]> git.lyx.org Git - lyx.git/blobdiff - src/tex2lyx/text.C
* support/qstring_helpers.h: erase ucs4_to_qstring() method.
[lyx.git] / src / tex2lyx / text.C
index 8a9f7d23ba2dbba5742aa2c90b6bf4dfca158250..d28fdea83bc4c58cec6b3edf4510df10cff40df2 100644 (file)
 #include <sstream>
 #include <vector>
 
-using lyx::support::ChangeExtension;
-using lyx::support::MakeAbsPath;
-using lyx::support::MakeRelPath;
-using lyx::support::rtrim;
-using lyx::support::suffixIs;
-using lyx::support::contains;
-using lyx::support::subst;
+
+namespace lyx {
+
+using support::addExtension;
+using support::changeExtension;
+using support::FileName;
+using support::makeAbsPath;
+using support::makeRelPath;
+using support::rtrim;
+using support::suffixIs;
+using support::contains;
+using support::subst;
 
 using std::cerr;
 using std::endl;
@@ -50,18 +55,8 @@ using std::vector;
 namespace fs = boost::filesystem;
 
 
-/// thin wrapper around parse_text using a string
-string parse_text(Parser & p, unsigned flags, const bool outer,
-                 Context & context)
-{
-       ostringstream os;
-       parse_text(p, os, flags, outer, context);
-       return os.str();
-}
-
-
 void parse_text_in_inset(Parser & p, ostream & os, unsigned flags, bool outer,
-               Context & context)
+               Context const & context)
 {
        Context newcontext(true, context.textclass);
        newcontext.font = context.font;
@@ -70,19 +65,45 @@ void parse_text_in_inset(Parser & p, ostream & os, unsigned flags, bool outer,
 }
 
 
+namespace {
+
 /// parses a paragraph snippet, useful for example for \\emph{...}
 void parse_text_snippet(Parser & p, ostream & os, unsigned flags, bool outer,
                Context & context)
 {
-       Context newcontext(false, context.textclass);
-       newcontext.font = context.font;
+       Context newcontext(context);
+       // Don't inherit the extra stuff
+       newcontext.extra_stuff.clear();
        parse_text(p, os, flags, outer, newcontext);
-       // should not be needed
-       newcontext.check_end_layout(os);
+       // Make sure that we don't create invalid .lyx files
+       context.need_layout = newcontext.need_layout;
+       context.need_end_layout = newcontext.need_end_layout;
 }
 
 
-namespace {
+/*!
+ * Thin wrapper around parse_text_snippet() using a string.
+ *
+ * We completely ignore \c context.need_layout and \c context.need_end_layout,
+ * because our return value is not used directly (otherwise the stream version
+ * of parse_text_snippet() could be used). That means that the caller needs
+ * to do layout management manually.
+ * This is intended to parse text that does not create any layout changes.
+ */
+string parse_text_snippet(Parser & p, unsigned flags, const bool outer,
+                 Context & context)
+{
+       Context newcontext(context);
+       newcontext.need_layout = false;
+       newcontext.need_end_layout = false;
+       newcontext.new_layout_allowed = false;
+       // Avoid warning by Context::~Context()
+       newcontext.extra_stuff.clear();
+       ostringstream os;
+       parse_text_snippet(p, os, flags, outer, newcontext);
+       return os.str();
+}
+
 
 char const * const known_latex_commands[] = { "ref", "cite", "label", "index",
 "printindex", "pageref", "url", "vref", "vpageref", "prettyref", "eqref", 0 };
@@ -104,8 +125,9 @@ char const * const known_natbib_commands[] = { "cite", "citet", "citep",
  * No starred form other than "cite*" known.
  */
 char const * const known_jurabib_commands[] = { "cite", "citet", "citep",
-"citealt", "citealp", "citeauthor", "citeyear", "citeyearpar", "fullcite",
+"citealt", "citealp", "citeauthor", "citeyear", "citeyearpar",
 // jurabib commands not (yet) supported by LyX:
+// "fullcite",
 // "footcite", "footcitet", "footcitep", "footcitealt", "footcitealp",
 // "footciteauthor", "footciteyear", "footciteyearpar",
 "citefield", "citetitle", "cite*", 0 };
@@ -116,7 +138,7 @@ char const * const known_quotes[] = { "glqq", "grqq", "quotedblbase",
 
