2 * \file tex2lyx/text.cpp
3 * This file is part of LyX, the document processor.
4 * Licence details can be found in the file COPYING.
7 * \author Jean-Marc Lasgouttes
10 * Full author contact details are available in file CREDITS.
21 #include "FloatList.h"
22 #include "LaTeXPackages.h"
27 #include "insets/ExternalTemplate.h"
29 #include "support/lassert.h"
30 #include "support/convert.h"
31 #include "support/FileName.h"
32 #include "support/filetools.h"
33 #include "support/lstrings.h"
34 #include "support/lyxtime.h"
43 using namespace lyx::support;
48 void parse_text_in_inset(Parser & p, ostream & os, unsigned flags, bool outer,
49 Context const & context, InsetLayout const * layout)
51 bool const forcePlainLayout =
52 layout ? layout->forcePlainLayout() : false;
53 Context newcontext(true, context.textclass);
55 newcontext.layout = &context.textclass.plainLayout();
57 newcontext.font = context.font;
58 parse_text(p, os, flags, outer, newcontext);
59 newcontext.check_end_layout(os);
65 void parse_text_in_inset(Parser & p, ostream & os, unsigned flags, bool outer,
66 Context const & context, string const & name)
68 InsetLayout const * layout = 0;
69 DocumentClass::InsetLayouts::const_iterator it =
70 context.textclass.insetLayouts().find(from_ascii(name));
71 if (it != context.textclass.insetLayouts().end())
72 layout = &(it->second);
73 parse_text_in_inset(p, os, flags, outer, context, layout);
76 /// parses a paragraph snippet, useful for example for \\emph{...}
77 void parse_text_snippet(Parser & p, ostream & os, unsigned flags, bool outer,
80 Context newcontext(context);
81 // Don't inherit the paragraph-level extra stuff
82 newcontext.par_extra_stuff.clear();
83 parse_text(p, os, flags, outer, newcontext);
84 // Make sure that we don't create invalid .lyx files
85 context.need_layout = newcontext.need_layout;
86 context.need_end_layout = newcontext.need_end_layout;
91 * Thin wrapper around parse_text_snippet() using a string.
93 * We completely ignore \c context.need_layout and \c context.need_end_layout,
94 * because our return value is not used directly (otherwise the stream version
95 * of parse_text_snippet() could be used). That means that the caller needs
96 * to do layout management manually.
97 * This is intended to parse text that does not create any layout changes.
99 string parse_text_snippet(Parser & p, unsigned flags, const bool outer,
102 Context newcontext(context);
103 newcontext.need_layout = false;
104 newcontext.need_end_layout = false;
105 newcontext.new_layout_allowed = false;
106 // Avoid warning by Context::~Context()
107 newcontext.par_extra_stuff.clear();
109 parse_text_snippet(p, os, flags, outer, newcontext);
114 char const * const known_ref_commands[] = { "ref", "pageref", "vref",
115 "vpageref", "prettyref", "eqref", 0 };
117 char const * const known_coded_ref_commands[] = { "ref", "pageref", "vref",
118 "vpageref", "formatted", "eqref", 0 };
120 char const * const known_refstyle_commands[] = { "algref", "chapref", "corref",
121 "eqref", "enuref", "figref", "fnref", "lemref", "parref", "partref", "propref",
122 "secref", "subref", "tabref", "thmref", 0 };
124 char const * const known_refstyle_prefixes[] = { "alg", "chap", "cor",
125 "eq", "enu", "fig", "fn", "lem", "par", "part", "prop",
126 "sec", "sub", "tab", "thm", 0 };
130 * supported CJK encodings
131 * JIS does not work with LyX's encoding conversion
133 const char * const supported_CJK_encodings[] = {
134 "EUC-JP", "KS", "GB", "UTF8",
135 "Bg5", /*"JIS",*/ "SJIS", 0};
138 * the same as supported_CJK_encodings with their corresponding LyX language name
139 * FIXME: The mapping "UTF8" => "chinese-traditional" is only correct for files
141 * NOTE: "Bg5", "JIS" and "SJIS" are not supported by LyX, on re-export the
142 * encodings "UTF8", "EUC-JP" and "EUC-JP" will be used.
143 * please keep this in sync with supported_CJK_encodings line by line!
145 const char * const supported_CJK_languages[] = {
146 "japanese-cjk", "korean", "chinese-simplified", "chinese-traditional",
147 "chinese-traditional", /*"japanese-cjk",*/ "japanese-cjk", 0};
151 * The starred forms are also known except for "citefullauthor",
152 * "citeyear" and "citeyearpar".
154 char const * const known_natbib_commands[] = { "cite", "citet", "citep",
155 "citealt", "citealp", "citeauthor", "citeyear", "citeyearpar",
156 "citefullauthor", "Citet", "Citep", "Citealt", "Citealp", "Citeauthor", 0 };
160 * No starred form other than "cite*" known.
162 char const * const known_jurabib_commands[] = { "cite", "citet", "citep",
163 "citealt", "citealp", "citeauthor", "citeyear", "citeyearpar",
164 // jurabib commands not (yet) supported by LyX:
166 // "footcite", "footcitet", "footcitep", "footcitealt", "footcitealp",
167 // "footciteauthor", "footciteyear", "footciteyearpar",
168 "citefield", "citetitle", 0 };
170 /// LaTeX names for quotes
171 char const * const known_quotes[] = { "dq", "guillemotleft", "flqq", "og",
172 "guillemotright", "frqq", "fg", "glq", "glqq", "textquoteleft", "grq", "grqq",
173 "quotedblbase", "textquotedblleft", "quotesinglbase", "textquoteright", "flq",
174 "guilsinglleft", "frq", "guilsinglright", 0};
176 /// the same as known_quotes with .lyx names
177 char const * const known_coded_quotes[] = { "prd", "ard", "ard", "ard",
178 "ald", "ald", "ald", "gls", "gld", "els", "els", "grd",
179 "gld", "grd", "gls", "ers", "fls",
180 "fls", "frs", "frs", 0};
182 /// LaTeX names for font sizes
183 char const * const known_sizes[] = { "tiny", "scriptsize", "footnotesize",
184 "small", "normalsize", "large", "Large", "LARGE", "huge", "Huge", 0};
186 /// the same as known_sizes with .lyx names
187 char const * const known_coded_sizes[] = { "tiny", "scriptsize", "footnotesize",
188 "small", "normal", "large", "larger", "largest", "huge", "giant", 0};
190 /// LaTeX 2.09 names for font families
191 char const * const known_old_font_families[] = { "rm", "sf", "tt", 0};
193 /// LaTeX names for font families
194 char const * const known_font_families[] = { "rmfamily", "sffamily",
197 /// LaTeX names for font family changing commands
198 char const * const known_text_font_families[] = { "textrm", "textsf",
201 /// The same as known_old_font_families, known_font_families and
202 /// known_text_font_families with .lyx names
203 char const * const known_coded_font_families[] = { "roman", "sans",
206 /// LaTeX 2.09 names for font series
207 char const * const known_old_font_series[] = { "bf", 0};
209 /// LaTeX names for font series
210 char const * const known_font_series[] = { "bfseries", "mdseries", 0};
212 /// LaTeX names for font series changing commands
213 char const * const known_text_font_series[] = { "textbf", "textmd", 0};
215 /// The same as known_old_font_series, known_font_series and
216 /// known_text_font_series with .lyx names
217 char const * const known_coded_font_series[] = { "bold", "medium", 0};
219 /// LaTeX 2.09 names for font shapes
220 char const * const known_old_font_shapes[] = { "it", "sl", "sc", 0};
222 /// LaTeX names for font shapes
223 char const * const known_font_shapes[] = { "itshape", "slshape", "scshape",
226 /// LaTeX names for font shape changing commands
227 char const * const known_text_font_shapes[] = { "textit", "textsl", "textsc",
230 /// The same as known_old_font_shapes, known_font_shapes and
231 /// known_text_font_shapes with .lyx names
232 char const * const known_coded_font_shapes[] = { "italic", "slanted",
233 "smallcaps", "up", 0};
235 /// Known special characters which need skip_spaces_braces() afterwards
236 char const * const known_special_chars[] = {"ldots", "lyxarrow",
237 "textcompwordmark", "slash", 0};
239 /// the same as known_special_chars with .lyx names
240 char const * const known_coded_special_chars[] = {"ldots{}", "menuseparator",
241 "textcompwordmark{}", "slash{}", 0};
244 * Graphics file extensions known by the dvips driver of the graphics package.
245 * These extensions are used to complete the filename of an included
246 * graphics file if it does not contain an extension.
247 * The order must be the same that latex uses to find a file, because we
248 * will use the first extension that matches.
249 * This is only an approximation for the common cases. If we would want to
250 * do it right in all cases, we would need to know which graphics driver is
251 * used and know the extensions of every driver of the graphics package.
253 char const * const known_dvips_graphics_formats[] = {"eps", "ps", "eps.gz",
254 "ps.gz", "eps.Z", "ps.Z", 0};
257 * Graphics file extensions known by the pdftex driver of the graphics package.
258 * \sa known_dvips_graphics_formats
260 char const * const known_pdftex_graphics_formats[] = {"png", "pdf", "jpg",
264 * Known file extensions for TeX files as used by \\include.
266 char const * const known_tex_extensions[] = {"tex", 0};
268 /// spaces known by InsetSpace
269 char const * const known_spaces[] = { " ", "space", ",",
270 "thinspace", "quad", "qquad", "enspace", "enskip",
271 "negthinspace", "negmedspace", "negthickspace", "textvisiblespace",
272 "hfill", "dotfill", "hrulefill", "leftarrowfill", "rightarrowfill",
273 "upbracefill", "downbracefill", 0};
275 /// the same as known_spaces with .lyx names
276 char const * const known_coded_spaces[] = { "space{}", "space{}",
277 "thinspace{}", "thinspace{}", "quad{}", "qquad{}", "enspace{}", "enskip{}",
278 "negthinspace{}", "negmedspace{}", "negthickspace{}", "textvisiblespace{}",
279 "hfill{}", "dotfill{}", "hrulefill{}", "leftarrowfill{}", "rightarrowfill{}",
280 "upbracefill{}", "downbracefill{}", 0};
282 /// These are translated by LyX to commands like "\\LyX{}", so we have to put
283 /// them in ERT. "LaTeXe" must come before "LaTeX"!
284 char const * const known_phrases[] = {"LyX", "TeX", "LaTeXe", "LaTeX", 0};
285 char const * const known_coded_phrases[] = {"LyX", "TeX", "LaTeX2e", "LaTeX", 0};
286 int const known_phrase_lengths[] = {3, 5, 7, 0};
288 /// known TIPA combining diacritical marks
289 char const * const known_tipa_marks[] = {"textsubwedge", "textsubumlaut",
290 "textsubtilde", "textseagull", "textsubbridge", "textinvsubbridge",
291 "textsubsquare", "textsubrhalfring", "textsublhalfring", "textsubplus",
292 "textovercross", "textsubarch", "textsuperimposetilde", "textraising",
293 "textlowering", "textadvancing", "textretracting", "textdoublegrave",
294 "texthighrise", "textlowrise", "textrisefall", "textsyllabic",
297 /// TIPA tones that need special handling
298 char const * const known_tones[] = {"15", "51", "45", "12", "454", 0};
300 // string to store the float type to be able to determine the type of subfloats
301 string float_type = "";
304 /// splits "x=z, y=b" into a map and an ordered keyword vector
305 void split_map(string const & s, map<string, string> & res, vector<string> & keys)
310 keys.resize(v.size());
311 for (size_t i = 0; i < v.size(); ++i) {
312 size_t const pos = v[i].find('=');
313 string const index = trimSpaceAndEol(v[i].substr(0, pos));
314 string const value = trimSpaceAndEol(v[i].substr(pos + 1, string::npos));
322 * Split a LaTeX length into value and unit.
323 * The latter can be a real unit like "pt", or a latex length variable
324 * like "\textwidth". The unit may contain additional stuff like glue
325 * lengths, but we don't care, because such lengths are ERT anyway.
326 * \returns true if \p value and \p unit are valid.
328 bool splitLatexLength(string const & len, string & value, string & unit)
332 const string::size_type i = len.find_first_not_of(" -+0123456789.,");
333 //'4,5' is a valid LaTeX length number. Change it to '4.5'
334 string const length = subst(len, ',', '.');
335 if (i == string::npos)
338 if (len[0] == '\\') {
339 // We had something like \textwidth without a factor
345 value = trimSpaceAndEol(string(length, 0, i));
349 // 'cM' is a valid LaTeX length unit. Change it to 'cm'
350 if (contains(len, '\\'))
351 unit = trimSpaceAndEol(string(len, i));
353 unit = ascii_lowercase(trimSpaceAndEol(string(len, i)));
358 /// A simple function to translate a latex length to something LyX can
359 /// understand. Not perfect, but rather best-effort.
360 bool translate_len(string const & length, string & valstring, string & unit)
362 if (!splitLatexLength(length, valstring, unit))
364 // LyX uses percent values
366 istringstream iss(valstring);
371 string const percentval = oss.str();
373 if (unit.empty() || unit[0] != '\\')
375 string::size_type const i = unit.find(' ');
376 string const endlen = (i == string::npos) ? string() : string(unit, i);
377 if (unit == "\\textwidth") {
378 valstring = percentval;
379 unit = "text%" + endlen;
380 } else if (unit == "\\columnwidth") {
381 valstring = percentval;
382 unit = "col%" + endlen;
383 } else if (unit == "\\paperwidth") {
384 valstring = percentval;
385 unit = "page%" + endlen;
386 } else if (unit == "\\linewidth") {
387 valstring = percentval;
388 unit = "line%" + endlen;
389 } else if (unit == "\\paperheight") {
390 valstring = percentval;
391 unit = "pheight%" + endlen;
392 } else if (unit == "\\textheight") {
393 valstring = percentval;
394 unit = "theight%" + endlen;
402 string translate_len(string const & length)
406 if (translate_len(length, value, unit))
408 // If the input is invalid, return what we have.
416 * Translates a LaTeX length into \p value, \p unit and
417 * \p special parts suitable for a box inset.
418 * The difference from translate_len() is that a box inset knows about
419 * some special "units" that are stored in \p special.
421 void translate_box_len(string const & length, string & value, string & unit, string & special)
423 if (translate_len(length, value, unit)) {
424 if (unit == "\\height" || unit == "\\depth" ||
425 unit == "\\totalheight" || unit == "\\width") {
426 special = unit.substr(1);
427 // The unit is not used, but LyX requires a dummy setting
440 * Find a file with basename \p name in path \p path and an extension
443 string find_file(string const & name, string const & path,
444 char const * const * extensions)
446 for (char const * const * what = extensions; *what; ++what) {
447 string const trial = addExtension(name, *what);
448 if (makeAbsPath(trial, path).exists())
455 void begin_inset(ostream & os, string const & name)
457 os << "\n\\begin_inset " << name;
461 void begin_command_inset(ostream & os, string const & name,
462 string const & latexname)
464 begin_inset(os, "CommandInset ");
465 os << name << "\nLatexCommand " << latexname << '\n';
469 void end_inset(ostream & os)
471 os << "\n\\end_inset\n\n";
475 bool skip_braces(Parser & p)
477 if (p.next_token().cat() != catBegin)
480 if (p.next_token().cat() == catEnd) {
489 /// replace LaTeX commands in \p s from the unicodesymbols file with their
491 docstring convert_unicodesymbols(docstring s)
494 for (size_t i = 0; i < s.size();) {
503 docstring parsed = encodings.fromLaTeXCommand(s,
504 Encodings::TEXT_CMD, termination, rem, &req);
505 set<string>::const_iterator it = req.begin();
506 set<string>::const_iterator en = req.end();
507 for (; it != en; ++it)
508 preamble.registerAutomaticallyLoadedPackage(*it);
511 if (s.empty() || s[0] != '\\')
520 /// try to convert \p s to a valid InsetCommand argument
521 string convert_command_inset_arg(string s)
524 // since we don't know the input encoding we can't use from_utf8
525 s = to_utf8(convert_unicodesymbols(from_ascii(s)));
526 // LyX cannot handle newlines in a latex command
527 return subst(s, "\n", " ");
531 void output_ert(ostream & os, string const & s, Context & context)
533 context.check_layout(os);
534 for (string::const_iterator it = s.begin(), et = s.end(); it != et; ++it) {
536 os << "\n\\backslash\n";
537 else if (*it == '\n') {
538 context.new_paragraph(os);
539 context.check_layout(os);
543 context.check_end_layout(os);
547 void output_ert_inset(ostream & os, string const & s, Context & context)
549 // We must have a valid layout before outputting the ERT inset.
550 context.check_layout(os);
551 Context newcontext(true, context.textclass);
552 InsetLayout const & layout = context.textclass.insetLayout(from_ascii("ERT"));
553 if (layout.forcePlainLayout())
554 newcontext.layout = &context.textclass.plainLayout();
555 begin_inset(os, "ERT");
556 os << "\nstatus collapsed\n";
557 output_ert(os, s, newcontext);
562 Layout const * findLayout(TextClass const & textclass, string const & name, bool command)
564 Layout const * layout = findLayoutWithoutModule(textclass, name, command);
567 if (checkModule(name, command))
568 return findLayoutWithoutModule(textclass, name, command);
573 InsetLayout const * findInsetLayout(TextClass const & textclass, string const & name, bool command)
575 InsetLayout const * insetlayout = findInsetLayoutWithoutModule(textclass, name, command);
578 if (checkModule(name, command))
579 return findInsetLayoutWithoutModule(textclass, name, command);
584 void eat_whitespace(Parser &, ostream &, Context &, bool);
588 * Skips whitespace and braces.
589 * This should be called after a command has been parsed that is not put into
590 * ERT, and where LyX adds "{}" if needed.
592 void skip_spaces_braces(Parser & p, bool keepws = false)
594 /* The following four examples produce the same typeset output and
595 should be handled by this function:
603 // Unfortunately we need to skip comments, too.
604 // We can't use eat_whitespace since writing them after the {}
605 // results in different output in some cases.
606 bool const skipped_spaces = p.skip_spaces(true);
607 bool const skipped_braces = skip_braces(p);
608 if (keepws && skipped_spaces && !skipped_braces)
609 // put back the space (it is better handled by check_space)
610 p.unskip_spaces(true);
614 void output_command_layout(ostream & os, Parser & p, bool outer,
615 Context & parent_context,
616 Layout const * newlayout)
618 TeXFont const oldFont = parent_context.font;
619 // save the current font size
620 string const size = oldFont.size;
621 // reset the font size to default, because the font size switches
622 // don't affect section headings and the like
623 parent_context.font.size = Context::normalfont.size;
624 // we only need to write the font change if we have an open layout
625 if (!parent_context.atParagraphStart())
626 output_font_change(os, oldFont, parent_context.font);
627 parent_context.check_end_layout(os);
628 Context context(true, parent_context.textclass, newlayout,
629 parent_context.layout, parent_context.font);
630 if (parent_context.deeper_paragraph) {
631 // We are beginning a nested environment after a
632 // deeper paragraph inside the outer list environment.
