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
9 * Full author contact details are available in file CREDITS.
18 #include "FloatList.h"
19 #include "lengthcommon.h"
20 #include "support/lstrings.h"
21 #include "support/convert.h"
22 #include "support/filetools.h"
24 #include <boost/filesystem/operations.hpp>
25 #include <boost/tuple/tuple.hpp>
37 using std::ostringstream;
38 using std::istringstream;
44 using support::addExtension;
45 using support::changeExtension;
46 using support::FileName;
47 using support::makeAbsPath;
48 using support::makeRelPath;
50 using support::suffixIs;
51 using support::contains;
54 namespace fs = boost::filesystem;
57 void parse_text_in_inset(Parser & p, ostream & os, unsigned flags, bool outer,
58 Context const & context)
60 Context newcontext(true, context.textclass);
61 newcontext.font = context.font;
62 parse_text(p, os, flags, outer, newcontext);
63 newcontext.check_end_layout(os);
69 /// parses a paragraph snippet, useful for example for \\emph{...}
70 void parse_text_snippet(Parser & p, ostream & os, unsigned flags, bool outer,
73 Context newcontext(context);
74 // Don't inherit the extra stuff
75 newcontext.extra_stuff.clear();
76 parse_text(p, os, flags, outer, newcontext);
77 // Make sure that we don't create invalid .lyx files
78 context.need_layout = newcontext.need_layout;
79 context.need_end_layout = newcontext.need_end_layout;
84 * Thin wrapper around parse_text_snippet() using a string.
86 * We completely ignore \c context.need_layout and \c context.need_end_layout,
87 * because our return value is not used directly (otherwise the stream version
88 * of parse_text_snippet() could be used). That means that the caller needs
89 * to do layout management manually.
90 * This is intended to parse text that does not create any layout changes.
92 string parse_text_snippet(Parser & p, unsigned flags, const bool outer,
95 Context newcontext(context);
96 newcontext.need_layout = false;
97 newcontext.need_end_layout = false;
98 newcontext.new_layout_allowed = false;
99 // Avoid warning by Context::~Context()
100 newcontext.extra_stuff.clear();
102 parse_text_snippet(p, os, flags, outer, newcontext);
107 char const * const known_latex_commands[] = { "ref", "cite", "label", "index",
108 "printindex", "pageref", "url", "vref", "vpageref", "prettyref", "eqref", 0 };
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[] = { "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 (fs::exists(makeAbsPath(trial, path).toFilesystemEncoding()))
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<Layout_ptr, bool> {
438 isLayout(string const name) : name_(name) {}
439 bool operator()(Layout_ptr const & ptr) const {
440 return ptr->latexname() == name_;
447 Layout_ptr 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) ? Layout_ptr() : *it;
460 void eat_whitespace(Parser &, ostream &, Context &, bool);
463 void output_command_layout(ostream & os, Parser & p, bool outer,
464 Context & parent_context,
465 Layout_ptr newlayout)
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)
721 Layout_ptr newlayout;
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 (!parent_context.new_layout_allowed)
790 parse_unknown_environment(p, name, os, FLAG_END, outer,
793 // Alignment settings
794 else if (name == "center" || name == "flushleft" || name == "flushright" ||
795 name == "centering" || name == "raggedright" || name == "raggedleft") {
796 eat_whitespace(p, os, parent_context, false);
797 // We must begin a new paragraph if not already done
798 if (! parent_context.atParagraphStart()) {
799 parent_context.check_end_layout(os);
800 parent_context.new_paragraph(os);
802 if (name == "flushleft" || name == "raggedright")
803 parent_context.add_extra_stuff("\\align left\n");
804 else if (name == "flushright" || name == "raggedleft")
805 parent_context.add_extra_stuff("\\align right\n");
807 parent_context.add_extra_stuff("\\align center\n");
808 parse_text(p, os, FLAG_END, outer, parent_context);
809 // Just in case the environment is empty ..
810 parent_context.extra_stuff.erase();
811 // We must begin a new paragraph to reset the alignment
812 parent_context.new_paragraph(os);
816 // The single '=' is meant here.
817 else if ((newlayout = findLayout(parent_context.textclass, name)).get() &&
818 newlayout->isEnvironment()) {
819 eat_whitespace(p, os, parent_context, false);
820 Context context(true, parent_context.textclass, newlayout,
821 parent_context.layout, parent_context.font);
822 if (parent_context.deeper_paragraph) {
823 // We are beginning a nested environment after a
824 // deeper paragraph inside the outer list environment.
825 // Therefore we don't need to output a "begin deeper".
826 context.need_end_deeper = true;
828 parent_context.check_end_layout(os);
829 switch (context.layout->latextype) {
830 case LATEX_LIST_ENVIRONMENT:
831 context.extra_stuff = "\\labelwidthstring "
832 + p.verbatim_item() + '\n';
835 case LATEX_BIB_ENVIRONMENT:
836 p.verbatim_item(); // swallow next arg
842 context.check_deeper(os);
843 parse_text(p, os, FLAG_END, outer, context);
844 context.check_end_layout(os);
845 if (parent_context.deeper_paragraph) {
846 // We must suppress the "end deeper" because we
847 // suppressed the "begin deeper" above.
848 context.need_end_deeper = false;
850 context.check_end_deeper(os);
851 parent_context.new_paragraph(os);
855 else if (name == "appendix") {
856 // This is no good latex style, but it works and is used in some documents...
857 eat_whitespace(p, os, parent_context, false);
858 parent_context.check_end_layout(os);
859 Context context(true, parent_context.textclass, parent_context.layout,
860 parent_context.layout, parent_context.font);
861 context.check_layout(os);
862 os << "\\start_of_appendix\n";
863 parse_text(p, os, FLAG_END, outer, context);
864 context.check_end_layout(os);
868 else if (known_environments.find(name) != known_environments.end()) {
869 vector<ArgumentType> arguments = known_environments[name];
870 // The last "argument" denotes wether we may translate the
871 // environment contents to LyX
872 // The default required if no argument is given makes us
873 // compatible with the reLyXre environment.
874 ArgumentType contents = arguments.empty() ?
877 if (!arguments.empty())
878 arguments.pop_back();
879 // See comment in parse_unknown_environment()
880 bool const specialfont =
881 (parent_context.font != parent_context.normalfont);
882 bool const new_layout_allowed =
883 parent_context.new_layout_allowed;
885 parent_context.new_layout_allowed = false;
886 parse_arguments("\\begin{" + name + "}", arguments, p, os,
887 outer, parent_context);
888 if (contents == verbatim)
889 handle_ert(os, p.verbatimEnvironment(name),
892 parse_text_snippet(p, os, FLAG_END, outer,
894 handle_ert(os, "\\end{" + name + "}", parent_context);
896 parent_context.new_layout_allowed = new_layout_allowed;
900 parse_unknown_environment(p, name, os, FLAG_END, outer,
903 active_environments.pop_back();
907 /// parses a comment and outputs it to \p os.