 /// the same as known_quotes with .lyx names
 char const * const known_coded_quotes[] = { "gld", "grd", "gld",
-"grd", "gls", "fls", "frd", 0};
+"grd", "gls", "fls", "frs", 0};
 
 /// LaTeX names for font sizes
 char const * const known_sizes[] = { "tiny", "scriptsize", "footnotesize",
@@ -182,6 +204,15 @@ char const * const known_pdftex_graphics_formats[] = {"png", "pdf", "jpg",
  */
 char const * const known_tex_extensions[] = {"tex", 0};
 
+/// spaces known by InsetSpace
+char const * const known_spaces[] = { " ", "space", ",", "thinspace", "quad",
+"qquad", "enspace", "enskip", "negthinspace", 0};
+
+/// the same as known_spaces with .lyx names
+char const * const known_coded_spaces[] = { "space{}", "space{}",
+"thinspace{}", "thinspace{}", "quad{}", "qquad{}", "enspace{}", "enskip{}",
+"negthinspace{}", 0};
+
 
 /// splits "x=z, y=b" into a map
 map<string, string> split_map(string const & s)
@@ -277,6 +308,8 @@ bool translate_len(string const & length, string & valstring, string & unit)
        return true;
 }
 
+}
+
 
 string translate_len(string const & length)
 {
@@ -289,6 +322,8 @@ string translate_len(string const & length)
 }
 
 
+namespace {
+
 /*!
  * Translates a LaTeX length into \p value, \p unit and
  * \p special parts suitable for a box inset.
@@ -320,11 +355,11 @@ void translate_box_len(string const & length, string & value, string & unit, str
 string find_file(string const & name, string const & path,
                 char const * const * extensions)
 {
+       // FIXME UNICODE encoding of name and path may be wrong (makeAbsPath
+       // expects utf8)
        for (char const * const * what = extensions; *what; ++what) {
-               // We don't use ChangeExtension() because it does the wrong
-               // thing if name contains a dot.
-               string const trial = name + '.' + (*what);
-               if (fs::exists(MakeAbsPath(trial, path)))
+               string const trial = addExtension(name, *what);
+               if (fs::exists(makeAbsPath(trial, path).toFilesystemEncoding()))
                        return trial;
        }
        return string();
@@ -356,14 +391,10 @@ void skip_braces(Parser & p)
 }
 
 
-
-void handle_ert(ostream & os, string const & s, Context & context,
-                bool check_layout = true)
+void handle_ert(ostream & os, string const & s, Context & context)
 {
-       if (check_layout) {
-               // We must have a valid layout before outputting the ERT inset.
-               context.check_layout(os);
-       }
+       // We must have a valid layout before outputting the ERT inset.
+       context.check_layout(os);
        Context newcontext(true, context.textclass);
        begin_inset(os, "ERT");
        os << "\nstatus collapsed\n";
@@ -417,7 +448,7 @@ private:
 LyXLayout_ptr findLayout(LyXTextClass const & textclass,
                         string const & name)
 {
-       LyXTextClass::const_iterator beg  = textclass.begin();
+       LyXTextClass::const_iterator beg = textclass.begin();
        LyXTextClass::const_iterator end = textclass.end();
 