633 // Therefore we don't need to output a "begin deeper".
634 context.need_end_deeper = true;
636 context.check_deeper(os);
637 context.check_layout(os);
638 // FIXME: Adjust to format 446!
639 // Since format 446, layouts do not require anymore all optional
640 // arguments before the required ones. Needs to be implemented!
642 while (optargs < context.layout->optArgs()) {
643 eat_whitespace(p, os, context, false);
644 if (p.next_token().cat() == catEscape ||
645 p.next_token().character() != '[')
647 p.get_token(); // eat '['
648 // FIXME: Just a workaround. InsetArgument::updateBuffer
649 // will compute a proper ID for all "999" Arguments
650 // (which is also what lyx2lyx produces).
651 // However, tex2lyx should be able to output proper IDs
653 begin_inset(os, "Argument 999\n");
654 os << "status collapsed\n\n";
655 parse_text_in_inset(p, os, FLAG_BRACK_LAST, outer, context);
657 eat_whitespace(p, os, context, false);
661 while (reqargs < context.layout->requiredArgs()) {
662 eat_whitespace(p, os, context, false);
663 if (p.next_token().cat() != catBegin)
665 p.get_token(); // eat '{'
666 // FIXME: Just a workaround. InsetArgument::updateBuffer
667 // will compute a proper ID for all "999" Arguments
668 // (which is also what lyx2lyx produces).
669 // However, tex2lyx should be able to output proper IDs
671 begin_inset(os, "Argument 999\n");
672 os << "status collapsed\n\n";
673 parse_text_in_inset(p, os, FLAG_BRACE_LAST, outer, context);
675 eat_whitespace(p, os, context, false);
678 parse_text(p, os, FLAG_ITEM, outer, context);
679 context.check_end_layout(os);
680 if (parent_context.deeper_paragraph) {
681 // We must suppress the "end deeper" because we
682 // suppressed the "begin deeper" above.
683 context.need_end_deeper = false;
685 context.check_end_deeper(os);
686 // We don't need really a new paragraph, but
687 // we must make sure that the next item gets a \begin_layout.
688 parent_context.new_paragraph(os);
689 // Set the font size to the original value. No need to output it here
690 // (Context::begin_layout() will do that if needed)
691 parent_context.font.size = size;
696 * Output a space if necessary.
697 * This function gets called for every whitespace token.
699 * We have three cases here:
700 * 1. A space must be suppressed. Example: The lyxcode case below
701 * 2. A space may be suppressed. Example: Spaces before "\par"
702 * 3. A space must not be suppressed. Example: A space between two words
704 * We currently handle only 1. and 3 and from 2. only the case of
705 * spaces before newlines as a side effect.
707 * 2. could be used to suppress as many spaces as possible. This has two effects:
708 * - Reimporting LyX generated LaTeX files changes almost no whitespace
709 * - Superflous whitespace from non LyX generated LaTeX files is removed.
710 * The drawback is that the logic inside the function becomes
711 * complicated, and that is the reason why it is not implemented.
713 void check_space(Parser & p, ostream & os, Context & context)
715 Token const next = p.next_token();
716 Token const curr = p.curr_token();
717 // A space before a single newline and vice versa must be ignored
718 // LyX emits a newline before \end{lyxcode}.
719 // This newline must be ignored,
720 // otherwise LyX will add an additional protected space.
721 if (next.cat() == catSpace ||
722 next.cat() == catNewline ||
723 (next.cs() == "end" && context.layout->free_spacing && curr.cat() == catNewline)) {
726 context.check_layout(os);
732 * Parse all arguments of \p command
734 void parse_arguments(string const & command,
735 vector<ArgumentType> const & template_arguments,
736 Parser & p, ostream & os, bool outer, Context & context)
738 string ert = command;
739 size_t no_arguments = template_arguments.size();
740 for (size_t i = 0; i < no_arguments; ++i) {
741 switch (template_arguments[i]) {
744 // This argument contains regular LaTeX
745 output_ert_inset(os, ert + '{', context);
746 eat_whitespace(p, os, context, false);
747 if (template_arguments[i] == required)
748 parse_text(p, os, FLAG_ITEM, outer, context);
750 parse_text_snippet(p, os, FLAG_ITEM, outer, context);
754 // This argument consists only of a single item.
755 // The presence of '{' or not must be preserved.
757 if (p.next_token().cat() == catBegin)
758 ert += '{' + p.verbatim_item() + '}';
760 ert += p.verbatim_item();
764 // This argument may contain special characters
765 ert += '{' + p.verbatim_item() + '}';
769 // true because we must not eat whitespace
770 // if an optional arg follows we must not strip the
771 // brackets from this one
772 if (i < no_arguments - 1 &&
773 template_arguments[i+1] == optional)
774 ert += p.getFullOpt(true);
776 ert += p.getOpt(true);
780 output_ert_inset(os, ert, context);
785 * Check whether \p command is a known command. If yes,
786 * handle the command with all arguments.
787 * \return true if the command was parsed, false otherwise.
789 bool parse_command(string const & command, Parser & p, ostream & os,
790 bool outer, Context & context)
792 if (known_commands.find(command) != known_commands.end()) {
793 parse_arguments(command, known_commands[command], p, os,
801 /// Parses a minipage or parbox
802 void parse_box(Parser & p, ostream & os, unsigned outer_flags,
803 unsigned inner_flags, bool outer, Context & parent_context,
804 string const & outer_type, string const & special,
805 string const & inner_type)
809 string hor_pos = "c";
810 // We need to set the height to the LaTeX default of 1\\totalheight
811 // for the case when no height argument is given
812 string height_value = "1";
813 string height_unit = "in";
814 string height_special = "totalheight";
819 string width_special = "none";
820 if (!inner_type.empty() && p.hasOpt()) {
821 if (inner_type != "makebox")
822 position = p.getArg('[', ']');
824 latex_width = p.getArg('[', ']');
825 translate_box_len(latex_width, width_value, width_unit, width_special);
828 if (position != "t" && position != "c" && position != "b") {
829 cerr << "invalid position " << position << " for "
830 << inner_type << endl;
834 if (inner_type != "makebox") {
835 latex_height = p.getArg('[', ']');
836 translate_box_len(latex_height, height_value, height_unit, height_special);
838 string const opt = p.getArg('[', ']');
841 if (hor_pos != "l" && hor_pos != "c" &&
842 hor_pos != "r" && hor_pos != "s") {
843 cerr << "invalid hor_pos " << hor_pos
844 << " for " << inner_type << endl;
851 inner_pos = p.getArg('[', ']');
852 if (inner_pos != "c" && inner_pos != "t" &&
853 inner_pos != "b" && inner_pos != "s") {
854 cerr << "invalid inner_pos "
855 << inner_pos << " for "
856 << inner_type << endl;
857 inner_pos = position;
862 if (inner_type.empty()) {
863 if (special.empty() && outer_type != "framebox")
864 latex_width = "1\\columnwidth";
867 latex_width = p2.getArg('[', ']');
868 string const opt = p2.getArg('[', ']');
871 if (hor_pos != "l" && hor_pos != "c" &&
872 hor_pos != "r" && hor_pos != "s") {
873 cerr << "invalid hor_pos " << hor_pos
874 << " for " << outer_type << endl;
879 } else if (inner_type != "makebox")
880 latex_width = p.verbatim_item();
881 // if e.g. only \ovalbox{content} was used, set the width to 1\columnwidth
882 // as this is LyX's standard for such cases (except for makebox)
883 // \framebox is more special and handled below
884 if (latex_width.empty() && inner_type != "makebox"
885 && outer_type != "framebox")
886 latex_width = "1\\columnwidth";
888 translate_len(latex_width, width_value, width_unit);
890 bool shadedparbox = false;
891 if (inner_type == "shaded") {
892 eat_whitespace(p, os, parent_context, false);
893 if (outer_type == "parbox") {
895 if (p.next_token().cat() == catBegin)
897 eat_whitespace(p, os, parent_context, false);
903 // If we already read the inner box we have to push the inner env
904 if (!outer_type.empty() && !inner_type.empty() &&
905 (inner_flags & FLAG_END))
906 active_environments.push_back(inner_type);
907 // LyX can't handle length variables
908 bool use_ert = contains(width_unit, '\\') || contains(height_unit, '\\');
909 if (!use_ert && !outer_type.empty() && !inner_type.empty()) {
910 // Look whether there is some content after the end of the
911 // inner box, but before the end of the outer box.
912 // If yes, we need to output ERT.
914 if (inner_flags & FLAG_END)
915 p.ertEnvironment(inner_type);
919 bool const outer_env(outer_type == "framed" || outer_type == "minipage");
920 if ((outer_env && p.next_token().asInput() != "\\end") ||
921 (!outer_env && p.next_token().cat() != catEnd)) {
922 // something is between the end of the inner box and
923 // the end of the outer box, so we need to use ERT.
928 // if only \makebox{content} was used we can set its width to 1\width
929 // because this identic and also identic to \mbox
930 // this doesn't work for \framebox{content}, thus we have to use ERT for this
931 if (latex_width.empty() && inner_type == "makebox") {
934 width_special = "width";
935 } else if (latex_width.empty() && outer_type == "framebox") {
940 if (!outer_type.empty()) {
941 if (outer_flags & FLAG_END)
942 ss << "\\begin{" << outer_type << '}';
944 ss << '\\' << outer_type << '{';
945 if (!special.empty())
949 if (!inner_type.empty()) {
950 if (inner_type != "shaded") {
951 if (inner_flags & FLAG_END)
952 ss << "\\begin{" << inner_type << '}';
954 ss << '\\' << inner_type;
956 if (!position.empty())
957 ss << '[' << position << ']';
958 if (!latex_height.empty())
959 ss << '[' << latex_height << ']';
960 if (!inner_pos.empty())
961 ss << '[' << inner_pos << ']';
962 ss << '{' << latex_width << '}';
963 if (!(inner_flags & FLAG_END))
966 if (inner_type == "shaded")
967 ss << "\\begin{shaded}";
968 output_ert_inset(os, ss.str(), parent_context);
969 if (!inner_type.empty()) {
970 parse_text(p, os, inner_flags, outer, parent_context);
971 if (inner_flags & FLAG_END)
972 output_ert_inset(os, "\\end{" + inner_type + '}',
975 output_ert_inset(os, "}", parent_context);
977 if (!outer_type.empty()) {
978 // If we already read the inner box we have to pop
980 if (!inner_type.empty() && (inner_flags & FLAG_END))
981 active_environments.pop_back();
983 // Ensure that the end of the outer box is parsed correctly:
984 // The opening brace has been eaten by parse_outer_box()
985 if (!outer_type.empty() && (outer_flags & FLAG_ITEM)) {
986 outer_flags &= ~FLAG_ITEM;
987 outer_flags |= FLAG_BRACE_LAST;
989 parse_text(p, os, outer_flags, outer, parent_context);
990 if (outer_flags & FLAG_END)
991 output_ert_inset(os, "\\end{" + outer_type + '}',
993 else if (inner_type.empty() && outer_type == "framebox")
994 // in this case it is already closed later
997 output_ert_inset(os, "}", parent_context);
1000 // LyX does not like empty positions, so we have
1001 // to set them to the LaTeX default values here.
1002 if (position.empty())
1004 if (inner_pos.empty())
1005 inner_pos = position;
1006 parent_context.check_layout(os);
1007 begin_inset(os, "Box ");
1008 if (outer_type == "framed")
1010 else if (outer_type == "framebox")
1012 else if (outer_type == "shadowbox")
1013 os << "Shadowbox\n";
1014 else if ((outer_type == "shaded" && inner_type.empty()) ||
1015 (outer_type == "minipage" && inner_type == "shaded") ||
1016 (outer_type == "parbox" && inner_type == "shaded")) {
1018 preamble.registerAutomaticallyLoadedPackage("color");
1019 } else if (outer_type == "doublebox")
1020 os << "Doublebox\n";
1021 else if (outer_type.empty())
1022 os << "Frameless\n";
1024 os << outer_type << '\n';
1025 os << "position \"" << position << "\"\n";
1026 os << "hor_pos \"" << hor_pos << "\"\n";
1027 os << "has_inner_box " << !inner_type.empty() << "\n";
1028 os << "inner_pos \"" << inner_pos << "\"\n";
1029 os << "use_parbox " << (inner_type == "parbox" || shadedparbox)
1031 os << "use_makebox " << (inner_type == "makebox") << '\n';
1032 os << "width \"" << width_value << width_unit << "\"\n";
1033 os << "special \"" << width_special << "\"\n";
1034 os << "height \"" << height_value << height_unit << "\"\n";
1035 os << "height_special \"" << height_special << "\"\n";
1036 os << "status open\n\n";
1038 // Unfortunately we can't use parse_text_in_inset:
1039 // InsetBox::forcePlainLayout() is hard coded and does not
1040 // use the inset layout. Apart from that do we call parse_text
1041 // up to two times, but need only one check_end_layout.
1042 bool const forcePlainLayout =
1043 (!inner_type.empty() || inner_type == "makebox") &&
1044 outer_type != "shaded" && outer_type != "framed";
1045 Context context(true, parent_context.textclass);
1046 if (forcePlainLayout)
1047 context.layout = &context.textclass.plainLayout();
1049 context.font = parent_context.font;
1051 // If we have no inner box the contents will be read with the outer box
1052 if (!inner_type.empty())
1053 parse_text(p, os, inner_flags, outer, context);
1055 // Ensure that the end of the outer box is parsed correctly:
1056 // The opening brace has been eaten by parse_outer_box()
1057 if (!outer_type.empty() && (outer_flags & FLAG_ITEM)) {
1058 outer_flags &= ~FLAG_ITEM;
1059 outer_flags |= FLAG_BRACE_LAST;
1062 // Find end of outer box, output contents if inner_type is
1063 // empty and output possible comments
1064 if (!outer_type.empty()) {
1065 // If we already read the inner box we have to pop
1067 if (!inner_type.empty() && (inner_flags & FLAG_END))
1068 active_environments.pop_back();
1069 // This does not output anything but comments if
1070 // inner_type is not empty (see use_ert)
1071 parse_text(p, os, outer_flags, outer, context);
1074 context.check_end_layout(os);
1076 #ifdef PRESERVE_LAYOUT
1077 // LyX puts a % after the end of the minipage
1078 if (p.next_token().cat() == catNewline && p.next_token().cs().size() > 1) {
1080 //output_ert_inset(os, "%dummy", parent_context);
1083 parent_context.new_paragraph(os);
1085 else if (p.next_token().cat() == catSpace || p.next_token().cat() == catNewline) {
1086 //output_ert_inset(os, "%dummy", parent_context);
1089 // We add a protected space if something real follows
1090 if (p.good() && p.next_token().cat() != catComment) {
1091 begin_inset(os, "space ~\n");
1100 void parse_outer_box(Parser & p, ostream & os, unsigned flags, bool outer,
1101 Context & parent_context, string const & outer_type,
1102 string const & special)
1104 eat_whitespace(p, os, parent_context, false);
1105 if (flags & FLAG_ITEM) {
1107 if (p.next_token().cat() == catBegin)
1110 cerr << "Warning: Ignoring missing '{' after \\"
1111 << outer_type << '.' << endl;
1112 eat_whitespace(p, os, parent_context, false);
1115 unsigned int inner_flags = 0;
1117 if (outer_type == "minipage" || outer_type == "parbox") {
1118 p.skip_spaces(true);
1119 while (p.hasOpt()) {
1121 p.skip_spaces(true);
1124 p.skip_spaces(true);
1125 if (outer_type == "parbox") {
1127 if (p.next_token().cat() == catBegin)
1129 p.skip_spaces(true);
1132 if (outer_type == "shaded") {
1133 // These boxes never have an inner box
1135 } else if (p.next_token().asInput() == "\\parbox") {
1136 inner = p.get_token().cs();
1137 inner_flags = FLAG_ITEM;
1138 } else if (p.next_token().asInput() == "\\begin") {
1139 // Is this a minipage or shaded box?
1142 inner = p.getArg('{', '}');
1144 if (inner == "minipage" || inner == "shaded")
1145 inner_flags = FLAG_END;
1150 if (inner_flags == FLAG_END) {
1151 if (inner != "shaded")
1155 eat_whitespace(p, os, parent_context, false);
1157 parse_box(p, os, flags, FLAG_END, outer, parent_context,
1158 outer_type, special, inner);
1160 if (inner_flags == FLAG_ITEM) {
1162 eat_whitespace(p, os, parent_context, false);
1164 parse_box(p, os, flags, inner_flags, outer, parent_context,
1165 outer_type, special, inner);
1170 void parse_listings(Parser & p, ostream & os, Context & parent_context, bool in_line)
1172 parent_context.check_layout(os);
1173 begin_inset(os, "listings\n");
1175 string arg = p.verbatimOption();
1176 os << "lstparams " << '"' << arg << '"' << '\n';
1177 if (arg.find("\\color") != string::npos)
1178 preamble.registerAutomaticallyLoadedPackage("color");
1181 os << "inline true\n";
1183 os << "inline false\n";
1184 os << "status collapsed\n";
1185 Context context(true, parent_context.textclass);
1186 context.layout = &parent_context.textclass.plainLayout();
1189 // set catcodes to verbatim early, just in case.
1190 p.setCatcodes(VERBATIM_CATCODES);
1191 string delim = p.get_token().asInput();
1192 //FIXME: handler error condition
1193 s = p.verbatimStuff(delim).second;
1194 // context.new_paragraph(os);
1196 s = p.verbatimEnvironment("lstlisting");
1197 output_ert(os, s, context);
1202 /// parse an unknown environment
1203 void parse_unknown_environment(Parser & p, string const & name, ostream & os,
1204 unsigned flags, bool outer,
1205 Context & parent_context)
1207 if (name == "tabbing")
1208 // We need to remember that we have to handle '\=' specially
1209 flags |= FLAG_TABBING;
1211 // We need to translate font changes and paragraphs inside the
1212 // environment to ERT if we have a non standard font.
1213 // Otherwise things like
1214 // \large\begin{foo}\huge bar\end{foo}
1216 bool const specialfont =
1217 (parent_context.font != parent_context.normalfont);
1218 bool const new_layout_allowed = parent_context.new_layout_allowed;
1220 parent_context.new_layout_allowed = false;
1221 output_ert_inset(os, "\\begin{" + name + "}", parent_context);
1222 parse_text_snippet(p, os, flags, outer, parent_context);
1223 output_ert_inset(os, "\\end{" + name + "}", parent_context);
1225 parent_context.new_layout_allowed = new_layout_allowed;
1229 void parse_environment(Parser & p, ostream & os, bool outer,
1230 string & last_env, Context & parent_context)
1232 Layout const * newlayout;
1233 InsetLayout const * newinsetlayout = 0;
1234 string const name = p.getArg('{', '}');
1235 const bool is_starred = suffixIs(name, '*');
1236 string const unstarred_name = rtrim(name, "*");
1237 active_environments.push_back(name);
1239 if (is_math_env(name)) {
1240 parent_context.check_layout(os);
1241 begin_inset(os, "Formula ");
1242 os << "\\begin{" << name << "}";
1243 parse_math(p, os, FLAG_END, MATH_MODE);
1244 os << "\\end{" << name << "}";
1246 if (is_display_math_env(name)) {
1247 // Prevent the conversion of a line break to a space
1248 // (bug 7668). This does not change the output, but
1249 // looks ugly in LyX.