908 void parse_comment(Parser & p, ostream & os, Token const & t, Context & context)
910 BOOST_ASSERT(t.cat() == catComment);
911 if (!t.cs().empty()) {
912 context.check_layout(os);
913 handle_comment(os, '%' + t.cs(), context);
914 if (p.next_token().cat() == catNewline) {
915 // A newline after a comment line starts a new
917 if (context.new_layout_allowed) {
918 if(!context.atParagraphStart())
919 // Only start a new paragraph if not already
920 // done (we might get called recursively)
921 context.new_paragraph(os);
923 handle_ert(os, "\n", context);
924 eat_whitespace(p, os, context, true);
934 * Reads spaces and comments until the first non-space, non-comment token.
935 * New paragraphs (double newlines or \\par) are handled like simple spaces
936 * if \p eatParagraph is true.
937 * Spaces are skipped, but comments are written to \p os.
939 void eat_whitespace(Parser & p, ostream & os, Context & context,
943 Token const & t = p.get_token();
944 if (t.cat() == catComment)
945 parse_comment(p, os, t, context);
946 else if ((! eatParagraph && p.isParagraph()) ||
947 (t.cat() != catSpace && t.cat() != catNewline)) {
956 * Set a font attribute, parse text and reset the font attribute.
957 * \param attribute Attribute name (e.g. \\family, \\shape etc.)
958 * \param currentvalue Current value of the attribute. Is set to the new
959 * value during parsing.
960 * \param newvalue New value of the attribute
962 void parse_text_attributes(Parser & p, ostream & os, unsigned flags, bool outer,
963 Context & context, string const & attribute,
964 string & currentvalue, string const & newvalue)
966 context.check_layout(os);
967 string const oldvalue = currentvalue;
968 currentvalue = newvalue;
969 os << '\n' << attribute << ' ' << newvalue << "\n";
970 parse_text_snippet(p, os, flags, outer, context);
971 context.check_layout(os);
972 os << '\n' << attribute << ' ' << oldvalue << "\n";
973 currentvalue = oldvalue;
977 /// get the arguments of a natbib or jurabib citation command
978 std::pair<string, string> getCiteArguments(Parser & p, bool natbibOrder)
980 // We need to distinguish "" and "[]", so we can't use p.getOpt().
982 // text before the citation
984 // text after the citation
985 string after = p.getFullOpt();
987 if (!after.empty()) {
988 before = p.getFullOpt();
989 if (natbibOrder && !before.empty())
990 std::swap(before, after);
992 return std::make_pair(before, after);
996 /// Convert filenames with TeX macros and/or quotes to something LyX can understand
997 string const normalize_filename(string const & name)
999 Parser p(trim(name, "\""));
1002 Token const & t = p.get_token();
1003 if (t.cat() != catEscape)
1005 else if (t.cs() == "lyxdot") {
1006 // This is used by LyX for simple dots in relative
1010 } else if (t.cs() == "space") {
1020 /// Convert \p name from TeX convention (relative to master file) to LyX
1021 /// convention (relative to .lyx file) if it is relative
1022 void fix_relative_filename(string & name)
1024 if (lyx::support::absolutePath(name))
1026 // FIXME UNICODE encoding of name may be wrong (makeAbsPath expects
1028 name = to_utf8(makeRelPath(from_utf8(makeAbsPath(name, getMasterFilePath()).absFilename()),
1029 from_utf8(getParentFilePath())));
1033 /// Parse a NoWeb Scrap section. The initial "<<" is already parsed.
1034 void parse_noweb(Parser & p, ostream & os, Context & context)
1036 // assemble the rest of the keyword
1040 Token const & t = p.get_token();
1041 if (t.asInput() == ">" && p.next_token().asInput() == ">") {
1044 scrap = (p.good() && p.next_token().asInput() == "=");
1046 name += p.get_token().asInput();
1049 name += t.asInput();
1052 if (!scrap || !context.new_layout_allowed ||
1053 !context.textclass.hasLayout(from_ascii("Scrap"))) {
1054 cerr << "Warning: Could not interpret '" << name
1055 << "'. Ignoring it." << endl;
1059 // We use new_paragraph instead of check_end_layout because the stuff
1060 // following the noweb chunk needs to start with a \begin_layout.
1061 // This may create a new paragraph even if there was none in the
1062 // noweb file, but the alternative is an invalid LyX file. Since
1063 // noweb code chunks are implemented with a layout style in LyX they
1064 // always must be in an own paragraph.
1065 context.new_paragraph(os);
1066 Context newcontext(true, context.textclass,
1067 context.textclass[from_ascii("Scrap")]);
1068 newcontext.check_layout(os);
1071 Token const & t = p.get_token();
1072 // We abuse the parser a bit, because this is no TeX syntax
1074 if (t.cat() == catEscape)
1075 os << subst(t.asInput(), "\\", "\n\\backslash\n");
1077 os << subst(t.asInput(), "\n", "\n\\newline\n");
1078 // The scrap chunk is ended by an @ at the beginning of a line.
1079 // After the @ the line may contain a comment and/or
1080 // whitespace, but nothing else.
