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.
20 #include "FloatList.h"
24 #include "support/lstrings.h"
25 #include "support/convert.h"
26 #include "support/filetools.h"
38 using std::ostringstream;
39 using std::istringstream;
45 using support::addExtension;
46 using support::changeExtension;
47 using support::FileName;
48 using support::makeAbsPath;
49 using support::makeRelPath;
51 using support::suffixIs;
52 using support::contains;
56 void parse_text_in_inset(Parser & p, ostream & os, unsigned flags, bool outer,
57 Context const & context)
59 Context newcontext(true, context.textclass);
60 newcontext.font = context.font;
61 parse_text(p, os, flags, outer, newcontext);
62 newcontext.check_end_layout(os);
68 /// parses a paragraph snippet, useful for example for \\emph{...}
69 void parse_text_snippet(Parser & p, ostream & os, unsigned flags, bool outer,
72 Context newcontext(context);
73 // Don't inherit the extra stuff
74 newcontext.extra_stuff.clear();
75 parse_text(p, os, flags, outer, newcontext);
76 // Make sure that we don't create invalid .lyx files
77 context.need_layout = newcontext.need_layout;
78 context.need_end_layout = newcontext.need_end_layout;
83 * Thin wrapper around parse_text_snippet() using a string.
85 * We completely ignore \c context.need_layout and \c context.need_end_layout,
86 * because our return value is not used directly (otherwise the stream version
87 * of parse_text_snippet() could be used). That means that the caller needs
88 * to do layout management manually.
89 * This is intended to parse text that does not create any layout changes.
91 string parse_text_snippet(Parser & p, unsigned flags, const bool outer,
94 Context newcontext(context);
95 newcontext.need_layout = false;
96 newcontext.need_end_layout = false;
97 newcontext.new_layout_allowed = false;
98 // Avoid warning by Context::~Context()
99 newcontext.extra_stuff.clear();
101 parse_text_snippet(p, os, flags, outer, newcontext);
106 char const * const known_latex_commands[] = { "ref", "cite", "label",
107 "index", "pageref", "url", "vref", "vpageref", "prettyref",
112 * We can't put these into known_latex_commands because the argument order
113 * is reversed in lyx if there are 2 arguments.
114 * The starred forms are also known.
116 char const * const known_natbib_commands[] = { "cite", "citet", "citep",
117 "citealt", "citealp", "citeauthor", "citeyear", "citeyearpar",
118 "citefullauthor", "Citet", "Citep", "Citealt", "Citealp", "Citeauthor", 0 };
122 * We can't put these into known_latex_commands because the argument order
123 * is reversed in lyx if there are 2 arguments.
124 * No starred form other than "cite*" known.
126 char const * const known_jurabib_commands[] = { "cite", "citet", "citep",
127 "citealt", "citealp", "citeauthor", "citeyear", "citeyearpar",
128 // jurabib commands not (yet) supported by LyX:
130 // "footcite", "footcitet", "footcitep", "footcitealt", "footcitealp",
131 // "footciteauthor", "footciteyear", "footciteyearpar",
132 "citefield", "citetitle", "cite*", 0 };
134 /// LaTeX names for quotes
135 char const * const known_quotes[] = { "glqq", "grqq", "quotedblbase",
136 "textquotedblleft", "quotesinglbase", "guilsinglleft", "guilsinglright", 0};
138 /// the same as known_quotes with .lyx names
139 char const * const known_coded_quotes[] = { "gld", "grd", "gld",
140 "grd", "gls", "fls", "frs", 0};
142 /// LaTeX names for font sizes
143 char const * const known_sizes[] = { "tiny", "scriptsize", "footnotesize",
144 "small", "normalsize", "large", "Large", "LARGE", "huge", "Huge", 0};
146 /// the same as known_sizes with .lyx names
147 char const * const known_coded_sizes[] = { "default", "tiny", "scriptsize", "footnotesize",
148 "small", "normal", "large", "larger", "largest", "huge", "giant", 0};
150 /// LaTeX 2.09 names for font families
151 char const * const known_old_font_families[] = { "rm", "sf", "tt", 0};
153 /// LaTeX names for font families
154 char const * const known_font_families[] = { "rmfamily", "sffamily",
157 /// the same as known_old_font_families and known_font_families with .lyx names
158 char const * const known_coded_font_families[] = { "roman", "sans",
161 /// LaTeX 2.09 names for font series
162 char const * const known_old_font_series[] = { "bf", 0};
164 /// LaTeX names for font series
165 char const * const known_font_series[] = { "bfseries", "mdseries", 0};
167 /// the same as known_old_font_series and known_font_series with .lyx names
168 char const * const known_coded_font_series[] = { "bold", "medium", 0};
170 /// LaTeX 2.09 names for font shapes
171 char const * const known_old_font_shapes[] = { "it", "sl", "sc", 0};
173 /// LaTeX names for font shapes
174 char const * const known_font_shapes[] = { "itshape", "slshape", "scshape",
177 /// the same as known_old_font_shapes and known_font_shapes with .lyx names
178 char const * const known_coded_font_shapes[] = { "italic", "slanted",
179 "smallcaps", "up", 0};
182 * Graphics file extensions known by the dvips driver of the graphics package.
183 * These extensions are used to complete the filename of an included
184 * graphics file if it does not contain an extension.
185 * The order must be the same that latex uses to find a file, because we
186 * will use the first extension that matches.
187 * This is only an approximation for the common cases. If we would want to
188 * do it right in all cases, we would need to know which graphics driver is
189 * used and know the extensions of every driver of the graphics package.
191 char const * const known_dvips_graphics_formats[] = {"eps", "ps", "eps.gz",
192 "ps.gz", "eps.Z", "ps.Z", 0};
195 * Graphics file extensions known by the pdftex driver of the graphics package.
196 * \sa known_dvips_graphics_formats
198 char const * const known_pdftex_graphics_formats[] = {"png", "pdf", "jpg",
202 * Known file extensions for TeX files as used by \\include.
204 char const * const known_tex_extensions[] = {"tex", 0};
206 /// spaces known by InsetSpace
207 char const * const known_spaces[] = { " ", "space", ",", "thinspace", "quad",
208 "qquad", "enspace", "enskip", "negthinspace", 0};
210 /// the same as known_spaces with .lyx names
211 char const * const known_coded_spaces[] = { "space{}", "space{}",
212 "thinspace{}", "thinspace{}", "quad{}", "qquad{}", "enspace{}", "enskip{}",
213 "negthinspace{}", 0};
216 /// splits "x=z, y=b" into a map
217 map<string, string> split_map(string const & s)
219 map<string, string> res;
222 for (size_t i = 0; i < v.size(); ++i) {
223 size_t const pos = v[i].find('=');
224 string const index = v[i].substr(0, pos);
225 string const value = v[i].substr(pos + 1, string::npos);
226 res[trim(index)] = trim(value);
233 * Split a LaTeX length into value and unit.
234 * The latter can be a real unit like "pt", or a latex length variable
235 * like "\textwidth". The unit may contain additional stuff like glue
236 * lengths, but we don't care, because such lengths are ERT anyway.
237 * \returns true if \p value and \p unit are valid.
239 bool splitLatexLength(string const & len, string & value, string & unit)
243 const string::size_type i = len.find_first_not_of(" -+0123456789.,");
244 //'4,5' is a valid LaTeX length number. Change it to '4.5'
245 string const length = subst(len, ',', '.');
246 if (i == string::npos)
249 if (len[0] == '\\') {
250 // We had something like \textwidth without a factor
256 value = trim(string(length, 0, i));
260 // 'cM' is a valid LaTeX length unit. Change it to 'cm'
261 if (contains(len, '\\'))
262 unit = trim(string(len, i));
264 unit = support::ascii_lowercase(trim(string(len, i)));
269 /// A simple function to translate a latex length to something lyx can
270 /// understand. Not perfect, but rather best-effort.
271 bool translate_len(string const & length, string & valstring, string & unit)
273 if (!splitLatexLength(length, valstring, unit))
275 // LyX uses percent values
277 istringstream iss(valstring);
282 string const percentval = oss.str();
284 if (unit.empty() || unit[0] != '\\')
286 string::size_type const i = unit.find(' ');
287 string const endlen = (i == string::npos) ? string() : string(unit, i);
288 if (unit == "\\textwidth") {
289 valstring = percentval;
290 unit = "text%" + endlen;
291 } else if (unit == "\\columnwidth") {
292 valstring = percentval;
293 unit = "col%" + endlen;
294 } else if (unit == "\\paperwidth") {
295 valstring = percentval;
296 unit = "page%" + endlen;
297 } else if (unit == "\\linewidth") {
298 valstring = percentval;
299 unit = "line%" + endlen;
300 } else if (unit == "\\paperheight") {
301 valstring = percentval;
302 unit = "pheight%" + endlen;
303 } else if (unit == "\\textheight") {
304 valstring = percentval;
305 unit = "theight%" + endlen;
313 string translate_len(string const & length)
317 if (translate_len(length, value, unit))
319 // If the input is invalid, return what we have.
327 * Translates a LaTeX length into \p value, \p unit and
328 * \p special parts suitable for a box inset.
329 * The difference from translate_len() is that a box inset knows about
330 * some special "units" that are stored in \p special.
