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", "printindex", "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);
779 else if (name == "lyxgreyedout") {
780 eat_whitespace(p, os, parent_context, false);
781 parent_context.check_layout(os);
782 begin_inset(os, "Note Greyedout\n");
783 os << "status open\n";
784 parse_text_in_inset(p, os, FLAG_END, outer, parent_context);
789 else if (name == "framed") {
790 eat_whitespace(p, os, parent_context, false);
791 parent_context.check_layout(os);
792 begin_inset(os, "Note Framed\n");
793 os << "status open\n";
794 parse_text_in_inset(p, os, FLAG_END, outer, parent_context);
799 else if (name == "shaded") {
800 eat_whitespace(p, os, parent_context, false);
801 parent_context.check_layout(os);
802 begin_inset(os, "Note Shaded\n");
803 os << "status open\n";
804 parse_text_in_inset(p, os, FLAG_END, outer, parent_context);
809 else if (!parent_context.new_layout_allowed)
810 parse_unknown_environment(p, name, os, FLAG_END, outer,
813 // Alignment settings
814 else if (name == "center" || name == "flushleft" || name == "flushright" ||
815 name == "centering" || name == "raggedright" || name == "raggedleft") {
816 eat_whitespace(p, os, parent_context, false);
817 // We must begin a new paragraph if not already done
818 if (! parent_context.atParagraphStart()) {
819 parent_context.check_end_layout(os);
820 parent_context.new_paragraph(os);
822 if (name == "flushleft" || name == "raggedright")
823 parent_context.add_extra_stuff("\\align left\n");
824 else if (name == "flushright" || name == "raggedleft")
825 parent_context.add_extra_stuff("\\align right\n");
827 parent_context.add_extra_stuff("\\align center\n");
828 parse_text(p, os, FLAG_END, outer, parent_context);
829 // Just in case the environment is empty ..
830 parent_context.extra_stuff.erase();
831 // We must begin a new paragraph to reset the alignment
832 parent_context.new_paragraph(os);
836 // The single '=' is meant here.
837 else if ((newlayout = findLayout(parent_context.textclass, name)).get() &&
838 newlayout->isEnvironment()) {
839 eat_whitespace(p, os, parent_context, false);
840 Context context(true, parent_context.textclass, newlayout,
841 parent_context.layout, parent_context.font);
842 if (parent_context.deeper_paragraph) {
843 // We are beginning a nested environment after a
844 // deeper paragraph inside the outer list environment.
845 // Therefore we don't need to output a "begin deeper".
846 context.need_end_deeper = true;
848 parent_context.check_end_layout(os);
849 switch (context.layout->latextype) {
850 case LATEX_LIST_ENVIRONMENT:
851 context.extra_stuff = "\\labelwidthstring "
852 + p.verbatim_item() + '\n';
855 case LATEX_BIB_ENVIRONMENT:
856 p.verbatim_item(); // swallow next arg
862 context.check_deeper(os);
863 parse_text(p, os, FLAG_END, outer, context);
864 context.check_end_layout(os);
865 if (parent_context.deeper_paragraph) {
866 // We must suppress the "end deeper" because we
867 // suppressed the "begin deeper" above.
868 context.need_end_deeper = false;
870 context.check_end_deeper(os);
871 parent_context.new_paragraph(os);
875 else if (name == "appendix") {
876 // This is no good latex style, but it works and is used in some documents...
877 eat_whitespace(p, os, parent_context, false);
878 parent_context.check_end_layout(os);
879 Context context(true, parent_context.textclass, parent_context.layout,
880 parent_context.layout, parent_context.font);
881 context.check_layout(os);
882 os << "\\start_of_appendix\n";
883 parse_text(p, os, FLAG_END, outer, context);
884 context.check_end_layout(os);
888 else if (known_environments.find(name) != known_environments.end()) {
889 vector<ArgumentType> arguments = known_environments[name];
890 // The last "argument" denotes wether we may translate the
891 // environment contents to LyX
892 // The default required if no argument is given makes us
893 // compatible with the reLyXre environment.
894 ArgumentType contents = arguments.empty() ?
897 if (!arguments.empty())
898 arguments.pop_back();
899 // See comment in parse_unknown_environment()
900 bool const specialfont =
901 (parent_context.font != parent_context.normalfont);
902 bool const new_layout_allowed =
903 parent_context.new_layout_allowed;
905 parent_context.new_layout_allowed = false;
906 parse_arguments("\\begin{" + name + "}", arguments, p, os,
907 outer, parent_context);
908 if (contents == verbatim)
909 handle_ert(os, p.verbatimEnvironment(name),
912 parse_text_snippet(p, os, FLAG_END, outer,
914 handle_ert(os, "\\end{" + name + "}", parent_context);
916 parent_context.new_layout_allowed = new_layout_allowed;
920 parse_unknown_environment(p, name, os, FLAG_END, outer,
923 active_environments.pop_back();
927 /// parses a comment and outputs it to \p os.
928 void parse_comment(Parser & p, ostream & os, Token const & t, Context & context)
930 BOOST_ASSERT(t.cat() == catComment);
931 if (!t.cs().empty()) {
932 context.check_layout(os);
933 handle_comment(os, '%' + t.cs(), context);
934 if (p.next_token().cat() == catNewline) {
935 // A newline after a comment line starts a new
937 if (context.new_layout_allowed) {
938 if(!context.atParagraphStart())
939 // Only start a new paragraph if not already
940 // done (we might get called recursively)
941 context.new_paragraph(os);
943 handle_ert(os, "\n", context);
944 eat_whitespace(p, os, context, true);
954 * Reads spaces and comments until the first non-space, non-comment token.
955 * New paragraphs (double newlines or \\par) are handled like simple spaces
956 * if \p eatParagraph is true.
957 * Spaces are skipped, but comments are written to \p os.
959 void eat_whitespace(Parser & p, ostream & os, Context & context,
963 Token const & t = p.get_token();
964 if (t.cat() == catComment)
965 parse_comment(p, os, t, context);
966 else if ((! eatParagraph && p.isParagraph()) ||
967 (t.cat() != catSpace && t.cat() != catNewline)) {
976 * Set a font attribute, parse text and reset the font attribute.
977 * \param attribute Attribute name (e.g. \\family, \\shape etc.)
978 * \param currentvalue Current value of the attribute. Is set to the new
979 * value during parsing.