        LyXTextClass::const_iterator
@@ -510,8 +541,8 @@ void check_space(Parser const & p, ostream & os, Context & context)
  * Parse all arguments of \p command
  */
 void parse_arguments(string const & command,
-                     vector<ArgumentType> const & template_arguments,
-                     Parser & p, ostream & os, bool outer, Context & context)
+                    vector<ArgumentType> const & template_arguments,
+                    Parser & p, ostream & os, bool outer, Context & context)
 {
        string ert = command;
        size_t no_arguments = template_arguments.size();
@@ -520,6 +551,7 @@ void parse_arguments(string const & command,
                case required:
                        // This argument contains regular LaTeX
                        handle_ert(os, ert + '{', context);
+                       eat_whitespace(p, os, context, false);
                        parse_text(p, os, FLAG_ITEM, outer, context);
                        ert = "}";
                        break;
@@ -546,7 +578,7 @@ bool parse_command(string const & command, Parser & p, ostream & os,
 {
        if (known_commands.find(command) != known_commands.end()) {
                parse_arguments(command, known_commands[command], p, os,
-                               outer, context);
+                               outer, context);
                return true;
        }
        return false;
@@ -655,6 +687,33 @@ void parse_box(Parser & p, ostream & os, unsigned flags, bool outer,
 }
 
 
+/// parse an unknown environment
+void parse_unknown_environment(Parser & p, string const & name, ostream & os,
+                              unsigned flags, bool outer,
+                              Context & parent_context)
+{
+       if (name == "tabbing")
+               // We need to remember that we have to handle '\=' specially
+               flags |= FLAG_TABBING;
+
+       // We need to translate font changes and paragraphs inside the
+       // environment to ERT if we have a non standard font.
+       // Otherwise things like
+       // \large\begin{foo}\huge bar\end{foo}
+       // will not work.
+       bool const specialfont =
+               (parent_context.font != parent_context.normalfont);
+       bool const new_layout_allowed = parent_context.new_layout_allowed;
+       if (specialfont)
+               parent_context.new_layout_allowed = false;
+       handle_ert(os, "\\begin{" + name + "}", parent_context);
+       parse_text_snippet(p, os, flags, outer, parent_context);
+       handle_ert(os, "\\end{" + name + "}", parent_context);
+       if (specialfont)
+               parent_context.new_layout_allowed = new_layout_allowed;
+}
+
+
 void parse_environment(Parser & p, ostream & os, bool outer,
                       Context & parent_context)
 {
@@ -662,7 +721,6 @@ void parse_environment(Parser & p, ostream & os, bool outer,
        string const name = p.getArg('{', '}');
        const bool is_starred = suffixIs(name, '*');
        string const unstarred_name = rtrim(name, "*");
-       eat_whitespace(p, os, parent_context, false);
        active_environments.push_back(name);
 
        if (is_math_env(name)) {
@@ -675,13 +733,16 @@ void parse_environment(Parser & p, ostream & os, bool outer,
        }
 
        else if (name == "tabular" || name == "longtable") {
+               eat_whitespace(p, os, parent_context, false);
                parent_context.check_layout(os);
                begin_inset(os, "Tabular ");
                handle_tabular(p, os, name == "longtable", parent_context);
                end_inset(os);
+               p.skip_spaces();
        }
 
        else if (parent_context.textclass.floats().typeExist(unstarred_name)) {
+               eat_whitespace(p, os, parent_context, false);
                parent_context.check_layout(os);
                begin_inset(os, "Float " + unstarred_name + "\n");
                if (p.next_token().asInput() == "[") {
@@ -695,14 +756,43 @@ void parse_environment(Parser & p, ostream & os, bool outer,
                // We don't need really a new paragraph, but
                // we must make sure that the next item gets a \begin_layout.
                parent_context.new_paragraph(os);
+               p.skip_spaces();
        }
 
-       else if (name == "minipage")
+       else if (name == "minipage") {
+               eat_whitespace(p, os, parent_context, false);
                parse_box(p, os, FLAG_END, outer, parent_context, false);
+               p.skip_spaces();
+       }
+
+       else if (name == "comment") {
+               eat_whitespace(p, os, parent_context, false);
+               parent_context.check_layout(os);
+               begin_inset(os, "Note Comment\n");
+               os << "status open\n";
+               parse_text_in_inset(p, os, FLAG_END, outer, parent_context);
+               end_inset(os);
+               p.skip_spaces();
+       }
+
+       else if (name == "lyxgreyedout") {
+               eat_whitespace(p, os, parent_context, false);
+               parent_context.check_layout(os);
+               begin_inset(os, "Note Greyedout\n");
+               os << "status open\n";
+               parse_text_in_inset(p, os, FLAG_END, outer, parent_context);
+               end_inset(os);
+               p.skip_spaces();
+       }
+
+       else if (!parent_context.new_layout_allowed)
+               parse_unknown_environment(p, name, os, FLAG_END, outer,
+                                         parent_context);
 