1250 eat_whitespace(p, os, parent_context, false);
1254 else if (is_known(name, preamble.polyglossia_languages)) {
1255 // We must begin a new paragraph if not already done
1256 if (! parent_context.atParagraphStart()) {
1257 parent_context.check_end_layout(os);
1258 parent_context.new_paragraph(os);
1260 // save the language in the context so that it is
1261 // handled by parse_text
1262 parent_context.font.language = preamble.polyglossia2lyx(name);
1263 parse_text(p, os, FLAG_END, outer, parent_context);
1264 // Just in case the environment is empty
1265 parent_context.extra_stuff.erase();
1266 // We must begin a new paragraph to reset the language
1267 parent_context.new_paragraph(os);
1271 else if (unstarred_name == "tabular" || name == "longtable") {
1272 eat_whitespace(p, os, parent_context, false);
1273 string width = "0pt";
1274 if (name == "tabular*") {
1275 width = lyx::translate_len(p.getArg('{', '}'));
1276 eat_whitespace(p, os, parent_context, false);
1278 parent_context.check_layout(os);
1279 begin_inset(os, "Tabular ");
1280 handle_tabular(p, os, name, width, parent_context);
1285 else if (parent_context.textclass.floats().typeExist(unstarred_name)) {
1286 eat_whitespace(p, os, parent_context, false);
1287 string const opt = p.hasOpt() ? p.getArg('[', ']') : string();
1288 eat_whitespace(p, os, parent_context, false);
1289 parent_context.check_layout(os);
1290 begin_inset(os, "Float " + unstarred_name + "\n");
1291 // store the float type for subfloats
1292 // subfloats only work with figures and tables
1293 if (unstarred_name == "figure")
1294 float_type = unstarred_name;
1295 else if (unstarred_name == "table")
1296 float_type = unstarred_name;
1300 os << "placement " << opt << '\n';
1301 if (contains(opt, "H"))
1302 preamble.registerAutomaticallyLoadedPackage("float");
1304 Floating const & fl = parent_context.textclass.floats()
1305 .getType(unstarred_name);
1306 if (!fl.floattype().empty() && fl.usesFloatPkg())
1307 preamble.registerAutomaticallyLoadedPackage("float");
1310 os << "wide " << convert<string>(is_starred)
1311 << "\nsideways false"
1312 << "\nstatus open\n\n";
1313 parse_text_in_inset(p, os, FLAG_END, outer, parent_context);
1315 // We don't need really a new paragraph, but
1316 // we must make sure that the next item gets a \begin_layout.
1317 parent_context.new_paragraph(os);
1319 // the float is parsed thus delete the type
1323 else if (unstarred_name == "sidewaysfigure"
1324 || unstarred_name == "sidewaystable") {
1325 eat_whitespace(p, os, parent_context, false);
1326 parent_context.check_layout(os);
1327 if (unstarred_name == "sidewaysfigure")
1328 begin_inset(os, "Float figure\n");
1330 begin_inset(os, "Float table\n");
1331 os << "wide " << convert<string>(is_starred)
1332 << "\nsideways true"
1333 << "\nstatus open\n\n";
1334 parse_text_in_inset(p, os, FLAG_END, outer, parent_context);
1336 // We don't need really a new paragraph, but
1337 // we must make sure that the next item gets a \begin_layout.
1338 parent_context.new_paragraph(os);
1340 preamble.registerAutomaticallyLoadedPackage("rotfloat");
1343 else if (name == "wrapfigure" || name == "wraptable") {
1344 // syntax is \begin{wrapfigure}[lines]{placement}[overhang]{width}
1345 eat_whitespace(p, os, parent_context, false);
1346 parent_context.check_layout(os);
1349 string overhang = "0col%";
1352 lines = p.getArg('[', ']');
1353 string const placement = p.getArg('{', '}');
1355 overhang = p.getArg('[', ']');
1356 string const width = p.getArg('{', '}');
1358 if (name == "wrapfigure")
1359 begin_inset(os, "Wrap figure\n");
1361 begin_inset(os, "Wrap table\n");
1362 os << "lines " << lines
1363 << "\nplacement " << placement
1364 << "\noverhang " << lyx::translate_len(overhang)
1365 << "\nwidth " << lyx::translate_len(width)
1366 << "\nstatus open\n\n";
1367 parse_text_in_inset(p, os, FLAG_END, outer, parent_context);
1369 // We don't need really a new paragraph, but
1370 // we must make sure that the next item gets a \begin_layout.
1371 parent_context.new_paragraph(os);
1373 preamble.registerAutomaticallyLoadedPackage("wrapfig");
1376 else if (name == "minipage") {
1377 eat_whitespace(p, os, parent_context, false);
1378 // Test whether this is an outer box of a shaded box
1380 // swallow arguments
1381 while (p.hasOpt()) {
1383 p.skip_spaces(true);
1386 p.skip_spaces(true);
1387 Token t = p.get_token();
1388 bool shaded = false;
1389 if (t.asInput() == "\\begin") {
1390 p.skip_spaces(true);
1391 if (p.getArg('{', '}') == "shaded")
1396 parse_outer_box(p, os, FLAG_END, outer,
1397 parent_context, name, "shaded");
1399 parse_box(p, os, 0, FLAG_END, outer, parent_context,
1404 else if (name == "comment") {
1405 eat_whitespace(p, os, parent_context, false);
1406 parent_context.check_layout(os);
1407 begin_inset(os, "Note Comment\n");
1408 os << "status open\n";
1409 parse_text_in_inset(p, os, FLAG_END, outer, parent_context);
1412 skip_braces(p); // eat {} that might by set by LyX behind comments
1413 preamble.registerAutomaticallyLoadedPackage("verbatim");
1416 else if (name == "verbatim") {
1417 // FIXME: this should go in the generic code that
1418 // handles environments defined in layout file that
1419 // have "PassThru 1". However, the code over there is
1420 // already too complicated for my taste.
1421 parent_context.new_paragraph(os);
1422 Context context(true, parent_context.textclass,
1423 &parent_context.textclass[from_ascii("Verbatim")]);
1424 string s = p.verbatimEnvironment("verbatim");
1425 output_ert(os, s, context);
1429 else if (name == "IPA") {
1430 eat_whitespace(p, os, parent_context, false);
1431 parent_context.check_layout(os);
1432 begin_inset(os, "IPA\n");
1433 parse_text_in_inset(p, os, FLAG_END, outer, parent_context);
1436 preamble.registerAutomaticallyLoadedPackage("tipa");
1437 preamble.registerAutomaticallyLoadedPackage("tipx");
1440 else if (name == "CJK") {
1441 // the scheme is \begin{CJK}{encoding}{mapping}text\end{CJK}
1442 // It is impossible to decide if a CJK environment was in its own paragraph or within
1443 // a line. We therefore always assume a paragraph since the latter is a rare case.
1444 eat_whitespace(p, os, parent_context, false);
1445 parent_context.check_end_layout(os);
1446 // store the encoding to be able to reset it
1447 string const encoding_old = p.getEncoding();
1448 string const encoding = p.getArg('{', '}');
1449 // FIXME: For some reason JIS does not work. Although the text
1450 // in tests/CJK.tex is identical with the SJIS version if you
1451 // convert both snippets using the recode command line utility,
1452 // the resulting .lyx file contains some extra characters if
1453 // you set buggy_encoding to false for JIS.
1454 bool const buggy_encoding = encoding == "JIS";
1455 if (!buggy_encoding)
1456 p.setEncoding(encoding, Encoding::CJK);
1458 // FIXME: This will read garbage, since the data is not encoded in utf8.
1459 p.setEncoding("UTF-8");
1461 // LyX only supports the same mapping for all CJK
1462 // environments, so we might need to output everything as ERT
1463 string const mapping = trim(p.getArg('{', '}'));
1464 char const * const * const where =
1465 is_known(encoding, supported_CJK_encodings);
1466 if (!buggy_encoding && !preamble.fontCJKSet())
1467 preamble.fontCJK(mapping);
1468 bool knownMapping = mapping == preamble.fontCJK();
1469 if (buggy_encoding || !knownMapping || !where) {
1470 parent_context.check_layout(os);
1471 output_ert_inset(os, "\\begin{" + name + "}{" + encoding + "}{" + mapping + "}",
1473 // we must parse the content as verbatim because e.g. JIS can contain
1474 // normally invalid characters
1475 // FIXME: This works only for the most simple cases.
1476 // Since TeX control characters are not parsed,
1477 // things like comments are completely wrong.
1478 string const s = p.plainEnvironment("CJK");
1479 for (string::const_iterator it = s.begin(), et = s.end(); it != et; ++it) {
1481 output_ert_inset(os, "\\", parent_context);
1482 else if (*it == '$')
1483 output_ert_inset(os, "$", parent_context);
1484 else if (*it == '\n' && it + 1 != et && s.begin() + 1 != it)
1489 output_ert_inset(os, "\\end{" + name + "}",
1493 supported_CJK_languages[where - supported_CJK_encodings];
1494 // store the language because we must reset it at the end
1495 string const lang_old = parent_context.font.language;
1496 parent_context.font.language = lang;
1497 parse_text_in_inset(p, os, FLAG_END, outer, parent_context);
1498 parent_context.font.language = lang_old;
1499 parent_context.new_paragraph(os);
1501 p.setEncoding(encoding_old);
1505 else if (name == "lyxgreyedout") {
1506 eat_whitespace(p, os, parent_context, false);
1507 parent_context.check_layout(os);
1508 begin_inset(os, "Note Greyedout\n");
1509 os << "status open\n";
1510 parse_text_in_inset(p, os, FLAG_END, outer, parent_context);
1513 if (!preamble.notefontcolor().empty())
1514 preamble.registerAutomaticallyLoadedPackage("color");
1517 else if (name == "framed" || name == "shaded") {
1518 eat_whitespace(p, os, parent_context, false);
1519 parse_outer_box(p, os, FLAG_END, outer, parent_context, name, "");
1523 else if (name == "lstlisting") {
1524 eat_whitespace(p, os, parent_context, false);
1525 parse_listings(p, os, parent_context, false);
1529 else if (!parent_context.new_layout_allowed)
1530 parse_unknown_environment(p, name, os, FLAG_END, outer,
1533 // Alignment and spacing settings
1534 // FIXME (bug xxxx): These settings can span multiple paragraphs and
1535 // therefore are totally broken!
1536 // Note that \centering, raggedright, and raggedleft cannot be handled, as
1537 // they are commands not environments. They are furthermore switches that
1538 // can be ended by another switches, but also by commands like \footnote or
1539 // \parbox. So the only safe way is to leave them untouched.
1540 else if (name == "center" || name == "centering" ||
1541 name == "flushleft" || name == "flushright" ||
1542 name == "singlespace" || name == "onehalfspace" ||
1543 name == "doublespace" || name == "spacing") {
1544 eat_whitespace(p, os, parent_context, false);
1545 // We must begin a new paragraph if not already done
1546 if (! parent_context.atParagraphStart()) {
1547 parent_context.check_end_layout(os);
1548 parent_context.new_paragraph(os);
1550 if (name == "flushleft")
1551 parent_context.add_extra_stuff("\\align left\n");
1552 else if (name == "flushright")
1553 parent_context.add_extra_stuff("\\align right\n");
1554 else if (name == "center" || name == "centering")
1555 parent_context.add_extra_stuff("\\align center\n");
1556 else if (name == "singlespace")
1557 parent_context.add_extra_stuff("\\paragraph_spacing single\n");
1558 else if (name == "onehalfspace") {
1559 parent_context.add_extra_stuff("\\paragraph_spacing onehalf\n");
1560 preamble.registerAutomaticallyLoadedPackage("setspace");
1561 } else if (name == "doublespace") {
1562 parent_context.add_extra_stuff("\\paragraph_spacing double\n");
1563 preamble.registerAutomaticallyLoadedPackage("setspace");
1564 } else if (name == "spacing") {
1565 parent_context.add_extra_stuff("\\paragraph_spacing other " + p.verbatim_item() + "\n");
1566 preamble.registerAutomaticallyLoadedPackage("setspace");
1568 parse_text(p, os, FLAG_END, outer, parent_context);
1569 // Just in case the environment is empty
1570 parent_context.extra_stuff.erase();
1571 // We must begin a new paragraph to reset the alignment
1572 parent_context.new_paragraph(os);
1576 // The single '=' is meant here.
1577 else if ((newlayout = findLayout(parent_context.textclass, name, false))) {
1578 eat_whitespace(p, os, parent_context, false);
1579 Context context(true, parent_context.textclass, newlayout,
1580 parent_context.layout, parent_context.font);
1581 if (parent_context.deeper_paragraph) {
1582 // We are beginning a nested environment after a
1583 // deeper paragraph inside the outer list environment.
1584 // Therefore we don't need to output a "begin deeper".
1585 context.need_end_deeper = true;
1587 parent_context.check_end_layout(os);
1588 if (last_env == name) {
1589 // we need to output a separator since LyX would export
1590 // the two environments as one otherwise (bug 5716)
1591 docstring const sep = from_ascii("--Separator--");
1592 TeX2LyXDocClass const & textclass(parent_context.textclass);
1593 if (textclass.hasLayout(sep)) {
1594 Context newcontext(parent_context);
1595 newcontext.layout = &(textclass[sep]);
1596 newcontext.check_layout(os);
1597 newcontext.check_end_layout(os);
1599 parent_context.check_layout(os);
1600 begin_inset(os, "Note Note\n");
1601 os << "status closed\n";
1602 Context newcontext(true, textclass,
1603 &(textclass.defaultLayout()));
1604 newcontext.check_layout(os);
1605 newcontext.check_end_layout(os);
1607 parent_context.check_end_layout(os);
1610 switch (context.layout->latextype) {
1611 case LATEX_LIST_ENVIRONMENT:
1612 context.add_par_extra_stuff("\\labelwidthstring "
1613 + p.verbatim_item() + '\n');
1616 case LATEX_BIB_ENVIRONMENT:
1617 p.verbatim_item(); // swallow next arg
1623 context.check_deeper(os);
1624 // handle known optional and required arguments
1625 // FIXME: Since format 446, layouts do not require anymore all optional
1626 // arguments before the required ones. Needs to be implemented!
1627 // Unfortunately LyX can't handle arguments of list arguments (bug 7468):
1628 // It is impossible to place anything after the environment name,
1629 // but before the first \\item.
1630 if (context.layout->latextype == LATEX_ENVIRONMENT) {
1631 bool need_layout = true;
1633 while (optargs < context.layout->optArgs()) {
1634 eat_whitespace(p, os, context, false);
1635 if (p.next_token().cat() == catEscape ||
1636 p.next_token().character() != '[')
1638 p.get_token(); // eat '['
1640 context.check_layout(os);
1641 need_layout = false;
1643 // FIXME: Just a workaround. InsetArgument::updateBuffer
1644 // will compute a proper ID for all "999" Arguments
1645 // (which is also what lyx2lyx produces).
1646 // However, tex2lyx should be able to output proper IDs
1648 begin_inset(os, "Argument 999\n");
1649 os << "status collapsed\n\n";
1650 parse_text_in_inset(p, os, FLAG_BRACK_LAST, outer, context);
1652 eat_whitespace(p, os, context, false);
1656 while (reqargs < context.layout->requiredArgs()) {
1657 eat_whitespace(p, os, context, false);
1658 if (p.next_token().cat() != catBegin)
1660 p.get_token(); // eat '{'
1662 context.check_layout(os);
1663 need_layout = false;
1665 // FIXME: Just a workaround. InsetArgument::updateBuffer
1666 // will compute a proper ID for all "999" Arguments
1667 // (which is also what lyx2lyx produces).
1668 // However, tex2lyx should be able to output proper IDs
1670 begin_inset(os, "Argument 999\n");
1671 os << "status collapsed\n\n";
1672 parse_text_in_inset(p, os, FLAG_BRACE_LAST, outer, context);
1674 eat_whitespace(p, os, context, false);
1678 parse_text(p, os, FLAG_END, outer, context);
1679 context.check_end_layout(os);
1680 if (parent_context.deeper_paragraph) {
1681 // We must suppress the "end deeper" because we
1682 // suppressed the "begin deeper" above.
1683 context.need_end_deeper = false;
1685 context.check_end_deeper(os);
1686 parent_context.new_paragraph(os);
1688 if (!preamble.titleLayoutFound())
1689 preamble.titleLayoutFound(newlayout->intitle);
1690 set<string> const & req = newlayout->requires();
1691 set<string>::const_iterator it = req.begin();
1692 set<string>::const_iterator en = req.end();
1693 for (; it != en; ++it)
1694 preamble.registerAutomaticallyLoadedPackage(*it);
1697 // The single '=' is meant here.
1698 else if ((newinsetlayout = findInsetLayout(parent_context.textclass, name, false))) {
1699 eat_whitespace(p, os, parent_context, false);
1700 parent_context.check_layout(os);
1701 begin_inset(os, "Flex ");
1702 os << to_utf8(newinsetlayout->name()) << '\n'
1703 << "status collapsed\n";
1704 if (newinsetlayout->isPassThru()) {
1705 string const arg = p.verbatimEnvironment(name);
1706 Context context(true, parent_context.textclass,
1707 &parent_context.textclass.plainLayout(),
1708 parent_context.layout);
1709 output_ert(os, arg, parent_context);
1711 parse_text_in_inset(p, os, FLAG_END, false, parent_context, newinsetlayout);
1715 else if (name == "appendix") {
1716 // This is no good latex style, but it works and is used in some documents...
1717 eat_whitespace(p, os, parent_context, false);
1718 parent_context.check_end_layout(os);
1719 Context context(true, parent_context.textclass, parent_context.layout,
1720 parent_context.layout, parent_context.font);
1721 context.check_layout(os);
1722 os << "\\start_of_appendix\n";
1723 parse_text(p, os, FLAG_END, outer, context);
1724 context.check_end_layout(os);
1728 else if (known_environments.find(name) != known_environments.end()) {
1729 vector<ArgumentType> arguments = known_environments[name];
1730 // The last "argument" denotes wether we may translate the
1731 // environment contents to LyX
1732 // The default required if no argument is given makes us
1733 // compatible with the reLyXre environment.