1081 if (t.asInput() == "@" && p.prev_token().cat() == catNewline &&
1082 (p.next_token().cat() == catSpace ||
1083 p.next_token().cat() == catNewline ||
1084 p.next_token().cat() == catComment)) {
1085 while (p.good() && p.next_token().cat() == catSpace)
1086 os << p.get_token().asInput();
1087 if (p.next_token().cat() == catComment)
1088 // The comment includes a final '\n'
1089 os << p.get_token().asInput();
1091 if (p.next_token().cat() == catNewline)
1098 newcontext.check_end_layout(os);
1101 } // anonymous namespace
1104 void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
1107 Layout_ptr newlayout;
1108 // Store the latest bibliographystyle (needed for bibtex inset)
1109 string bibliographystyle;
1110 bool const use_natbib = used_packages.find("natbib") != used_packages.end();
1111 bool const use_jurabib = used_packages.find("jurabib") != used_packages.end();
1113 Token const & t = p.get_token();
1116 cerr << "t: " << t << " flags: " << flags << "\n";
1119 if (flags & FLAG_ITEM) {
1120 if (t.cat() == catSpace)
1123 flags &= ~FLAG_ITEM;
1124 if (t.cat() == catBegin) {
1125 // skip the brace and collect everything to the next matching
1127 flags |= FLAG_BRACE_LAST;
1131 // handle only this single token, leave the loop if done
1132 flags |= FLAG_LEAVE;
1135 if (t.character() == ']' && (flags & FLAG_BRACK_LAST))
1141 if (t.cat() == catMath) {
1142 // we are inside some text mode thingy, so opening new math is allowed
1143 context.check_layout(os);
1144 begin_inset(os, "Formula ");
1145 Token const & n = p.get_token();
1146 if (n.cat() == catMath && outer) {
1147 // TeX's $$...$$ syntax for displayed math
1149 parse_math(p, os, FLAG_SIMPLE, MATH_MODE);
1151 p.get_token(); // skip the second '$' token
1153 // simple $...$ stuff
1156 parse_math(p, os, FLAG_SIMPLE, MATH_MODE);
1162 else if (t.cat() == catSuper || t.cat() == catSub)
1163 cerr << "catcode " << t << " illegal in text mode\n";
1165 // Basic support for english quotes. This should be
1166 // extended to other quotes, but is not so easy (a
1167 // left english quote is the same as a right german
1169 else if (t.asInput() == "`"
1170 && p.next_token().asInput() == "`") {
1171 context.check_layout(os);
1172 begin_inset(os, "Quotes ");
1178 else if (t.asInput() == "'"
1179 && p.next_token().asInput() == "'") {
1180 context.check_layout(os);
1181 begin_inset(os, "Quotes ");
1188 else if (t.asInput() == "<"
1189 && p.next_token().asInput() == "<" && noweb_mode) {
1191 parse_noweb(p, os, context);
1194 else if (t.cat() == catSpace || (t.cat() == catNewline && ! p.isParagraph()))
1195 check_space(p, os, context);
1197 else if (t.character() == '[' && noweb_mode &&
1198 p.next_token().character() == '[') {
1199 // These can contain underscores
1201 string const s = p.getFullOpt() + ']';
1202 if (p.next_token().character() == ']')
1205 cerr << "Warning: Inserting missing ']' in '"
1206 << s << "'." << endl;
1207 handle_ert(os, s, context);
1210 else if (t.cat() == catLetter ||
1211 t.cat() == catOther ||
1212 t.cat() == catAlign ||
1213 t.cat() == catParameter) {
1214 // This translates "&" to "\\&" which may be wrong...
1215 context.check_layout(os);
1216 os << t.character();
1219 else if (p.isParagraph()) {
1220 if (context.new_layout_allowed)
1221 context.new_paragraph(os);
1223 handle_ert(os, "\\par ", context);
1224 eat_whitespace(p, os, context, true);
1227 else if (t.cat() == catActive) {
1228 context.check_layout(os);
1229 if (t.character() == '~') {
1230 if (context.layout->free_spacing)
1233 os << "\\InsetSpace ~\n";
1235 os << t.character();
1238 else if (t.cat() == catBegin &&
1239 p.next_token().cat() == catEnd) {
1241 Token const prev = p.prev_token();
1243 if (p.next_token().character() == '`' ||
1244 (prev.character() == '-' &&
1245 p.next_token().character() == '-'))
1246 ; // ignore it in {}`` or -{}-
1248 handle_ert(os, "{}", context);
1252 else if (t.cat() == catBegin) {
1253 context.check_layout(os);
1254 // special handling of font attribute changes
1255 Token const prev = p.prev_token();
1256 Token const next = p.next_token();
1257 TeXFont const oldFont = context.font;
1258 if (next.character() == '[' ||
1259 next.character() == ']' ||
1260 next.character() == '*') {
1262 if (p.next_token().cat() == catEnd) {
1263 os << next.character();
1267 handle_ert(os, "{", context);
1268 parse_text_snippet(p, os,
1271 handle_ert(os, "}", context);
1273 } else if (! context.new_layout_allowed) {
1274 handle_ert(os, "{", context);
1275 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1277 handle_ert(os, "}", context);
1278 } else if (is_known(next.cs(), known_sizes)) {
1279 // next will change the size, so we must
1281 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1283 if (!context.atParagraphStart())
1285 << context.font.size << "\n";
1286 } else if (is_known(next.cs(), known_font_families)) {
1287 // next will change the font family, so we
1288 // must reset it here
1289 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1291 if (!context.atParagraphStart())
1293 << context.font.family << "\n";
1294 } else if (is_known(next.cs(), known_font_series)) {
1295 // next will change the font series, so we
1296 // must reset it here
1297 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1299 if (!context.atParagraphStart())
1301 << context.font.series << "\n";
1302 } else if (is_known(next.cs(), known_font_shapes)) {
1303 // next will change the font shape, so we
1304 // must reset it here
1305 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1307 if (!context.atParagraphStart())
1309 << context.font.shape << "\n";
1310 } else if (is_known(next.cs(), known_old_font_families) ||
1311 is_known(next.cs(), known_old_font_series) ||
1312 is_known(next.cs(), known_old_font_shapes)) {
1313 // next will change the font family, series
1314 // and shape, so we must reset it here
1315 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1317 if (!context.atParagraphStart())
1319 << context.font.family
1321 << context.font.series
1323 << context.font.shape << "\n";
1325 handle_ert(os, "{", context);
1326 parse_text_snippet(p, os, FLAG_BRACE_LAST,
1328 handle_ert(os, "}", context);
1332 else if (t.cat() == catEnd) {
1333 if (flags & FLAG_BRACE_LAST) {
1336 cerr << "stray '}' in text\n";
1337 handle_ert(os, "}", context);
1340 else if (t.cat() == catComment)
1341 parse_comment(p, os, t, context);
1344 // control sequences
1347 else if (t.cs() == "(") {
1348 context.check_layout(os);
1349 begin_inset(os, "Formula");
1351 parse_math(p, os, FLAG_SIMPLE2, MATH_MODE);
1356 else if (t.cs() == "[") {
1357 context.check_layout(os);
1358 begin_inset(os, "Formula");
1360 parse_math(p, os, FLAG_EQUATION, MATH_MODE);
1365 else if (t.cs() == "begin")
1366 parse_environment(p, os, outer, context);
1368 else if (t.cs() == "end") {
1369 if (flags & FLAG_END) {
1370 // eat environment name
1371 string const name = p.getArg('{', '}');
1372 if (name != active_environment())
1373 cerr << "\\end{" + name + "} does not match \\begin{"
1374 + active_environment() + "}\n";
1377 p.error("found 'end' unexpectedly");
1380 else if (t.cs() == "item") {
1383 bool optarg = false;
1384 if (p.next_token().character() == '[') {
1385 p.get_token(); // eat '['
1386 s = parse_text_snippet(p, FLAG_BRACK_LAST,
1391 context.check_layout(os);
1392 if (context.has_item) {
1393 // An item in an unknown list-like environment
1394 // FIXME: Do this in check_layout()!