        // Alignment settings
        else if (name == "center" || name == "flushleft" || name == "flushright" ||
-                name == "centering" || name == "raggedright" || name == "raggedleft") {
+                name == "centering" || name == "raggedright" || name == "raggedleft") {
+               eat_whitespace(p, os, parent_context, false);
                // We must begin a new paragraph if not already done
                if (! parent_context.atParagraphStart()) {
                        parent_context.check_end_layout(os);
@@ -719,11 +809,13 @@ void parse_environment(Parser & p, ostream & os, bool outer,
                parent_context.extra_stuff.erase();
                // We must begin a new paragraph to reset the alignment
                parent_context.new_paragraph(os);
+               p.skip_spaces();
        }
 
        // The single '=' is meant here.
        else if ((newlayout = findLayout(parent_context.textclass, name)).get() &&
                  newlayout->isEnvironment()) {
+               eat_whitespace(p, os, parent_context, false);
                Context context(true, parent_context.textclass, newlayout,
                                parent_context.layout, parent_context.font);
                if (parent_context.deeper_paragraph) {
@@ -756,10 +848,12 @@ void parse_environment(Parser & p, ostream & os, bool outer,
                }
                context.check_end_deeper(os);
                parent_context.new_paragraph(os);
+               p.skip_spaces();
        }
 
        else if (name == "appendix") {
                // This is no good latex style, but it works and is used in some documents...
+               eat_whitespace(p, os, parent_context, false);
                parent_context.check_end_layout(os);
                Context context(true, parent_context.textclass, parent_context.layout,
                                parent_context.layout, parent_context.font);
@@ -767,34 +861,7 @@ void parse_environment(Parser & p, ostream & os, bool outer,
                os << "\\start_of_appendix\n";
                parse_text(p, os, FLAG_END, outer, context);
                context.check_end_layout(os);
-       }
-
-       else if (name == "comment") {
-               parent_context.check_layout(os);
-               begin_inset(os, "Note Comment\n");
-               os << "status open\n";
-               parse_text_in_inset(p, os, FLAG_END, outer, parent_context);
-               end_inset(os);
-       }
-
-       else if (name == "lyxgreyedout") {
-               parent_context.check_layout(os);
-               begin_inset(os, "Note Greyedout\n");
-               os << "status open\n";
-               parse_text_in_inset(p, os, FLAG_END, outer, parent_context);
-               end_inset(os);
-       }
-
-       else if (name == "tabbing") {
-               // We need to remember that we have to handle '\=' specially
-               handle_ert(os, "\\begin{" + name + "}", parent_context);
-               // FIXME: Try whether parse_text instead of parse_text_snippet
-               // works. Then no manual layout checking would be needed.
-               parent_context.check_end_layout(os);
-               parse_text_snippet(p, os, FLAG_END | FLAG_TABBING, outer,
-                                  parent_context);
-               parent_context.need_layout = true;
-               handle_ert(os, "\\end{" + name + "}", parent_context);
+               p.skip_spaces();
        }
 
        else if (known_environments.find(name) != known_environments.end()) {
@@ -808,36 +875,31 @@ void parse_environment(Parser & p, ostream & os, bool outer,
                        arguments.back();
                if (!arguments.empty())
                        arguments.pop_back();
+               // See comment in parse_unknown_environment()
+               bool const specialfont =
+                       (parent_context.font != parent_context.normalfont);
+               bool const new_layout_allowed =
+                       parent_context.new_layout_allowed;
+               if (specialfont)
+                       parent_context.new_layout_allowed = false;
                parse_arguments("\\begin{" + name + "}", arguments, p, os,
-                               outer, parent_context);
+                               outer, parent_context);
                if (contents == verbatim)
                        handle_ert(os, p.verbatimEnvironment(name),
-                                  parent_context);
-               else {
-                       // FIXME: Try whether parse_text instead of
-                       // parse_text_snippet works. Then no manual layout
-                       // checking would be needed.
-                       parent_context.check_end_layout(os);
+                                  parent_context);
+               else
                        parse_text_snippet(p, os, FLAG_END, outer,
-                                          parent_context);
-                       parent_context.need_layout = true;
-               }
+                                          parent_context);
                handle_ert(os, "\\end{" + name + "}", parent_context);
+               if (specialfont)
+                       parent_context.new_layout_allowed = new_layout_allowed;
        }
 