1734 ArgumentType contents = arguments.empty() ?
1737 if (!arguments.empty())
1738 arguments.pop_back();
1739 // See comment in parse_unknown_environment()
1740 bool const specialfont =
1741 (parent_context.font != parent_context.normalfont);
1742 bool const new_layout_allowed =
1743 parent_context.new_layout_allowed;
1745 parent_context.new_layout_allowed = false;
1746 parse_arguments("\\begin{" + name + "}", arguments, p, os,
1747 outer, parent_context);
1748 if (contents == verbatim)
1749 output_ert_inset(os, p.ertEnvironment(name),
1752 parse_text_snippet(p, os, FLAG_END, outer,
1754 output_ert_inset(os, "\\end{" + name + "}", parent_context);
1756 parent_context.new_layout_allowed = new_layout_allowed;
1760 parse_unknown_environment(p, name, os, FLAG_END, outer,
1764 active_environments.pop_back();
1768 /// parses a comment and outputs it to \p os.
1769 void parse_comment(Parser & p, ostream & os, Token const & t, Context & context)
1771 LASSERT(t.cat() == catComment, return);
1772 if (!t.cs().empty()) {
1773 context.check_layout(os);
1774 output_ert_inset(os, '%' + t.cs(), context);
1775 if (p.next_token().cat() == catNewline) {
1776 // A newline after a comment line starts a new
1778 if (context.new_layout_allowed) {
1779 if(!context.atParagraphStart())
1780 // Only start a new paragraph if not already
1781 // done (we might get called recursively)
1782 context.new_paragraph(os);
1784 output_ert_inset(os, "\n", context);
1785 eat_whitespace(p, os, context, true);
1788 // "%\n" combination
1795 * Reads spaces and comments until the first non-space, non-comment token.
1796 * New paragraphs (double newlines or \\par) are handled like simple spaces
1797 * if \p eatParagraph is true.
1798 * Spaces are skipped, but comments are written to \p os.
1800 void eat_whitespace(Parser & p, ostream & os, Context & context,
1804 Token const & t = p.get_token();
1805 if (t.cat() == catComment)
1806 parse_comment(p, os, t, context);
1807 else if ((! eatParagraph && p.isParagraph()) ||
1808 (t.cat() != catSpace && t.cat() != catNewline)) {
1817 * Set a font attribute, parse text and reset the font attribute.
1818 * \param attribute Attribute name (e.g. \\family, \\shape etc.)
1819 * \param currentvalue Current value of the attribute. Is set to the new
1820 * value during parsing.
1821 * \param newvalue New value of the attribute
1823 void parse_text_attributes(Parser & p, ostream & os, unsigned flags, bool outer,
1824 Context & context, string const & attribute,
1825 string & currentvalue, string const & newvalue)
1827 context.check_layout(os);
1828 string const oldvalue = currentvalue;
1829 currentvalue = newvalue;
1830 os << '\n' << attribute << ' ' << newvalue << "\n";
1831 parse_text_snippet(p, os, flags, outer, context);
1832 context.check_layout(os);
1833 os << '\n' << attribute << ' ' << oldvalue << "\n";
1834 currentvalue = oldvalue;
1838 /// get the arguments of a natbib or jurabib citation command
1839 void get_cite_arguments(Parser & p, bool natbibOrder,
1840 string & before, string & after)
1842 // We need to distinguish "" and "[]", so we can't use p.getOpt().
1844 // text before the citation
1846 // text after the citation
1847 after = p.getFullOpt();
1849 if (!after.empty()) {
1850 before = p.getFullOpt();
1851 if (natbibOrder && !before.empty())
1852 swap(before, after);
1857 /// Convert filenames with TeX macros and/or quotes to something LyX
1859 string const normalize_filename(string const & name)
1864 Token const & t = p.get_token();
1865 if (t.cat() != catEscape)
1867 else if (t.cs() == "lyxdot") {
1868 // This is used by LyX for simple dots in relative
1872 } else if (t.cs() == "space") {
1875 } else if (t.cs() == "string") {
1876 // Convert \string" to " and \string~ to ~
1877 Token const & n = p.next_token();
1878 if (n.asInput() != "\"" && n.asInput() != "~")
1883 // Strip quotes. This is a bit complicated (see latex_path()).
1884 string full = os.str();
1885 if (!full.empty() && full[0] == '"') {
1886 string base = removeExtension(full);
1887 string ext = getExtension(full);
1888 if (!base.empty() && base[base.length()-1] == '"')
1891 return addExtension(trim(base, "\""), ext);
1892 if (full[full.length()-1] == '"')
1895 return trim(full, "\"");
1901 /// Convert \p name from TeX convention (relative to master file) to LyX
1902 /// convention (relative to .lyx file) if it is relative
1903 void fix_child_filename(string & name)
1905 string const absMasterTeX = getMasterFilePath(true);
1906 bool const isabs = FileName::isAbsolute(name);
1907 // convert from "relative to .tex master" to absolute original path
1909 name = makeAbsPath(name, absMasterTeX).absFileName();
1910 bool copyfile = copyFiles();
1911 string const absParentLyX = getParentFilePath(false);
1914 // convert from absolute original path to "relative to master file"
1915 string const rel = to_utf8(makeRelPath(from_utf8(name),
1916 from_utf8(absMasterTeX)));
1917 // re-interpret "relative to .tex file" as "relative to .lyx file"
1918 // (is different if the master .lyx file resides in a
1919 // different path than the master .tex file)
1920 string const absMasterLyX = getMasterFilePath(false);
1921 abs = makeAbsPath(rel, absMasterLyX).absFileName();
1922 // Do not copy if the new path is impossible to create. Example:
1923 // absMasterTeX = "/foo/bar/"
1924 // absMasterLyX = "/bar/"
1925 // name = "/baz.eps" => new absolute name would be "/../baz.eps"
1926 if (contains(name, "/../"))
1933 // convert from absolute original path to
1934 // "relative to .lyx file"
1935 name = to_utf8(makeRelPath(from_utf8(abs),
1936 from_utf8(absParentLyX)));
1940 // convert from absolute original path to "relative to .lyx file"
1941 name = to_utf8(makeRelPath(from_utf8(name),
1942 from_utf8(absParentLyX)));
1947 void copy_file(FileName const & src, string dstname)
1951 string const absParent = getParentFilePath(false);
1953 if (FileName::isAbsolute(dstname))
1954 dst = FileName(dstname);
1956 dst = makeAbsPath(dstname, absParent);
1957 string const absMaster = getMasterFilePath(false);
1958 FileName const srcpath = src.onlyPath();
1959 FileName const dstpath = dst.onlyPath();
1960 if (equivalent(srcpath, dstpath))
1962 if (!dstpath.isDirectory()) {
1963 if (!dstpath.createPath()) {
1964 cerr << "Warning: Could not create directory for file `"
1965 << dst.absFileName() << "´." << endl;
1969 if (dst.isReadableFile()) {
1970 if (overwriteFiles())
1971 cerr << "Warning: Overwriting existing file `"
1972 << dst.absFileName() << "´." << endl;
1974 cerr << "Warning: Not overwriting existing file `"
1975 << dst.absFileName() << "´." << endl;
1979 if (!src.copyTo(dst))
1980 cerr << "Warning: Could not copy file `" << src.absFileName()
1981 << "´ to `" << dst.absFileName() << "´." << endl;
1985 /// Parse a NoWeb Chunk section. The initial "<<" is already parsed.
1986 bool parse_noweb(Parser & p, ostream & os, Context & context)
1988 // check whether a chunk is possible here.
1989 if (!context.new_layout_allowed ||
1990 !context.textclass.hasLayout(from_ascii("Chunk"))) {
1996 // read the parameters
1997 Parser::Arg stuff = p.verbatimStuff(">>=", false);
2002 string chunk = "<<" + stuff.second + ">>="
2003 + p.verbatimStuff("\n").second + '\n';
2005 stuff = p.verbatimStuff("\n@");
2010 chunk += stuff.second + "\n@";
2011 string post_chunk = p.verbatimStuff("\n").second + '\n';
2012 if (post_chunk[0] != ' ' && post_chunk[0] != '\n') {
2016 chunk += post_chunk;
2018 context.new_paragraph(os);
2019 Context newcontext(true, context.textclass,
2020 &context.textclass[from_ascii("Chunk")]);
2021 output_ert(os, chunk, newcontext);
2028 /// detects \\def, \\long\\def and \\global\\long\\def with ws and comments
2029 bool is_macro(Parser & p)
2031 Token first = p.curr_token();
2032 if (first.cat() != catEscape || !p.good())
2034 if (first.cs() == "def")
2036 if (first.cs() != "global" && first.cs() != "long")
2038 Token second = p.get_token();
2040 while (p.good() && !p.isParagraph() && (second.cat() == catSpace ||
2041 second.cat() == catNewline || second.cat() == catComment)) {
2042 second = p.get_token();
2045 bool secondvalid = second.cat() == catEscape;
2047 bool thirdvalid = false;
2048 if (p.good() && first.cs() == "global" && secondvalid &&
2049 second.cs() == "long") {
2050 third = p.get_token();
2052 while (p.good() && !p.isParagraph() &&
2053 (third.cat() == catSpace ||
2054 third.cat() == catNewline ||
2055 third.cat() == catComment)) {
2056 third = p.get_token();
2059 thirdvalid = third.cat() == catEscape;
2061 for (int i = 0; i < pos; ++i)
2066 return (first.cs() == "global" || first.cs() == "long") &&
2067 second.cs() == "def";
2068 return first.cs() == "global" && second.cs() == "long" &&
2069 third.cs() == "def";
2073 /// Parse a macro definition (assumes that is_macro() returned true)
2074 void parse_macro(Parser & p, ostream & os, Context & context)
2076 context.check_layout(os);
2077 Token first = p.curr_token();
2080 string command = first.asInput();
2081 if (first.cs() != "def") {
2083 eat_whitespace(p, os, context, false);
2084 second = p.curr_token();
2085 command += second.asInput();
2086 if (second.cs() != "def") {
2088 eat_whitespace(p, os, context, false);
2089 third = p.curr_token();
2090 command += third.asInput();
2093 eat_whitespace(p, os, context, false);
2094 string const name = p.get_token().cs();
2095 eat_whitespace(p, os, context, false);
2101 while (p.next_token().cat() != catBegin) {
2102 if (p.next_token().cat() == catParameter) {
2107 // followed by number?
2108 if (p.next_token().cat() == catOther) {
2109 string s = p.get_token().asInput();
2111 // number = current arity + 1?
2112 if (s.size() == 1 && s[0] == arity + '0' + 1)
2117 paramtext += p.get_token().cs();
2119 paramtext += p.get_token().cs();
2124 // only output simple (i.e. compatible) macro as FormulaMacros
2125 string ert = '\\' + name + ' ' + paramtext + '{' + p.verbatim_item() + '}';
2127 context.check_layout(os);
2128 begin_inset(os, "FormulaMacro");
2129 os << "\n\\def" << ert;
2132 output_ert_inset(os, command + ert, context);
2136 void registerExternalTemplatePackages(string const & name)
2138 external::TemplateManager const & etm = external::TemplateManager::get();
2139 external::Template const * const et = etm.getTemplateByName(name);
2142 external::Template::Formats::const_iterator cit = et->formats.end();
2144 cit = et->formats.find("PDFLaTeX");
2145 if (cit == et->formats.end())
2146 // If the template has not specified a PDFLaTeX output,
2147 // we try the LaTeX format.
2148 cit = et->formats.find("LaTeX");
2149 if (cit == et->formats.end())
2151 vector<string>::const_iterator qit = cit->second.requirements.begin();
2152 vector<string>::const_iterator qend = cit->second.requirements.end();
2153 for (; qit != qend; ++qit)
2154 preamble.registerAutomaticallyLoadedPackage(*qit);
2157 } // anonymous namespace
2160 void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
2163 Layout const * newlayout = 0;
2164 InsetLayout const * newinsetlayout = 0;
2165 char const * const * where = 0;
2166 // Store the latest bibliographystyle, addcontentslineContent and
2167 // nocite{*} option (needed for bibtex inset)
2169 string contentslineContent;
2170 string bibliographystyle = "default";
2171 bool const use_natbib = isProvided("natbib");
2172 bool const use_jurabib = isProvided("jurabib");
2175 Token const & t = p.get_token();
2177 // it is impossible to determine the correct encoding for non-CJK Japanese.
2178 // Therefore write a note at the beginning of the document
2179 if (is_nonCJKJapanese) {
2180 context.check_layout(os);
2181 begin_inset(os, "Note Note\n");
2182 os << "status open\n\\begin_layout Plain Layout\n"
2183 << "\\series bold\n"
2184 << "Important information:\n"
2185 << "\\end_layout\n\n"
2186 << "\\begin_layout Plain Layout\n"
2187 << "The original LaTeX source for this document is in Japanese (pLaTeX).\n"
2188 << " It was therefore impossible for tex2lyx to determine the correct encoding.\n"
2189 << " The iconv encoding " << p.getEncoding() << " was used.\n"
2190 << " If this is incorrect, you must run the tex2lyx program on the command line\n"
2191 << " and specify the encoding using the -e command-line switch.\n"
2192 << " In addition, you might want to double check that the desired output encoding\n"
2193 << " is correctly selected in Document > Settings > Language.\n"
2194 << "\\end_layout\n";
2196 is_nonCJKJapanese = false;
2200 debugToken(cerr, t, flags);
2203 if (flags & FLAG_ITEM) {
2204 if (t.cat() == catSpace)
2207 flags &= ~FLAG_ITEM;
2208 if (t.cat() == catBegin) {
2209 // skip the brace and collect everything to the next matching
2211 flags |= FLAG_BRACE_LAST;
2215 // handle only this single token, leave the loop if done
2216 flags |= FLAG_LEAVE;
2219 if (t.cat() != catEscape && t.character() == ']' &&
2220 (flags & FLAG_BRACK_LAST))
2222 if (t.cat() == catEnd && (flags & FLAG_BRACE_LAST))
2225 // If there is anything between \end{env} and \begin{env} we
2226 // don't need to output a separator.
2227 if (t.cat() != catSpace && t.cat() != catNewline &&
2228 t.asInput() != "\\begin")
2234 if (t.cat() == catMath) {
2235 // we are inside some text mode thingy, so opening new math is allowed
2236 context.check_layout(os);
2237 begin_inset(os, "Formula ");
2238 Token const & n = p.get_token();
2239 bool const display(n.cat() == catMath && outer);
2241 // TeX's $$...$$ syntax for displayed math
2243 parse_math(p, os, FLAG_SIMPLE, MATH_MODE);
2245 p.get_token(); // skip the second '$' token
2247 // simple $...$ stuff
2250 parse_math(p, os, FLAG_SIMPLE, MATH_MODE);
2255 // Prevent the conversion of a line break to a
2256 // space (bug 7668). This does not change the
2257 // output, but looks ugly in LyX.
2258 eat_whitespace(p, os, context, false);
2262 else if (t.cat() == catSuper || t.cat() == catSub)
2263 cerr << "catcode " << t << " illegal in text mode\n";
2265 // Basic support for english quotes. This should be
2266 // extended to other quotes, but is not so easy (a
2267 // left english quote is the same as a right german
2269 else if (t.asInput() == "`" && p.next_token().asInput() == "`") {
2270 context.check_layout(os);
2271 begin_inset(os, "Quotes ");
2277 else if (t.asInput() == "'" && p.next_token().asInput() == "'") {
2278 context.check_layout(os);
2279 begin_inset(os, "Quotes ");
2286 else if (t.asInput() == ">" && p.next_token().asInput() == ">") {
2287 context.check_layout(os);
2288 begin_inset(os, "Quotes ");
2295 else if (t.asInput() == "<"
2296 && p.next_token().asInput() == "<") {
2297 bool has_noweb = false;
2301 has_noweb = parse_noweb(p, os, context);
2307 context.check_layout(os);
2308 begin_inset(os, "Quotes ");
2309 //FIXME: this is a right danish quote;
2310 // why not a left french quote?
2318 else if (t.cat() == catSpace || (t.cat() == catNewline && ! p.isParagraph()))
2319 check_space(p, os, context);
2321 else if (t.character() == '[' && noweb_mode &&
2322 p.next_token().character() == '[') {
2323 // These can contain underscores
2325 string const s = p.getFullOpt() + ']';
2326 if (p.next_token().character() == ']')
2329 cerr << "Warning: Inserting missing ']' in '"
2330 << s << "'." << endl;
2331 output_ert_inset(os, s, context);
2334 else if (t.cat() == catLetter) {
2335 context.check_layout(os);
2336 // Workaround for bug 4752.
2337 // FIXME: This whole code block needs to be removed
2338 // when the bug is fixed and tex2lyx produces
2339 // the updated file format.
2340 // The replacement algorithm in LyX is so stupid that
2341 // it even translates a phrase if it is part of a word.
2342 bool handled = false;
2343 for (int const * l = known_phrase_lengths; *l; ++l) {
2344 string phrase = t.cs();
2345 for (int i = 1; i < *l && p.next_token().isAlnumASCII(); ++i)
2346 phrase += p.get_token().cs();
2347 if (is_known(phrase, known_coded_phrases)) {
2348 output_ert_inset(os, phrase, context);
2352 for (size_t i = 1; i < phrase.length(); ++i)
2360 else if (t.cat() == catOther ||
2361 t.cat() == catAlign ||
2362 t.cat() == catParameter) {
2363 // This translates "&" to "\\&" which may be wrong...
2364 context.check_layout(os);
2368 else if (p.isParagraph()) {
2369 if (context.new_layout_allowed)
2370 context.new_paragraph(os);
2372 output_ert_inset(os, "\\par ", context);
2373 eat_whitespace(p, os, context, true);
2376 else if (t.cat() == catActive) {
2377 context.check_layout(os);
2378 if (t.character() == '~') {
2379 if (context.layout->free_spacing)
2382 begin_inset(os, "space ~\n");
2389 else if (t.cat() == catBegin) {
2390 Token const next = p.next_token();
2391 Token const end = p.next_next_token();
2392 if (next.cat() == catEnd) {
2394 Token const prev = p.prev_token();
2396 if (p.next_token().character() == '`' ||
2397 (prev.character() == '-' &&
2398 p.next_token().character() == '-'))
2399 ; // ignore it in {}`` or -{}-
2401 output_ert_inset(os, "{}", context);
2402 } else if (next.cat() == catEscape &&
2403 is_known(next.cs(), known_quotes) &&
2404 end.cat() == catEnd) {
2405 // Something like {\textquoteright} (e.g.
2406 // from writer2latex). LyX writes
2407 // \textquoteright{}, so we may skip the
2408 // braces here for better readability.