332 void translate_box_len(string const & length, string & value, string & unit, string & special)
334 if (translate_len(length, value, unit)) {
335 if (unit == "\\height" || unit == "\\depth" ||
336 unit == "\\totalheight" || unit == "\\width") {
337 special = unit.substr(1);
338 // The unit is not used, but LyX requires a dummy setting
351 * Find a file with basename \p name in path \p path and an extension
354 string find_file(string const & name, string const & path,
355 char const * const * extensions)
357 // FIXME UNICODE encoding of name and path may be wrong (makeAbsPath
359 for (char const * const * what = extensions; *what; ++what) {
360 string const trial = addExtension(name, *what);
361 if (makeAbsPath(trial, path).exists())
368 void begin_inset(ostream & os, string const & name)
370 os << "\n\\begin_inset " << name;
374 void end_inset(ostream & os)
376 os << "\n\\end_inset\n\n";
380 void skip_braces(Parser & p)
382 if (p.next_token().cat() != catBegin)
385 if (p.next_token().cat() == catEnd) {
393 void handle_ert(ostream & os, string const & s, Context & context)
395 // We must have a valid layout before outputting the ERT inset.
396 context.check_layout(os);
397 Context newcontext(true, context.textclass);
398 begin_inset(os, "ERT");
399 os << "\nstatus collapsed\n";
400 newcontext.check_layout(os);
401 for (string::const_iterator it = s.begin(), et = s.end(); it != et; ++it) {
403 os << "\n\\backslash\n";
404 else if (*it == '\n') {
405 newcontext.new_paragraph(os);
406 newcontext.check_layout(os);
410 newcontext.check_end_layout(os);
415 void handle_comment(ostream & os, string const & s, Context & context)
417 // TODO: Handle this better
418 Context newcontext(true, context.textclass);
419 begin_inset(os, "ERT");
420 os << "\nstatus collapsed\n";
421 newcontext.check_layout(os);
422 for (string::const_iterator it = s.begin(), et = s.end(); it != et; ++it) {
424 os << "\n\\backslash\n";
428 // make sure that our comment is the last thing on the line
429 newcontext.new_paragraph(os);
430 newcontext.check_layout(os);
431 newcontext.check_end_layout(os);
436 class isLayout : public std::unary_function<LayoutPtr, bool> {
438 isLayout(string const name) : name_(name) {}
439 bool operator()(LayoutPtr const & ptr) const {
440 return ptr->latexname() == name_;
447 LayoutPtr findLayout(TextClass const & textclass,
450 TextClass::const_iterator beg = textclass.begin();
451 TextClass::const_iterator end = textclass.end();
453 TextClass::const_iterator
454 it = std::find_if(beg, end, isLayout(name));
456 return (it == end) ? LayoutPtr() : *it;
460 void eat_whitespace(Parser &, ostream &, Context &, bool);
463 void output_command_layout(ostream & os, Parser & p, bool outer,
464 Context & parent_context,
467 parent_context.check_end_layout(os);
468 Context context(true, parent_context.textclass, newlayout,
469 parent_context.layout, parent_context.font);
470 if (parent_context.deeper_paragraph) {
471 // We are beginning a nested environment after a
472 // deeper paragraph inside the outer list environment.
473 // Therefore we don't need to output a "begin deeper".
474 context.need_end_deeper = true;
476 context.check_deeper(os);
477 context.check_layout(os);
478 if (context.layout->optionalargs > 0) {
479 eat_whitespace(p, os, context, false);
480 if (p.next_token().character() == '[') {
481 p.get_token(); // eat '['
482 begin_inset(os, "OptArg\n");
483 os << "status collapsed\n\n";
484 parse_text_in_inset(p, os, FLAG_BRACK_LAST, outer, context);
486 eat_whitespace(p, os, context, false);
489 parse_text(p, os, FLAG_ITEM, outer, context);
490 context.check_end_layout(os);
491 if (parent_context.deeper_paragraph) {
492 // We must suppress the "end deeper" because we
493 // suppressed the "begin deeper" above.
494 context.need_end_deeper = false;
496 context.check_end_deeper(os);
497 // We don't need really a new paragraph, but
498 // we must make sure that the next item gets a \begin_layout.
499 parent_context.new_paragraph(os);
504 * Output a space if necessary.
505 * This function gets called for every whitespace token.
507 * We have three cases here:
508 * 1. A space must be suppressed. Example: The lyxcode case below
509 * 2. A space may be suppressed. Example: Spaces before "\par"
510 * 3. A space must not be suppressed. Example: A space between two words
512 * We currently handle only 1. and 3 and from 2. only the case of
513 * spaces before newlines as a side effect.
515 * 2. could be used to suppress as many spaces as possible. This has two effects:
516 * - Reimporting LyX generated LaTeX files changes almost no whitespace
517 * - Superflous whitespace from non LyX generated LaTeX files is removed.
518 * The drawback is that the logic inside the function becomes
519 * complicated, and that is the reason why it is not implemented.
521 void check_space(Parser const & p, ostream & os, Context & context)
523 Token const next = p.next_token();
524 Token const curr = p.curr_token();
525 // A space before a single newline and vice versa must be ignored
526 // LyX emits a newline before \end{lyxcode}.
527 // This newline must be ignored,
528 // otherwise LyX will add an additional protected space.
529 if (next.cat() == catSpace ||
530 next.cat() == catNewline ||
531 (next.cs() == "end" && context.layout->free_spacing && curr.cat() == catNewline)) {
534 context.check_layout(os);
540 * Parse all arguments of \p command
542 void parse_arguments(string const & command,
543 vector<ArgumentType> const & template_arguments,
544 Parser & p, ostream & os, bool outer, Context & context)
546 string ert = command;
547 size_t no_arguments = template_arguments.size();
548 for (size_t i = 0; i < no_arguments; ++i) {
549 switch (template_arguments[i]) {
551 // This argument contains regular LaTeX
552 handle_ert(os, ert + '{', context);
553 eat_whitespace(p, os, context, false);
554 parse_text(p, os, FLAG_ITEM, outer, context);
558 // This argument may contain special characters
559 ert += '{' + p.verbatim_item() + '}';
566 handle_ert(os, ert, context);
571 * Check whether \p command is a known command. If yes,
572 * handle the command with all arguments.
573 * \return true if the command was parsed, false otherwise.
575 bool parse_command(string const & command, Parser & p, ostream & os,
576 bool outer, Context & context)
578 if (known_commands.find(command) != known_commands.end()) {
579 parse_arguments(command, known_commands[command], p, os,
587 /// Parses a minipage or parbox
588 void parse_box(Parser & p, ostream & os, unsigned flags, bool outer,
589 Context & parent_context, bool use_parbox)
593 // We need to set the height to the LaTeX default of 1\\totalheight
594 // for the case when no height argument is given
595 string height_value = "1";
596 string height_unit = "in";
597 string height_special = "totalheight";
599 if (p.next_token().asInput() == "[") {
600 position = p.getArg('[', ']');
601 if (position != "t" && position != "c" && position != "b") {
603 cerr << "invalid position for minipage/parbox" << endl;
605 if (p.next_token().asInput() == "[") {
606 latex_height = p.getArg('[', ']');
607 translate_box_len(latex_height, height_value, height_unit, height_special);
609 if (p.next_token().asInput() == "[") {
610 inner_pos = p.getArg('[', ']');
611 if (inner_pos != "c" && inner_pos != "t" &&
612 inner_pos != "b" && inner_pos != "s") {
613 inner_pos = position;
614 cerr << "invalid inner_pos for minipage/parbox"
622 string const latex_width = p.verbatim_item();
623 translate_len(latex_width, width_value, width_unit);
624 if (contains(width_unit, '\\') || contains(height_unit, '\\')) {
625 // LyX can't handle length variables
630 ss << "\\begin{minipage}";
631 if (!position.empty())
632 ss << '[' << position << ']';
633 if (!latex_height.empty())
634 ss << '[' << latex_height << ']';
635 if (!inner_pos.empty())
636 ss << '[' << inner_pos << ']';
637 ss << "{" << latex_width << "}";
640 handle_ert(os, ss.str(), parent_context);
641 parent_context.new_paragraph(os);
642 parse_text_in_inset(p, os, flags, outer, parent_context);
644 handle_ert(os, "}", parent_context);
646 handle_ert(os, "\\end{minipage}", parent_context);
648 // LyX does not like empty positions, so we have
649 // to set them to the LaTeX default values here.
650 if (position.empty())
652 if (inner_pos.empty())
653 inner_pos = position;
654 parent_context.check_layout(os);
655 begin_inset(os, "Box Frameless\n");
656 os << "position \"" << position << "\"\n";
657 os << "hor_pos \"c\"\n";
658 os << "has_inner_box 1\n";
659 os << "inner_pos \"" << inner_pos << "\"\n";
660 os << "use_parbox " << use_parbox << "\n";
661 os << "width \"" << width_value << width_unit << "\"\n";
662 os << "special \"none\"\n";
663 os << "height \"" << height_value << height_unit << "\"\n";
664 os << "height_special \"" << height_special << "\"\n";
665 os << "status open\n\n";
666 parse_text_in_inset(p, os, flags, outer, parent_context);
668 #ifdef PRESERVE_LAYOUT
669 // lyx puts a % after the end of the minipage
670 if (p.next_token().cat() == catNewline && p.next_token().cs().size() > 1) {
672 //handle_comment(os, "%dummy", parent_context);
675 parent_context.new_paragraph(os);
677 else if (p.next_token().cat() == catSpace || p.next_token().cat() == catNewline) {
678 //handle_comment(os, "%dummy", parent_context);
681 // We add a protected space if something real follows
682 if (p.good() && p.next_token().cat() != catComment) {
683 os << "\\InsetSpace ~\n";
691 /// parse an unknown environment
692 void parse_unknown_environment(Parser & p, string const & name, ostream & os,
693 unsigned flags, bool outer,
694 Context & parent_context)
696 if (name == "tabbing")
697 // We need to remember that we have to handle '\=' specially
698 flags |= FLAG_TABBING;
700 // We need to translate font changes and paragraphs inside the
701 // environment to ERT if we have a non standard font.