980 * \param newvalue New value of the attribute
982 void parse_text_attributes(Parser & p, ostream & os, unsigned flags, bool outer,
983 Context & context, string const & attribute,
984 string & currentvalue, string const & newvalue)
986 context.check_layout(os);
987 string const oldvalue = currentvalue;
988 currentvalue = newvalue;
989 os << '\n' << attribute << ' ' << newvalue << "\n";
990 parse_text_snippet(p, os, flags, outer, context);
991 context.check_layout(os);
992 os << '\n' << attribute << ' ' << oldvalue << "\n";
993 currentvalue = oldvalue;
997 /// get the arguments of a natbib or jurabib citation command
998 void get_cite_arguments(Parser & p, bool natbibOrder,
999 string & before, string & after)
1001 // We need to distinguish "" and "[]", so we can't use p.getOpt().
1003 // text before the citation
1005 // text after the citation
1006 after = p.getFullOpt();
1008 if (!after.empty()) {
1009 before = p.getFullOpt();
1010 if (natbibOrder && !before.empty())
1011 std::swap(before, after);
1016 /// Convert filenames with TeX macros and/or quotes to something LyX
1018 string const normalize_filename(string const & name)
1020 Parser p(trim(name, "\""));
1023 Token const & t = p.get_token();
1024 if (t.cat() != catEscape)
1026 else if (t.cs() == "lyxdot") {
1027 // This is used by LyX for simple dots in relative
1031 } else if (t.cs() == "space") {
1041 /// Convert \p name from TeX convention (relative to master file) to LyX
1042 /// convention (relative to .lyx file) if it is relative
1043 void fix_relative_filename(string & name)
1045 if (lyx::support::absolutePath(name))
1047 // FIXME UNICODE encoding of name may be wrong (makeAbsPath expects
1049 name = to_utf8(makeRelPath(from_utf8(makeAbsPath(name, getMasterFilePath()).absFilename()),
1050 from_utf8(getParentFilePath())));
1054 /// Parse a NoWeb Scrap section. The initial "<<" is already parsed.
1055 void parse_noweb(Parser & p, ostream & os, Context & context)
1057 // assemble the rest of the keyword
1061 Token const & t = p.get_token();
1062 if (t.asInput() == ">" && p.next_token().asInput() == ">") {
1065 scrap = (p.good() && p.next_token().asInput() == "=");
1067 name += p.get_token().asInput();
1070 name += t.asInput();
1073 if (!scrap || !context.new_layout_allowed ||
1074 !context.textclass.hasLayout(from_ascii("Scrap"))) {
1075 cerr << "Warning: Could not interpret '" << name
1076 << "'. Ignoring it." << endl;
1080 // We use new_paragraph instead of check_end_layout because the stuff
1081 // following the noweb chunk needs to start with a \begin_layout.
1082 // This may create a new paragraph even if there was none in the
1083 // noweb file, but the alternative is an invalid LyX file. Since
1084 // noweb code chunks are implemented with a layout style in LyX they
1085 // always must be in an own paragraph.
1086 context.new_paragraph(os);
1087 Context newcontext(true, context.textclass,
1088 context.textclass[from_ascii("Scrap")]);
1089 newcontext.check_layout(os);
1092 Token const & t = p.get_token();
1093 // We abuse the parser a bit, because this is no TeX syntax
1095 if (t.cat() == catEscape)
1096 os << subst(t.asInput(), "\\", "\n\\backslash\n");
1098 os << subst(t.asInput(), "\n", "\n\\newline\n");
1099 // The scrap chunk is ended by an @ at the beginning of a line.
1100 // After the @ the line may contain a comment and/or
1101 // whitespace, but nothing else.
1102 if (t.asInput() == "@" && p.prev_token().cat() == catNewline &&
1103 (p.next_token().cat() == catSpace ||
1104 p.next_token().cat() == catNewline ||
1105 p.next_token().cat() == catComment)) {
1106 while (p.good() && p.next_token().cat() == catSpace)
1107 os << p.get_token().asInput();
1108 if (p.next_token().cat() == catComment)
1109 // The comment includes a final '\n'
1110 os << p.get_token().asInput();
1112 if (p.next_token().cat() == catNewline)
1119 newcontext.check_end_layout(os);
1122 } // anonymous namespace
1125 void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
1128 LayoutPtr newlayout;
1129 // store the current selectlanguage to be used after \foreignlanguage
1131 // Store the latest bibliographystyle (needed for bibtex inset)
1132 string bibliographystyle;
1133 bool const use_natbib = used_packages.find("natbib") != used_packages.end();
1134 bool const use_jurabib = used_packages.find("jurabib") != used_packages.end();
1136 Token const & t = p.get_token();
1139 cerr << "t: " << t << " flags: " << flags << "\n";
1142 if (flags & FLAG_ITEM) {
1143 if (t.cat() == catSpace)
1146 flags &= ~FLAG_ITEM;
1147 if (t.cat() == catBegin) {
1148 // skip the brace and collect everything to the next matching
1150 flags |= FLAG_BRACE_LAST;
1154 // handle only this single token, leave the loop if done
1155 flags |= FLAG_LEAVE;
1158 if (t.character() == ']' && (flags & FLAG_BRACK_LAST))
1164 if (t.cat() == catMath) {
1165 // we are inside some text mode thingy, so opening new math is allowed
1166 context.check_layout(os);
1167 begin_inset(os, "Formula ");
1168 Token const & n = p.get_token();
1169 if (n.cat() == catMath && outer) {
1170 // TeX's $$...$$ syntax for displayed math
1172 parse_math(p, os, FLAG_SIMPLE, MATH_MODE);
1174 p.get_token(); // skip the second '$' token
1176 // simple $...$ stuff
1179 parse_math(p, os, FLAG_SIMPLE, MATH_MODE);
1185 else if (t.cat() == catSuper || t.cat() == catSub)
1186 cerr << "catcode " << t << " illegal in text mode\n";
1188 // Basic support for english quotes. This should be
1189 // extended to other quotes, but is not so easy (a
1190 // left english quote is the same as a right german
1192 else if (t.asInput() == "`"
1193 && p.next_token().asInput() == "`") {
1194 context.check_layout(os);
1195 begin_inset(os, "Quotes ");
1201 else if (t.asInput() == "'"
1202 && p.next_token().asInput() == "'") {
1203 context.check_layout(os);
1204 begin_inset(os, "Quotes ");
1211 else if (t.asInput() == "<"
1212 && p.next_token().asInput() == "<" && noweb_mode) {
1214 parse_noweb(p, os, context);
1217 else if (t.cat() == catSpace || (t.cat() == catNewline && ! p.isParagraph()))
1218 check_space(p, os, context);
1220 else if (t.character() == '[' && noweb_mode &&
1221 p.next_token().character() == '[') {
1222 // These can contain underscores
1224 string const s = p.getFullOpt() + ']';
1225 if (p.next_token().character() == ']')
1228 cerr << "Warning: Inserting missing ']' in '"
1229 << s << "'." << endl;
1230 handle_ert(os, s, context);
1233 else if (t.cat() == catLetter ||
1234 t.cat() == catOther ||
1235 t.cat() == catAlign ||
1236 t.cat() == catParameter) {
1237 // This translates "&" to "\\&" which may be wrong...