2409 parse_text_snippet(p, os, FLAG_BRACE_LAST,
2411 } else if (p.next_token().asInput() == "\\ascii") {
2412 // handle the \ascii characters
2413 // (the case without braces is handled later)
2414 // the code is "{\ascii\xxx}"
2415 p.get_token(); // eat \ascii
2416 string name2 = p.get_token().asInput();
2417 p.get_token(); // eat the final '}'
2418 string const name = "{\\ascii" + name2 + "}";
2422 // get the character from unicodesymbols
2423 docstring s = encodings.fromLaTeXCommand(from_utf8(name),
2424 Encodings::TEXT_CMD, termination, rem, &req);
2426 context.check_layout(os);
2429 output_ert_inset(os,
2430 to_utf8(rem), context);
2432 // we did not find a non-ert version
2433 output_ert_inset(os, name, context);
2435 context.check_layout(os);
2436 // special handling of font attribute changes
2437 Token const prev = p.prev_token();
2438 TeXFont const oldFont = context.font;
2439 if (next.character() == '[' ||
2440 next.character() == ']' ||
2441 next.character() == '*') {
2443 if (p.next_token().cat() == catEnd) {
2448 output_ert_inset(os, "{", context);
2449 parse_text_snippet(p, os,
2452 output_ert_inset(os, "}", context);
2454 } else if (! context.new_layout_allowed) {
2455 output_ert_inset(os, "{", context);
2456 parse_text_snippet(p, os, FLAG_BRACE_LAST,
2458 output_ert_inset(os, "}", context);
2459 } else if (is_known(next.cs(), known_sizes)) {
2460 // next will change the size, so we must
2462 parse_text_snippet(p, os, FLAG_BRACE_LAST,
2464 if (!context.atParagraphStart())
2466 << context.font.size << "\n";
2467 } else if (is_known(next.cs(), known_font_families)) {
2468 // next will change the font family, so we
2469 // must reset it here
2470 parse_text_snippet(p, os, FLAG_BRACE_LAST,
2472 if (!context.atParagraphStart())
2474 << context.font.family << "\n";
2475 } else if (is_known(next.cs(), known_font_series)) {
2476 // next will change the font series, so we
2477 // must reset it here
2478 parse_text_snippet(p, os, FLAG_BRACE_LAST,
2480 if (!context.atParagraphStart())
2482 << context.font.series << "\n";
2483 } else if (is_known(next.cs(), known_font_shapes)) {
2484 // next will change the font shape, so we
2485 // must reset it here
2486 parse_text_snippet(p, os, FLAG_BRACE_LAST,
2488 if (!context.atParagraphStart())
2490 << context.font.shape << "\n";
2491 } else if (is_known(next.cs(), known_old_font_families) ||
2492 is_known(next.cs(), known_old_font_series) ||
2493 is_known(next.cs(), known_old_font_shapes)) {
2494 // next will change the font family, series
2495 // and shape, so we must reset it here
2496 parse_text_snippet(p, os, FLAG_BRACE_LAST,
2498 if (!context.atParagraphStart())
2500 << context.font.family
2502 << context.font.series
2504 << context.font.shape << "\n";
2506 output_ert_inset(os, "{", context);
2507 parse_text_snippet(p, os, FLAG_BRACE_LAST,
2509 output_ert_inset(os, "}", context);
2514 else if (t.cat() == catEnd) {
2515 if (flags & FLAG_BRACE_LAST) {
2518 cerr << "stray '}' in text\n";
2519 output_ert_inset(os, "}", context);
2522 else if (t.cat() == catComment)
2523 parse_comment(p, os, t, context);
2526 // control sequences
2529 else if (t.cs() == "(") {
2530 context.check_layout(os);
2531 begin_inset(os, "Formula");
2533 parse_math(p, os, FLAG_SIMPLE2, MATH_MODE);
2538 else if (t.cs() == "[") {
2539 context.check_layout(os);
2540 begin_inset(os, "Formula");
2542 parse_math(p, os, FLAG_EQUATION, MATH_MODE);
2545 // Prevent the conversion of a line break to a space
2546 // (bug 7668). This does not change the output, but
2547 // looks ugly in LyX.
2548 eat_whitespace(p, os, context, false);
2551 else if (t.cs() == "begin")
2552 parse_environment(p, os, outer, last_env,
2555 else if (t.cs() == "end") {
2556 if (flags & FLAG_END) {
2557 // eat environment name
2558 string const name = p.getArg('{', '}');
2559 if (name != active_environment())
2560 cerr << "\\end{" + name + "} does not match \\begin{"
2561 + active_environment() + "}\n";
2564 p.error("found 'end' unexpectedly");
2567 else if (t.cs() == "item") {
2569 bool const optarg = p.hasOpt();
2571 // FIXME: This swallows comments, but we cannot use
2572 // eat_whitespace() since we must not output
2573 // anything before the item.
2574 p.skip_spaces(true);
2575 s = p.verbatimOption();
2577 p.skip_spaces(false);
2579 context.check_layout(os);
2580 if (context.has_item) {
2581 // An item in an unknown list-like environment
2582 // FIXME: Do this in check_layout()!
2583 context.has_item = false;
2585 output_ert_inset(os, "\\item", context);
2587 output_ert_inset(os, "\\item ", context);
2590 if (context.layout->labeltype != LABEL_MANUAL) {
2591 // LyX does not support \item[\mybullet]
2592 // in itemize environments
2594 os << parse_text_snippet(p2,
2595 FLAG_BRACK_LAST, outer, context);
2596 } else if (!s.empty()) {
2597 // LyX adds braces around the argument,
2598 // so we need to remove them here.
2599 if (s.size() > 2 && s[0] == '{' &&
2600 s[s.size()-1] == '}')
2601 s = s.substr(1, s.size()-2);
2602 // If the argument contains a space we
2603 // must put it into ERT: Otherwise LyX
2604 // would misinterpret the space as
2605 // item delimiter (bug 7663)
2606 if (contains(s, ' ')) {
2607 output_ert_inset(os, s, context);
2610 os << parse_text_snippet(p2,
2614 // The space is needed to separate the
2615 // item from the rest of the sentence.
2617 eat_whitespace(p, os, context, false);
2622 else if (t.cs() == "bibitem") {
2624 context.check_layout(os);
2625 eat_whitespace(p, os, context, false);
2626 string label = convert_command_inset_arg(p.verbatimOption());
2627 string key = convert_command_inset_arg(p.verbatim_item());
2628 if (contains(label, '\\') || contains(key, '\\')) {
2629 // LyX can't handle LaTeX commands in labels or keys
2630 output_ert_inset(os, t.asInput() + '[' + label +
2631 "]{" + p.verbatim_item() + '}',
2634 begin_command_inset(os, "bibitem", "bibitem");
2635 os << "label \"" << label << "\"\n"
2636 "key \"" << key << "\"\n";
2641 else if (is_macro(p)) {
2642 // catch the case of \def\inputGnumericTable
2644 if (t.cs() == "def") {
2645 Token second = p.next_token();
2646 if (second.cs() == "inputGnumericTable") {
2650 Token third = p.get_token();
2652 if (third.cs() == "input") {
2656 string name = normalize_filename(p.verbatim_item());
2657 string const path = getMasterFilePath(true);
2658 // We want to preserve relative / absolute filenames,
2659 // therefore path is only used for testing
2660 // The file extension is in every case ".tex".
2661 // So we need to remove this extension and check for
2662 // the original one.
2663 name = removeExtension(name);
2664 if (!makeAbsPath(name, path).exists()) {
2665 char const * const Gnumeric_formats[] = {"gnumeric",
2667 string const Gnumeric_name =
2668 find_file(name, path, Gnumeric_formats);
2669 if (!Gnumeric_name.empty())
2670 name = Gnumeric_name;
2672 FileName const absname = makeAbsPath(name, path);
2673 if (absname.exists()) {
2674 fix_child_filename(name);
2675 copy_file(absname, name);
2677 cerr << "Warning: Could not find file '"
2678 << name << "'." << endl;
2679 context.check_layout(os);
2680 begin_inset(os, "External\n\ttemplate ");
2681 os << "GnumericSpreadsheet\n\tfilename "
2684 context.check_layout(os);
2686 // register the packages that are automatically loaded
2687 // by the Gnumeric template
2688 registerExternalTemplatePackages("GnumericSpreadsheet");
2693 parse_macro(p, os, context);
2696 else if (t.cs() == "noindent") {
2698 context.add_par_extra_stuff("\\noindent\n");
2701 else if (t.cs() == "appendix") {
2702 context.add_par_extra_stuff("\\start_of_appendix\n");
2703 // We need to start a new paragraph. Otherwise the
2704 // appendix in 'bla\appendix\chapter{' would start
2706 context.new_paragraph(os);
2707 // We need to make sure that the paragraph is
2708 // generated even if it is empty. Otherwise the
2709 // appendix in '\par\appendix\par\chapter{' would
2711 context.check_layout(os);
2712 // FIXME: This is a hack to prevent paragraph
2713 // deletion if it is empty. Handle this better!
2714 output_ert_inset(os,
2715 "%dummy comment inserted by tex2lyx to "
2716 "ensure that this paragraph is not empty",
2718 // Both measures above may generate an additional
2719 // empty paragraph, but that does not hurt, because
2720 // whitespace does not matter here.
2721 eat_whitespace(p, os, context, true);
2724 // Must catch empty dates before findLayout is called below
2725 else if (t.cs() == "date") {
2726 eat_whitespace(p, os, context, false);
2728 string const date = p.verbatim_item();
2731 preamble.suppressDate(true);
2734 preamble.suppressDate(false);
2735 if (context.new_layout_allowed &&
2736 (newlayout = findLayout(context.textclass,
2739 output_command_layout(os, p, outer,
2740 context, newlayout);
2741 parse_text_snippet(p, os, FLAG_ITEM, outer, context);
2742 if (!preamble.titleLayoutFound())
2743 preamble.titleLayoutFound(newlayout->intitle);
2744 set<string> const & req = newlayout->requires();
2745 set<string>::const_iterator it = req.begin();
2746 set<string>::const_iterator en = req.end();
2747 for (; it != en; ++it)
2748 preamble.registerAutomaticallyLoadedPackage(*it);
2750 output_ert_inset(os,
2751 "\\date{" + p.verbatim_item() + '}',
2756 // Starred section headings
2757 // Must attempt to parse "Section*" before "Section".
2758 else if ((p.next_token().asInput() == "*") &&
2759 context.new_layout_allowed &&
2760 (newlayout = findLayout(context.textclass, t.cs() + '*', true))) {
2763 output_command_layout(os, p, outer, context, newlayout);
2765 if (!preamble.titleLayoutFound())
2766 preamble.titleLayoutFound(newlayout->intitle);
2767 set<string> const & req = newlayout->requires();
2768 for (set<string>::const_iterator it = req.begin(); it != req.end(); ++it)
2769 preamble.registerAutomaticallyLoadedPackage(*it);
2772 // Section headings and the like
2773 else if (context.new_layout_allowed &&
2774 (newlayout = findLayout(context.textclass, t.cs(), true))) {
2776 output_command_layout(os, p, outer, context, newlayout);
2778 if (!preamble.titleLayoutFound())
2779 preamble.titleLayoutFound(newlayout->intitle);
2780 set<string> const & req = newlayout->requires();
2781 for (set<string>::const_iterator it = req.begin(); it != req.end(); ++it)
2782 preamble.registerAutomaticallyLoadedPackage(*it);
2785 else if (t.cs() == "caption") {
2787 context.check_layout(os);
2789 begin_inset(os, "Caption Standard\n");
2790 Context newcontext(true, context.textclass, 0, 0, context.font);
2791 newcontext.check_layout(os);
2792 // FIXME InsetArgument is now properly implemented in InsetLayout
2793 // (for captions, but also for others)
2794 if (p.next_token().cat() != catEscape &&
2795 p.next_token().character() == '[') {
2796 p.get_token(); // eat '['
2797 begin_inset(os, "Argument 1\n");
2798 os << "status collapsed\n";
2799 parse_text_in_inset(p, os, FLAG_BRACK_LAST, outer, context);
2801 eat_whitespace(p, os, context, false);
2803 parse_text(p, os, FLAG_ITEM, outer, context);
2804 context.check_end_layout(os);
2805 // We don't need really a new paragraph, but
2806 // we must make sure that the next item gets a \begin_layout.
2807 context.new_paragraph(os);
2810 newcontext.check_end_layout(os);
2813 else if (t.cs() == "subfloat") {
2814 // the syntax is \subfloat[caption]{content}
2815 // if it is a table of figure depends on the surrounding float
2816 bool has_caption = false;
2818 // do nothing if there is no outer float
2819 if (!float_type.empty()) {
2820 context.check_layout(os);
2822 begin_inset(os, "Float " + float_type + "\n");
2824 << "\nsideways false"
2825 << "\nstatus collapsed\n\n";
2828 if (p.next_token().cat() != catEscape &&
2829 p.next_token().character() == '[') {
2830 p.get_token(); // eat '['
2831 caption = parse_text_snippet(p, FLAG_BRACK_LAST, outer, context);
2835 parse_text_in_inset(p, os, FLAG_ITEM, outer, context);
2836 // the caption comes always as the last
2838 // we must make sure that the caption gets a \begin_layout
2839 os << "\n\\begin_layout Plain Layout";
2841 begin_inset(os, "Caption Standard\n");
2842 Context newcontext(true, context.textclass,
2843 0, 0, context.font);
2844 newcontext.check_layout(os);
2845 os << caption << "\n";
2846 newcontext.check_end_layout(os);
2847 // We don't need really a new paragraph, but
2848 // we must make sure that the next item gets a \begin_layout.
2849 //newcontext.new_paragraph(os);
2853 // We don't need really a new paragraph, but
2854 // we must make sure that the next item gets a \begin_layout.
2856 context.new_paragraph(os);
2859 context.check_end_layout(os);
2860 // close the layout we opened
2862 os << "\n\\end_layout\n";
2864 // if the float type is not supported or there is no surrounding float
2867 string opt_arg = convert_command_inset_arg(p.getArg('[', ']'));
2868 output_ert_inset(os, t.asInput() + '[' + opt_arg +
2869 "]{" + p.verbatim_item() + '}', context);
2871 output_ert_inset(os, t.asInput() + "{" + p.verbatim_item() + '}', context);
2875 else if (t.cs() == "includegraphics") {
2876 bool const clip = p.next_token().asInput() == "*";
2879 string const arg = p.getArg('[', ']');
2880 map<string, string> opts;
2881 vector<string> keys;
2882 split_map(arg, opts, keys);
2884 opts["clip"] = string();
2885 string name = normalize_filename(p.verbatim_item());
2887 string const path = getMasterFilePath(true);
2888 // We want to preserve relative / absolute filenames,
2889 // therefore path is only used for testing
2890 if (!makeAbsPath(name, path).exists()) {
2891 // The file extension is probably missing.
2892 // Now try to find it out.
2893 string const dvips_name =
2894 find_file(name, path,
2895 known_dvips_graphics_formats);
2896 string const pdftex_name =
2897 find_file(name, path,
2898 known_pdftex_graphics_formats);
2899 if (!dvips_name.empty()) {
2900 if (!pdftex_name.empty()) {
2901 cerr << "This file contains the "
2903 "\"\\includegraphics{"
2905 "However, files\n\""
2906 << dvips_name << "\" and\n\""
2907 << pdftex_name << "\"\n"
2908 "both exist, so I had to make a "
2909 "choice and took the first one.\n"
2910 "Please move the unwanted one "
2911 "someplace else and try again\n"
2912 "if my choice was wrong."
2916 } else if (!pdftex_name.empty()) {
2922 FileName const absname = makeAbsPath(name, path);
2923 if (absname.exists()) {
2924 fix_child_filename(name);
2925 copy_file(absname, name);
2927 cerr << "Warning: Could not find graphics file '"
2928 << name << "'." << endl;
2930 context.check_layout(os);
2931 begin_inset(os, "Graphics ");
2932 os << "\n\tfilename " << name << '\n';
2933 if (opts.find("width") != opts.end())
2935 << translate_len(opts["width"]) << '\n';
2936 if (opts.find("height") != opts.end())
2938 << translate_len(opts["height"]) << '\n';
2939 if (opts.find("scale") != opts.end()) {
2940 istringstream iss(opts["scale"]);
2944 os << "\tscale " << val << '\n';
2946 if (opts.find("angle") != opts.end()) {
2947 os << "\trotateAngle "
2948 << opts["angle"] << '\n';
2949 vector<string>::const_iterator a =
2950 find(keys.begin(), keys.end(), "angle");
2951 vector<string>::const_iterator s =
2952 find(keys.begin(), keys.end(), "width");
2953 if (s == keys.end())
2954 s = find(keys.begin(), keys.end(), "height");
2955 if (s == keys.end())
2956 s = find(keys.begin(), keys.end(), "scale");
2957 if (s != keys.end() && distance(s, a) > 0)
2958 os << "\tscaleBeforeRotation\n";
2960 if (opts.find("origin") != opts.end()) {
2962 string const opt = opts["origin"];
2963 if (opt.find('l') != string::npos) ss << "left";
2964 if (opt.find('r') != string::npos) ss << "right";
2965 if (opt.find('c') != string::npos) ss << "center";
2966 if (opt.find('t') != string::npos) ss << "Top";
2967 if (opt.find('b') != string::npos) ss << "Bottom";
2968 if (opt.find('B') != string::npos) ss << "Baseline";
2969 if (!ss.str().empty())
2970 os << "\trotateOrigin " << ss.str() << '\n';
2972 cerr << "Warning: Ignoring unknown includegraphics origin argument '" << opt << "'\n";
2974 if (opts.find("keepaspectratio") != opts.end())
2975 os << "\tkeepAspectRatio\n";
2976 if (opts.find("clip") != opts.end())
2978 if (opts.find("draft") != opts.end())
2980 if (opts.find("bb") != opts.end())
2981 os << "\tBoundingBox "
2982 << opts["bb"] << '\n';
2983 int numberOfbbOptions = 0;
2984 if (opts.find("bbllx") != opts.end())
2985 numberOfbbOptions++;
2986 if (opts.find("bblly") != opts.end())
2987 numberOfbbOptions++;
2988 if (opts.find("bburx") != opts.end())
2989 numberOfbbOptions++;
2990 if (opts.find("bbury") != opts.end())
2991 numberOfbbOptions++;
2992 if (numberOfbbOptions == 4)
2993 os << "\tBoundingBox "
2994 << opts["bbllx"] << " " << opts["bblly"] << " "
2995 << opts["bburx"] << " " << opts["bbury"] << '\n';
2996 else if (numberOfbbOptions > 0)
2997 cerr << "Warning: Ignoring incomplete includegraphics boundingbox arguments.\n";
2998 numberOfbbOptions = 0;
2999 if (opts.find("natwidth") != opts.end())
3000 numberOfbbOptions++;
3001 if (opts.find("natheight") != opts.end())
3002 numberOfbbOptions++;
3003 if (numberOfbbOptions == 2)
3004 os << "\tBoundingBox 0bp 0bp "
3005 << opts["natwidth"] << " " << opts["natheight"] << '\n';
3006 else if (numberOfbbOptions > 0)
3007 cerr << "Warning: Ignoring incomplete includegraphics boundingbox arguments.\n";
3008 ostringstream special;
3009 if (opts.find("hiresbb") != opts.end())
3010 special << "hiresbb,";
3011 if (opts.find("trim") != opts.end())
3013 if (opts.find("viewport") != opts.end())
3014 special << "viewport=" << opts["viewport"] << ',';
3015 if (opts.find("totalheight") != opts.end())
3016 special << "totalheight=" << opts["totalheight"] << ',';
3017 if (opts.find("type") != opts.end())
3018 special << "type=" << opts["type"] << ',';
3019 if (opts.find("ext") != opts.end())
3020 special << "ext=" << opts["ext"] << ',';
3021 if (opts.find("read") != opts.end())
3022 special << "read=" << opts["read"] << ',';
3023 if (opts.find("command") != opts.end())
3024 special << "command=" << opts["command"] << ',';
3025 string s_special = special.str();
3026 if (!s_special.empty()) {
3027 // We had special arguments. Remove the trailing ','.