1395 context.has_item = false;
1397 handle_ert(os, "\\item", context);
1399 handle_ert(os, "\\item ", context);
1402 if (context.layout->labeltype != LABEL_MANUAL) {
1403 // lyx does not support \item[\mybullet]
1404 // in itemize environments
1405 handle_ert(os, "[", context);
1407 handle_ert(os, "]", context);
1408 } else if (!s.empty()) {
1409 // The space is needed to separate the
1410 // item from the rest of the sentence.
1412 eat_whitespace(p, os, context, false);
1417 else if (t.cs() == "bibitem") {
1419 context.check_layout(os);
1422 os << '{' << p.verbatim_item() << '}' << "\n";
1425 else if (t.cs() == "def") {
1426 context.check_layout(os);
1427 eat_whitespace(p, os, context, false);
1428 string name = p.get_token().cs();
1429 eat_whitespace(p, os, context, false);
1435 while (p.next_token().cat() != catBegin) {
1436 if (p.next_token().cat() == catParameter) {
1441 // followed by number?
1442 if (p.next_token().cat() == catOther) {
1443 char c = p.getChar();
1445 // number = current arity + 1?
1446 if (c == arity + '0' + 1)
1451 paramtext += p.get_token().asString();
1453 paramtext += p.get_token().asString();
1458 // only output simple (i.e. compatible) macro as FormulaMacros
1459 string ert = "\\def\\" + name + ' ' + paramtext + '{' + p.verbatim_item() + '}';
1461 context.check_layout(os);
1462 begin_inset(os, "FormulaMacro");
1466 handle_ert(os, ert, context);
1469 else if (t.cs() == "noindent") {
1471 context.add_extra_stuff("\\noindent\n");
1474 else if (t.cs() == "appendix") {
1475 context.add_extra_stuff("\\start_of_appendix\n");
1476 // We need to start a new paragraph. Otherwise the
1477 // appendix in 'bla\appendix\chapter{' would start
1479 context.new_paragraph(os);
1480 // We need to make sure that the paragraph is
1481 // generated even if it is empty. Otherwise the
1482 // appendix in '\par\appendix\par\chapter{' would
1484 context.check_layout(os);
1485 // FIXME: This is a hack to prevent paragraph
1486 // deletion if it is empty. Handle this better!
1488 "%dummy comment inserted by tex2lyx to "
1489 "ensure that this paragraph is not empty",
1491 // Both measures above may generate an additional
1492 // empty paragraph, but that does not hurt, because
1493 // whitespace does not matter here.
1494 eat_whitespace(p, os, context, true);
1497 // Must attempt to parse "Section*" before "Section".
1498 else if ((p.next_token().asInput() == "*") &&
1499 context.new_layout_allowed &&
1500 // The single '=' is meant here.
1501 (newlayout = findLayout(context.textclass,
1502 t.cs() + '*')).get() &&
1503 newlayout->isCommand()) {
1505 output_command_layout(os, p, outer, context, newlayout);
1509 // The single '=' is meant here.
1510 else if (context.new_layout_allowed &&
1511 (newlayout = findLayout(context.textclass, t.cs())).get() &&
1512 newlayout->isCommand()) {
1513 output_command_layout(os, p, outer, context, newlayout);
1517 // Special handling for \caption
1518 // FIXME: remove this when InsetCaption is supported.
1519 else if (context.new_layout_allowed &&
1520 t.cs() == captionlayout->latexname()) {
1521 output_command_layout(os, p, outer, context,
1526 else if (t.cs() == "includegraphics") {
1527 bool const clip = p.next_token().asInput() == "*";
1530 map<string, string> opts = split_map(p.getArg('[', ']'));
1532 opts["clip"] = string();
1533 string name = normalize_filename(p.verbatim_item());
1535 string const path = getMasterFilePath();
1536 // We want to preserve relative / absolute filenames,
1537 // therefore path is only used for testing
1538 // FIXME UNICODE encoding of name and path may be
1539 // wrong (makeAbsPath expects utf8)
1540 if (!fs::exists(makeAbsPath(name, path).toFilesystemEncoding())) {
1541 // The file extension is probably missing.
1542 // Now try to find it out.
1543 string const dvips_name =
1544 find_file(name, path,
1545 known_dvips_graphics_formats);
1546 string const pdftex_name =
1547 find_file(name, path,
1548 known_pdftex_graphics_formats);
1549 if (!dvips_name.empty()) {
1550 if (!pdftex_name.empty()) {
1551 cerr << "This file contains the "
1553 "\"\\includegraphics{"
1555 "However, files\n\""
1556 << dvips_name << "\" and\n\""
1557 << pdftex_name << "\"\n"
1558 "both exist, so I had to make a "
1559 "choice and took the first one.\n"
1560 "Please move the unwanted one "
1561 "someplace else and try again\n"
1562 "if my choice was wrong."