702 // Otherwise things like
703 // \large\begin{foo}\huge bar\end{foo}
705 bool const specialfont =
706 (parent_context.font != parent_context.normalfont);
707 bool const new_layout_allowed = parent_context.new_layout_allowed;
709 parent_context.new_layout_allowed = false;
710 handle_ert(os, "\\begin{" + name + "}", parent_context);
711 parse_text_snippet(p, os, flags, outer, parent_context);
712 handle_ert(os, "\\end{" + name + "}", parent_context);
714 parent_context.new_layout_allowed = new_layout_allowed;
718 void parse_environment(Parser & p, ostream & os, bool outer,
719 Context & parent_context)
722 string const name = p.getArg('{', '}');
723 const bool is_starred = suffixIs(name, '*');
724 string const unstarred_name = rtrim(name, "*");
725 active_environments.push_back(name);
727 if (is_math_env(name)) {
728 parent_context.check_layout(os);
729 begin_inset(os, "Formula ");
730 os << "\\begin{" << name << "}";
731 parse_math(p, os, FLAG_END, MATH_MODE);
732 os << "\\end{" << name << "}";
736 else if (name == "tabular" || name == "longtable") {
737 eat_whitespace(p, os, parent_context, false);
738 parent_context.check_layout(os);
739 begin_inset(os, "Tabular ");
740 handle_tabular(p, os, name == "longtable", parent_context);
745 else if (parent_context.textclass.floats().typeExist(unstarred_name)) {
746 eat_whitespace(p, os, parent_context, false);
747 parent_context.check_layout(os);
748 begin_inset(os, "Float " + unstarred_name + "\n");
749 if (p.next_token().asInput() == "[") {
750 os << "placement " << p.getArg('[', ']') << '\n';
752 os << "wide " << convert<string>(is_starred)
753 << "\nsideways false"
754 << "\nstatus open\n\n";
755 parse_text_in_inset(p, os, FLAG_END, outer, parent_context);
757 // We don't need really a new paragraph, but
758 // we must make sure that the next item gets a \begin_layout.
759 parent_context.new_paragraph(os);
763 else if (name == "minipage") {
764 eat_whitespace(p, os, parent_context, false);
765 parse_box(p, os, FLAG_END, outer, parent_context, false);
769 else if (name == "comment") {
770 eat_whitespace(p, os, parent_context, false);
771 parent_context.check_layout(os);
772 begin_inset(os, "Note Comment\n");
773 os << "status open\n";
774 parse_text_in_inset(p, os, FLAG_END, outer, parent_context);
777 // FIXME the comment note inset has a trailing "{}" pair
781 else if (name == "lyxgreyedout") {
782 eat_whitespace(p, os, parent_context, false);
783 parent_context.check_layout(os);
784 begin_inset(os, "Note Greyedout\n");
785 os << "status open\n";
786 parse_text_in_inset(p, os, FLAG_END, outer, parent_context);
791 else if (name == "framed") {
792 eat_whitespace(p, os, parent_context, false);
793 parent_context.check_layout(os);
794 begin_inset(os, "Note Framed\n");
795 os << "status open\n";
796 parse_text_in_inset(p, os, FLAG_END, outer, parent_context);
801 else if (name == "shaded") {
802 eat_whitespace(p, os, parent_context, false);
803 parent_context.check_layout(os);
804 begin_inset(os, "Note Shaded\n");
805 os << "status open\n";
806 parse_text_in_inset(p, os, FLAG_END, outer, parent_context);
811 else if (!parent_context.new_layout_allowed)
812 parse_unknown_environment(p, name, os, FLAG_END, outer,
815 // Alignment settings
816 else if (name == "center" || name == "flushleft" || name == "flushright" ||
817 name == "centering" || name == "raggedright" || name == "raggedleft") {
818 eat_whitespace(p, os, parent_context, false);
819 // We must begin a new paragraph if not already done
820 if (! parent_context.atParagraphStart()) {
821 parent_context.check_end_layout(os);
822 parent_context.new_paragraph(os);
824 if (name == "flushleft" || name == "raggedright")
825 parent_context.add_extra_stuff("\\align left\n");
826 else if (name == "flushright" || name == "raggedleft")
827 parent_context.add_extra_stuff("\\align right\n");
829 parent_context.add_extra_stuff("\\align center\n");
830 parse_text(p, os, FLAG_END, outer, parent_context);
831 // Just in case the environment is empty ..
832 parent_context.extra_stuff.erase();
833 // We must begin a new paragraph to reset the alignment
834 parent_context.new_paragraph(os);
838 // The single '=' is meant here.
839 else if ((newlayout = findLayout(parent_context.textclass, name)).get() &&
840 newlayout->isEnvironment()) {
841 eat_whitespace(p, os, parent_context, false);
842 Context context(true, parent_context.textclass, newlayout,
843 parent_context.layout, parent_context.font);
844 if (parent_context.deeper_paragraph) {
845 // We are beginning a nested environment after a
846 // deeper paragraph inside the outer list environment.
847 // Therefore we don't need to output a "begin deeper".
848 context.need_end_deeper = true;
850 parent_context.check_end_layout(os);
851 switch (context.layout->latextype) {
852 case LATEX_LIST_ENVIRONMENT:
853 context.extra_stuff = "\\labelwidthstring "
854 + p.verbatim_item() + '\n';
857 case LATEX_BIB_ENVIRONMENT:
858 p.verbatim_item(); // swallow next arg
864 context.check_deeper(os);
865 parse_text(p, os, FLAG_END, outer, context);
866 context.check_end_layout(os);
867 if (parent_context.deeper_paragraph) {
868 // We must suppress the "end deeper" because we
869 // suppressed the "begin deeper" above.
870 context.need_end_deeper = false;
872 context.check_end_deeper(os);
873 parent_context.new_paragraph(os);
877 else if (name == "appendix") {
878 // This is no good latex style, but it works and is used in some documents...
879 eat_whitespace(p, os, parent_context, false);
880 parent_context.check_end_layout(os);
881 Context context(true, parent_context.textclass, parent_context.layout,
882 parent_context.layout, parent_context.font);
883 context.check_layout(os);
884 os << "\\start_of_appendix\n";
885 parse_text(p, os, FLAG_END, outer, context);
886 context.check_end_layout(os);
890 else if (known_environments.find(name) != known_environments.end()) {
891 vector<ArgumentType> arguments = known_environments[name];
892 // The last "argument" denotes wether we may translate the
893 // environment contents to LyX
894 // The default required if no argument is given makes us
895 // compatible with the reLyXre environment.
896 ArgumentType contents = arguments.empty() ?
899 if (!arguments.empty())
900 arguments.pop_back();
901 // See comment in parse_unknown_environment()
902 bool const specialfont =
903 (parent_context.font != parent_context.normalfont);
904 bool const new_layout_allowed =
905 parent_context.new_layout_allowed;
907 parent_context.new_layout_allowed = false;
908 parse_arguments("\\begin{" + name + "}", arguments, p, os,
909 outer, parent_context);
910 if (contents == verbatim)
911 handle_ert(os, p.verbatimEnvironment(name),
914 parse_text_snippet(p, os, FLAG_END, outer,
916 handle_ert(os, "\\end{" + name + "}", parent_context);
918 parent_context.new_layout_allowed = new_layout_allowed;
922 parse_unknown_environment(p, name, os, FLAG_END, outer,
925 active_environments.pop_back();
929 /// parses a comment and outputs it to \p os.
930 void parse_comment(Parser & p, ostream & os, Token const & t, Context & context)
932 BOOST_ASSERT(t.cat() == catComment);
933 if (!t.cs().empty()) {
934 context.check_layout(os);
935 handle_comment(os, '%' + t.cs(), context);
936 if (p.next_token().cat() == catNewline) {
937 // A newline after a comment line starts a new
939 if (context.new_layout_allowed) {
940 if(!context.atParagraphStart())
941 // Only start a new paragraph if not already
942 // done (we might get called recursively)
943 context.new_paragraph(os);
945 handle_ert(os, "\n", context);
946 eat_whitespace(p, os, context, true);
956 * Reads spaces and comments until the first non-space, non-comment token.
957 * New paragraphs (double newlines or \\par) are handled like simple spaces
958 * if \p eatParagraph is true.
959 * Spaces are skipped, but comments are written to \p os.
961 void eat_whitespace(Parser & p, ostream & os, Context & context,
965 Token const & t = p.get_token();
966 if (t.cat() == catComment)
967 parse_comment(p, os, t, context);
968 else if ((! eatParagraph && p.isParagraph()) ||
969 (t.cat() != catSpace && t.cat() != catNewline)) {
978 * Set a font attribute, parse text and reset the font attribute.
979 * \param attribute Attribute name (e.g. \\family, \\shape etc.)