1238 context.check_layout(os);
1239 os << t.character();
1242 else if (p.isParagraph()) {
1243 if (context.new_layout_allowed)
1244 context.new_paragraph(os);
1246 handle_ert(os, "\\par ", context);
1247 eat_whitespace(p, os, context, true);
1250 else if (t.cat() == catActive) {
1251 context.check_layout(os);
1252 if (t.character() == '~') {
1253 if (context.layout->free_spacing)
1256 os << "\\InsetSpace ~\n";
1258 os << t.character();
1261 else if (t.cat() == catBegin &&
1262 p.next_token().cat() == catEnd) {
1264 Token const prev = p.prev_token();
1266 if (p.next_token().character() == '`' ||
1267 (prev.character() == '-' &&
1268 p.next_token().character() == '-'))
1269 ; // ignore it in {}`` or -{}-
1271 handle_ert(os, "{}", context);
1275 else if (t.cat() == catBegin) {
1276 context.check_layout(os);
1277 // special handling of font attribute changes
1278 Token const prev = p.prev_token();
1279 Token const next = p.next_token();
1280 TeXFont const oldFont = context.font;
1281 if (next.character() == '[' ||
1282 next.character() == ']' ||
1283 next.character() == '*') {
1285 if (p.next_token().cat() == catEnd) {
1286 os << next.character();
1290 handle_ert(os, "{", context);
1291 parse_text_snippet(p, os,
1294 handle_ert(os, "}", context);
1296 } else if (! context.new_layout_allowed) {
1297 handle_ert(os, "{", context);
1298 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1300 handle_ert(os, "}", context);
1301 } else if (is_known(next.cs(), known_sizes)) {
1302 // next will change the size, so we must
1304 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1306 if (!context.atParagraphStart())
1308 << context.font.size << "\n";
1309 } else if (is_known(next.cs(), known_font_families)) {
1310 // next will change the font family, so we
1311 // must reset it here
1312 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1314 if (!context.atParagraphStart())
1316 << context.font.family << "\n";
1317 } else if (is_known(next.cs(), known_font_series)) {
1318 // next will change the font series, so we
1319 // must reset it here
1320 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1322 if (!context.atParagraphStart())
1324 << context.font.series << "\n";
1325 } else if (is_known(next.cs(), known_font_shapes)) {
1326 // next will change the font shape, so we
1327 // must reset it here
1328 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1330 if (!context.atParagraphStart())
1332 << context.font.shape << "\n";
1333 } else if (is_known(next.cs(), known_old_font_families) ||
1334 is_known(next.cs(), known_old_font_series) ||
1335 is_known(next.cs(), known_old_font_shapes)) {
1336 // next will change the font family, series
1337 // and shape, so we must reset it here
1338 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1340 if (!context.atParagraphStart())
1342 << context.font.family
1344 << context.font.series
1346 << context.font.shape << "\n";
1348 handle_ert(os, "{", context);
1349 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1351 handle_ert(os, "}", context);
1355 else if (t.cat() == catEnd) {
1356 if (flags & FLAG_BRACE_LAST) {
1359 cerr << "stray '}' in text\n";
1360 handle_ert(os, "}", context);
1363 else if (t.cat() == catComment)
1364 parse_comment(p, os, t, context);
1367 // control sequences
1370 else if (t.cs() == "(") {
1371 context.check_layout(os);
1372 begin_inset(os, "Formula");
1374 parse_math(p, os, FLAG_SIMPLE2, MATH_MODE);
1379 else if (t.cs() == "[") {
1380 context.check_layout(os);
1381 begin_inset(os, "Formula");
1383 parse_math(p, os, FLAG_EQUATION, MATH_MODE);
1388 else if (t.cs() == "begin")
1389 parse_environment(p, os, outer, context);
1391 else if (t.cs() == "end") {
1392 if (flags & FLAG_END) {
1393 // eat environment name
1394 string const name = p.getArg('{', '}');
1395 if (name != active_environment())
1396 cerr << "\\end{" + name + "} does not match \\begin{"
1397 + active_environment() + "}\n";
1400 p.error("found 'end' unexpectedly");
1403 else if (t.cs() == "item") {
1406 bool optarg = false;
1407 if (p.next_token().character() == '[') {
1408 p.get_token(); // eat '['
1409 s = parse_text_snippet(p, FLAG_BRACK_LAST,
1414 context.check_layout(os);
1415 if (context.has_item) {
1416 // An item in an unknown list-like environment
1417 // FIXME: Do this in check_layout()!
1418 context.has_item = false;
1420 handle_ert(os, "\\item", context);
1422 handle_ert(os, "\\item ", context);
1425 if (context.layout->labeltype != LABEL_MANUAL) {
1426 // lyx does not support \item[\mybullet]
1427 // in itemize environments
1428 handle_ert(os, "[", context);
1430 handle_ert(os, "]", context);
1431 } else if (!s.empty()) {
1432 // The space is needed to separate the
1433 // item from the rest of the sentence.
1435 eat_whitespace(p, os, context, false);
1440 else if (t.cs() == "bibitem") {
1442 context.check_layout(os);
1445 os << '{' << p.verbatim_item() << '}' << "\n";
1448 else if (t.cs() == "def") {
1449 context.check_layout(os);
1450 eat_whitespace(p, os, context, false);
1451 string name = p.get_token().cs();
1452 eat_whitespace(p, os, context, false);
1458 while (p.next_token().cat() != catBegin) {
1459 if (p.next_token().cat() == catParameter) {
1464 // followed by number?
1465 if (p.next_token().cat() == catOther) {
1466 char c = p.getChar();
1468 // number = current arity + 1?