1566 } else if (!pdftex_name.empty())
1570 // FIXME UNICODE encoding of name and path may be
1571 // wrong (makeAbsPath expects utf8)
1572 if (fs::exists(makeAbsPath(name, path).toFilesystemEncoding()))
1573 fix_relative_filename(name);
1575 cerr << "Warning: Could not find graphics file '"
1576 << name << "'." << endl;
1578 context.check_layout(os);
1579 begin_inset(os, "Graphics ");
1580 os << "\n\tfilename " << name << '\n';
1581 if (opts.find("width") != opts.end())
1583 << translate_len(opts["width"]) << '\n';
1584 if (opts.find("height") != opts.end())
1586 << translate_len(opts["height"]) << '\n';
1587 if (opts.find("scale") != opts.end()) {
1588 istringstream iss(opts["scale"]);
1592 os << "\tscale " << val << '\n';
1594 if (opts.find("angle") != opts.end())
1595 os << "\trotateAngle "
1596 << opts["angle"] << '\n';
1597 if (opts.find("origin") != opts.end()) {
1599 string const opt = opts["origin"];
1600 if (opt.find('l') != string::npos) ss << "left";
1601 if (opt.find('r') != string::npos) ss << "right";
1602 if (opt.find('c') != string::npos) ss << "center";
1603 if (opt.find('t') != string::npos) ss << "Top";
1604 if (opt.find('b') != string::npos) ss << "Bottom";
1605 if (opt.find('B') != string::npos) ss << "Baseline";
1606 if (!ss.str().empty())
1607 os << "\trotateOrigin " << ss.str() << '\n';
1609 cerr << "Warning: Ignoring unknown includegraphics origin argument '" << opt << "'\n";
1611 if (opts.find("keepaspectratio") != opts.end())
1612 os << "\tkeepAspectRatio\n";
1613 if (opts.find("clip") != opts.end())
1615 if (opts.find("draft") != opts.end())
1617 if (opts.find("bb") != opts.end())
1618 os << "\tBoundingBox "
1619 << opts["bb"] << '\n';
1620 int numberOfbbOptions = 0;
1621 if (opts.find("bbllx") != opts.end())
1622 numberOfbbOptions++;
1623 if (opts.find("bblly") != opts.end())
1624 numberOfbbOptions++;
1625 if (opts.find("bburx") != opts.end())
1626 numberOfbbOptions++;
1627 if (opts.find("bbury") != opts.end())
1628 numberOfbbOptions++;
1629 if (numberOfbbOptions == 4)
1630 os << "\tBoundingBox "
1631 << opts["bbllx"] << opts["bblly"]
1632 << opts["bburx"] << opts["bbury"] << '\n';
1633 else if (numberOfbbOptions > 0)
1634 cerr << "Warning: Ignoring incomplete includegraphics boundingbox arguments.\n";
1635 numberOfbbOptions = 0;
1636 if (opts.find("natwidth") != opts.end())
1637 numberOfbbOptions++;
1638 if (opts.find("natheight") != opts.end())
1639 numberOfbbOptions++;
1640 if (numberOfbbOptions == 2)
1641 os << "\tBoundingBox 0bp 0bp "
1642 << opts["natwidth"] << opts["natheight"] << '\n';
1643 else if (numberOfbbOptions > 0)
1644 cerr << "Warning: Ignoring incomplete includegraphics boundingbox arguments.\n";
1645 ostringstream special;
1646 if (opts.find("hiresbb") != opts.end())
1647 special << "hiresbb,";
1648 if (opts.find("trim") != opts.end())
1650 if (opts.find("viewport") != opts.end())
1651 special << "viewport=" << opts["viewport"] << ',';
1652 if (opts.find("totalheight") != opts.end())
1653 special << "totalheight=" << opts["totalheight"] << ',';
1654 if (opts.find("type") != opts.end())
1655 special << "type=" << opts["type"] << ',';
1656 if (opts.find("ext") != opts.end())
1657 special << "ext=" << opts["ext"] << ',';
1658 if (opts.find("read") != opts.end())
1659 special << "read=" << opts["read"] << ',';
1660 if (opts.find("command") != opts.end())
1661 special << "command=" << opts["command"] << ',';
1662 string s_special = special.str();
1663 if (!s_special.empty()) {
1664 // We had special arguments. Remove the trailing ','.
1665 os << "\tspecial " << s_special.substr(0, s_special.size() - 1) << '\n';
1667 // TODO: Handle the unknown settings better.
1668 // Warn about invalid options.
1669 // Check whether some option was given twice.
1673 else if (t.cs() == "footnote" ||
1674 (t.cs() == "thanks" && context.layout->intitle)) {
1676 context.check_layout(os);
1677 begin_inset(os, "Foot\n");
1678 os << "status collapsed\n\n";
1679 parse_text_in_inset(p, os, FLAG_ITEM, false, context);
1683 else if (t.cs() == "marginpar") {
1685 context.check_layout(os);
1686 begin_inset(os, "Marginal\n");
1687 os << "status collapsed\n\n";
1688 parse_text_in_inset(p, os, FLAG_ITEM, false, context);
1692 else if (t.cs() == "ensuremath") {
1694 context.check_layout(os);
1695 string const s = p.verbatim_item();
1696 if (s == "\xb1" || s == "\xb3" || s == "\xb2" || s == "\xb5")
1699 handle_ert(os, "\\ensuremath{" + s + "}",
1703 else if (t.cs() == "hfill") {
1704 context.check_layout(os);
1705 os << "\n\\hfill\n";
1710 else if (t.cs() == "makeindex" || t.cs() == "maketitle") {
1711 // FIXME: Somehow prevent title layouts if
1712 // "maketitle" was not found
1714 skip_braces(p); // swallow this
1717 else if (t.cs() == "tableofcontents") {
1719 context.check_layout(os);
1720 begin_inset(os, "LatexCommand \\tableofcontents\n");
1722 skip_braces(p); // swallow this
1725 else if (t.cs() == "listoffigures") {
1727 context.check_layout(os);
1728 begin_inset(os, "FloatList figure\n");
1730 skip_braces(p); // swallow this
1733 else if (t.cs() == "listoftables") {
1735 context.check_layout(os);
1736 begin_inset(os, "FloatList table\n");
1738 skip_braces(p); // swallow this
1741 else if (t.cs() == "listof") {
1742 p.skip_spaces(true);
1743 string const name = p.get_token().asString();
1744 if (context.textclass.floats().typeExist(name)) {
1745 context.check_layout(os);
1746 begin_inset(os, "FloatList ");
1749 p.get_token(); // swallow second arg
1751 handle_ert(os, "\\listof{" + name + "}", context);
1754 else if (t.cs() == "textrm")
1755 parse_text_attributes(p, os, FLAG_ITEM, outer,
1756 context, "\\family",
1757 context.font.family, "roman");
1759 else if (t.cs() == "textsf")
1760 parse_text_attributes(p, os, FLAG_ITEM, outer,
1761 context, "\\family",
1762 context.font.family, "sans");
1764 else if (t.cs() == "texttt")
1765 parse_text_attributes(p, os, FLAG_ITEM, outer,
1766 context, "\\family",
1767 context.font.family, "typewriter");
1769 else if (t.cs() == "textmd")
1770 parse_text_attributes(p, os, FLAG_ITEM, outer,
1771 context, "\\series",
1772 context.font.series, "medium");
1774 else if (t.cs() == "textbf")
1775 parse_text_attributes(p, os, FLAG_ITEM, outer,
1776 context, "\\series",
1777 context.font.series, "bold");
1779 else if (t.cs() == "textup")
1780 parse_text_attributes(p, os, FLAG_ITEM, outer,
1782 context.font.shape, "up");
1784 else if (t.cs() == "textit")
1785 parse_text_attributes(p, os, FLAG_ITEM, outer,
1787 context.font.shape, "italic");
1789 else if (t.cs() == "textsl")
1790 parse_text_attributes(p, os, FLAG_ITEM, outer,
1792 context.font.shape, "slanted");
1794 else if (t.cs() == "textsc")
1795 parse_text_attributes(p, os, FLAG_ITEM, outer,
1797 context.font.shape, "smallcaps");
1799 else if (t.cs() == "textnormal" || t.cs() == "normalfont") {
1800 context.check_layout(os);
1801 TeXFont oldFont = context.font;
1802 context.font.init();
1803 context.font.size = oldFont.size;
1804 os << "\n\\family " << context.font.family << "\n";
1805 os << "\n\\series " << context.font.series << "\n";
1806 os << "\n\\shape " << context.font.shape << "\n";
1807 if (t.cs() == "textnormal") {
1808 parse_text_snippet(p, os, FLAG_ITEM, outer, context);
1809 output_font_change(os, context.font, oldFont);
1810 context.font = oldFont;
1812 eat_whitespace(p, os, context, false);
1815 else if (t.cs() == "underbar") {
1816 // Do NOT handle \underline.