980 * \param currentvalue Current value of the attribute. Is set to the new
981 * value during parsing.
982 * \param newvalue New value of the attribute
984 void parse_text_attributes(Parser & p, ostream & os, unsigned flags, bool outer,
985 Context & context, string const & attribute,
986 string & currentvalue, string const & newvalue)
988 context.check_layout(os);
989 string const oldvalue = currentvalue;
990 currentvalue = newvalue;
991 os << '\n' << attribute << ' ' << newvalue << "\n";
992 parse_text_snippet(p, os, flags, outer, context);
993 context.check_layout(os);
994 os << '\n' << attribute << ' ' << oldvalue << "\n";
995 currentvalue = oldvalue;
999 /// get the arguments of a natbib or jurabib citation command
1000 void get_cite_arguments(Parser & p, bool natbibOrder,
1001 string & before, string & after)
1003 // We need to distinguish "" and "[]", so we can't use p.getOpt().
1005 // text before the citation
1007 // text after the citation
1008 after = p.getFullOpt();
1010 if (!after.empty()) {
1011 before = p.getFullOpt();
1012 if (natbibOrder && !before.empty())
1013 std::swap(before, after);
1018 /// Convert filenames with TeX macros and/or quotes to something LyX
1020 string const normalize_filename(string const & name)
1022 Parser p(trim(name, "\""));
1025 Token const & t = p.get_token();
1026 if (t.cat() != catEscape)
1028 else if (t.cs() == "lyxdot") {
1029 // This is used by LyX for simple dots in relative
1033 } else if (t.cs() == "space") {
1043 /// Convert \p name from TeX convention (relative to master file) to LyX
1044 /// convention (relative to .lyx file) if it is relative
1045 void fix_relative_filename(string & name)
1047 if (lyx::support::absolutePath(name))
1049 // FIXME UNICODE encoding of name may be wrong (makeAbsPath expects
1051 name = to_utf8(makeRelPath(from_utf8(makeAbsPath(name, getMasterFilePath()).absFilename()),
1052 from_utf8(getParentFilePath())));
1056 /// Parse a NoWeb Scrap section. The initial "<<" is already parsed.
1057 void parse_noweb(Parser & p, ostream & os, Context & context)
1059 // assemble the rest of the keyword
1063 Token const & t = p.get_token();
1064 if (t.asInput() == ">" && p.next_token().asInput() == ">") {
1067 scrap = (p.good() && p.next_token().asInput() == "=");
1069 name += p.get_token().asInput();
1072 name += t.asInput();
1075 if (!scrap || !context.new_layout_allowed ||
1076 !context.textclass.hasLayout(from_ascii("Scrap"))) {
1077 cerr << "Warning: Could not interpret '" << name
1078 << "'. Ignoring it." << endl;
1082 // We use new_paragraph instead of check_end_layout because the stuff
1083 // following the noweb chunk needs to start with a \begin_layout.
1084 // This may create a new paragraph even if there was none in the
1085 // noweb file, but the alternative is an invalid LyX file. Since
1086 // noweb code chunks are implemented with a layout style in LyX they
1087 // always must be in an own paragraph.
1088 context.new_paragraph(os);
1089 Context newcontext(true, context.textclass,
1090 context.textclass[from_ascii("Scrap")]);
1091 newcontext.check_layout(os);
1094 Token const & t = p.get_token();
1095 // We abuse the parser a bit, because this is no TeX syntax
1097 if (t.cat() == catEscape)
1098 os << subst(t.asInput(), "\\", "\n\\backslash\n");
1100 os << subst(t.asInput(), "\n", "\n\\newline\n");
1101 // The scrap chunk is ended by an @ at the beginning of a line.
1102 // After the @ the line may contain a comment and/or
1103 // whitespace, but nothing else.
1104 if (t.asInput() == "@" && p.prev_token().cat() == catNewline &&
1105 (p.next_token().cat() == catSpace ||
1106 p.next_token().cat() == catNewline ||
1107 p.next_token().cat() == catComment)) {
1108 while (p.good() && p.next_token().cat() == catSpace)
1109 os << p.get_token().asInput();
1110 if (p.next_token().cat() == catComment)
1111 // The comment includes a final '\n'
1112 os << p.get_token().asInput();
1114 if (p.next_token().cat() == catNewline)
1121 newcontext.check_end_layout(os);
1124 } // anonymous namespace
1127 void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
1130 LayoutPtr newlayout;
1131 // store the current selectlanguage to be used after \foreignlanguage
1133 // Store the latest bibliographystyle (needed for bibtex inset)
1134 string bibliographystyle;
1135 bool const use_natbib = used_packages.find("natbib") != used_packages.end();
1136 bool const use_jurabib = used_packages.find("jurabib") != used_packages.end();
1138 Token const & t = p.get_token();
1141 cerr << "t: " << t << " flags: " << flags << "\n";
1144 if (flags & FLAG_ITEM) {
1145 if (t.cat() == catSpace)
1148 flags &= ~FLAG_ITEM;
1149 if (t.cat() == catBegin) {
1150 // skip the brace and collect everything to the next matching
1152 flags |= FLAG_BRACE_LAST;
1156 // handle only this single token, leave the loop if done
1157 flags |= FLAG_LEAVE;
1160 if (t.character() == ']' && (flags & FLAG_BRACK_LAST))
1166 if (t.cat() == catMath) {
1167 // we are inside some text mode thingy, so opening new math is allowed
1168 context.check_layout(os);
1169 begin_inset(os, "Formula ");
1170 Token const & n = p.get_token();
1171 if (n.cat() == catMath && outer) {
1172 // TeX's $$...$$ syntax for displayed math
1174 parse_math(p, os, FLAG_SIMPLE, MATH_MODE);
1176 p.get_token(); // skip the second '$' token
1178 // simple $...$ stuff
1181 parse_math(p, os, FLAG_SIMPLE, MATH_MODE);
1187 else if (t.cat() == catSuper || t.cat() == catSub)
1188 cerr << "catcode " << t << " illegal in text mode\n";
1190 // Basic support for english quotes. This should be
1191 // extended to other quotes, but is not so easy (a
1192 // left english quote is the same as a right german
1194 else if (t.asInput() == "`"
1195 && p.next_token().asInput() == "`") {
1196 context.check_layout(os);
1197 begin_inset(os, "Quotes ");
1203 else if (t.asInput() == "'"
1204 && p.next_token().asInput() == "'") {
1205 context.check_layout(os);
1206 begin_inset(os, "Quotes ");
1213 else if (t.asInput() == "<"
1214 && p.next_token().asInput() == "<" && noweb_mode) {
1216 parse_noweb(p, os, context);
1219 else if (t.cat() == catSpace || (t.cat() == catNewline && ! p.isParagraph()))
1220 check_space(p, os, context);
1222 else if (t.character() == '[' && noweb_mode &&
1223 p.next_token().character() == '[') {
1224 // These can contain underscores
1226 string const s = p.getFullOpt() + ']';
1227 if (p.next_token().character() == ']')
1230 cerr << "Warning: Inserting missing ']' in '"
1231 << s << "'." << endl;
1232 handle_ert(os, s, context);
1235 else if (t.cat() == catLetter ||
1236 t.cat() == catOther ||
1237 t.cat() == catAlign ||
1238 t.cat() == catParameter) {
1239 // This translates "&" to "\\&" which may be wrong...