1469 if (c == arity + '0' + 1)
1474 paramtext += p.get_token().asString();
1476 paramtext += p.get_token().asString();
1481 // only output simple (i.e. compatible) macro as FormulaMacros
1482 string ert = "\\def\\" + name + ' ' + paramtext + '{' + p.verbatim_item() + '}';
1484 context.check_layout(os);
1485 begin_inset(os, "FormulaMacro");
1489 handle_ert(os, ert, context);
1492 else if (t.cs() == "noindent") {
1494 context.add_extra_stuff("\\noindent\n");
1497 else if (t.cs() == "appendix") {
1498 context.add_extra_stuff("\\start_of_appendix\n");
1499 // We need to start a new paragraph. Otherwise the
1500 // appendix in 'bla\appendix\chapter{' would start
1502 context.new_paragraph(os);
1503 // We need to make sure that the paragraph is
1504 // generated even if it is empty. Otherwise the
1505 // appendix in '\par\appendix\par\chapter{' would
1507 context.check_layout(os);
1508 // FIXME: This is a hack to prevent paragraph
1509 // deletion if it is empty. Handle this better!
1511 "%dummy comment inserted by tex2lyx to "
1512 "ensure that this paragraph is not empty",
1514 // Both measures above may generate an additional
1515 // empty paragraph, but that does not hurt, because
1516 // whitespace does not matter here.
1517 eat_whitespace(p, os, context, true);
1520 // Must attempt to parse "Section*" before "Section".
1521 else if ((p.next_token().asInput() == "*") &&
1522 context.new_layout_allowed &&
1523 // The single '=' is meant here.
1524 (newlayout = findLayout(context.textclass,
1525 t.cs() + '*')).get() &&
1526 newlayout->isCommand()) {
1528 output_command_layout(os, p, outer, context, newlayout);
1532 // The single '=' is meant here.
1533 else if (context.new_layout_allowed &&
1534 (newlayout = findLayout(context.textclass, t.cs())).get() &&
1535 newlayout->isCommand()) {
1536 output_command_layout(os, p, outer, context, newlayout);
1540 // Special handling for \caption
1541 // FIXME: remove this when InsetCaption is supported.
1542 else if (context.new_layout_allowed &&
1543 t.cs() == captionlayout->latexname()) {
1544 output_command_layout(os, p, outer, context,
1549 else if (t.cs() == "includegraphics") {
1550 bool const clip = p.next_token().asInput() == "*";
1553 map<string, string> opts = split_map(p.getArg('[', ']'));
1555 opts["clip"] = string();
1556 string name = normalize_filename(p.verbatim_item());
1558 string const path = getMasterFilePath();
1559 // We want to preserve relative / absolute filenames,
1560 // therefore path is only used for testing
1561 // FIXME UNICODE encoding of name and path may be
1562 // wrong (makeAbsPath expects utf8)
1563 if (!makeAbsPath(name, path).exists()) {
1564 // The file extension is probably missing.
1565 // Now try to find it out.
1566 string const dvips_name =
1567 find_file(name, path,
1568 known_dvips_graphics_formats);
1569 string const pdftex_name =
1570 find_file(name, path,
1571 known_pdftex_graphics_formats);
1572 if (!dvips_name.empty()) {
1573 if (!pdftex_name.empty()) {
1574 cerr << "This file contains the "
1576 "\"\\includegraphics{"
1578 "However, files\n\""
1579 << dvips_name << "\" and\n\""
1580 << pdftex_name << "\"\n"
1581 "both exist, so I had to make a "
1582 "choice and took the first one.\n"
1583 "Please move the unwanted one "
1584 "someplace else and try again\n"
1585 "if my choice was wrong."
1589 } else if (!pdftex_name.empty())
1593 // FIXME UNICODE encoding of name and path may be
1594 // wrong (makeAbsPath expects utf8)
1595 if (makeAbsPath(name, path).exists())
1596 fix_relative_filename(name);
1598 cerr << "Warning: Could not find graphics file '"
1599 << name << "'." << endl;
1601 context.check_layout(os);
1602 begin_inset(os, "Graphics ");
1603 os << "\n\tfilename " << name << '\n';
1604 if (opts.find("width") != opts.end())
1606 << translate_len(opts["width"]) << '\n';
1607 if (opts.find("height") != opts.end())
1609 << translate_len(opts["height"]) << '\n';
1610 if (opts.find("scale") != opts.end()) {
1611 istringstream iss(opts["scale"]);
1615 os << "\tscale " << val << '\n';
1617 if (opts.find("angle") != opts.end())
1618 os << "\trotateAngle "
1619 << opts["angle"] << '\n';
1620 if (opts.find("origin") != opts.end()) {
1622 string const opt = opts["origin"];
1623 if (opt.find('l') != string::npos) ss << "left";
1624 if (opt.find('r') != string::npos) ss << "right";
1625 if (opt.find('c') != string::npos) ss << "center";
1626 if (opt.find('t') != string::npos) ss << "Top";
1627 if (opt.find('b') != string::npos) ss << "Bottom";
1628 if (opt.find('B') != string::npos) ss << "Baseline";
1629 if (!ss.str().empty())
1630 os << "\trotateOrigin " << ss.str() << '\n';
1632 cerr << "Warning: Ignoring unknown includegraphics origin argument '" << opt << "'\n";
1634 if (opts.find("keepaspectratio") != opts.end())
1635 os << "\tkeepAspectRatio\n";
1636 if (opts.find("clip") != opts.end())
1638 if (opts.find("draft") != opts.end())
1640 if (opts.find("bb") != opts.end())
1641 os << "\tBoundingBox "
1642 << opts["bb"] << '\n';
1643 int numberOfbbOptions = 0;
1644 if (opts.find("bbllx") != opts.end())
1645 numberOfbbOptions++;
1646 if (opts.find("bblly") != opts.end())
1647 numberOfbbOptions++;
1648 if (opts.find("bburx") != opts.end())
1649 numberOfbbOptions++;
1650 if (opts.find("bbury") != opts.end())
1651 numberOfbbOptions++;
1652 if (numberOfbbOptions == 4)
1653 os << "\tBoundingBox "
1654 << opts["bbllx"] << opts["bblly"]
1655 << opts["bburx"] << opts["bbury"] << '\n';
1656 else if (numberOfbbOptions > 0)
1657 cerr << "Warning: Ignoring incomplete includegraphics boundingbox arguments.\n";
1658 numberOfbbOptions = 0;
1659 if (opts.find("natwidth") != opts.end())
1660 numberOfbbOptions++;
1661 if (opts.find("natheight") != opts.end())
1662 numberOfbbOptions++;
1663 if (numberOfbbOptions == 2)
1664 os << "\tBoundingBox 0bp 0bp "
1665 << opts["natwidth"] << opts["natheight"] << '\n';
1666 else if (numberOfbbOptions > 0)
1667 cerr << "Warning: Ignoring incomplete includegraphics boundingbox arguments.\n";
1668 ostringstream special;
1669 if (opts.find("hiresbb") != opts.end())
1670 special << "hiresbb,";
1671 if (opts.find("trim") != opts.end())
1673 if (opts.find("viewport") != opts.end())
1674 special << "viewport=" << opts["viewport"] << ',';
1675 if (opts.find("totalheight") != opts.end())
1676 special << "totalheight=" << opts["totalheight"] << ',';
1677 if (opts.find("type") != opts.end())
1678 special << "type=" << opts["type"] << ',';
1679 if (opts.find("ext") != opts.end())
1680 special << "ext=" << opts["ext"] << ',';
1681 if (opts.find("read") != opts.end())
1682 special << "read=" << opts["read"] << ',';
1683 if (opts.find("command") != opts.end())
1684 special << "command=" << opts["command"] << ',';
1685 string s_special = special.str();
1686 if (!s_special.empty()) {
1687 // We had special arguments. Remove the trailing ','.