-       else {
-               handle_ert(os, "\\begin{" + name + "}", parent_context);
-               // FIXME: Try whether parse_text instead of parse_text_snippet
-               // works. Then no manual layout checking would be needed.
-               parent_context.check_end_layout(os);
-               parse_text_snippet(p, os, FLAG_END, outer, parent_context);
-               parent_context.need_layout = true;
-               handle_ert(os, "\\end{" + name + "}", parent_context);
-       }
+       else
+               parse_unknown_environment(p, name, os, FLAG_END, outer,
+                                         parent_context);
 
        active_environments.pop_back();
-       if (name != "math")
-               p.skip_spaces();
 }
 
 
@@ -851,11 +913,13 @@ void parse_comment(Parser & p, ostream & os, Token const & t, Context & context)
                if (p.next_token().cat() == catNewline) {
                        // A newline after a comment line starts a new
                        // paragraph
-                       if(!context.atParagraphStart()) {
-                               // Only start a new paragraph if not already
-                               // done (we might get called recursively)
-                               context.new_paragraph(os);
-                       }
+                       if (context.new_layout_allowed) {
+                               if(!context.atParagraphStart())
+                                       // Only start a new paragraph if not already
+                                       // done (we might get called recursively)
+                                       context.new_paragraph(os);
+                       } else
+                               handle_ert(os, "\n", context);
                        eat_whitespace(p, os, context, true);
                }
        } else {
@@ -879,7 +943,7 @@ void eat_whitespace(Parser & p, ostream & os, Context & context,
                if (t.cat() == catComment)
                        parse_comment(p, os, t, context);
                else if ((! eatParagraph && p.isParagraph()) ||
-                        (t.cat() != catSpace && t.cat() != catNewline)) {
+                        (t.cat() != catSpace && t.cat() != catNewline)) {
                        p.putback();
                        return;
                }
@@ -899,12 +963,13 @@ void parse_text_attributes(Parser & p, ostream & os, unsigned flags, bool outer,
                           string & currentvalue, string const & newvalue)
 {
        context.check_layout(os);
-       string oldvalue = currentvalue;
+       string const oldvalue = currentvalue;
        currentvalue = newvalue;
        os << '\n' << attribute << ' ' << newvalue << "\n";
        parse_text_snippet(p, os, flags, outer, context);
-       currentvalue = oldvalue;
+       context.check_layout(os);
        os << '\n' << attribute << ' ' << oldvalue << "\n";
+       currentvalue = oldvalue;
 }
 