1240 context.check_layout(os);
1241 os << t.character();
1244 else if (p.isParagraph()) {
1245 if (context.new_layout_allowed)
1246 context.new_paragraph(os);
1248 handle_ert(os, "\\par ", context);
1249 eat_whitespace(p, os, context, true);
1252 else if (t.cat() == catActive) {
1253 context.check_layout(os);
1254 if (t.character() == '~') {
1255 if (context.layout->free_spacing)
1258 os << "\\InsetSpace ~\n";
1260 os << t.character();
1263 else if (t.cat() == catBegin &&
1264 p.next_token().cat() == catEnd) {
1266 Token const prev = p.prev_token();
1268 if (p.next_token().character() == '`' ||
1269 (prev.character() == '-' &&
1270 p.next_token().character() == '-'))
1271 ; // ignore it in {}`` or -{}-
1273 handle_ert(os, "{}", context);
1277 else if (t.cat() == catBegin) {
1278 context.check_layout(os);
1279 // special handling of font attribute changes
1280 Token const prev = p.prev_token();
1281 Token const next = p.next_token();
1282 TeXFont const oldFont = context.font;
1283 if (next.character() == '[' ||
1284 next.character() == ']' ||
1285 next.character() == '*') {
1287 if (p.next_token().cat() == catEnd) {
1288 os << next.character();
1292 handle_ert(os, "{", context);
1293 parse_text_snippet(p, os,
1296 handle_ert(os, "}", context);
1298 } else if (! context.new_layout_allowed) {
1299 handle_ert(os, "{", context);
1300 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1302 handle_ert(os, "}", context);
1303 } else if (is_known(next.cs(), known_sizes)) {
1304 // next will change the size, so we must
1306 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1308 if (!context.atParagraphStart())
1310 << context.font.size << "\n";
1311 } else if (is_known(next.cs(), known_font_families)) {
1312 // next will change the font family, so we
1313 // must reset it here
1314 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1316 if (!context.atParagraphStart())
1318 << context.font.family << "\n";
1319 } else if (is_known(next.cs(), known_font_series)) {
1320 // next will change the font series, so we
1321 // must reset it here
1322 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1324 if (!context.atParagraphStart())
1326 << context.font.series << "\n";
1327 } else if (is_known(next.cs(), known_font_shapes)) {
1328 // next will change the font shape, so we
1329 // must reset it here
1330 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1332 if (!context.atParagraphStart())
1334 << context.font.shape << "\n";
1335 } else if (is_known(next.cs(), known_old_font_families) ||
1336 is_known(next.cs(), known_old_font_series) ||
1337 is_known(next.cs(), known_old_font_shapes)) {
1338 // next will change the font family, series
1339 // and shape, so we must reset it here
1340 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1342 if (!context.atParagraphStart())
1344 << context.font.family
1346 << context.font.series
1348 << context.font.shape << "\n";
1350 handle_ert(os, "{", context);
1351 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1353 handle_ert(os, "}", context);
1357 else if (t.cat() == catEnd) {
1358 if (flags & FLAG_BRACE_LAST) {
1361 cerr << "stray '}' in text\n";
1362 handle_ert(os, "}", context);
1365 else if (t.cat() == catComment)
1366 parse_comment(p, os, t, context);
1369 // control sequences
1372 else if (t.cs() == "(") {
1373 context.check_layout(os);
1374 begin_inset(os, "Formula");
1376 parse_math(p, os, FLAG_SIMPLE2, MATH_MODE);
1381 else if (t.cs() == "[") {
1382 context.check_layout(os);
1383 begin_inset(os, "Formula");
1385 parse_math(p, os, FLAG_EQUATION, MATH_MODE);
1390 else if (t.cs() == "begin")
1391 parse_environment(p, os, outer, context);
1393 else if (t.cs() == "end") {
1394 if (flags & FLAG_END) {
1395 // eat environment name
1396 string const name = p.getArg('{', '}');
1397 if (name != active_environment())
1398 cerr << "\\end{" + name + "} does not match \\begin{"
1399 + active_environment() + "}\n";
1402 p.error("found 'end' unexpectedly");
1405 else if (t.cs() == "item") {
1408 bool optarg = false;
1409 if (p.next_token().character() == '[') {
1410 p.get_token(); // eat '['
1411 s = parse_text_snippet(p, FLAG_BRACK_LAST,
1416 context.check_layout(os);
1417 if (context.has_item) {
1418 // An item in an unknown list-like environment
1419 // FIXME: Do this in check_layout()!
1420 context.has_item = false;
1422 handle_ert(os, "\\item", context);
1424 handle_ert(os, "\\item ", context);
1427 if (context.layout->labeltype != LABEL_MANUAL) {
1428 // lyx does not support \item[\mybullet]
1429 // in itemize environments
1430 handle_ert(os, "[", context);
1432 handle_ert(os, "]", context);
1433 } else if (!s.empty()) {
1434 // The space is needed to separate the
1435 // item from the rest of the sentence.
1437 eat_whitespace(p, os, context, false);
1442 else if (t.cs() == "bibitem") {
1444 context.check_layout(os);
1447 os << '{' << p.verbatim_item() << '}' << "\n";
1450 else if (t.cs() == "def") {
1451 context.check_layout(os);
1452 eat_whitespace(p, os, context, false);
1453 string name = p.get_token().cs();
1454 eat_whitespace(p, os, context, false);
1460 while (p.next_token().cat() != catBegin) {
1461 if (p.next_token().cat() == catParameter) {
1466 // followed by number?
1467 if (p.next_token().cat() == catOther) {
1468 char c = p.getChar();
1470 // number = current arity + 1?
1471 if (c == arity + '0' + 1)
1476 paramtext += p.get_token().asString();
1478 paramtext += p.get_token().asString();
1483 // only output simple (i.e. compatible) macro as FormulaMacros
1484 string ert = "\\def\\" + name + ' ' + paramtext + '{' + p.verbatim_item() + '}';
1486 context.check_layout(os);
1487 begin_inset(os, "FormulaMacro");
1491 handle_ert(os, ert, context);
1494 else if (t.cs() == "noindent") {
1496 context.add_extra_stuff("\\noindent\n");
1499 else if (t.cs() == "appendix") {
1500 context.add_extra_stuff("\\start_of_appendix\n");
1501 // We need to start a new paragraph. Otherwise the
1502 // appendix in 'bla\appendix\chapter{' would start
1504 context.new_paragraph(os);
1505 // We need to make sure that the paragraph is
1506 // generated even if it is empty. Otherwise the
1507 // appendix in '\par\appendix\par\chapter{' would
1509 context.check_layout(os);
1510 // FIXME: This is a hack to prevent paragraph
1511 // deletion if it is empty. Handle this better!
1513 "%dummy comment inserted by tex2lyx to "
1514 "ensure that this paragraph is not empty",
1516 // Both measures above may generate an additional
1517 // empty paragraph, but that does not hurt, because
1518 // whitespace does not matter here.
1519 eat_whitespace(p, os, context, true);
1522 // Must attempt to parse "Section*" before "Section".
1523 else if ((p.next_token().asInput() == "*") &&
1524 context.new_layout_allowed &&
1525 // The single '=' is meant here.
1526 (newlayout = findLayout(context.textclass,
1527 t.cs() + '*')).get() &&
1528 newlayout->isCommand()) {
1530 output_command_layout(os, p, outer, context, newlayout);
1534 // The single '=' is meant here.
1535 else if (context.new_layout_allowed &&
1536 (newlayout = findLayout(context.textclass, t.cs())).get() &&
1537 newlayout->isCommand()) {
1538 output_command_layout(os, p, outer, context, newlayout);
1542 // Special handling for \caption
1543 // FIXME: remove this when InsetCaption is supported.
1544 else if (context.new_layout_allowed &&
1545 t.cs() == captionlayout->latexname()) {
1546 output_command_layout(os, p, outer, context,
1551 else if (t.cs() == "includegraphics") {
1552 bool const clip = p.next_token().asInput() == "*";
1555 map<string, string> opts = split_map(p.getArg('[', ']'));
1557 opts["clip"] = string();
1558 string name = normalize_filename(p.verbatim_item());
1560 string const path = getMasterFilePath();
1561 // We want to preserve relative / absolute filenames,
1562 // therefore path is only used for testing
1563 // FIXME UNICODE encoding of name and path may be
1564 // wrong (makeAbsPath expects utf8)
1565 if (!makeAbsPath(name, path).exists()) {
1566 // The file extension is probably missing.
1567 // Now try to find it out.
1568 string const dvips_name =
1569 find_file(name, path,
1570 known_dvips_graphics_formats);
1571 string const pdftex_name =
1572 find_file(name, path,
1573 known_pdftex_graphics_formats);
1574 if (!dvips_name.empty()) {
1575 if (!pdftex_name.empty()) {
1576 cerr << "This file contains the "
1578 "\"\\includegraphics{"
1580 "However, files\n\""
1581 << dvips_name << "\" and\n\""
1582 << pdftex_name << "\"\n"
1583 "both exist, so I had to make a "
1584 "choice and took the first one.\n"
1585 "Please move the unwanted one "
1586 "someplace else and try again\n"
1587 "if my choice was wrong."
1591 } else if (!pdftex_name.empty())
1595 // FIXME UNICODE encoding of name and path may be
1596 // wrong (makeAbsPath expects utf8)
1597 if (makeAbsPath(name, path).exists())
1598 fix_relative_filename(name);
1600 cerr << "Warning: Could not find graphics file '"
1601 << name << "'." << endl;
1603 context.check_layout(os);
1604 begin_inset(os, "Graphics ");
1605 os << "\n\tfilename " << name << '\n';
1606 if (opts.find("width") != opts.end())
1608 << translate_len(opts["width"]) << '\n';
1609 if (opts.find("height") != opts.end())
1611 << translate_len(opts["height"]) << '\n';
1612 if (opts.find("scale") != opts.end()) {
1613 istringstream iss(opts["scale"]);
1617 os << "\tscale " << val << '\n';
1619 if (opts.find("angle") != opts.end())
1620 os << "\trotateAngle "
1621 << opts["angle"] << '\n';
1622 if (opts.find("origin") != opts.end()) {
1624 string const opt = opts["origin"];
1625 if (opt.find('l') != string::npos) ss << "left";
1626 if (opt.find('r') != string::npos) ss << "right";
1627 if (opt.find('c') != string::npos) ss << "center";
1628 if (opt.find('t') != string::npos) ss << "Top";
1629 if (opt.find('b') != string::npos) ss << "Bottom";
1630 if (opt.find('B') != string::npos) ss << "Baseline";
1631 if (!ss.str().empty())
1632 os << "\trotateOrigin " << ss.str() << '\n';
1634 cerr << "Warning: Ignoring unknown includegraphics origin argument '" << opt << "'\n";
1636 if (opts.find("keepaspectratio") != opts.end())
1637 os << "\tkeepAspectRatio\n";
1638 if (opts.find("clip") != opts.end())
1640 if (opts.find("draft") != opts.end())
1642 if (opts.find("bb") != opts.end())
1643 os << "\tBoundingBox "
1644 << opts["bb"] << '\n';
1645 int numberOfbbOptions = 0;
1646 if (opts.find("bbllx") != opts.end())
1647 numberOfbbOptions++;
1648 if (opts.find("bblly") != opts.end())
1649 numberOfbbOptions++;
1650 if (opts.find("bburx") != opts.end())
1651 numberOfbbOptions++;
1652 if (opts.find("bbury") != opts.end())
1653 numberOfbbOptions++;
1654 if (numberOfbbOptions == 4)
1655 os << "\tBoundingBox "
1656 << opts["bbllx"] << opts["bblly"]
1657 << opts["bburx"] << opts["bbury"] << '\n';
1658 else if (numberOfbbOptions > 0)
1659 cerr << "Warning: Ignoring incomplete includegraphics boundingbox arguments.\n";
1660 numberOfbbOptions = 0;
1661 if (opts.find("natwidth") != opts.end())
1662 numberOfbbOptions++;
1663 if (opts.find("natheight") != opts.end())
1664 numberOfbbOptions++;
1665 if (numberOfbbOptions == 2)
1666 os << "\tBoundingBox 0bp 0bp "
1667 << opts["natwidth"] << opts["natheight"] << '\n';
1668 else if (numberOfbbOptions > 0)
1669 cerr << "Warning: Ignoring incomplete includegraphics boundingbox arguments.\n";
1670 ostringstream special;
1671 if (opts.find("hiresbb") != opts.end())
1672 special << "hiresbb,";
1673 if (opts.find("trim") != opts.end())
1675 if (opts.find("viewport") != opts.end())
1676 special << "viewport=" << opts["viewport"] << ',';
1677 if (opts.find("totalheight") != opts.end())
1678 special << "totalheight=" << opts["totalheight"] << ',';
1679 if (opts.find("type") != opts.end())
1680 special << "type=" << opts["type"] << ',';
1681 if (opts.find("ext") != opts.end())
1682 special << "ext=" << opts["ext"] << ',';
1683 if (opts.find("read") != opts.end())
1684 special << "read=" << opts["read"] << ',';
1685 if (opts.find("command") != opts.end())
1686 special << "command=" << opts["command"] << ',';
1687 string s_special = special.str();
1688 if (!s_special.empty()) {
1689 // We had special arguments. Remove the trailing ','.