1688 os << "\tspecial " << s_special.substr(0, s_special.size() - 1) << '\n';
1690 // TODO: Handle the unknown settings better.
1691 // Warn about invalid options.
1692 // Check whether some option was given twice.
1696 else if (t.cs() == "footnote" ||
1697 (t.cs() == "thanks" && context.layout->intitle)) {
1699 context.check_layout(os);
1700 begin_inset(os, "Foot\n");
1701 os << "status collapsed\n\n";
1702 parse_text_in_inset(p, os, FLAG_ITEM, false, context);
1706 else if (t.cs() == "marginpar") {
1708 context.check_layout(os);
1709 begin_inset(os, "Marginal\n");
1710 os << "status collapsed\n\n";
1711 parse_text_in_inset(p, os, FLAG_ITEM, false, context);
1715 else if (t.cs() == "ensuremath") {
1717 context.check_layout(os);
1718 string const s = p.verbatim_item();
1719 if (s == "\xb1" || s == "\xb3" || s == "\xb2" || s == "\xb5")
1722 handle_ert(os, "\\ensuremath{" + s + "}",
1726 else if (t.cs() == "hfill") {
1727 context.check_layout(os);
1728 os << "\n\\hfill\n";
1733 else if (t.cs() == "makeindex" || t.cs() == "maketitle") {
1734 // FIXME: Somehow prevent title layouts if
1735 // "maketitle" was not found
1737 skip_braces(p); // swallow this
1740 else if (t.cs() == "tableofcontents") {
1742 context.check_layout(os);
1743 begin_inset(os, "LatexCommand \\tableofcontents\n");
1745 skip_braces(p); // swallow this
1748 else if (t.cs() == "listoffigures") {
1750 context.check_layout(os);
1751 begin_inset(os, "FloatList figure\n");
1753 skip_braces(p); // swallow this
1756 else if (t.cs() == "listoftables") {
1758 context.check_layout(os);
1759 begin_inset(os, "FloatList table\n");
1761 skip_braces(p); // swallow this
1764 else if (t.cs() == "listof") {
1765 p.skip_spaces(true);
1766 string const name = p.get_token().asString();
1767 if (context.textclass.floats().typeExist(name)) {
1768 context.check_layout(os);
1769 begin_inset(os, "FloatList ");
1772 p.get_token(); // swallow second arg
1774 handle_ert(os, "\\listof{" + name + "}", context);
1777 else if (t.cs() == "textrm")
1778 parse_text_attributes(p, os, FLAG_ITEM, outer,
1779 context, "\\family",
1780 context.font.family, "roman");
1782 else if (t.cs() == "textsf")
1783 parse_text_attributes(p, os, FLAG_ITEM, outer,
1784 context, "\\family",
1785 context.font.family, "sans");
1787 else if (t.cs() == "texttt")
1788 parse_text_attributes(p, os, FLAG_ITEM, outer,
1789 context, "\\family",
1790 context.font.family, "typewriter");
1792 else if (t.cs() == "textmd")
1793 parse_text_attributes(p, os, FLAG_ITEM, outer,
1794 context, "\\series",
1795 context.font.series, "medium");
1797 else if (t.cs() == "textbf")
1798 parse_text_attributes(p, os, FLAG_ITEM, outer,
1799 context, "\\series",
1800 context.font.series, "bold");
1802 else if (t.cs() == "textup")
1803 parse_text_attributes(p, os, FLAG_ITEM, outer,
1805 context.font.shape, "up");
1807 else if (t.cs() == "textit")
1808 parse_text_attributes(p, os, FLAG_ITEM, outer,
1810 context.font.shape, "italic");
1812 else if (t.cs() == "textsl")
1813 parse_text_attributes(p, os, FLAG_ITEM, outer,
1815 context.font.shape, "slanted");
1817 else if (t.cs() == "textsc")
1818 parse_text_attributes(p, os, FLAG_ITEM, outer,
1820 context.font.shape, "smallcaps");
1822 else if (t.cs() == "textnormal" || t.cs() == "normalfont") {
1823 context.check_layout(os);
1824 TeXFont oldFont = context.font;
1825 context.font.init();
1826 context.font.size = oldFont.size;
1827 os << "\n\\family " << context.font.family << "\n";
1828 os << "\n\\series " << context.font.series << "\n";
1829 os << "\n\\shape " << context.font.shape << "\n";
1830 if (t.cs() == "textnormal") {
1831 parse_text_snippet(p, os, FLAG_ITEM, outer, context);
1832 output_font_change(os, context.font, oldFont);
1833 context.font = oldFont;
1835 eat_whitespace(p, os, context, false);
1838 else if (t.cs() == "underbar") {
1839 // Do NOT handle \underline.
1840 // \underbar cuts through y, g, q, p etc.,
1841 // \underline does not.
1842 context.check_layout(os);
1843 os << "\n\\bar under\n";
1844 parse_text_snippet(p, os, FLAG_ITEM, outer, context);
1845 context.check_layout(os);
1846 os << "\n\\bar default\n";
1849 else if (t.cs() == "emph" || t.cs() == "noun") {
1850 context.check_layout(os);
1851 os << "\n\\" << t.cs() << " on\n";
1852 parse_text_snippet(p, os, FLAG_ITEM, outer, context);
1853 context.check_layout(os);
1854 os << "\n\\" << t.cs() << " default\n";
1857 else if (use_natbib &&
1858 is_known(t.cs(), known_natbib_commands) &&
1859 ((t.cs() != "citefullauthor" &&
1860 t.cs() != "citeyear" &&
1861 t.cs() != "citeyearpar") ||
1862 p.next_token().asInput() != "*")) {
1863 context.check_layout(os);
1865 // \citet[before][after]{a} \citet[after][before]{a}
1866 // \citet[before][]{a} \citet[][before]{a}
1867 // \citet[after]{a} \citet[after]{a}
1868 // \citet{a} \citet{a}
1869 string command = '\\' + t.cs();
1870 if (p.next_token().asInput() == "*") {
1874 if (command == "\\citefullauthor")
1875 // alternative name for "\\citeauthor*"
1876 command = "\\citeauthor*";
1878 // text before the citation
1880 // text after the citation
1882 get_cite_arguments(p, true, before, after);
1884 if (command == "\\cite") {
1885 // \cite without optional argument means
1886 // \citet, \cite with at least one optional
1887 // argument means \citep.