 
@@ -955,10 +1020,80 @@ string const normalize_filename(string const & name)
 /// convention (relative to .lyx file) if it is relative
 void fix_relative_filename(string & name)
 {
-       if (lyx::support::AbsolutePath(name))
+       if (lyx::support::absolutePath(name))
+               return;
+       // FIXME UNICODE encoding of name may be wrong (makeAbsPath expects
+       // utf8)
+       name = makeRelPath(makeAbsPath(name, getMasterFilePath()).absFilename(),
+                          getParentFilePath());
+}
+
+
+/// Parse a NoWeb Scrap section. The initial "<<" is already parsed.
+void parse_noweb(Parser & p, ostream & os, Context & context)
+{
+       // assemble the rest of the keyword
+       string name("<<");
+       bool scrap = false;
+       while (p.good()) {
+               Token const & t = p.get_token();
+               if (t.asInput() == ">" && p.next_token().asInput() == ">") {
+                       name += ">>";
+                       p.get_token();
+                       scrap = (p.good() && p.next_token().asInput() == "=");
+                       if (scrap)
+                               name += p.get_token().asInput();
+                       break;
+               }
+               name += t.asInput();
+       }
+
+       if (!scrap || !context.new_layout_allowed ||
+           !context.textclass.hasLayout("Scrap")) {
+               cerr << "Warning: Could not interpret '" << name
+                    << "'. Ignoring it." << endl;
                return;
-       name = MakeRelPath(MakeAbsPath(name, getMasterFilePath()),
-                          getParentFilePath());
+       }
+
+       // We use new_paragraph instead of check_end_layout because the stuff
+       // following the noweb chunk needs to start with a \begin_layout.
+       // This may create a new paragraph even if there was none in the
+       // noweb file, but the alternative is an invalid LyX file. Since
+       // noweb code chunks are implemented with a layout style in LyX they
+       // always must be in an own paragraph.
+       context.new_paragraph(os);
+       Context newcontext(true, context.textclass, context.textclass["Scrap"]);
+       newcontext.check_layout(os);
+       os << name;
+       while (p.good()) {
+               Token const & t = p.get_token();
+               // We abuse the parser a bit, because this is no TeX syntax
+               // at all.
+               if (t.cat() == catEscape)
+                       os << subst(t.asInput(), "\\", "\n\\backslash\n");
+               else
+                       os << subst(t.asInput(), "\n", "\n\\newline\n");
+               // The scrap chunk is ended by an @ at the beginning of a line.
+               // After the @ the line may contain a comment and/or
+               // whitespace, but nothing else.
+               if (t.asInput() == "@" && p.prev_token().cat() == catNewline &&
+                   (p.next_token().cat() == catSpace ||
+                    p.next_token().cat() == catNewline ||
+                    p.next_token().cat() == catComment)) {
+                       while (p.good() && p.next_token().cat() == catSpace)
+                               os << p.get_token().asInput();
+                       if (p.next_token().cat() == catComment)
+                               // The comment includes a final '\n'
+                               os << p.get_token().asInput();
+                       else {
+                               if (p.next_token().cat() == catNewline)
+                                       p.get_token();
+                               os << '\n';
+                       }
+                       break;
+               }
+       }
+       newcontext.check_end_layout(os);
 }
 
 } // anonymous namespace
@@ -1048,9 +1183,28 @@ void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
                        skip_braces(p);
                }
 
+               else if (t.asInput() == "<"
+                        && p.next_token().asInput() == "<" && noweb_mode) {
+                       p.get_token();
+                       parse_noweb(p, os, context);
+               }
+
                else if (t.cat() == catSpace || (t.cat() == catNewline && ! p.isParagraph()))
                        check_space(p, os, context);
 
+               else if (t.character() == '[' && noweb_mode &&
+                        p.next_token().character() == '[') {
+                       // These can contain underscores
+                       p.putback();
+                       string const s = p.getFullOpt() + ']';
+                       if (p.next_token().character() == ']')
+                               p.get_token();
+                       else
+                               cerr << "Warning: Inserting missing ']' in '"
+                                    << s << "'." << endl;
+                       handle_ert(os, s, context);
+               }
+
                else if (t.cat() == catLetter ||
                               t.cat() == catOther ||
                               t.cat() == catAlign ||
@@ -1061,7 +1215,10 @@ void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
                }
 
                else if (p.isParagraph()) {
-                       context.new_paragraph(os);
+                       if (context.new_layout_allowed)
+                               context.new_paragraph(os);
+                       else
+                               handle_ert(os, "\\par ", context);
                        eat_whitespace(p, os, context, true);
                }
 
@@ -1076,55 +1233,85 @@ void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
                                os << t.character();
                }
 