1690 os << "\tspecial " << s_special.substr(0, s_special.size() - 1) << '\n';
1692 // TODO: Handle the unknown settings better.
1693 // Warn about invalid options.
1694 // Check whether some option was given twice.
1698 else if (t.cs() == "footnote" ||
1699 (t.cs() == "thanks" && context.layout->intitle)) {
1701 context.check_layout(os);
1702 begin_inset(os, "Foot\n");
1703 os << "status collapsed\n\n";
1704 parse_text_in_inset(p, os, FLAG_ITEM, false, context);
1708 else if (t.cs() == "marginpar") {
1710 context.check_layout(os);
1711 begin_inset(os, "Marginal\n");
1712 os << "status collapsed\n\n";
1713 parse_text_in_inset(p, os, FLAG_ITEM, false, context);
1717 else if (t.cs() == "ensuremath") {
1719 context.check_layout(os);
1720 string const s = p.verbatim_item();
1721 if (s == "\xb1" || s == "\xb3" || s == "\xb2" || s == "\xb5")
1724 handle_ert(os, "\\ensuremath{" + s + "}",
1728 else if (t.cs() == "hfill") {
1729 context.check_layout(os);
1730 os << "\n\\hfill\n";
1735 else if (t.cs() == "makeindex" || t.cs() == "maketitle") {
1736 // FIXME: Somehow prevent title layouts if
1737 // "maketitle" was not found
1739 skip_braces(p); // swallow this
1742 else if (t.cs() == "tableofcontents") {
1744 context.check_layout(os);
1745 begin_inset(os, "CommandInset toc\n");
1746 os << "LatexCommand tableofcontents\n";
1748 skip_braces(p); // swallow this
1751 else if (t.cs() == "listoffigures") {
1753 context.check_layout(os);
1754 begin_inset(os, "FloatList figure\n");
1756 skip_braces(p); // swallow this
1759 else if (t.cs() == "listoftables") {
1761 context.check_layout(os);
1762 begin_inset(os, "FloatList table\n");
1764 skip_braces(p); // swallow this
1767 else if (t.cs() == "listof") {
1768 p.skip_spaces(true);
1769 string const name = p.verbatim_item();
1770 string const name2 = subst(p.verbatim_item(), "\n", " ");
1771 if (context.textclass.floats().typeExist(name)) {
1772 context.check_layout(os);
1773 begin_inset(os, "FloatList ");
1776 // the second argument is not needed
1778 handle_ert(os, "\\listof{" + name + "}{" + name2 + "}", context);
1781 else if (t.cs() == "textrm")
1782 parse_text_attributes(p, os, FLAG_ITEM, outer,
1783 context, "\\family",
1784 context.font.family, "roman");
1786 else if (t.cs() == "textsf")
1787 parse_text_attributes(p, os, FLAG_ITEM, outer,
1788 context, "\\family",
1789 context.font.family, "sans");
1791 else if (t.cs() == "texttt")
1792 parse_text_attributes(p, os, FLAG_ITEM, outer,
1793 context, "\\family",
1794 context.font.family, "typewriter");
1796 else if (t.cs() == "textmd")
1797 parse_text_attributes(p, os, FLAG_ITEM, outer,
1798 context, "\\series",
1799 context.font.series, "medium");
1801 else if (t.cs() == "textbf")
1802 parse_text_attributes(p, os, FLAG_ITEM, outer,
1803 context, "\\series",
1804 context.font.series, "bold");
1806 else if (t.cs() == "textup")
1807 parse_text_attributes(p, os, FLAG_ITEM, outer,
1809 context.font.shape, "up");
1811 else if (t.cs() == "textit")
1812 parse_text_attributes(p, os, FLAG_ITEM, outer,
1814 context.font.shape, "italic");
1816 else if (t.cs() == "textsl")
1817 parse_text_attributes(p, os, FLAG_ITEM, outer,
1819 context.font.shape, "slanted");
1821 else if (t.cs() == "textsc")
1822 parse_text_attributes(p, os, FLAG_ITEM, outer,
1824 context.font.shape, "smallcaps");
1826 else if (t.cs() == "textnormal" || t.cs() == "normalfont") {
1827 context.check_layout(os);
1828 TeXFont oldFont = context.font;
1829 context.font.init();
1830 context.font.size = oldFont.size;
1831 os << "\n\\family " << context.font.family << "\n";
1832 os << "\n\\series " << context.font.series << "\n";
1833 os << "\n\\shape " << context.font.shape << "\n";
1834 if (t.cs() == "textnormal") {
1835 parse_text_snippet(p, os, FLAG_ITEM, outer, context);
1836 output_font_change(os, context.font, oldFont);
1837 context.font = oldFont;
1839 eat_whitespace(p, os, context, false);
1842 else if (t.cs() == "underbar") {
1843 // Do NOT handle \underline.
1844 // \underbar cuts through y, g, q, p etc.,
1845 // \underline does not.
1846 context.check_layout(os);
1847 os << "\n\\bar under\n";
1848 parse_text_snippet(p, os, FLAG_ITEM, outer, context);
1849 context.check_layout(os);
1850 os << "\n\\bar default\n";
1853 else if (t.cs() == "emph" || t.cs() == "noun") {
1854 context.check_layout(os);
1855 os << "\n\\" << t.cs() << " on\n";
1856 parse_text_snippet(p, os, FLAG_ITEM, outer, context);
1857 context.check_layout(os);
1858 os << "\n\\" << t.cs() << " default\n";
1861 else if (use_natbib &&
1862 is_known(t.cs(), known_natbib_commands) &&
1863 ((t.cs() != "citefullauthor" &&
1864 t.cs() != "citeyear" &&
1865 t.cs() != "citeyearpar") ||
1866 p.next_token().asInput() != "*")) {
1867 context.check_layout(os);
1869 // \citet[before][after]{a} \citet[after][before]{a}
1870 // \citet[before][]{a} \citet[][before]{a}
1871 // \citet[after]{a} \citet[after]{a}
1872 // \citet{a} \citet{a}
1873 string command = '\\' + t.cs();
1874 if (p.next_token().asInput() == "*") {
1878 if (command == "\\citefullauthor")
1879 // alternative name for "\\citeauthor*"
1880 command = "\\citeauthor*";
1882 // text before the citation
1884 // text after the citation
1886 get_cite_arguments(p, true, before, after);
1888 if (command == "\\cite") {
1889 // \cite without optional argument means
1890 // \citet, \cite with at least one optional
1891 // argument means \citep.
1892 if (before.empty() && after.empty())
1893 command = "\\citet";
1895 command = "\\citep";
1897 if (before.empty() && after == "[]")
1898 // avoid \citet[]{a}
1900 else if (before == "[]" && after == "[]") {
1901 // avoid \citet[][]{a}
1905 begin_inset(os, "LatexCommand ");
1906 os << command << after << before
1907 << '{' << p.verbatim_item() << "}\n";
1911 else if (use_jurabib &&
1912 is_known(t.cs(), known_jurabib_commands)) {
1913 context.check_layout(os);
1914 string const command = '\\' + t.cs();
1915 char argumentOrder = '\0';
1916 vector<string> const & options = used_packages["jurabib"];
1917 if (std::find(options.begin(), options.end(),
1918 "natbiborder") != options.end())
1919 argumentOrder = 'n';
1920 else if (std::find(options.begin(), options.end(),
1921 "jurabiborder") != options.end())
1922 argumentOrder = 'j';
1924 // text before the citation
1926 // text after the citation
1928 get_cite_arguments(p, argumentOrder != 'j', before, after);
1930 string const citation = p.verbatim_item();
1931 if (!before.empty() && argumentOrder == '\0') {
1932 cerr << "Warning: Assuming argument order "
1933 "of jurabib version 0.6 for\n'"
1934 << command << before << after << '{'
1935 << citation << "}'.\n"
1936 "Add 'jurabiborder' to the jurabib "
1937 "package options if you used an\n"
1938 "earlier jurabib version." << endl;
1940 begin_inset(os, "LatexCommand ");
1941 os << command << after << before
1942 << '{' << citation << "}\n";
1946 else if (is_known(t.cs(), known_latex_commands)) {
1947 // This needs to be after the check for natbib and
1948 // jurabib commands, because "cite" has different
1949 // arguments with natbib and jurabib.