1888 if (before.empty() && after.empty())
1889 command = "\\citet";
1891 command = "\\citep";
1893 if (before.empty() && after == "[]")
1894 // avoid \citet[]{a}
1896 else if (before == "[]" && after == "[]") {
1897 // avoid \citet[][]{a}
1901 begin_inset(os, "LatexCommand ");
1902 os << command << after << before
1903 << '{' << p.verbatim_item() << "}\n";
1907 else if (use_jurabib &&
1908 is_known(t.cs(), known_jurabib_commands)) {
1909 context.check_layout(os);
1910 string const command = '\\' + t.cs();
1911 char argumentOrder = '\0';
1912 vector<string> const & options = used_packages["jurabib"];
1913 if (std::find(options.begin(), options.end(),
1914 "natbiborder") != options.end())
1915 argumentOrder = 'n';
1916 else if (std::find(options.begin(), options.end(),
1917 "jurabiborder") != options.end())
1918 argumentOrder = 'j';
1920 // text before the citation
1922 // text after the citation
1924 get_cite_arguments(p, argumentOrder != 'j', before, after);
1926 string const citation = p.verbatim_item();
1927 if (!before.empty() && argumentOrder == '\0') {
1928 cerr << "Warning: Assuming argument order "
1929 "of jurabib version 0.6 for\n'"
1930 << command << before << after << '{'
1931 << citation << "}'.\n"
1932 "Add 'jurabiborder' to the jurabib "
1933 "package options if you used an\n"
1934 "earlier jurabib version." << endl;
1936 begin_inset(os, "LatexCommand ");
1937 os << command << after << before
1938 << '{' << citation << "}\n";
1942 else if (is_known(t.cs(), known_latex_commands)) {
1943 // This needs to be after the check for natbib and
1944 // jurabib commands, because "cite" has different
1945 // arguments with natbib and jurabib.
1946 context.check_layout(os);
1947 begin_inset(os, "LatexCommand ");
1948 os << '\\' << t.cs();
1949 // lyx cannot handle newlines in a latex command
1950 // FIXME: Move the substitution into parser::getOpt()?
1951 os << subst(p.getOpt(), "\n", " ");
1952 os << subst(p.getOpt(), "\n", " ");
1953 os << '{' << subst(p.verbatim_item(), "\n", " ") << "}\n";
1957 else if (is_known(t.cs(), known_quotes)) {
1958 char const * const * where = is_known(t.cs(), known_quotes);
1959 context.check_layout(os);
1960 begin_inset(os, "Quotes ");
1961 os << known_coded_quotes[where - known_quotes];
1963 // LyX adds {} after the quote, so we have to eat
1964 // spaces here if there are any before a possible
1966 eat_whitespace(p, os, context, false);
1970 else if (is_known(t.cs(), known_sizes) &&
1971 context.new_layout_allowed) {
1972 char const * const * where = is_known(t.cs(), known_sizes);
1973 context.check_layout(os);
1974 TeXFont const oldFont = context.font;
1975 context.font.size = known_coded_sizes[where - known_sizes];
1976 output_font_change(os, oldFont, context.font);
1977 eat_whitespace(p, os, context, false);
1980 else if (is_known(t.cs(), known_font_families) &&
1981 context.new_layout_allowed) {
1982 char const * const * where =
1983 is_known(t.cs(), known_font_families);
1984 context.check_layout(os);
1985 TeXFont const oldFont = context.font;
1986 context.font.family =
1987 known_coded_font_families[where - known_font_families];
1988 output_font_change(os, oldFont, context.font);
1989 eat_whitespace(p, os, context, false);
1992 else if (is_known(t.cs(), known_font_series) &&
1993 context.new_layout_allowed) {
1994 char const * const * where =
1995 is_known(t.cs(), known_font_series);
1996 context.check_layout(os);
1997 TeXFont const oldFont = context.font;
1998 context.font.series =
1999 known_coded_font_series[where - known_font_series];
2000 output_font_change(os, oldFont, context.font);
2001 eat_whitespace(p, os, context, false);
2004 else if (is_known(t.cs(), known_font_shapes) &&
2005 context.new_layout_allowed) {
2006 char const * const * where =
2007 is_known(t.cs(), known_font_shapes);
2008 context.check_layout(os);
2009 TeXFont const oldFont = context.font;
2010 context.font.shape =
2011 known_coded_font_shapes[where - known_font_shapes];
2012 output_font_change(os, oldFont, context.font);
2013 eat_whitespace(p, os, context, false);
2015 else if (is_known(t.cs(), known_old_font_families) &&
2016 context.new_layout_allowed) {
2017 char const * const * where =
2018 is_known(t.cs(), known_old_font_families);
2019 context.check_layout(os);
2020 TeXFont const oldFont = context.font;
2021 context.font.init();
2022 context.font.size = oldFont.size;
2023 context.font.family =
2024 known_coded_font_families[where - known_old_font_families];
2025 output_font_change(os, oldFont, context.font);
2026 eat_whitespace(p, os, context, false);
2029 else if (is_known(t.cs(), known_old_font_series) &&
2030 context.new_layout_allowed) {
2031 char const * const * where =
2032 is_known(t.cs(), known_old_font_series);
2033 context.check_layout(os);
2034 TeXFont const oldFont = context.font;
2035 context.font.init();
2036 context.font.size = oldFont.size;
2037 context.font.series =
2038 known_coded_font_series[where - known_old_font_series];
2039 output_font_change(os, oldFont, context.font);
2040 eat_whitespace(p, os, context, false);
2043 else if (is_known(t.cs(), known_old_font_shapes) &&
2044 context.new_layout_allowed) {
2045 char const * const * where =
2046 is_known(t.cs(), known_old_font_shapes);
2047 context.check_layout(os);
2048 TeXFont const oldFont = context.font;
2049 context.font.init();
2050 context.font.size = oldFont.size;
2051 context.font.shape =
2052 known_coded_font_shapes[where - known_old_font_shapes];
2053 output_font_change(os, oldFont, context.font);
2054 eat_whitespace(p, os, context, false);
2057 else if (t.cs() == "selectlanguage") {
2058 context.check_layout(os);
2059 // save the language for the case that a \foreignlanguage is used
2060 selectlang = subst(p.verbatim_item(), "\n", " ");
2061 os << "\\lang " << selectlang << "\n";