3028 os << "\tspecial " << s_special.substr(0, s_special.size() - 1) << '\n';
3030 // TODO: Handle the unknown settings better.
3031 // Warn about invalid options.
3032 // Check whether some option was given twice.
3034 preamble.registerAutomaticallyLoadedPackage("graphicx");
3037 else if (t.cs() == "footnote" ||
3038 (t.cs() == "thanks" && context.layout->intitle)) {
3040 context.check_layout(os);
3041 begin_inset(os, "Foot\n");
3042 os << "status collapsed\n\n";
3043 parse_text_in_inset(p, os, FLAG_ITEM, false, context);
3047 else if (t.cs() == "marginpar") {
3049 context.check_layout(os);
3050 begin_inset(os, "Marginal\n");
3051 os << "status collapsed\n\n";
3052 parse_text_in_inset(p, os, FLAG_ITEM, false, context);
3056 else if (t.cs() == "lstinline") {
3058 parse_listings(p, os, context, true);
3061 else if (t.cs() == "ensuremath") {
3063 context.check_layout(os);
3064 string const s = p.verbatim_item();
3065 //FIXME: this never triggers in UTF8
3066 if (s == "\xb1" || s == "\xb3" || s == "\xb2" || s == "\xb5")
3069 output_ert_inset(os, "\\ensuremath{" + s + "}",
3073 else if (t.cs() == "makeindex" || t.cs() == "maketitle") {
3074 if (preamble.titleLayoutFound()) {
3076 skip_spaces_braces(p);
3078 output_ert_inset(os, t.asInput(), context);
3081 else if (t.cs() == "tableofcontents" || t.cs() == "lstlistoflistings") {
3082 context.check_layout(os);
3083 begin_command_inset(os, "toc", t.cs());
3085 skip_spaces_braces(p);
3086 if (t.cs() == "lstlistoflistings")
3087 preamble.registerAutomaticallyLoadedPackage("listings");
3090 else if (t.cs() == "listoffigures") {
3091 context.check_layout(os);
3092 begin_inset(os, "FloatList figure\n");
3094 skip_spaces_braces(p);
3097 else if (t.cs() == "listoftables") {
3098 context.check_layout(os);
3099 begin_inset(os, "FloatList table\n");
3101 skip_spaces_braces(p);
3104 else if (t.cs() == "listof") {
3105 p.skip_spaces(true);
3106 string const name = p.get_token().cs();
3107 if (context.textclass.floats().typeExist(name)) {
3108 context.check_layout(os);
3109 begin_inset(os, "FloatList ");
3112 p.get_token(); // swallow second arg
3114 output_ert_inset(os, "\\listof{" + name + "}", context);
3117 else if ((where = is_known(t.cs(), known_text_font_families)))
3118 parse_text_attributes(p, os, FLAG_ITEM, outer,
3119 context, "\\family", context.font.family,
3120 known_coded_font_families[where - known_text_font_families]);
3122 else if ((where = is_known(t.cs(), known_text_font_series)))
3123 parse_text_attributes(p, os, FLAG_ITEM, outer,
3124 context, "\\series", context.font.series,
3125 known_coded_font_series[where - known_text_font_series]);
3127 else if ((where = is_known(t.cs(), known_text_font_shapes)))
3128 parse_text_attributes(p, os, FLAG_ITEM, outer,
3129 context, "\\shape", context.font.shape,
3130 known_coded_font_shapes[where - known_text_font_shapes]);
3132 else if (t.cs() == "textnormal" || t.cs() == "normalfont") {
3133 context.check_layout(os);
3134 TeXFont oldFont = context.font;
3135 context.font.init();
3136 context.font.size = oldFont.size;
3137 os << "\n\\family " << context.font.family << "\n";
3138 os << "\n\\series " << context.font.series << "\n";
3139 os << "\n\\shape " << context.font.shape << "\n";
3140 if (t.cs() == "textnormal") {
3141 parse_text_snippet(p, os, FLAG_ITEM, outer, context);
3142 output_font_change(os, context.font, oldFont);
3143 context.font = oldFont;
3145 eat_whitespace(p, os, context, false);
3148 else if (t.cs() == "textcolor") {
3149 // scheme is \textcolor{color name}{text}
3150 string const color = p.verbatim_item();
3151 // we only support the predefined colors of the color package
3152 if (color == "black" || color == "blue" || color == "cyan"
3153 || color == "green" || color == "magenta" || color == "red"
3154 || color == "white" || color == "yellow") {
3155 context.check_layout(os);
3156 os << "\n\\color " << color << "\n";
3157 parse_text_snippet(p, os, FLAG_ITEM, outer, context);
3158 context.check_layout(os);
3159 os << "\n\\color inherit\n";
3160 preamble.registerAutomaticallyLoadedPackage("color");
3162 // for custom defined colors
3163 output_ert_inset(os, t.asInput() + "{" + color + "}", context);
3166 else if (t.cs() == "underbar" || t.cs() == "uline") {
3167 // \underbar is not 100% correct (LyX outputs \uline
3168 // of ulem.sty). The difference is that \ulem allows
3169 // line breaks, and \underbar does not.
3170 // Do NOT handle \underline.
3171 // \underbar cuts through y, g, q, p etc.,
3172 // \underline does not.
3173 context.check_layout(os);
3174 os << "\n\\bar under\n";
3175 parse_text_snippet(p, os, FLAG_ITEM, outer, context);
3176 context.check_layout(os);
3177 os << "\n\\bar default\n";
3178 preamble.registerAutomaticallyLoadedPackage("ulem");
3181 else if (t.cs() == "sout") {
3182 context.check_layout(os);
3183 os << "\n\\strikeout on\n";
3184 parse_text_snippet(p, os, FLAG_ITEM, outer, context);
3185 context.check_layout(os);
3186 os << "\n\\strikeout default\n";
3187 preamble.registerAutomaticallyLoadedPackage("ulem");
3190 else if (t.cs() == "uuline" || t.cs() == "uwave" ||
3191 t.cs() == "emph" || t.cs() == "noun") {
3192 context.check_layout(os);
3193 os << "\n\\" << t.cs() << " on\n";
3194 parse_text_snippet(p, os, FLAG_ITEM, outer, context);
3195 context.check_layout(os);
3196 os << "\n\\" << t.cs() << " default\n";
3197 if (t.cs() == "uuline" || t.cs() == "uwave")
3198 preamble.registerAutomaticallyLoadedPackage("ulem");
3201 else if (t.cs() == "lyxadded" || t.cs() == "lyxdeleted") {
3202 context.check_layout(os);
3203 string name = p.getArg('{', '}');
3204 string localtime = p.getArg('{', '}');
3205 preamble.registerAuthor(name);
3206 Author const & author = preamble.getAuthor(name);
3207 // from_asctime_utc() will fail if LyX decides to output the
3208 // time in the text language.
3209 time_t ptime = from_asctime_utc(localtime);
3210 if (ptime == static_cast<time_t>(-1)) {
3211 cerr << "Warning: Could not parse time `" << localtime
3212 << "´ for change tracking, using current time instead.\n";
3213 ptime = current_time();
3215 if (t.cs() == "lyxadded")
3216 os << "\n\\change_inserted ";
3218 os << "\n\\change_deleted ";
3219 os << author.bufferId() << ' ' << ptime << '\n';
3220 parse_text_snippet(p, os, FLAG_ITEM, outer, context);
3221 bool dvipost = LaTeXPackages::isAvailable("dvipost");
3222 bool xcolorulem = LaTeXPackages::isAvailable("ulem") &&
3223 LaTeXPackages::isAvailable("xcolor");
3224 // No need to test for luatex, since luatex comes in
3225 // two flavours (dvi and pdf), like latex, and those
3226 // are detected by pdflatex.
3227 if (pdflatex || xetex) {
3229 preamble.registerAutomaticallyLoadedPackage("ulem");
3230 preamble.registerAutomaticallyLoadedPackage("xcolor");
3231 preamble.registerAutomaticallyLoadedPackage("pdfcolmk");
3235 preamble.registerAutomaticallyLoadedPackage("dvipost");
3236 } else if (xcolorulem) {
3237 preamble.registerAutomaticallyLoadedPackage("ulem");
3238 preamble.registerAutomaticallyLoadedPackage("xcolor");
3243 else if (t.cs() == "textipa") {
3244 context.check_layout(os);
3245 begin_inset(os, "IPA\n");
3246 parse_text_in_inset(p, os, FLAG_ITEM, outer, context);
3248 preamble.registerAutomaticallyLoadedPackage("tipa");
3249 preamble.registerAutomaticallyLoadedPackage("tipx");
3252 else if (t.cs() == "texttoptiebar" || t.cs() == "textbottomtiebar") {
3253 context.check_layout(os);
3254 begin_inset(os, "IPADeco " + t.cs().substr(4) + "\n");
3255 os << "status open\n";
3256 parse_text_in_inset(p, os, FLAG_ITEM, outer, context);
3261 else if (t.cs() == "textvertline") {
3262 // FIXME: This is not correct, \textvertline is higher than |
3268 else if (t.cs() == "tone" ) {
3269 context.check_layout(os);
3270 // register the tone package
3271 preamble.registerAutomaticallyLoadedPackage("tone");
3272 string content = trimSpaceAndEol(p.verbatim_item());
3273 string command = t.asInput() + "{" + content + "}";
3274 // some tones can be detected by unicodesymbols, some need special code
3275 if (is_known(content, known_tones)) {
3276 os << "\\IPAChar " << command << "\n";
3279 // try to see whether the string is in unicodesymbols
3283 docstring s = encodings.fromLaTeXCommand(from_utf8(command),
3284 Encodings::TEXT_CMD | Encodings::MATH_CMD,
3285 termination, rem, &req);
3289 output_ert_inset(os, to_utf8(rem), context);
3291 // we did not find a non-ert version
3292 output_ert_inset(os, command, context);
3295 else if (t.cs() == "phantom" || t.cs() == "hphantom" ||
3296 t.cs() == "vphantom") {
3297 context.check_layout(os);
3298 if (t.cs() == "phantom")
3299 begin_inset(os, "Phantom Phantom\n");
3300 if (t.cs() == "hphantom")
3301 begin_inset(os, "Phantom HPhantom\n");
3302 if (t.cs() == "vphantom")
3303 begin_inset(os, "Phantom VPhantom\n");
3304 os << "status open\n";
3305 parse_text_in_inset(p, os, FLAG_ITEM, outer, context,
3310 else if (t.cs() == "href") {
3311 context.check_layout(os);
3312 string target = convert_command_inset_arg(p.verbatim_item());
3313 string name = convert_command_inset_arg(p.verbatim_item());
3315 size_t i = target.find(':');
3316 if (i != string::npos) {
3317 type = target.substr(0, i + 1);
3318 if (type == "mailto:" || type == "file:")
3319 target = target.substr(i + 1);
3320 // handle the case that name is equal to target, except of "http://"
3321 else if (target.substr(i + 3) == name && type == "http:")
3324 begin_command_inset(os, "href", "href");
3326 os << "name \"" << name << "\"\n";
3327 os << "target \"" << target << "\"\n";
3328 if (type == "mailto:" || type == "file:")
3329 os << "type \"" << type << "\"\n";
3331 skip_spaces_braces(p);
3334 else if (t.cs() == "lyxline") {
3335 // swallow size argument (it is not used anyway)
3337 if (!context.atParagraphStart()) {
3338 // so our line is in the middle of a paragraph
3339 // we need to add a new line, lest this line
3340 // follow the other content on that line and
3341 // run off the side of the page
3342 // FIXME: This may create an empty paragraph,
3343 // but without that it would not be
3344 // possible to set noindent below.
3345 // Fortunately LaTeX does not care
3346 // about the empty paragraph.
3347 context.new_paragraph(os);
3349 if (preamble.indentParagraphs()) {
3350 // we need to unindent, lest the line be too long
3351 context.add_par_extra_stuff("\\noindent\n");
3353 context.check_layout(os);
3354 begin_command_inset(os, "line", "rule");
3355 os << "offset \"0.5ex\"\n"
3356 "width \"100line%\"\n"
3361 else if (t.cs() == "rule") {
3362 string const offset = (p.hasOpt() ? p.getArg('[', ']') : string());
3363 string const width = p.getArg('{', '}');
3364 string const thickness = p.getArg('{', '}');
3365 context.check_layout(os);
3366 begin_command_inset(os, "line", "rule");
3367 if (!offset.empty())
3368 os << "offset \"" << translate_len(offset) << "\"\n";
3369 os << "width \"" << translate_len(width) << "\"\n"
3370 "height \"" << translate_len(thickness) << "\"\n";
3374 else if (is_known(t.cs(), known_phrases) ||
3375 (t.cs() == "protect" &&
3376 p.next_token().cat() == catEscape &&
3377 is_known(p.next_token().cs(), known_phrases))) {
3378 // LyX sometimes puts a \protect in front, so we have to ignore it
3379 // FIXME: This needs to be changed when bug 4752 is fixed.
3381 t.cs() == "protect" ? p.get_token().cs() : t.cs(),
3383 context.check_layout(os);
3384 os << known_coded_phrases[where - known_phrases];
3385 skip_spaces_braces(p);
3388 // handle refstyle first to catch \eqref which can also occur
3389 // without refstyle. Only recognize these commands if
3390 // refstyle.sty was found in the preamble (otherwise \eqref
3391 // and user defined ref commands could be misdetected).
3392 else if ((where = is_known(t.cs(), known_refstyle_commands)) &&
3393 preamble.refstyle()) {
3394 context.check_layout(os);
3395 begin_command_inset(os, "ref", "formatted");
3396 os << "reference \"";
3397 os << known_refstyle_prefixes[where - known_refstyle_commands]
3399 os << convert_command_inset_arg(p.verbatim_item())
3402 preamble.registerAutomaticallyLoadedPackage("refstyle");
3405 // if refstyle is used, we must not convert \prettyref to a
3406 // formatted reference, since that would result in a refstyle command.
3407 else if ((where = is_known(t.cs(), known_ref_commands)) &&
3408 (t.cs() != "prettyref" || !preamble.refstyle())) {
3409 string const opt = p.getOpt();
3411 context.check_layout(os);
3412 begin_command_inset(os, "ref",
3413 known_coded_ref_commands[where - known_ref_commands]);
3414 os << "reference \""
3415 << convert_command_inset_arg(p.verbatim_item())
3418 if (t.cs() == "vref" || t.cs() == "vpageref")
3419 preamble.registerAutomaticallyLoadedPackage("varioref");
3420 else if (t.cs() == "prettyref")
3421 preamble.registerAutomaticallyLoadedPackage("prettyref");
3423 // LyX does not yet support optional arguments of ref commands
3424 output_ert_inset(os, t.asInput() + '[' + opt + "]{" +
3425 p.verbatim_item() + '}', context);
3429 else if (use_natbib &&
3430 is_known(t.cs(), known_natbib_commands) &&
3431 ((t.cs() != "citefullauthor" &&
3432 t.cs() != "citeyear" &&
3433 t.cs() != "citeyearpar") ||
3434 p.next_token().asInput() != "*")) {
3435 context.check_layout(os);
3436 string command = t.cs();
3437 if (p.next_token().asInput() == "*") {
3441 if (command == "citefullauthor")
3442 // alternative name for "\\citeauthor*"
3443 command = "citeauthor*";
3445 // text before the citation
3447 // text after the citation
3449 get_cite_arguments(p, true, before, after);
3451 if (command == "cite") {
3452 // \cite without optional argument means
3453 // \citet, \cite with at least one optional
3454 // argument means \citep.
3455 if (before.empty() && after.empty())
3460 if (before.empty() && after == "[]")
3461 // avoid \citet[]{a}
3463 else if (before == "[]" && after == "[]") {
3464 // avoid \citet[][]{a}
3468 // remove the brackets around after and before
3469 if (!after.empty()) {
3471 after.erase(after.length() - 1, 1);
3472 after = convert_command_inset_arg(after);
3474 if (!before.empty()) {
3476 before.erase(before.length() - 1, 1);
3477 before = convert_command_inset_arg(before);
3479 begin_command_inset(os, "citation", command);
3480 os << "after " << '"' << after << '"' << "\n";
3481 os << "before " << '"' << before << '"' << "\n";
3483 << convert_command_inset_arg(p.verbatim_item())
3486 // Need to set the cite engine if natbib is loaded by
3487 // the document class directly
3488 if (preamble.citeEngine() == "basic")
3489 preamble.citeEngine("natbib");
3492 else if (use_jurabib &&
3493 is_known(t.cs(), known_jurabib_commands) &&
3494 (t.cs() == "cite" || p.next_token().asInput() != "*")) {
3495 context.check_layout(os);
3496 string command = t.cs();
3497 if (p.next_token().asInput() == "*") {
3501 char argumentOrder = '\0';
3502 vector<string> const options =
3503 preamble.getPackageOptions("jurabib");
3504 if (find(options.begin(), options.end(),
3505 "natbiborder") != options.end())
3506 argumentOrder = 'n';
3507 else if (find(options.begin(), options.end(),
3508 "jurabiborder") != options.end())
3509 argumentOrder = 'j';
3511 // text before the citation
3513 // text after the citation
3515 get_cite_arguments(p, argumentOrder != 'j', before, after);
3517 string const citation = p.verbatim_item();
3518 if (!before.empty() && argumentOrder == '\0') {
3519 cerr << "Warning: Assuming argument order "
3520 "of jurabib version 0.6 for\n'"
3521 << command << before << after << '{'
3522 << citation << "}'.\n"
3523 "Add 'jurabiborder' to the jurabib "
3524 "package options if you used an\n"
3525 "earlier jurabib version." << endl;
3527 if (!after.empty()) {
3529 after.erase(after.length() - 1, 1);
3531 if (!before.empty()) {
3533 before.erase(before.length() - 1, 1);
3535 begin_command_inset(os, "citation", command);
3536 os << "after " << '"' << after << '"' << "\n";
3537 os << "before " << '"' << before << '"' << "\n";
3538 os << "key " << '"' << citation << '"' << "\n";
3540 // Need to set the cite engine if jurabib is loaded by
3541 // the document class directly
3542 if (preamble.citeEngine() == "basic")
3543 preamble.citeEngine("jurabib");
3546 else if (t.cs() == "cite"
3547 || t.cs() == "nocite") {
3548 context.check_layout(os);
3549 string after = convert_command_inset_arg(p.getArg('[', ']'));
3550 string key = convert_command_inset_arg(p.verbatim_item());
3551 // store the case that it is "\nocite{*}" to use it later for
3554 begin_command_inset(os, "citation", t.cs());
3555 os << "after " << '"' << after << '"' << "\n";
3556 os << "key " << '"' << key << '"' << "\n";
3558 } else if (t.cs() == "nocite")
3562 else if (t.cs() == "index" ||
3563 (t.cs() == "sindex" && preamble.use_indices() == "true")) {
3564 context.check_layout(os);
3565 string const arg = (t.cs() == "sindex" && p.hasOpt()) ?