+               else if (t.cat() == catBegin &&
+                        p.next_token().cat() == catEnd) {
+                       // {}
+                       Token const prev = p.prev_token();
+                       p.get_token();
+                       if (p.next_token().character() == '`' ||
+                           (prev.character() == '-' &&
+                            p.next_token().character() == '-'))
+                               ; // ignore it in {}`` or -{}-
+                       else
+                               handle_ert(os, "{}", context);
+
+               }
+
                else if (t.cat() == catBegin) {
                        context.check_layout(os);
                        // special handling of font attribute changes
                        Token const prev = p.prev_token();
                        Token const next = p.next_token();
                        Font const oldFont = context.font;
-                       string const s = parse_text(p, FLAG_BRACE_LAST, outer,
-                                                   context);
-                       context.font = oldFont;
-                       if (s.empty() && (p.next_token().character() == '`' ||
-                                         (prev.character() == '-' &&
-                                          p.next_token().character() == '-')))
-                               ; // ignore it in {}`` or -{}-
-                       else if (s == "[" || s == "]" || s == "*")
-                               os << s;
-                       else if (is_known(next.cs(), known_sizes)) {
-                               // s will change the size, so we must reset
-                               // it here
-                               os << s;
+                       if (next.character() == '[' ||
+                           next.character() == ']' ||
+                           next.character() == '*') {
+                               p.get_token();
+                               if (p.next_token().cat() == catEnd) {
+                                       os << next.character();
+                                       p.get_token();
+                               } else {
+                                       p.putback();
+                                       handle_ert(os, "{", context);
+                                       parse_text_snippet(p, os,
+                                                       FLAG_BRACE_LAST,
+                                                       outer, context);
+                                       handle_ert(os, "}", context);
+                               }
+                       } else if (! context.new_layout_allowed) {
+                               handle_ert(os, "{", context);
+                               parse_text_snippet(p, os, FLAG_BRACE_LAST,
+                                                  outer, context);
+                               handle_ert(os, "}", context);
+                       } else if (is_known(next.cs(), known_sizes)) {
+                               // next will change the size, so we must
+                               // reset it here
+                               parse_text_snippet(p, os, FLAG_BRACE_LAST,
+                                                  outer, context);
                                if (!context.atParagraphStart())
                                        os << "\n\\size "
                                           << context.font.size << "\n";
                        } else if (is_known(next.cs(), known_font_families)) {
-                               // s will change the font family, so we must
-                               // reset it here
-                               os << s;
+                               // next will change the font family, so we
+                               // must reset it here
+                               parse_text_snippet(p, os, FLAG_BRACE_LAST,
+                                                  outer, context);
                                if (!context.atParagraphStart())
                                        os << "\n\\family "
                                           << context.font.family << "\n";
                        } else if (is_known(next.cs(), known_font_series)) {
-                               // s will change the font series, so we must
-                               // reset it here
-                               os << s;
+                               // next will change the font series, so we
+                               // must reset it here
+                               parse_text_snippet(p, os, FLAG_BRACE_LAST,
+                                                  outer, context);
                                if (!context.atParagraphStart())
                                        os << "\n\\series "
                                           << context.font.series << "\n";
                        } else if (is_known(next.cs(), known_font_shapes)) {
-                               // s will change the font shape, so we must
-                               // reset it here
-                               os << s;
+                               // next will change the font shape, so we
+                               // must reset it here
+                               parse_text_snippet(p, os, FLAG_BRACE_LAST,
+                                                  outer, context);
                                if (!context.atParagraphStart())
                                        os << "\n\\shape "
                                           << context.font.shape << "\n";
                        } else if (is_known(next.cs(), known_old_font_families) ||
-                                  is_known(next.cs(), known_old_font_series) ||
-                                  is_known(next.cs(), known_old_font_shapes)) {
-                               // s will change the font family, series
+                                  is_known(next.cs(), known_old_font_series) ||
+                                  is_known(next.cs(), known_old_font_shapes)) {
+                               // next will change the font family, series
                                // and shape, so we must reset it here
-                               os << s;
+                               parse_text_snippet(p, os, FLAG_BRACE_LAST,
+                                                  outer, context);
                                if (!context.atParagraphStart())
                                        os <<  "\n\\family "
                                           << context.font.family
@@ -1133,10 +1320,9 @@ void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
                                           << "\n\\shape "
                                           << context.font.shape << "\n";
                        } else {
-                               handle_ert(os, "{", context, false);
-                               // s will end the current layout and begin a
-                               // new one if necessary
-                               os << s;
+                               handle_ert(os, "{", context);
+                               parse_text_snippet(p, os, FLAG_BRACE_LAST,
+                                                  outer, context);
                                handle_ert(os, "}", context);
                        }
                }
@@ -1195,13 +1381,21 @@ void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
                        bool optarg = false;
                        if (p.next_token().character() == '[') {
                                p.get_token(); // eat '['
-                               Context newcontext(false, context.textclass);
-                               newcontext.font = context.font;
-                               s = parse_text(p, FLAG_BRACK_LAST, outer, newcontext);
+                               s = parse_text_snippet(p, FLAG_BRACK_LAST,
+                                                      outer, context);
                                optarg = true;
                        }
                        context.set_item();
                        context.check_layout(os);
+                       if (context.has_item) {
+                               // An item in an unknown list-like environment
+                               // FIXME: Do this in check_layout()!
+                               context.has_item = false;
+                               if (optarg)
+                                       handle_ert(os, "\\item", context);
+                               else
+                                       handle_ert(os, "\\item ", context);
+                       }
                        if (optarg) {
                                if (context.layout->labeltype != LABEL_MANUAL) {
                                        // lyx does not support \item[\mybullet]
@@ -1265,6 +1459,7 @@ void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
 