2065 else if (t.cs() == "foreignlanguage") {
2066 context.check_layout(os);
2067 os << "\n\\lang " << subst(p.verbatim_item(), "\n", " ") << "\n";
2068 os << subst(p.verbatim_item(), "\n", " ");
2069 // set back to last selectlanguage
2070 os << "\n\\lang " << selectlang << "\n";
2073 else if (t.cs() == "inputencoding")
2074 // write nothing because this is done by LyX using the "\lang"
2075 // information given by selectlanguage and foreignlanguage
2076 subst(p.verbatim_item(), "\n", " ");
2078 else if (t.cs() == "LyX" || t.cs() == "TeX"
2079 || t.cs() == "LaTeX") {
2080 context.check_layout(os);
2082 skip_braces(p); // eat {}
2085 else if (t.cs() == "LaTeXe") {
2086 context.check_layout(os);
2088 skip_braces(p); // eat {}
2091 else if (t.cs() == "ldots") {
2092 context.check_layout(os);
2094 os << "\\SpecialChar \\ldots{}\n";
2097 else if (t.cs() == "lyxarrow") {
2098 context.check_layout(os);
2099 os << "\\SpecialChar \\menuseparator\n";
2103 else if (t.cs() == "textcompwordmark") {
2104 context.check_layout(os);
2105 os << "\\SpecialChar \\textcompwordmark{}\n";
2109 else if (t.cs() == "@" && p.next_token().asInput() == ".") {
2110 context.check_layout(os);
2111 os << "\\SpecialChar \\@.\n";
2115 else if (t.cs() == "-") {
2116 context.check_layout(os);
2117 os << "\\SpecialChar \\-\n";
2120 else if (t.cs() == "textasciitilde") {
2121 context.check_layout(os);
2126 else if (t.cs() == "textasciicircum") {
2127 context.check_layout(os);
2132 else if (t.cs() == "textbackslash") {
2133 context.check_layout(os);
2134 os << "\n\\backslash\n";
2138 else if (t.cs() == "_" || t.cs() == "&" || t.cs() == "#"
2139 || t.cs() == "$" || t.cs() == "{" || t.cs() == "}"
2141 context.check_layout(os);
2145 else if (t.cs() == "char") {
2146 context.check_layout(os);
2147 if (p.next_token().character() == '`') {
2149 if (p.next_token().cs() == "\"") {
2154 handle_ert(os, "\\char`", context);
2157 handle_ert(os, "\\char", context);
2161 else if (t.cs() == "verb") {
2162 context.check_layout(os);
2163 char const delimiter = p.next_token().character();
2164 string const arg = p.getArg(delimiter, delimiter);
2166 oss << "\\verb" << delimiter << arg << delimiter;
2167 handle_ert(os, oss.str(), context);
2170 else if (t.cs() == "\"") {
2171 context.check_layout(os);
2172 string const name = p.verbatim_item();
2173 if (name == "a") os << '\xe4';
2174 else if (name == "o") os << '\xf6';
2175 else if (name == "u") os << '\xfc';
2176 else if (name == "A") os << '\xc4';
2177 else if (name == "O") os << '\xd6';
2178 else if (name == "U") os << '\xdc';
2179 else handle_ert(os, "\"{" + name + "}", context);
2182 // Problem: \= creates a tabstop inside the tabbing environment
2183 // and else an accent. In the latter case we really would want
2184 // \={o} instead of \= o.
2185 else if (t.cs() == "=" && (flags & FLAG_TABBING))
2186 handle_ert(os, t.asInput(), context);
2188 else if (t.cs() == "H" || t.cs() == "c" || t.cs() == "^"
2189 || t.cs() == "'" || t.cs() == "`"
2190 || t.cs() == "~" || t.cs() == "." || t.cs() == "=") {
2191 // we need the trim as the LyX parser chokes on such spaces
2192 // The argument of InsetLatexAccent is parsed as a
2193 // subset of LaTeX, so don't parse anything here,
2194 // but use the raw argument.
2195 // Otherwise we would convert \~{\i} wrongly.
2196 // This will of course not translate \~{\ss} to \~{ß},
2197 // but that does at least compile and does only look
2198 // strange on screen.
2199 context.check_layout(os);
2200 os << "\\i \\" << t.cs() << "{"
2201 << trim(p.verbatim_item(), " ")
2205 else if (t.cs() == "ss") {
2206 context.check_layout(os);
2208 skip_braces(p); // eat {}
2211 else if (t.cs() == "i" || t.cs() == "j" || t.cs() == "l" ||
2213 context.check_layout(os);
2214 os << "\\i \\" << t.cs() << "{}\n";
2215 skip_braces(p); // eat {}
2218 else if (t.cs() == "\\") {
2219 context.check_layout(os);
2220 string const next = p.next_token().asInput();
2222 handle_ert(os, "\\\\" + p.getOpt(), context);
2223 else if (next == "*") {
2225 handle_ert(os, "\\\\*" + p.getOpt(), context);
2228 os << "\n\\newline\n";
2232 else if (t.cs() == "newline" ||
2233 t.cs() == "linebreak") {
2234 context.check_layout(os);
2235 os << "\n\\" << t.cs() << "\n";
2236 skip_braces(p); // eat {}
2239 else if (t.cs() == "href") {
2240 context.check_layout(os);
2241 begin_inset(os, "CommandInset ");
2242 os << t.cs() << "\n";
2243 os << "LatexCommand " << t.cs() << "\n";
2246 // the first argument is "type:target", "type:" is optional
2247 // the second argument the name
2248 string href_target = subst(p.verbatim_item(), "\n", " ");
2249 string href_name = subst(p.verbatim_item(), "\n", " ");
2251 // serach for the ":" to divide type from target
2252 if ((pos = href_target.find(":", 0)) != string::npos){
2253 href_type = href_target;
2254 href_type.erase(pos + 1, href_type.length());
2255 href_target.erase(0, pos + 1);
2258 os << "name " << '"' << href_name << '"' << "\n";
2259 os << "target " << '"' << href_target << '"' << "\n";
2261 os << "type " << '"' << href_type << '"' << "\n";
2265 else if (t.cs() == "input" || t.cs() == "include"
2266 || t.cs() == "verbatiminput") {
2267 string name = '\\' + t.cs();
2268 if (t.cs() == "verbatiminput"
2269 && p.next_token().asInput() == "*")
2270 name += p.get_token().asInput();
2271 context.check_layout(os);
2272 begin_inset(os, "Include ");
2273 string filename(normalize_filename(p.getArg('{', '}')));
2274 string const path = getMasterFilePath();
2275 // We want to preserve relative / absolute filenames,
2276 // therefore path is only used for testing
2277 // FIXME UNICODE encoding of filename and path may be
2278 // wrong (makeAbsPath expects utf8)
2279 if ((t.cs() == "include" || t.cs() == "input") &&
2280 !makeAbsPath(filename, path).exists()) {
2281 // The file extension is probably missing.