1950 context.check_layout(os);
1951 begin_inset(os, "LatexCommand ");
1952 os << '\\' << t.cs();
1953 // lyx cannot handle newlines in a latex command
1954 // FIXME: Move the substitution into parser::getOpt()?
1955 os << subst(p.getOpt(), "\n", " ");
1956 os << subst(p.getOpt(), "\n", " ");
1957 os << '{' << subst(p.verbatim_item(), "\n", " ") << "}\n";
1961 else if (is_known(t.cs(), known_quotes)) {
1962 char const * const * where = is_known(t.cs(), known_quotes);
1963 context.check_layout(os);
1964 begin_inset(os, "Quotes ");
1965 os << known_coded_quotes[where - known_quotes];
1967 // LyX adds {} after the quote, so we have to eat
1968 // spaces here if there are any before a possible
1970 eat_whitespace(p, os, context, false);
1974 else if (is_known(t.cs(), known_sizes) &&
1975 context.new_layout_allowed) {
1976 char const * const * where = is_known(t.cs(), known_sizes);
1977 context.check_layout(os);
1978 TeXFont const oldFont = context.font;
1979 context.font.size = known_coded_sizes[where - known_sizes];
1980 output_font_change(os, oldFont, context.font);
1981 eat_whitespace(p, os, context, false);
1984 else if (is_known(t.cs(), known_font_families) &&
1985 context.new_layout_allowed) {
1986 char const * const * where =
1987 is_known(t.cs(), known_font_families);
1988 context.check_layout(os);
1989 TeXFont const oldFont = context.font;
1990 context.font.family =
1991 known_coded_font_families[where - known_font_families];
1992 output_font_change(os, oldFont, context.font);
1993 eat_whitespace(p, os, context, false);
1996 else if (is_known(t.cs(), known_font_series) &&
1997 context.new_layout_allowed) {
1998 char const * const * where =
1999 is_known(t.cs(), known_font_series);
2000 context.check_layout(os);
2001 TeXFont const oldFont = context.font;
2002 context.font.series =
2003 known_coded_font_series[where - known_font_series];
2004 output_font_change(os, oldFont, context.font);
2005 eat_whitespace(p, os, context, false);
2008 else if (is_known(t.cs(), known_font_shapes) &&
2009 context.new_layout_allowed) {
2010 char const * const * where =
2011 is_known(t.cs(), known_font_shapes);
2012 context.check_layout(os);
2013 TeXFont const oldFont = context.font;
2014 context.font.shape =
2015 known_coded_font_shapes[where - known_font_shapes];
2016 output_font_change(os, oldFont, context.font);
2017 eat_whitespace(p, os, context, false);
2019 else if (is_known(t.cs(), known_old_font_families) &&
2020 context.new_layout_allowed) {
2021 char const * const * where =
2022 is_known(t.cs(), known_old_font_families);
2023 context.check_layout(os);
2024 TeXFont const oldFont = context.font;
2025 context.font.init();
2026 context.font.size = oldFont.size;
2027 context.font.family =
2028 known_coded_font_families[where - known_old_font_families];
2029 output_font_change(os, oldFont, context.font);
2030 eat_whitespace(p, os, context, false);
2033 else if (is_known(t.cs(), known_old_font_series) &&
2034 context.new_layout_allowed) {
2035 char const * const * where =
2036 is_known(t.cs(), known_old_font_series);
2037 context.check_layout(os);
2038 TeXFont const oldFont = context.font;
2039 context.font.init();
2040 context.font.size = oldFont.size;
2041 context.font.series =
2042 known_coded_font_series[where - known_old_font_series];
2043 output_font_change(os, oldFont, context.font);
2044 eat_whitespace(p, os, context, false);
2047 else if (is_known(t.cs(), known_old_font_shapes) &&
2048 context.new_layout_allowed) {
2049 char const * const * where =
2050 is_known(t.cs(), known_old_font_shapes);
2051 context.check_layout(os);
2052 TeXFont const oldFont = context.font;
2053 context.font.init();
2054 context.font.size = oldFont.size;
2055 context.font.shape =
2056 known_coded_font_shapes[where - known_old_font_shapes];
2057 output_font_change(os, oldFont, context.font);
2058 eat_whitespace(p, os, context, false);
2061 else if (t.cs() == "selectlanguage") {
2062 context.check_layout(os);
2063 // save the language for the case that a \foreignlanguage is used
2064 selectlang = subst(p.verbatim_item(), "\n", " ");
2065 os << "\\lang " << selectlang << "\n";
2069 else if (t.cs() == "foreignlanguage") {
2070 context.check_layout(os);
2071 os << "\n\\lang " << subst(p.verbatim_item(), "\n", " ") << "\n";
2072 os << subst(p.verbatim_item(), "\n", " ");
2073 // set back to last selectlanguage
2074 os << "\n\\lang " << selectlang << "\n";
2077 else if (t.cs() == "inputencoding")
2078 // write nothing because this is done by LyX using the "\lang"
2079 // information given by selectlanguage and foreignlanguage
2080 subst(p.verbatim_item(), "\n", " ");
2082 else if (t.cs() == "LyX" || t.cs() == "TeX"
2083 || t.cs() == "LaTeX") {
2084 context.check_layout(os);
2086 skip_braces(p); // eat {}
2089 else if (t.cs() == "LaTeXe") {
2090 context.check_layout(os);
2092 skip_braces(p); // eat {}
2095 else if (t.cs() == "ldots") {
2096 context.check_layout(os);
2098 os << "\\SpecialChar \\ldots{}\n";
2101 else if (t.cs() == "lyxarrow") {
2102 context.check_layout(os);
2103 os << "\\SpecialChar \\menuseparator\n";
2107 else if (t.cs() == "lyxline") {
2108 context.check_layout(os);
2109 // the argument can be omitted, is handled by LyX
2110 subst(p.verbatim_item(), "\n", " ");
2111 os << "\\lyxline\n ";
2115 else if (t.cs() == "textcompwordmark") {
2116 context.check_layout(os);
2117 os << "\\SpecialChar \\textcompwordmark{}\n";
2121 else if (t.cs() == "@" && p.next_token().asInput() == ".") {
2122 context.check_layout(os);
2123 os << "\\SpecialChar \\@.\n";
2127 else if (t.cs() == "-") {
2128 context.check_layout(os);
2129 os << "\\SpecialChar \\-\n";
2132 else if (t.cs() == "textasciitilde") {
2133 context.check_layout(os);
2138 else if (t.cs() == "textasciicircum") {
2139 context.check_layout(os);
2144 else if (t.cs() == "textbackslash") {
2145 context.check_layout(os);
2146 os << "\n\\backslash\n";
2150 else if (t.cs() == "_" || t.cs() == "&" || t.cs() == "#"
2151 || t.cs() == "$" || t.cs() == "{" || t.cs() == "}"
2153 context.check_layout(os);
2157 else if (t.cs() == "char") {
2158 context.check_layout(os);
2159 if (p.next_token().character() == '`') {
2161 if (p.next_token().cs() == "\"") {
2166 handle_ert(os, "\\char`", context);
2169 handle_ert(os, "\\char", context);
2173 else if (t.cs() == "verb") {
2174 context.check_layout(os);
2175 char const delimiter = p.next_token().character();
2176 string const arg = p.getArg(delimiter, delimiter);
2178 oss << "\\verb" << delimiter << arg << delimiter;
2179 handle_ert(os, oss.str(), context);
2182 else if (t.cs() == "\"") {
2183 context.check_layout(os);
2184 string const name = p.verbatim_item();
2185 if (name == "a") os << '\xe4';
2186 else if (name == "o") os << '\xf6';
2187 else if (name == "u") os << '\xfc';
2188 else if (name == "A") os << '\xc4';
2189 else if (name == "O") os << '\xd6';
2190 else if (name == "U") os << '\xdc';
2191 else handle_ert(os, "\"{" + name + "}", context);
2194 // Problem: \= creates a tabstop inside the tabbing environment
2195 // and else an accent. In the latter case we really would want
2196 // \={o} instead of \= o.
2197 else if (t.cs() == "=" && (flags & FLAG_TABBING))
2198 handle_ert(os, t.asInput(), context);
2200 else if (t.cs() == "H" || t.cs() == "c" || t.cs() == "^"
2201 || t.cs() == "'" || t.cs() == "`"
2202 || t.cs() == "~" || t.cs() == "." || t.cs() == "=") {
2203 // we need the trim as the LyX parser chokes on such spaces
2204 // The argument of InsetLatexAccent is parsed as a
2205 // subset of LaTeX, so don't parse anything here,
2206 // but use the raw argument.
