);
3008 begin_inset(os, "space ");
3009 os << '\\' << known_coded_spaces[where - known_spaces]
3012 // LaTeX swallows whitespace after all spaces except
3013 // "\\,". We have to do that here, too, because LyX
3014 // adds "{}" which would make the spaces significant.
3016 eat_whitespace(p, os, context, false);
3017 // LyX adds "{}" after all spaces except "\\ " and
3018 // "\\,", so we have to remove "{}".
3019 // "\\,{}" is equivalent to "\\," in LaTeX, so we
3020 // remove the braces after "\\,", too.
3025 else if (t.cs() == "newpage" ||
3026 (t.cs() == "pagebreak" && !p.hasOpt()) ||
3027 t.cs() == "clearpage" ||
3028 t.cs() == "cleardoublepage") {
3029 context.check_layout(os);
3030 begin_inset(os, "Newpage ");
3033 skip_spaces_braces(p);
3036 else if (t.cs() == "DeclareRobustCommand" ||
3037 t.cs() == "DeclareRobustCommandx" ||
3038 t.cs() == "newcommand" ||
3039 t.cs() == "newcommandx" ||
3040 t.cs() == "providecommand" ||
3041 t.cs() == "providecommandx" ||
3042 t.cs() == "renewcommand" ||
3043 t.cs() == "renewcommandx") {
3044 // DeclareRobustCommand, DeclareRobustCommandx,
3045 // providecommand and providecommandx could be handled
3046 // by parse_command(), but we need to call
3047 // add_known_command() here.
3048 string name = t.asInput();
3049 if (p.next_token().asInput() == "*") {
3050 // Starred form. Eat '*'
3054 string const command = p.verbatim_item();
3055 string const opt1 = p.getFullOpt();
3056 string const opt2 = p.getFullOpt();
3057 add_known_command(command, opt1, !opt2.empty());
3058 string const ert = name + '{' + command + '}' +
3060 '{' + p.verbatim_item() + '}';
3062 if (t.cs() == "DeclareRobustCommand" ||
3063 t.cs() == "DeclareRobustCommandx" ||
3064 t.cs() == "providecommand" ||
3065 t.cs() == "providecommandx" ||
3066 name[name.length()-1] == '*')
3067 handle_ert(os, ert, context);
3069 context.check_layout(os);
3070 begin_inset(os, "FormulaMacro");
3076 else if (t.cs() == "let" && p.next_token().asInput() != "*") {
3077 // let could be handled by parse_command(),
3078 // but we need to call add_known_command() here.
3079 string ert = t.asInput();
3082 if (p.next_token().cat() == catBegin) {
3083 name = p.verbatim_item();
3084 ert += '{' + name + '}';
3086 name = p.verbatim_item();
3091 if (p.next_token().cat() == catBegin) {
3092 command = p.verbatim_item();
3093 ert += '{' + command + '}';
3095 command = p.verbatim_item();
3098 // If command is known, make name known too, to parse
3099 // its arguments correctly. For this reason we also
3100 // have commands in syntax.default that are hardcoded.
3101 CommandMap::iterator it = known_commands.find(command);
3102 if (it != known_commands.end())
3103 known_commands[t.asInput()] = it->second;
3104 handle_ert(os, ert, context);
3107 else if (t.cs() == "hspace" || t.cs() == "vspace") {
3108 bool starred = false;
3109 if (p.next_token().asInput() == "*") {
3113 string name = t.asInput();
3114 string const length = p.verbatim_item();
3117 bool valid = splitLatexLength(length, valstring, unit);
3118 bool known_hspace = false;
3119 bool known_vspace = false;
3120 bool known_unit = false;
3123 istringstream iss(valstring);
3126 if (t.cs()[0] == 'h') {
3127 if (unit == "\\fill") {
3132 known_hspace = true;
3135 if (unit == "\\smallskipamount") {
3137 known_vspace = true;
3138 } else if (unit == "\\medskipamount") {
3140 known_vspace = true;
3141 } else if (unit == "\\bigskipamount") {
3143 known_vspace = true;
3144 } else if (unit == "\\fill") {
3146 known_vspace = true;
3150 if (!known_hspace && !known_vspace) {
3151 switch (unitFromString(unit)) {
3172 if (t.cs()[0] == 'h' && (known_unit || known_hspace)) {
3173 // Literal horizontal length or known variable
3174 context.check_layout(os);
3175 begin_inset(os, "space ");
3183 if (known_unit && !known_hspace)
3185 << translate_len(length);
3187 } else if (known_unit || known_vspace) {
3188 // Literal vertical length or known variable
3189 context.check_layout(os);
3190 begin_inset(os, "VSpace ");
3198 // LyX can't handle other length variables in Inset VSpace/space
3203 handle_ert(os, name + '{' + unit + '}', context);
3204 else if (value == -1.0)
3205 handle_ert(os, name + "{-" + unit + '}', context);
3207 handle_ert(os, name + '{' + valstring + unit + '}', context);
3209 handle_ert(os, name + '{' + length + '}', context);
3213 // The single '=' is meant here.
3214 else if ((newinsetlayout = findInsetLayout(context.textclass, t.cs(), true))) {
3216 context.check_layout(os);
3217 begin_inset(os, "Flex ");
3218 os << to_utf8(newinsetlayout->name()) << '\n'
3219 << "status collapsed\n";
3220 parse_text_in_inset(p, os, FLAG_ITEM, false, context, newinsetlayout);
3225 // try to see whether the string is in unicodesymbols
3226 // Only use text mode commands, since we are in text mode here,
3227 // and math commands may be invalid (bug 6797)
3229 docstring s = encodings.fromLaTeXCommand(from_utf8(t.asInput()),
3230 rem, Encodings::TEXT_CMD);
3233 cerr << "When parsing " << t.cs()
3234 << ", result is " << to_utf8(s)
3235 << "+" << to_utf8(rem) << endl;
3236 context.check_layout(os);
3238 skip_spaces_braces(p);
3240 //cerr << "#: " << t << " mode: " << mode << endl;
3241 // heuristic: read up to next non-nested space
3243 string s = t.asInput();
3244 string z = p.verbatim_item();
3245 while (p.good() && z != " " && z.size()) {
3246 //cerr << "read: " << z << endl;
3248 z = p.verbatim_item();
3250 cerr << "found ERT: " << s << endl;
3251 handle_ert(os, s + ' ', context);
3254 string name = t.asInput();
3255 if (p.next_token().asInput() == "*") {
3256 // Starred commands like \vspace*{}
3257 p.get_token(); // Eat '*'
3260 if (!parse_command(name, p, os, outer, context))
3261 handle_ert(os, name, context);
3265 if (flags & FLAG_LEAVE) {
3266 flags &= ~FLAG_LEAVE;