3566 p.getArg('[', ']') : "";
3567 string const kind = arg.empty() ? "idx" : arg;
3568 begin_inset(os, "Index ");
3569 os << kind << "\nstatus collapsed\n";
3570 parse_text_in_inset(p, os, FLAG_ITEM, false, context, "Index");
3573 preamble.registerAutomaticallyLoadedPackage("splitidx");
3576 else if (t.cs() == "nomenclature") {
3577 context.check_layout(os);
3578 begin_command_inset(os, "nomenclature", "nomenclature");
3579 string prefix = convert_command_inset_arg(p.getArg('[', ']'));
3580 if (!prefix.empty())
3581 os << "prefix " << '"' << prefix << '"' << "\n";
3582 os << "symbol " << '"'
3583 << convert_command_inset_arg(p.verbatim_item());
3584 os << "\"\ndescription \""
3585 << convert_command_inset_arg(p.verbatim_item())
3588 preamble.registerAutomaticallyLoadedPackage("nomencl");
3591 else if (t.cs() == "label") {
3592 context.check_layout(os);
3593 begin_command_inset(os, "label", "label");
3595 << convert_command_inset_arg(p.verbatim_item())
3600 else if (t.cs() == "printindex" || t.cs() == "printsubindex") {
3601 context.check_layout(os);
3602 string commandname = t.cs();
3604 if (p.next_token().asInput() == "*") {
3609 begin_command_inset(os, "index_print", commandname);
3610 string const indexname = p.getArg('[', ']');
3612 if (indexname.empty())
3613 os << "type \"idx\"\n";
3615 os << "type \"" << indexname << "\"\n";
3618 skip_spaces_braces(p);
3619 preamble.registerAutomaticallyLoadedPackage("makeidx");
3620 if (preamble.use_indices() == "true")
3621 preamble.registerAutomaticallyLoadedPackage("splitidx");
3624 else if (t.cs() == "printnomenclature") {
3626 string width_type = "";
3627 context.check_layout(os);
3628 begin_command_inset(os, "nomencl_print", "printnomenclature");
3629 // case of a custom width
3631 width = p.getArg('[', ']');
3632 width = translate_len(width);
3633 width_type = "custom";
3635 // case of no custom width
3636 // the case of no custom width but the width set
3637 // via \settowidth{\nomlabelwidth}{***} cannot be supported
3638 // because the user could have set anything, not only the width
3639 // of the longest label (which would be width_type = "auto")
3640 string label = convert_command_inset_arg(p.getArg('{', '}'));
3641 if (label.empty() && width_type.empty())
3642 width_type = "none";
3643 os << "set_width \"" << width_type << "\"\n";
3644 if (width_type == "custom")
3645 os << "width \"" << width << '\"';
3647 skip_spaces_braces(p);
3648 preamble.registerAutomaticallyLoadedPackage("nomencl");
3651 else if ((t.cs() == "textsuperscript" || t.cs() == "textsubscript")) {
3652 context.check_layout(os);
3653 begin_inset(os, "script ");
3654 os << t.cs().substr(4) << '\n';
3655 parse_text_in_inset(p, os, FLAG_ITEM, false, context);
3657 if (t.cs() == "textsubscript")
3658 preamble.registerAutomaticallyLoadedPackage("subscript");
3661 else if ((where = is_known(t.cs(), known_quotes))) {
3662 context.check_layout(os);
3663 begin_inset(os, "Quotes ");
3664 os << known_coded_quotes[where - known_quotes];
3666 // LyX adds {} after the quote, so we have to eat
3667 // spaces here if there are any before a possible
3669 eat_whitespace(p, os, context, false);
3673 else if ((where = is_known(t.cs(), known_sizes)) &&
3674 context.new_layout_allowed) {
3675 context.check_layout(os);
3676 TeXFont const oldFont = context.font;
3677 context.font.size = known_coded_sizes[where - known_sizes];
3678 output_font_change(os, oldFont, context.font);
3679 eat_whitespace(p, os, context, false);
3682 else if ((where = is_known(t.cs(), known_font_families)) &&
3683 context.new_layout_allowed) {
3684 context.check_layout(os);
3685 TeXFont const oldFont = context.font;
3686 context.font.family =
3687 known_coded_font_families[where - known_font_families];
3688 output_font_change(os, oldFont, context.font);
3689 eat_whitespace(p, os, context, false);
3692 else if ((where = is_known(t.cs(), known_font_series)) &&
3693 context.new_layout_allowed) {
3694 context.check_layout(os);
3695 TeXFont const oldFont = context.font;
3696 context.font.series =
3697 known_coded_font_series[where - known_font_series];
3698 output_font_change(os, oldFont, context.font);
3699 eat_whitespace(p, os, context, false);
3702 else if ((where = is_known(t.cs(), known_font_shapes)) &&
3703 context.new_layout_allowed) {
3704 context.check_layout(os);
3705 TeXFont const oldFont = context.font;
3706 context.font.shape =
3707 known_coded_font_shapes[where - known_font_shapes];
3708 output_font_change(os, oldFont, context.font);
3709 eat_whitespace(p, os, context, false);
3711 else if ((where = is_known(t.cs(), known_old_font_families)) &&
3712 context.new_layout_allowed) {
3713 context.check_layout(os);
3714 TeXFont const oldFont = context.font;
3715 context.font.init();
3716 context.font.size = oldFont.size;
3717 context.font.family =
3718 known_coded_font_families[where - known_old_font_families];
3719 output_font_change(os, oldFont, context.font);
3720 eat_whitespace(p, os, context, false);
3723 else if ((where = is_known(t.cs(), known_old_font_series)) &&
3724 context.new_layout_allowed) {
3725 context.check_layout(os);
3726 TeXFont const oldFont = context.font;
3727 context.font.init();
3728 context.font.size = oldFont.size;
3729 context.font.series =
3730 known_coded_font_series[where - known_old_font_series];
3731 output_font_change(os, oldFont, context.font);
3732 eat_whitespace(p, os, context, false);
3735 else if ((where = is_known(t.cs(), known_old_font_shapes)) &&
3736 context.new_layout_allowed) {
3737 context.check_layout(os);
3738 TeXFont const oldFont = context.font;
3739 context.font.init();
3740 context.font.size = oldFont.size;
3741 context.font.shape =
3742 known_coded_font_shapes[where - known_old_font_shapes];
3743 output_font_change(os, oldFont, context.font);
3744 eat_whitespace(p, os, context, false);
3747 else if (t.cs() == "selectlanguage") {
3748 context.check_layout(os);
3749 // save the language for the case that a
3750 // \foreignlanguage is used
3751 context.font.language = babel2lyx(p.verbatim_item());
3752 os << "\n\\lang " << context.font.language << "\n";
3755 else if (t.cs() == "foreignlanguage") {
3756 string const lang = babel2lyx(p.verbatim_item());
3757 parse_text_attributes(p, os, FLAG_ITEM, outer,
3759 context.font.language, lang);
3762 else if (prefixIs(t.cs(), "text")
3763 && is_known(t.cs().substr(4), preamble.polyglossia_languages)) {
3764 // scheme is \textLANGUAGE{text} where LANGUAGE is in polyglossia_languages[]
3766 // We have to output the whole command if it has an option
3767 // because LyX doesn't support this yet, see bug #8214,
3768 // only if there is a single option specifying a variant, we can handle it.
3770 string langopts = p.getOpt();
3771 // check if the option contains a variant, if yes, extract it
3772 string::size_type pos_var = langopts.find("variant");
3773 string::size_type i = langopts.find(',');
3774 string::size_type k = langopts.find('=', pos_var);
3775 if (pos_var != string::npos && i == string::npos) {
3777 variant = langopts.substr(k + 1, langopts.length() - k - 2);
3778 lang = preamble.polyglossia2lyx(variant);
3779 parse_text_attributes(p, os, FLAG_ITEM, outer,
3781 context.font.language, lang);
3783 output_ert_inset(os, t.asInput() + langopts, context);
3785 lang = preamble.polyglossia2lyx(t.cs().substr(4, string::npos));
3786 parse_text_attributes(p, os, FLAG_ITEM, outer,
3788 context.font.language, lang);
3792 else if (t.cs() == "inputencoding") {
3793 // nothing to write here
3794 string const enc = subst(p.verbatim_item(), "\n", " ");
3795 p.setEncoding(enc, Encoding::inputenc);
3798 else if ((where = is_known(t.cs(), known_special_chars))) {
3799 context.check_layout(os);
3800 os << "\\SpecialChar \\"
3801 << known_coded_special_chars[where - known_special_chars]
3803 skip_spaces_braces(p);
3806 else if (t.cs() == "nobreakdash" && p.next_token().asInput() == "-") {
3807 context.check_layout(os);
3808 os << "\\SpecialChar \\nobreakdash-\n";
3812 else if (t.cs() == "textquotedbl") {
3813 context.check_layout(os);
3818 else if (t.cs() == "@" && p.next_token().asInput() == ".") {
3819 context.check_layout(os);
3820 os << "\\SpecialChar \\@.\n";
3824 else if (t.cs() == "-") {
3825 context.check_layout(os);
3826 os << "\\SpecialChar \\-\n";
3829 else if (t.cs() == "textasciitilde") {
3830 context.check_layout(os);
3832 skip_spaces_braces(p);
3835 else if (t.cs() == "textasciicircum") {
3836 context.check_layout(os);
3838 skip_spaces_braces(p);
3841 else if (t.cs() == "textbackslash") {
3842 context.check_layout(os);
3843 os << "\n\\backslash\n";
3844 skip_spaces_braces(p);
3847 else if (t.cs() == "_" || t.cs() == "&" || t.cs() == "#"
3848 || t.cs() == "$" || t.cs() == "{" || t.cs() == "}"
3850 context.check_layout(os);
3854 else if (t.cs() == "char") {
3855 context.check_layout(os);
3856 if (p.next_token().character() == '`') {
3858 if (p.next_token().cs() == "\"") {
3863 output_ert_inset(os, "\\char`", context);
3866 output_ert_inset(os, "\\char", context);
3870 else if (t.cs() == "verb") {
3871 context.check_layout(os);
3872 // set catcodes to verbatim early, just in case.
3873 p.setCatcodes(VERBATIM_CATCODES);
3874 string delim = p.get_token().asInput();
3875 Parser::Arg arg = p.verbatimStuff(delim);
3877 output_ert_inset(os, "\\verb" + delim
3878 + arg.second + delim, context);
3880 cerr << "invalid \\verb command. Skipping" << endl;
3883 // Problem: \= creates a tabstop inside the tabbing environment
3884 // and else an accent. In the latter case we really would want
3885 // \={o} instead of \= o.
3886 else if (t.cs() == "=" && (flags & FLAG_TABBING))
3887 output_ert_inset(os, t.asInput(), context);
3889 else if (t.cs() == "\\") {
3890 context.check_layout(os);
3892 output_ert_inset(os, "\\\\" + p.getOpt(), context);
3893 else if (p.next_token().asInput() == "*") {
3895 // getOpt() eats the following space if there
3896 // is no optional argument, but that is OK
3897 // here since it has no effect in the output.
3898 output_ert_inset(os, "\\\\*" + p.getOpt(), context);
3901 begin_inset(os, "Newline newline");
3906 else if (t.cs() == "newline" ||
3907 (t.cs() == "linebreak" && !p.hasOpt())) {
3908 context.check_layout(os);
3909 begin_inset(os, "Newline ");
3912 skip_spaces_braces(p);
3915 else if (t.cs() == "input" || t.cs() == "include"
3916 || t.cs() == "verbatiminput") {
3917 string name = t.cs();
3918 if (t.cs() == "verbatiminput"
3919 && p.next_token().asInput() == "*")
3920 name += p.get_token().asInput();
3921 context.check_layout(os);
3922 string filename(normalize_filename(p.getArg('{', '}')));
3923 string const path = getMasterFilePath(true);
3924 // We want to preserve relative / absolute filenames,
3925 // therefore path is only used for testing
3926 if ((t.cs() == "include" || t.cs() == "input") &&
3927 !makeAbsPath(filename, path).exists()) {
3928 // The file extension is probably missing.
3929 // Now try to find it out.
3930 string const tex_name =
3931 find_file(filename, path,
3932 known_tex_extensions);
3933 if (!tex_name.empty())
3934 filename = tex_name;
3936 bool external = false;
3938 if (makeAbsPath(filename, path).exists()) {
3939 string const abstexname =
3940 makeAbsPath(filename, path).absFileName();
3941 string const absfigname =
3942 changeExtension(abstexname, ".fig");
3943 fix_child_filename(filename);
3944 string const lyxname =
3945 changeExtension(filename, ".lyx");
3946 string const abslyxname = makeAbsPath(
3947 lyxname, getParentFilePath(false)).absFileName();
3949 if (!skipChildren())
3950 external = FileName(absfigname).exists();
3951 if (t.cs() == "input" && !skipChildren()) {
3952 string const ext = getExtension(abstexname);
3954 // Combined PS/LaTeX:
3955 // x.eps, x.pstex_t (old xfig)
3956 // x.pstex, x.pstex_t (new xfig, e.g. 3.2.5)
3957 FileName const absepsname(
3958 changeExtension(abstexname, ".eps"));
3959 FileName const abspstexname(
3960 changeExtension(abstexname, ".pstex"));
3961 bool const xfigeps =
3962 (absepsname.exists() ||
3963 abspstexname.exists()) &&
3966 // Combined PDF/LaTeX:
3967 // x.pdf, x.pdftex_t (old xfig)
3968 // x.pdf, x.pdf_t (new xfig, e.g. 3.2.5)
3969 FileName const abspdfname(
3970 changeExtension(abstexname, ".pdf"));
3971 bool const xfigpdf =
3972 abspdfname.exists() &&
3973 (ext == "pdftex_t" || ext == "pdf_t");
3977 // Combined PS/PDF/LaTeX:
3978 // x_pspdftex.eps, x_pspdftex.pdf, x.pspdftex
3979 string const absbase2(
3980 removeExtension(abstexname) + "_pspdftex");
3981 FileName const abseps2name(
3982 addExtension(absbase2, ".eps"));
3983 FileName const abspdf2name(
3984 addExtension(absbase2, ".pdf"));
3985 bool const xfigboth =
3986 abspdf2name.exists() &&
3987 abseps2name.exists() && ext == "pspdftex";
3989 xfig = xfigpdf || xfigeps || xfigboth;
3990 external = external && xfig;
3993 outname = changeExtension(filename, ".fig");
3994 FileName abssrc(changeExtension(abstexname, ".fig"));
3995 copy_file(abssrc, outname);
3997 // Don't try to convert, the result
3998 // would be full of ERT.
4000 FileName abssrc(abstexname);
4001 copy_file(abssrc, outname);
4002 } else if (t.cs() != "verbatiminput" &&
4004 tex2lyx(abstexname, FileName(abslyxname),
4007 // no need to call copy_file
4008 // tex2lyx creates the file
4011 FileName abssrc(abstexname);
4012 copy_file(abssrc, outname);
4015 cerr << "Warning: Could not find included file '"
4016 << filename << "'." << endl;
4020 begin_inset(os, "External\n");
4021 os << "\ttemplate XFig\n"
4022 << "\tfilename " << outname << '\n';
4023 registerExternalTemplatePackages("XFig");
4025 begin_command_inset(os, "include", name);
4026 outname = subst(outname, "\"", "\\\"");
4027 os << "preview false\n"
4028 "filename \"" << outname << "\"\n";
4029 if (t.cs() == "verbatiminput")
4030 preamble.registerAutomaticallyLoadedPackage("verbatim");
4035 else if (t.cs() == "bibliographystyle") {
4036 // store new bibliographystyle
4037 bibliographystyle = p.verbatim_item();
4038 // If any other command than \bibliography, \addcontentsline
4039 // and \nocite{*} follows, we need to output the style
4040 // (because it might be used by that command).
4041 // Otherwise, it will automatically be output by LyX.
4044 for (Token t2 = p.get_token(); p.good(); t2 = p.get_token()) {
4045 if (t2.cat() == catBegin)
4047 if (t2.cat() != catEscape)
4049 if (t2.cs() == "nocite") {
4050 if (p.getArg('{', '}') == "*")
4052 } else if (t2.cs() == "bibliography")
4054 else if (t2.cs() == "phantomsection") {
4058 else if (t2.cs() == "addcontentsline") {
4059 // get the 3 arguments of \addcontentsline
4062 contentslineContent = p.getArg('{', '}');
4063 // if the last argument is not \refname we must output
4064 if (contentslineContent == "\\refname")
4071 output_ert_inset(os,
4072 "\\bibliographystyle{" + bibliographystyle + '}',
4077 else if (t.cs() == "phantomsection") {
4078 // we only support this if it occurs between
4079 // \bibliographystyle and \bibliography
4080 if (bibliographystyle.empty())
4081 output_ert_inset(os, "\\phantomsection", context);
4084 else if (t.cs() == "addcontentsline") {
4085 context.check_layout(os);
4086 // get the 3 arguments of \addcontentsline
4087 string const one = p.getArg('{', '}');
4088 string const two = p.getArg('{', '}');
4089 string const three = p.getArg('{', '}');
4090 // only if it is a \refname, we support if for the bibtex inset
4091 if (contentslineContent != "\\refname") {
4092 output_ert_inset(os,
4093 "\\addcontentsline{" + one + "}{" + two + "}{"+ three + '}',
4098 else if (t.cs() == "bibliography") {
4099 context.check_layout(os);
4101 begin_command_inset(os, "bibtex", "bibtex");
4102 if (!btprint.empty()) {
4103 os << "btprint " << '"' << "btPrintAll" << '"' << "\n";
4104 // clear the string because the next BibTeX inset can be without the
4105 // \nocite{*} option
4108 os << "bibfiles " << '"' << p.verbatim_item() << '"' << "\n";
4109 // Do we have addcontentsline?
4110 if (contentslineContent == "\\refname") {
4111 BibOpts = "bibtotoc";
4112 // clear string because next BibTeX inset can be without addcontentsline
4113 contentslineContent.clear();
4115 // Do we have a bibliographystyle set?