1817 // \underbar cuts through y, g, q, p etc.,
1818 // \underline does not.
1819 context.check_layout(os);
1820 os << "\n\\bar under\n";
1821 parse_text_snippet(p, os, FLAG_ITEM, outer, context);
1822 context.check_layout(os);
1823 os << "\n\\bar default\n";
1826 else if (t.cs() == "emph" || t.cs() == "noun") {
1827 context.check_layout(os);
1828 os << "\n\\" << t.cs() << " on\n";
1829 parse_text_snippet(p, os, FLAG_ITEM, outer, context);
1830 context.check_layout(os);
1831 os << "\n\\" << t.cs() << " default\n";
1834 else if (use_natbib &&
1835 is_known(t.cs(), known_natbib_commands) &&
1836 ((t.cs() != "citefullauthor" &&
1837 t.cs() != "citeyear" &&
1838 t.cs() != "citeyearpar") ||
1839 p.next_token().asInput() != "*")) {
1840 context.check_layout(os);
1842 // \citet[before][after]{a} \citet[after][before]{a}
1843 // \citet[before][]{a} \citet[][before]{a}
1844 // \citet[after]{a} \citet[after]{a}
1845 // \citet{a} \citet{a}
1846 string command = '\\' + t.cs();
1847 if (p.next_token().asInput() == "*") {
1851 if (command == "\\citefullauthor")
1852 // alternative name for "\\citeauthor*"
1853 command = "\\citeauthor*";
1855 // text before the citation
1857 // text after the citation
1860 boost::tie(before, after) = getCiteArguments(p, true);
1861 if (command == "\\cite") {
1862 // \cite without optional argument means
1863 // \citet, \cite with at least one optional
1864 // argument means \citep.
1865 if (before.empty() && after.empty())
1866 command = "\\citet";
1868 command = "\\citep";
1870 if (before.empty() && after == "[]")
1871 // avoid \citet[]{a}
1873 else if (before == "[]" && after == "[]") {
1874 // avoid \citet[][]{a}
1878 begin_inset(os, "LatexCommand ");
1879 os << command << after << before
1880 << '{' << p.verbatim_item() << "}\n";
1884 else if (use_jurabib &&
1885 is_known(t.cs(), known_jurabib_commands)) {
1886 context.check_layout(os);
1887 string const command = '\\' + t.cs();
1888 char argumentOrder = '\0';
1889 vector<string> const & options = used_packages["jurabib"];
1890 if (std::find(options.begin(), options.end(),
1891 "natbiborder") != options.end())
1892 argumentOrder = 'n';
1893 else if (std::find(options.begin(), options.end(),
1894 "jurabiborder") != options.end())
1895 argumentOrder = 'j';
1897 // text before the citation
1899 // text after the citation
1902 boost::tie(before, after) =
1903 getCiteArguments(p, argumentOrder != 'j');
1904 string const citation = p.verbatim_item();
1905 if (!before.empty() && argumentOrder == '\0') {
1906 cerr << "Warning: Assuming argument order "
1907 "of jurabib version 0.6 for\n'"
1908 << command << before << after << '{'
1909 << citation << "}'.\n"
1910 "Add 'jurabiborder' to the jurabib "
1911 "package options if you used an\n"
1912 "earlier jurabib version." << endl;
1914 begin_inset(os, "LatexCommand ");
1915 os << command << after << before
1916 << '{' << citation << "}\n";
1920 else if (is_known(t.cs(), known_latex_commands)) {
1921 // This needs to be after the check for natbib and
1922 // jurabib commands, because "cite" has different
1923 // arguments with natbib and jurabib.
1924 context.check_layout(os);
1925 begin_inset(os, "LatexCommand ");
1926 os << '\\' << t.cs();
1927 // lyx cannot handle newlines in a latex command
1928 // FIXME: Move the substitution into parser::getOpt()?
1929 os << subst(p.getOpt(), "\n", " ");
1930 os << subst(p.getOpt(), "\n", " ");
1931 os << '{' << subst(p.verbatim_item(), "\n", " ") << "}\n";
1935 else if (is_known(t.cs(), known_quotes)) {
1936 char const * const * where = is_known(t.cs(), known_quotes);
1937 context.check_layout(os);
1938 begin_inset(os, "Quotes ");
1939 os << known_coded_quotes[where - known_quotes];
1941 // LyX adds {} after the quote, so we have to eat
1942 // spaces here if there are any before a possible
1944 eat_whitespace(p, os, context, false);
1948 else if (is_known(t.cs(), known_sizes) &&
1949 context.new_layout_allowed) {
1950 char const * const * where = is_known(t.cs(), known_sizes);
1951 context.check_layout(os);
1952 TeXFont const oldFont = context.font;
1953 context.font.size = known_coded_sizes[where - known_sizes];
1954 output_font_change(os, oldFont, context.font);
1955 eat_whitespace(p, os, context, false);
1958 else if (is_known(t.cs(), known_font_families) &&
1959 context.new_layout_allowed) {
1960 char const * const * where =
1961 is_known(t.cs(), known_font_families);
1962 context.check_layout(os);
1963 TeXFont const oldFont = context.font;
1964 context.font.family =
1965 known_coded_font_families[where - known_font_families];
1966 output_font_change(os, oldFont, context.font);
1967 eat_whitespace(p, os, context, false);
1970 else if (is_known(t.cs(), known_font_series) &&
1971 context.new_layout_allowed) {
1972 char const * const * where =
1973 is_known(t.cs(), known_font_series);
1974 context.check_layout(os);
1975 TeXFont const oldFont = context.font;
1976 context.font.series =
1977 known_coded_font_series[where - known_font_series];
1978 output_font_change(os, oldFont, context.font);
1979 eat_whitespace(p, os, context, false);
1982 else if (is_known(t.cs(), known_font_shapes) &&
1983 context.new_layout_allowed) {
1984 char const * const * where =
1985 is_known(t.cs(), known_font_shapes);
1986 context.check_layout(os);
1987 TeXFont const oldFont = context.font;
1988 context.font.shape =
1989 known_coded_font_shapes[where - known_font_shapes];
1990 output_font_change(os, oldFont, context.font);
1991 eat_whitespace(p, os, context, false);
1993 else if (is_known(t.cs(), known_old_font_families) &&
1994 context.new_layout_allowed) {
1995 char const * const * where =
1996 is_known(t.cs(), known_old_font_families);
1997 context.check_layout(os);