2282 // Now try to find it out.
2283 string const tex_name =
2284 find_file(filename, path,
2285 known_tex_extensions);
2286 if (!tex_name.empty())
2287 filename = tex_name;
2289 // FIXME UNICODE encoding of filename and path may be
2290 // wrong (makeAbsPath expects utf8)
2291 if (makeAbsPath(filename, path).exists()) {
2292 string const abstexname =
2293 makeAbsPath(filename, path).absFilename();
2294 string const abslyxname =
2295 changeExtension(abstexname, ".lyx");
2296 fix_relative_filename(filename);
2297 string const lyxname =
2298 changeExtension(filename, ".lyx");
2299 if (t.cs() != "verbatiminput" &&
2300 tex2lyx(abstexname, FileName(abslyxname))) {
2301 os << name << '{' << lyxname << "}\n";
2303 os << name << '{' << filename << "}\n";
2306 cerr << "Warning: Could not find included file '"
2307 << filename << "'." << endl;
2308 os << name << '{' << filename << "}\n";
2310 os << "preview false\n";
2314 else if (t.cs() == "bibliographystyle") {
2315 // store new bibliographystyle
2316 bibliographystyle = p.verbatim_item();
2317 // output new bibliographystyle.
2318 // This is only necessary if used in some other macro than \bibliography.
2319 handle_ert(os, "\\bibliographystyle{" + bibliographystyle + "}", context);
2322 else if (t.cs() == "bibliography") {
2323 context.check_layout(os);
2324 begin_inset(os, "LatexCommand ");
2326 // Do we have a bibliographystyle set?
2327 if (!bibliographystyle.empty()) {
2328 os << '[' << bibliographystyle << ']';
2330 os << '{' << p.verbatim_item() << "}\n";
2334 else if (t.cs() == "parbox")
2335 parse_box(p, os, FLAG_ITEM, outer, context, true);
2337 else if (t.cs() == "smallskip" ||
2338 t.cs() == "medskip" ||
2339 t.cs() == "bigskip" ||
2340 t.cs() == "vfill") {
2341 context.check_layout(os);
2342 begin_inset(os, "VSpace ");
2348 else if (is_known(t.cs(), known_spaces)) {
2349 char const * const * where = is_known(t.cs(), known_spaces);
2350 context.check_layout(os);
2351 begin_inset(os, "InsetSpace ");
2352 os << '\\' << known_coded_spaces[where - known_spaces]
2354 // LaTeX swallows whitespace after all spaces except
2355 // "\\,". We have to do that here, too, because LyX
2356 // adds "{}" which would make the spaces significant.
2358 eat_whitespace(p, os, context, false);
2359 // LyX adds "{}" after all spaces except "\\ " and
2360 // "\\,", so we have to remove "{}".
2361 // "\\,{}" is equivalent to "\\," in LaTeX, so we
2362 // remove the braces after "\\,", too.
2367 else if (t.cs() == "newpage" ||
2368 t.cs() == "pagebreak" ||
2369 t.cs() == "clearpage" ||
2370 t.cs() == "cleardoublepage") {
2371 context.check_layout(os);
2372 os << "\n\\" << t.cs() << "\n";
2373 skip_braces(p); // eat {}
2376 else if (t.cs() == "newcommand" ||
2377 t.cs() == "providecommand" ||
2378 t.cs() == "renewcommand") {
2379 // these could be handled by parse_command(), but
2380 // we need to call add_known_command() here.
2381 string name = t.asInput();
2382 if (p.next_token().asInput() == "*") {
2383 // Starred form. Eat '*'
2387 string const command = p.verbatim_item();
2388 string const opt1 = p.getOpt();
2389 string const opt2 = p.getFullOpt();
2390 add_known_command(command, opt1, !opt2.empty());
2391 string const ert = name + '{' + command + '}' +
2393 '{' + p.verbatim_item() + '}';
2395 context.check_layout(os);
2396 begin_inset(os, "FormulaMacro");
2401 else if (t.cs() == "vspace") {
2402 bool starred = false;
2403 if (p.next_token().asInput() == "*") {
2407 string const length = p.verbatim_item();
2410 bool valid = splitLatexLength(length, valstring, unit);
2411 bool known_vspace = false;
2412 bool known_unit = false;
2415 istringstream iss(valstring);
2418 if (unit == "\\smallskipamount") {
2420 known_vspace = true;
2421 } else if (unit == "\\medskipamount") {
2423 known_vspace = true;
2424 } else if (unit == "\\bigskipamount") {
2426 known_vspace = true;
2427 } else if (unit == "\\fill") {
2429 known_vspace = true;
2432 if (!known_vspace) {
2433 switch (unitFromString(unit)) {
2454 if (known_unit || known_vspace) {
2455 // Literal length or known variable
2456 context.check_layout(os);
2457 begin_inset(os, "VSpace ");
2465 // LyX can't handle other length variables in Inset VSpace
2466 string name = t.asInput();
2471 handle_ert(os, name + '{' + unit + '}', context);
2472 else if (value == -1.0)
2473 handle_ert(os, name + "{-" + unit + '}', context);
2475 handle_ert(os, name + '{' + valstring + unit + '}', context);
2477 handle_ert(os, name + '{' + length + '}', context);
2482 //cerr << "#: " << t << " mode: " << mode << endl;
2483 // heuristic: read up to next non-nested space
2485 string s = t.asInput();
2486 string z = p.verbatim_item();
2487 while (p.good() && z != " " && z.size()) {
2488 //cerr << "read: " << z << endl;
2490 z = p.verbatim_item();
2492 cerr << "found ERT: " << s << endl;
2493 handle_ert(os, s + ' ', context);
2495 string name = t.asInput();
2496 if (p.next_token().asInput() == "*") {
2497 // Starred commands like \vspace*{}
2498 p.get_token(); // Eat '*'
2501 if (! parse_command(name, p, os, outer, context))
2502 handle_ert(os, name, context);
2505 if (flags & FLAG_LEAVE) {
2506 flags &= ~FLAG_LEAVE;