4116 if (!bibliographystyle.empty()) {
4117 if (BibOpts.empty())
4118 BibOpts = bibliographystyle;
4120 BibOpts = BibOpts + ',' + bibliographystyle;
4121 // clear it because each bibtex entry has its style
4122 // and we need an empty string to handle \phantomsection
4123 bibliographystyle.clear();
4125 os << "options " << '"' << BibOpts << '"' << "\n";
4129 else if (t.cs() == "parbox") {
4130 // Test whether this is an outer box of a shaded box
4132 // swallow arguments
4133 while (p.hasOpt()) {
4135 p.skip_spaces(true);
4138 p.skip_spaces(true);
4140 if (p.next_token().cat() == catBegin)
4142 p.skip_spaces(true);
4143 Token to = p.get_token();
4144 bool shaded = false;
4145 if (to.asInput() == "\\begin") {
4146 p.skip_spaces(true);
4147 if (p.getArg('{', '}') == "shaded")
4152 parse_outer_box(p, os, FLAG_ITEM, outer,
4153 context, "parbox", "shaded");
4155 parse_box(p, os, 0, FLAG_ITEM, outer, context,
4159 else if (t.cs() == "ovalbox" || t.cs() == "Ovalbox" ||
4160 t.cs() == "shadowbox" || t.cs() == "doublebox")
4161 parse_outer_box(p, os, FLAG_ITEM, outer, context, t.cs(), "");
4163 else if (t.cs() == "framebox") {
4164 if (p.next_token().character() == '(') {
4165 //the syntax is: \framebox(x,y)[position]{content}
4166 string arg = t.asInput();
4167 arg += p.getFullParentheseArg();
4168 arg += p.getFullOpt();
4169 eat_whitespace(p, os, context, false);
4170 output_ert_inset(os, arg + '{', context);
4171 parse_text(p, os, FLAG_ITEM, outer, context);
4172 output_ert_inset(os, "}", context);
4174 string special = p.getFullOpt();
4175 special += p.getOpt();
4176 // LyX does not yet support \framebox without any option
4177 if (!special.empty())
4178 parse_outer_box(p, os, FLAG_ITEM, outer,
4179 context, t.cs(), special);
4181 eat_whitespace(p, os, context, false);
4182 output_ert_inset(os, "\\framebox{", context);
4183 parse_text(p, os, FLAG_ITEM, outer, context);
4184 output_ert_inset(os, "}", context);
4189 //\makebox() is part of the picture environment and different from \makebox{}
4190 //\makebox{} will be parsed by parse_box
4191 else if (t.cs() == "makebox") {
4192 if (p.next_token().character() == '(') {
4193 //the syntax is: \makebox(x,y)[position]{content}
4194 string arg = t.asInput();
4195 arg += p.getFullParentheseArg();
4196 arg += p.getFullOpt();
4197 eat_whitespace(p, os, context, false);
4198 output_ert_inset(os, arg + '{', context);
4199 parse_text(p, os, FLAG_ITEM, outer, context);
4200 output_ert_inset(os, "}", context);
4202 //the syntax is: \makebox[width][position]{content}
4203 parse_box(p, os, 0, FLAG_ITEM, outer, context,
4207 else if (t.cs() == "smallskip" ||
4208 t.cs() == "medskip" ||
4209 t.cs() == "bigskip" ||
4210 t.cs() == "vfill") {
4211 context.check_layout(os);
4212 begin_inset(os, "VSpace ");
4215 skip_spaces_braces(p);
4218 else if ((where = is_known(t.cs(), known_spaces))) {
4219 context.check_layout(os);
4220 begin_inset(os, "space ");
4221 os << '\\' << known_coded_spaces[where - known_spaces]
4224 // LaTeX swallows whitespace after all spaces except
4225 // "\\,". We have to do that here, too, because LyX
4226 // adds "{}" which would make the spaces significant.
4228 eat_whitespace(p, os, context, false);
4229 // LyX adds "{}" after all spaces except "\\ " and
4230 // "\\,", so we have to remove "{}".
4231 // "\\,{}" is equivalent to "\\," in LaTeX, so we
4232 // remove the braces after "\\,", too.
4237 else if (t.cs() == "newpage" ||
4238 (t.cs() == "pagebreak" && !p.hasOpt()) ||
4239 t.cs() == "clearpage" ||
4240 t.cs() == "cleardoublepage") {
4241 context.check_layout(os);
4242 begin_inset(os, "Newpage ");
4245 skip_spaces_braces(p);
4248 else if (t.cs() == "DeclareRobustCommand" ||
4249 t.cs() == "DeclareRobustCommandx" ||
4250 t.cs() == "newcommand" ||
4251 t.cs() == "newcommandx" ||
4252 t.cs() == "providecommand" ||
4253 t.cs() == "providecommandx" ||
4254 t.cs() == "renewcommand" ||
4255 t.cs() == "renewcommandx") {
4256 // DeclareRobustCommand, DeclareRobustCommandx,
4257 // providecommand and providecommandx could be handled
4258 // by parse_command(), but we need to call
4259 // add_known_command() here.
4260 string name = t.asInput();
4261 if (p.next_token().asInput() == "*") {
4262 // Starred form. Eat '*'
4266 string const command = p.verbatim_item();
4267 string const opt1 = p.getFullOpt();
4268 string const opt2 = p.getFullOpt();
4269 add_known_command(command, opt1, !opt2.empty());
4270 string const ert = name + '{' + command + '}' +
4272 '{' + p.verbatim_item() + '}';
4274 if (t.cs() == "DeclareRobustCommand" ||
4275 t.cs() == "DeclareRobustCommandx" ||
4276 t.cs() == "providecommand" ||
4277 t.cs() == "providecommandx" ||
4278 name[name.length()-1] == '*')
4279 output_ert_inset(os, ert, context);
4281 context.check_layout(os);
4282 begin_inset(os, "FormulaMacro");
4288 else if (t.cs() == "let" && p.next_token().asInput() != "*") {
4289 // let could be handled by parse_command(),
4290 // but we need to call add_known_command() here.
4291 string ert = t.asInput();
4294 if (p.next_token().cat() == catBegin) {
4295 name = p.verbatim_item();
4296 ert += '{' + name + '}';
4298 name = p.verbatim_item();
4303 if (p.next_token().cat() == catBegin) {
4304 command = p.verbatim_item();
4305 ert += '{' + command + '}';
4307 command = p.verbatim_item();
4310 // If command is known, make name known too, to parse
4311 // its arguments correctly. For this reason we also
4312 // have commands in syntax.default that are hardcoded.
4313 CommandMap::iterator it = known_commands.find(command);
4314 if (it != known_commands.end())
4315 known_commands[t.asInput()] = it->second;
4316 output_ert_inset(os, ert, context);
4319 else if (t.cs() == "hspace" || t.cs() == "vspace") {
4320 bool starred = false;
4321 if (p.next_token().asInput() == "*") {
4325 string name = t.asInput();
4326 string const length = p.verbatim_item();
4329 bool valid = splitLatexLength(length, valstring, unit);
4330 bool known_hspace = false;
4331 bool known_vspace = false;
4332 bool known_unit = false;
4335 istringstream iss(valstring);
4338 if (t.cs()[0] == 'h') {
4339 if (unit == "\\fill") {
4344 known_hspace = true;
4347 if (unit == "\\smallskipamount") {
4349 known_vspace = true;
4350 } else if (unit == "\\medskipamount") {
4352 known_vspace = true;
4353 } else if (unit == "\\bigskipamount") {
4355 known_vspace = true;
4356 } else if (unit == "\\fill") {
4358 known_vspace = true;
4362 if (!known_hspace && !known_vspace) {
4363 switch (unitFromString(unit)) {
4384 if (t.cs()[0] == 'h' && (known_unit || known_hspace)) {
4385 // Literal horizontal length or known variable
4386 context.check_layout(os);
4387 begin_inset(os, "space ");
4395 if (known_unit && !known_hspace)
4397 << translate_len(length);
4399 } else if (known_unit || known_vspace) {
4400 // Literal vertical length or known variable
4401 context.check_layout(os);
4402 begin_inset(os, "VSpace ");
4410 // LyX can't handle other length variables in Inset VSpace/space
4415 output_ert_inset(os, name + '{' + unit + '}', context);
4416 else if (value == -1.0)
4417 output_ert_inset(os, name + "{-" + unit + '}', context);
4419 output_ert_inset(os, name + '{' + valstring + unit + '}', context);
4421 output_ert_inset(os, name + '{' + length + '}', context);
4425 // The single '=' is meant here.
4426 else if ((newinsetlayout = findInsetLayout(context.textclass, t.cs(), true))) {
4428 context.check_layout(os);
4429 begin_inset(os, "Flex ");
4430 os << to_utf8(newinsetlayout->name()) << '\n'
4431 << "status collapsed\n";
4432 if (newinsetlayout->isPassThru()) {
4433 // set catcodes to verbatim early, just in case.
4434 p.setCatcodes(VERBATIM_CATCODES);
4435 string delim = p.get_token().asInput();
4437 cerr << "Warning: bad delimiter for command " << t.asInput() << endl;
4438 //FIXME: handle error condition
4439 string const arg = p.verbatimStuff("}").second;
4440 Context newcontext(true, context.textclass);
4441 if (newinsetlayout->forcePlainLayout())
4442 newcontext.layout = &context.textclass.plainLayout();
4443 output_ert(os, arg, newcontext);
4446 parse_text_in_inset(p, os, FLAG_ITEM, false, context, newinsetlayout);
4450 else if (t.cs() == "includepdf") {
4452 string const arg = p.getArg('[', ']');
4453 map<string, string> opts;
4454 vector<string> keys;
4455 split_map(arg, opts, keys);
4456 string name = normalize_filename(p.verbatim_item());
4457 string const path = getMasterFilePath(true);
4458 // We want to preserve relative / absolute filenames,
4459 // therefore path is only used for testing
4460 if (!makeAbsPath(name, path).exists()) {
4461 // The file extension is probably missing.
4462 // Now try to find it out.
4463 char const * const pdfpages_format[] = {"pdf", 0};
4464 string const pdftex_name =
4465 find_file(name, path, pdfpages_format);
4466 if (!pdftex_name.empty()) {
4471 FileName const absname = makeAbsPath(name, path);
4472 if (absname.exists())
4474 fix_child_filename(name);
4475 copy_file(absname, name);
4477 cerr << "Warning: Could not find file '"
4478 << name << "'." << endl;
4480 context.check_layout(os);
4481 begin_inset(os, "External\n\ttemplate ");
4482 os << "PDFPages\n\tfilename "
4484 // parse the options
4485 if (opts.find("pages") != opts.end())
4486 os << "\textra LaTeX \"pages="
4487 << opts["pages"] << "\"\n";
4488 if (opts.find("angle") != opts.end())
4489 os << "\trotateAngle "
4490 << opts["angle"] << '\n';
4491 if (opts.find("origin") != opts.end()) {
4493 string const opt = opts["origin"];
4494 if (opt == "tl") ss << "topleft";
4495 if (opt == "bl") ss << "bottomleft";
4496 if (opt == "Bl") ss << "baselineleft";
4497 if (opt == "c") ss << "center";
4498 if (opt == "tc") ss << "topcenter";
4499 if (opt == "bc") ss << "bottomcenter";
4500 if (opt == "Bc") ss << "baselinecenter";
4501 if (opt == "tr") ss << "topright";
4502 if (opt == "br") ss << "bottomright";
4503 if (opt == "Br") ss << "baselineright";
4504 if (!ss.str().empty())
4505 os << "\trotateOrigin " << ss.str() << '\n';
4507 cerr << "Warning: Ignoring unknown includegraphics origin argument '" << opt << "'\n";
4509 if (opts.find("width") != opts.end())
4511 << translate_len(opts["width"]) << '\n';
4512 if (opts.find("height") != opts.end())
4514 << translate_len(opts["height"]) << '\n';
4515 if (opts.find("keepaspectratio") != opts.end())
4516 os << "\tkeepAspectRatio\n";
4518 context.check_layout(os);
4519 registerExternalTemplatePackages("PDFPages");
4522 else if (t.cs() == "loadgame") {
4524 string name = normalize_filename(p.verbatim_item());
4525 string const path = getMasterFilePath(true);
4526 // We want to preserve relative / absolute filenames,
4527 // therefore path is only used for testing
4528 if (!makeAbsPath(name, path).exists()) {
4529 // The file extension is probably missing.
4530 // Now try to find it out.
4531 char const * const lyxskak_format[] = {"fen", 0};
4532 string const lyxskak_name =
4533 find_file(name, path, lyxskak_format);
4534 if (!lyxskak_name.empty())
4535 name = lyxskak_name;
4537 FileName const absname = makeAbsPath(name, path);
4538 if (absname.exists())
4540 fix_child_filename(name);
4541 copy_file(absname, name);
4543 cerr << "Warning: Could not find file '"
4544 << name << "'." << endl;
4545 context.check_layout(os);
4546 begin_inset(os, "External\n\ttemplate ");
4547 os << "ChessDiagram\n\tfilename "
4550 context.check_layout(os);
4551 // after a \loadgame follows a \showboard
4552 if (p.get_token().asInput() == "showboard")
4554 registerExternalTemplatePackages("ChessDiagram");
4558 // try to see whether the string is in unicodesymbols
4559 // Only use text mode commands, since we are in text mode here,
4560 // and math commands may be invalid (bug 6797)
4561 string name = t.asInput();
4562 // handle the dingbats and Cyrillic
4563 if (name == "\\ding" || name == "\\textcyr")
4564 name = name + '{' + p.getArg('{', '}') + '}';
4565 // handle the ifsym characters
4566 else if (name == "\\textifsymbol") {
4567 string const optif = p.getFullOpt();
4568 string const argif = p.getArg('{', '}');
4569 name = name + optif + '{' + argif + '}';
4571 // handle the \ascii characters
4572 // the case of \ascii within braces, as LyX outputs it, is already
4573 // handled for t.cat() == catBegin
4574 else if (name == "\\ascii") {
4575 // the code is "\asci\xxx"
4576 name = "{" + name + p.get_token().asInput() + "}";
4579 // handle some TIPA special characters
4580 else if (preamble.isPackageUsed("tipa")) {
4581 if (name == "\\textglobfall") {
4584 } else if (name == "\\s") {
4585 // fromLaTeXCommand() does not yet
4586 // recognize tipa short cuts
4587 name = "\\textsyllabic";
4588 } else if (name == "\\=" &&
4589 p.next_token().asInput() == "*") {
4590 // fromLaTeXCommand() does not yet
4591 // recognize tipa short cuts
4594 } else if (name == "\\textdoublevertline") {
4595 // FIXME: This is not correct,
4596 // \textvertline is higher than \textbardbl
4597 name = "\\textbardbl";
4599 } else if (name == "\\!" ) {
4600 if (p.next_token().asInput() == "b") {
4601 p.get_token(); // eat 'b'
4604 } else if (p.next_token().asInput() == "d") {
4608 } else if (p.next_token().asInput() == "g") {
4612 } else if (p.next_token().asInput() == "G") {
4614 name = "\\texthtscg";
4616 } else if (p.next_token().asInput() == "j") {
4618 name = "\\texthtbardotlessj";
4620 } else if (p.next_token().asInput() == "o") {
4622 name = "\\textbullseye";
4625 } else if (name == "\\*" ) {
4626 if (p.next_token().asInput() == "k") {
4628 name = "\\textturnk";
4630 } else if (p.next_token().asInput() == "r") {
4631 p.get_token(); // eat 'b'
4632 name = "\\textturnr";
4634 } else if (p.next_token().asInput() == "t") {
4636 name = "\\textturnt";
4638 } else if (p.next_token().asInput() == "w") {
4640 name = "\\textturnw";
4645 if ((name.size() == 2 &&
4646 contains("\"'.=^`bcdHkrtuv~", name[1]) &&
4647 p.next_token().asInput() != "*") ||
4648 is_known(name.substr(1), known_tipa_marks)) {
4649 // name is a command that corresponds to a
4650 // combining character in unicodesymbols.
4651 // Append the argument, fromLaTeXCommand()
4652 // will either convert it to a single
4653 // character or a combining sequence.
4654 name += '{' + p.verbatim_item() + '}';
4656 // now get the character from unicodesymbols
4660 docstring s = encodings.fromLaTeXCommand(from_utf8(name),
4661 Encodings::TEXT_CMD, termination, rem, &req);
4663 context.check_layout(os);
4666 output_ert_inset(os, to_utf8(rem), context);
4668 skip_spaces_braces(p);
4669 for (set<string>::const_iterator it = req.begin(); it != req.end(); ++it)
4670 preamble.registerAutomaticallyLoadedPackage(*it);
4672 //cerr << "#: " << t << " mode: " << mode << endl;
4673 // heuristic: read up to next non-nested space
4675 string s = t.asInput();
4676 string z = p.verbatim_item();
4677 while (p.good() && z != " " && !z.empty()) {
4678 //cerr << "read: " << z << endl;
4680 z = p.verbatim_item();
4682 cerr << "found ERT: " << s << endl;
4683 output_ert_inset(os, s + ' ', context);
4686 if (t.asInput() == name &&
4687 p.next_token().asInput() == "*") {
4688 // Starred commands like \vspace*{}
4689 p.get_token(); // Eat '*'
4692 if (!parse_command(name, p, os, outer, context))
4693 output_ert_inset(os, name, context);
4697 if (flags & FLAG_LEAVE) {
4698 flags &= ~FLAG_LEAVE;
4705 string guessLanguage(Parser & p, string const & lang)
4707 typedef std::map<std::string, size_t> LangMap;
4708 // map from language names to number of characters
4711 for (char const * const * i = supported_CJK_languages; *i; i++)
4712 used[string(*i)] = 0;
4715 Token const t = p.get_token();
4716 // comments are not counted for any language
4717 if (t.cat() == catComment)
4719 // commands are not counted as well, but we need to detect
4720 // \begin{CJK} and switch encoding if needed
4721 if (t.cat() == catEscape) {
4722 if (t.cs() == "inputencoding") {
4723 string const enc = subst(p.verbatim_item(), "\n", " ");
4724 p.setEncoding(enc, Encoding::inputenc);
4727 if (t.cs() != "begin")
4730 // Non-CJK content is counted for lang.
4731 // We do not care about the real language here:
4732 // If we have more non-CJK contents than CJK contents,
4733 // we simply use the language that was specified as
4734 // babel main language.
4735 used[lang] += t.asInput().length();
4738 // Now we are starting an environment
4740 string const name = p.getArg('{', '}');
4741 if (name != "CJK") {
4745 // It is a CJK environment
4747 /* name = */ p.getArg('{', '}');
4748 string const encoding = p.getArg('{', '}');
4749 /* mapping = */ p.getArg('{', '}');
4750 string const encoding_old = p.getEncoding();
4751 char const * const * const where =
4752 is_known(encoding, supported_CJK_encodings);
4754 p.setEncoding(encoding, Encoding::CJK);
4756 p.setEncoding("UTF-8");
4757 string const text = p.ertEnvironment("CJK");
4758 p.setEncoding(encoding_old);
4761 // ignore contents in unknown CJK encoding
4764 // the language of the text
4766 supported_CJK_languages[where - supported_CJK_encodings];
4767 used[cjk] += text.length();
4769 LangMap::const_iterator use = used.begin();
4770 for (LangMap::const_iterator it = used.begin(); it != used.end(); ++it) {
4771 if (it->second > use->second)