1998 TeXFont const oldFont = context.font;
1999 context.font.init();
2000 context.font.size = oldFont.size;
2001 context.font.family =
2002 known_coded_font_families[where - known_old_font_families];
2003 output_font_change(os, oldFont, context.font);
2004 eat_whitespace(p, os, context, false);
2007 else if (is_known(t.cs(), known_old_font_series) &&
2008 context.new_layout_allowed) {
2009 char const * const * where =
2010 is_known(t.cs(), known_old_font_series);
2011 context.check_layout(os);
2012 TeXFont const oldFont = context.font;
2013 context.font.init();
2014 context.font.size = oldFont.size;
2015 context.font.series =
2016 known_coded_font_series[where - known_old_font_series];
2017 output_font_change(os, oldFont, context.font);
2018 eat_whitespace(p, os, context, false);
2021 else if (is_known(t.cs(), known_old_font_shapes) &&
2022 context.new_layout_allowed) {
2023 char const * const * where =
2024 is_known(t.cs(), known_old_font_shapes);
2025 context.check_layout(os);
2026 TeXFont const oldFont = context.font;
2027 context.font.init();
2028 context.font.size = oldFont.size;
2029 context.font.shape =
2030 known_coded_font_shapes[where - known_old_font_shapes];
2031 output_font_change(os, oldFont, context.font);
2032 eat_whitespace(p, os, context, false);
2035 else if (t.cs() == "LyX" || t.cs() == "TeX"
2036 || t.cs() == "LaTeX") {
2037 context.check_layout(os);
2039 skip_braces(p); // eat {}
2042 else if (t.cs() == "LaTeXe") {
2043 context.check_layout(os);
2045 skip_braces(p); // eat {}
2048 else if (t.cs() == "ldots") {
2049 context.check_layout(os);
2051 os << "\\SpecialChar \\ldots{}\n";
2054 else if (t.cs() == "lyxarrow") {
2055 context.check_layout(os);
2056 os << "\\SpecialChar \\menuseparator\n";
2060 else if (t.cs() == "textcompwordmark") {
2061 context.check_layout(os);
2062 os << "\\SpecialChar \\textcompwordmark{}\n";
2066 else if (t.cs() == "@" && p.next_token().asInput() == ".") {
2067 context.check_layout(os);
2068 os << "\\SpecialChar \\@.\n";
2072 else if (t.cs() == "-") {
2073 context.check_layout(os);
2074 os << "\\SpecialChar \\-\n";
2077 else if (t.cs() == "textasciitilde") {
2078 context.check_layout(os);
2083 else if (t.cs() == "textasciicircum") {
2084 context.check_layout(os);
2089 else if (t.cs() == "textbackslash") {
2090 context.check_layout(os);
2091 os << "\n\\backslash\n";
2095 else if (t.cs() == "_" || t.cs() == "&" || t.cs() == "#"
2096 || t.cs() == "$" || t.cs() == "{" || t.cs() == "}"
2098 context.check_layout(os);
2102 else if (t.cs() == "char") {
2103 context.check_layout(os);
2104 if (p.next_token().character() == '`') {
2106 if (p.next_token().cs() == "\"") {
2111 handle_ert(os, "\\char`", context);
2114 handle_ert(os, "\\char", context);
2118 else if (t.cs() == "verb") {
2119 context.check_layout(os);
2120 char const delimiter = p.next_token().character();
2121 string const arg = p.getArg(delimiter, delimiter);
2123 oss << "\\verb" << delimiter << arg << delimiter;
2124 handle_ert(os, oss.str(), context);
2127 else if (t.cs() == "\"") {
2128 context.check_layout(os);
2129 string const name = p.verbatim_item();
2130 if (name == "a") os << '\xe4';
2131 else if (name == "o") os << '\xf6';
2132 else if (name == "u") os << '\xfc';
2133 else if (name == "A") os << '\xc4';
2134 else if (name == "O") os << '\xd6';
2135 else if (name == "U") os << '\xdc';
2136 else handle_ert(os, "\"{" + name + "}", context);
2139 // Problem: \= creates a tabstop inside the tabbing environment
2140 // and else an accent. In the latter case we really would want
2141 // \={o} instead of \= o.
2142 else if (t.cs() == "=" && (flags & FLAG_TABBING))
2143 handle_ert(os, t.asInput(), context);
2145 else if (t.cs() == "H" || t.cs() == "c" || t.cs() == "^"
2146 || t.cs() == "'" || t.cs() == "`"
2147 || t.cs() == "~" || t.cs() == "." || t.cs() == "=") {
2148 // we need the trim as the LyX parser chokes on such spaces
2149 // The argument of InsetLatexAccent is parsed as a
2150 // subset of LaTeX, so don't parse anything here,
2151 // but use the raw argument.
2152 // Otherwise we would convert \~{\i} wrongly.
2153 // This will of course not translate \~{\ss} to \~{ß},
2154 // but that does at least compile and does only look
2155 // strange on screen.
2156 context.check_layout(os);
2157 os << "\\i \\" << t.cs() << "{"
2158 << trim(p.verbatim_item(), " ")
2162 else if (t.cs() == "ss") {
2163 context.check_layout(os);
2165 skip_braces(p); // eat {}
2168 else if (t.cs() == "i" || t.cs() == "j" || t.cs() == "l" ||
2170 context.check_layout(os);
2171 os << "\\i \\" << t.cs() << "{}\n";
2172 skip_braces(p); // eat {}
2175 else if (t.cs() == "\\") {
2176 context.check_layout(os);
2177 string const next = p.next_token().asInput();
2179 handle_ert(os, "\\\\" + p.getOpt(), context);
2180 else if (next == "*") {
2182 handle_ert(os, "\\\\*" + p.getOpt(), context);
2185 os << "\n\\newline\n";
2189 else if (t.cs() == "input" || t.cs() == "include"
2190 || t.cs() == "verbatiminput") {
2191 string name = '\\' + t.cs();
2192 if (t.cs() == "verbatiminput"
2193 && p.next_token().asInput() == "*")
2194 name += p.get_token().asInput();
2195 context.check_layout(os);
2196 begin_inset(os, "Include ");
2197 string filename(normalize_filename(p.getArg('{', '}')));
2198 string const path = getMasterFilePath();
2199 // We want to preserve relative / absolute filenames,
2200 // therefore path is only used for testing
2201 // FIXME UNICODE encoding of filename and path may be
2202 // wrong (makeAbsPath expects utf8)
2203 if ((t.cs() == "include" || t.cs() == "input") &&
2204 !fs::exists(makeAbsPath(filename, path).toFilesystemEncoding())) {
2205 // The file extension is probably missing.