2207 // Otherwise we would convert \~{\i} wrongly.
2208 // This will of course not translate \~{\ss} to \~{ß},
2209 // but that does at least compile and does only look
2210 // strange on screen.
2211 context.check_layout(os);
2212 os << "\\i \\" << t.cs() << "{"
2213 << trim(p.verbatim_item(), " ")
2217 else if (t.cs() == "ss") {
2218 context.check_layout(os);
2220 skip_braces(p); // eat {}
2223 else if (t.cs() == "i" || t.cs() == "j" || t.cs() == "l" ||
2225 context.check_layout(os);
2226 os << "\\i \\" << t.cs() << "{}\n";
2227 skip_braces(p); // eat {}
2230 else if (t.cs() == "\\") {
2231 context.check_layout(os);
2232 string const next = p.next_token().asInput();
2234 handle_ert(os, "\\\\" + p.getOpt(), context);
2235 else if (next == "*") {
2237 handle_ert(os, "\\\\*" + p.getOpt(), context);
2240 os << "\n\\newline\n";
2244 else if (t.cs() == "newline" ||
2245 t.cs() == "linebreak") {
2246 context.check_layout(os);
2247 os << "\n\\" << t.cs() << "\n";
2248 skip_braces(p); // eat {}
2251 else if (t.cs() == "href") {
2252 context.check_layout(os);
2253 begin_inset(os, "CommandInset ");
2254 os << t.cs() << "\n";
2255 os << "LatexCommand " << t.cs() << "\n";
2258 // the first argument is "type:target", "type:" is optional
2259 // the second argument the name
2260 string href_target = subst(p.verbatim_item(), "\n", " ");
2261 string href_name = subst(p.verbatim_item(), "\n", " ");
2263 // serach for the ":" to divide type from target
2264 if ((pos = href_target.find(":", 0)) != string::npos){
2265 href_type = href_target;
2266 href_type.erase(pos + 1, href_type.length());
2267 href_target.erase(0, pos + 1);
2270 os << "name " << '"' << href_name << '"' << "\n";
2271 os << "target " << '"' << href_target << '"' << "\n";
2273 os << "type " << '"' << href_type << '"' << "\n";
2277 else if (t.cs() == "input" || t.cs() == "include"
2278 || t.cs() == "verbatiminput") {
2279 string name = '\\' + t.cs();
2280 if (t.cs() == "verbatiminput"
2281 && p.next_token().asInput() == "*")
2282 name += p.get_token().asInput();
2283 context.check_layout(os);
2284 begin_inset(os, "Include ");
2285 string filename(normalize_filename(p.getArg('{', '}')));
2286 string const path = getMasterFilePath();
2287 // We want to preserve relative / absolute filenames,
2288 // therefore path is only used for testing
2289 // FIXME UNICODE encoding of filename and path may be
2290 // wrong (makeAbsPath expects utf8)
2291 if ((t.cs() == "include" || t.cs() == "input") &&
2292 !makeAbsPath(filename, path).exists()) {
2293 // The file extension is probably missing.
2294 // Now try to find it out.
2295 string const tex_name =
2296 find_file(filename, path,
2297 known_tex_extensions);
2298 if (!tex_name.empty())
2299 filename = tex_name;
2301 // FIXME UNICODE encoding of filename and path may be
2302 // wrong (makeAbsPath expects utf8)
2303 if (makeAbsPath(filename, path).exists()) {
2304 string const abstexname =
2305 makeAbsPath(filename, path).absFilename();
2306 string const abslyxname =
2307 changeExtension(abstexname, ".lyx");
2308 fix_relative_filename(filename);
2309 string const lyxname =
2310 changeExtension(filename, ".lyx");
2311 if (t.cs() != "verbatiminput" &&
2312 tex2lyx(abstexname, FileName(abslyxname))) {
2313 os << name << '{' << lyxname << "}\n";
2315 os << name << '{' << filename << "}\n";
2318 cerr << "Warning: Could not find included file '"
2319 << filename << "'." << endl;
2320 os << name << '{' << filename << "}\n";
2322 os << "preview false\n";
2326 else if (t.cs() == "bibliographystyle") {
2327 // store new bibliographystyle
2328 bibliographystyle = p.verbatim_item();
2329 // output new bibliographystyle.
2330 // This is only necessary if used in some other macro than \bibliography.
2331 handle_ert(os, "\\bibliographystyle{" + bibliographystyle + "}", context);
2334 else if (t.cs() == "bibliography") {
2335 context.check_layout(os);
2336 begin_inset(os, "CommandInset bibtex");
2337 os << "\nLatexCommand bibtex";
2338 // Do we have a bibliographystyle set?
2339 if (!bibliographystyle.empty())
2340 os << "\noptions " << '"' << bibliographystyle << '"';
2341 os << "\nbibfiles " << '"' << p.verbatim_item() << '"';
2345 else if (t.cs() == "printindex") {
2346 context.check_layout(os);
2347 begin_inset(os, "CommandInset index_print\n");
2348 os << "LatexCommand printindex";
2353 else if (t.cs() == "printnomenclature") {
2354 context.check_layout(os);
2355 begin_inset(os, "CommandInset nomencl_print\n");
2356 os << "LatexCommand printnomenclature";
2361 else if (t.cs() == "parbox")
2362 parse_box(p, os, FLAG_ITEM, outer, context, true);
2364 else if (t.cs() == "smallskip" ||
2365 t.cs() == "medskip" ||
2366 t.cs() == "bigskip" ||
2367 t.cs() == "vfill") {
2368 context.check_layout(os);
2369 begin_inset(os, "VSpace ");
2375 else if (is_known(t.cs(), known_spaces)) {
2376 char const * const * where = is_known(t.cs(), known_spaces);
2377 context.check_layout(os);
2378 begin_inset(os, "InsetSpace ");
2379 os << '\\' << known_coded_spaces[where - known_spaces]
2381 // LaTeX swallows whitespace after all spaces except
2382 // "\\,". We have to do that here, too, because LyX
2383 // adds "{}" which would make the spaces significant.
2385 eat_whitespace(p, os, context, false);
2386 // LyX adds "{}" after all spaces except "\\ " and
2387 // "\\,", so we have to remove "{}".
2388 // "\\,{}" is equivalent to "\\," in LaTeX, so we
2389 // remove the braces after "\\,", too.
2394 else if (t.cs() == "newpage" ||
2395 t.cs() == "pagebreak" ||
2396 t.cs() == "clearpage" ||
2397 t.cs() == "cleardoublepage") {
2398 context.check_layout(os);
2399 os << "\n\\" << t.cs() << "\n";
2400 skip_braces(p); // eat {}
2403 else if (t.cs() == "newcommand" ||
2404 t.cs() == "providecommand" ||
2405 t.cs() == "renewcommand") {
2406 // these could be handled by parse_command(), but
2407 // we need to call add_known_command() here.
2408 string name = t.asInput();
2409 if (p.next_token().asInput() == "*") {
2410 // Starred form. Eat '*'
2414 string const command = p.verbatim_item();
2415 string const opt1 = p.getOpt();
2416 string const opt2 = p.getFullOpt();
2417 add_known_command(command, opt1, !opt2.empty());
2418 string const ert = name + '{' + command + '}' +
2420 '{' + p.verbatim_item() + '}';
2422 context.check_layout(os);
2423 begin_inset(os, "FormulaMacro");
2428 else if (t.cs() == "vspace") {
2429 bool starred = false;
2430 if (p.next_token().asInput() == "*") {
2434 string const length = p.verbatim_item();
2437 bool valid = splitLatexLength(length, valstring, unit);
2438 bool known_vspace = false;
2439 bool known_unit = false;
2442 istringstream iss(valstring);
2445 if (unit == "\\smallskipamount") {
2447 known_vspace = true;
2448 } else if (unit == "\\medskipamount") {
2450 known_vspace = true;
2451 } else if (unit == "\\bigskipamount") {
2453 known_vspace = true;
2454 } else if (unit == "\\fill") {
2456 known_vspace = true;
2459 if (!known_vspace) {
2460 switch (unitFromString(unit)) {
2481 if (known_unit || known_vspace) {
2482 // Literal length or known variable
2483 context.check_layout(os);
2484 begin_inset(os, "VSpace ");
2492 // LyX can't handle other length variables in Inset VSpace
2493 string name = t.asInput();
2498 handle_ert(os, name + '{' + unit + '}', context);
2499 else if (value == -1.0)
2500 handle_ert(os, name + "{-" + unit + '}', context);
2502 handle_ert(os, name + '{' + valstring + unit + '}', context);
2504 handle_ert(os, name + '{' + length + '}', context);
2509 //cerr << "#: " << t << " mode: " << mode << endl;
2510 // heuristic: read up to next non-nested space
2512 string s = t.asInput();
2513 string z = p.verbatim_item();
2514 while (p.good() && z != " " && z.size()) {
2515 //cerr << "read: " << z << endl;
2517 z = p.verbatim_item();
2519 cerr << "found ERT: " << s << endl;
2520 handle_ert(os, s + ' ', context);
2522 string name = t.asInput();
2523 if (p.next_token().asInput() == "*") {
2524 // Starred commands like \vspace*{}
2525 p.get_token(); // Eat '*'
2528 if (! parse_command(name, p, os, outer, context))
2529 handle_ert(os, name, context);
2532 if (flags & FLAG_LEAVE) {
2533 flags &= ~FLAG_LEAVE;