2206 // Now try to find it out.
2207 string const tex_name =
2208 find_file(filename, path,
2209 known_tex_extensions);
2210 if (!tex_name.empty())
2211 filename = tex_name;
2213 // FIXME UNICODE encoding of filename and path may be
2214 // wrong (makeAbsPath expects utf8)
2215 if (fs::exists(makeAbsPath(filename, path).toFilesystemEncoding())) {
2216 string const abstexname =
2217 makeAbsPath(filename, path).absFilename();
2218 string const abslyxname =
2219 changeExtension(abstexname, ".lyx");
2220 fix_relative_filename(filename);
2221 string const lyxname =
2222 changeExtension(filename, ".lyx");
2223 if (t.cs() != "verbatiminput" &&
2224 tex2lyx(abstexname, FileName(abslyxname))) {
2225 os << name << '{' << lyxname << "}\n";
2227 os << name << '{' << filename << "}\n";
2230 cerr << "Warning: Could not find included file '"
2231 << filename << "'." << endl;
2232 os << name << '{' << filename << "}\n";
2234 os << "preview false\n";
2238 else if (t.cs() == "bibliographystyle") {
2239 // store new bibliographystyle
2240 bibliographystyle = p.verbatim_item();
2241 // output new bibliographystyle.
2242 // This is only necessary if used in some other macro than \bibliography.
2243 handle_ert(os, "\\bibliographystyle{" + bibliographystyle + "}", context);
2246 else if (t.cs() == "bibliography") {
2247 context.check_layout(os);
2248 begin_inset(os, "LatexCommand ");
2250 // Do we have a bibliographystyle set?
2251 if (!bibliographystyle.empty()) {
2252 os << '[' << bibliographystyle << ']';
2254 os << '{' << p.verbatim_item() << "}\n";
2258 else if (t.cs() == "parbox")
2259 parse_box(p, os, FLAG_ITEM, outer, context, true);
2261 else if (t.cs() == "smallskip" ||
2262 t.cs() == "medskip" ||
2263 t.cs() == "bigskip" ||
2264 t.cs() == "vfill") {
2265 context.check_layout(os);
2266 begin_inset(os, "VSpace ");
2272 else if (is_known(t.cs(), known_spaces)) {
2273 char const * const * where = is_known(t.cs(), known_spaces);
2274 context.check_layout(os);
2275 begin_inset(os, "InsetSpace ");
2276 os << '\\' << known_coded_spaces[where - known_spaces]
2278 // LaTeX swallows whitespace after all spaces except
2279 // "\\,". We have to do that here, too, because LyX
2280 // adds "{}" which would make the spaces significant.
2282 eat_whitespace(p, os, context, false);
2283 // LyX adds "{}" after all spaces except "\\ " and
2284 // "\\,", so we have to remove "{}".
2285 // "\\,{}" is equivalent to "\\," in LaTeX, so we
2286 // remove the braces after "\\,", too.
2291 else if (t.cs() == "newpage" ||
2292 t.cs() == "clearpage" ||
2293 t.cs() == "cleardoublepage") {
2294 context.check_layout(os);
2295 // FIXME: what about \\pagebreak?
2296 os << "\n\\" << t.cs() << "\n";
2297 skip_braces(p); // eat {}
2300 else if (t.cs() == "newcommand" ||
2301 t.cs() == "providecommand" ||
2302 t.cs() == "renewcommand") {
2303 // these could be handled by parse_command(), but
2304 // we need to call add_known_command() here.
2305 string name = t.asInput();
2306 if (p.next_token().asInput() == "*") {
2307 // Starred form. Eat '*'
2311 string const command = p.verbatim_item();
2312 string const opt1 = p.getOpt();
2313 string const opt2 = p.getFullOpt();
2314 add_known_command(command, opt1, !opt2.empty());
2315 string const ert = name + '{' + command + '}' +
2317 '{' + p.verbatim_item() + '}';
2320 context.check_layout(os);
2321 begin_inset(os, "FormulaMacro");
2325 // we cannot handle optional argument, so only output ERT
2326 handle_ert(os, ert, context);
2329 else if (t.cs() == "vspace") {
2330 bool starred = false;
2331 if (p.next_token().asInput() == "*") {
2335 string const length = p.verbatim_item();
2338 bool valid = splitLatexLength(length, valstring, unit);
2339 bool known_vspace = false;
2340 bool known_unit = false;
2343 istringstream iss(valstring);
2346 if (unit == "\\smallskipamount") {
2348 known_vspace = true;
2349 } else if (unit == "\\medskipamount") {
2351 known_vspace = true;
2352 } else if (unit == "\\bigskipamount") {
2354 known_vspace = true;
2355 } else if (unit == "\\fill") {
2357 known_vspace = true;
2360 if (!known_vspace) {
2361 switch (unitFromString(unit)) {
2382 if (known_unit || known_vspace) {
2383 // Literal length or known variable
2384 context.check_layout(os);
2385 begin_inset(os, "VSpace ");
2393 // LyX can't handle other length variables in Inset VSpace
2394 string name = t.asInput();
2399 handle_ert(os, name + '{' + unit + '}', context);
2400 else if (value == -1.0)
2401 handle_ert(os, name + "{-" + unit + '}', context);
2403 handle_ert(os, name + '{' + valstring + unit + '}', context);
2405 handle_ert(os, name + '{' + length + '}', context);
2410 //cerr << "#: " << t << " mode: " << mode << endl;
2411 // heuristic: read up to next non-nested space
2413 string s = t.asInput();
2414 string z = p.verbatim_item();
2415 while (p.good() && z != " " && z.size()) {
2416 //cerr << "read: " << z << endl;
2418 z = p.verbatim_item();
2420 cerr << "found ERT: " << s << endl;
2421 handle_ert(os, s + ' ', context);
2423 string name = t.asInput();
2424 if (p.next_token().asInput() == "*") {
2425 // Starred commands like \vspace*{}
2426 p.get_token(); // Eat '*'
2429 if (! parse_command(name, p, os, outer, context))
2430 handle_ert(os, name, context);
2433 if (flags & FLAG_LEAVE) {
2434 flags &= ~FLAG_LEAVE;