                // Must attempt to parse "Section*" before "Section".
                else if ((p.next_token().asInput() == "*") &&
+                        context.new_layout_allowed &&
                         // The single '=' is meant here.
                         (newlayout = findLayout(context.textclass,
                                                 t.cs() + '*')).get() &&
@@ -1275,7 +1470,8 @@ void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
                }
 
                // The single '=' is meant here.
-               else if ((newlayout = findLayout(context.textclass, t.cs())).get() &&
+               else if (context.new_layout_allowed &&
+                        (newlayout = findLayout(context.textclass, t.cs())).get() &&
                         newlayout->isCommand()) {
                        output_command_layout(os, p, outer, context, newlayout);
                        p.skip_spaces();
@@ -1293,29 +1489,31 @@ void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
                        string const path = getMasterFilePath();
                        // We want to preserve relative / absolute filenames,
                        // therefore path is only used for testing
-                       if (!fs::exists(MakeAbsPath(name, path))) {
+                       // FIXME UNICODE encoding of name and path may be
+                       // wrong (makeAbsPath expects utf8)
+                       if (!fs::exists(makeAbsPath(name, path).toFilesystemEncoding())) {
                                // The file extension is probably missing.
                                // Now try to find it out.
                                string const dvips_name =
                                        find_file(name, path,
-                                                 known_dvips_graphics_formats);
+                                                 known_dvips_graphics_formats);
                                string const pdftex_name =
                                        find_file(name, path,
-                                                 known_pdftex_graphics_formats);
+                                                 known_pdftex_graphics_formats);
                                if (!dvips_name.empty()) {
                                        if (!pdftex_name.empty()) {
                                                cerr << "This file contains the "
-                                                       "latex snippet\n"
-                                                       "\"\\includegraphics{"
+                                                       "latex snippet\n"
+                                                       "\"\\includegraphics{"
                                                     << name << "}\".\n"
-                                                       "However, files\n\""
+                                                       "However, files\n\""
                                                     << dvips_name << "\" and\n\""
                                                     << pdftex_name << "\"\n"
-                                                       "both exist, so I had to make a "
-                                                       "choice and took the first one.\n"
-                                                       "Please move the unwanted one "
-                                                       "someplace else and try again\n"
-                                                       "if my choice was wrong."
+                                                       "both exist, so I had to make a "
+                                                       "choice and took the first one.\n"
+                                                       "Please move the unwanted one "
+                                                       "someplace else and try again\n"
+                                                       "if my choice was wrong."
                                                     << endl;
                                        }
                                        name = dvips_name;
@@ -1323,7 +1521,9 @@ void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
                                        name = pdftex_name;
                        }
 
-                       if (fs::exists(MakeAbsPath(name, path)))
+                       // FIXME UNICODE encoding of name and path may be
+                       // wrong (makeAbsPath expects utf8)
+                       if (fs::exists(makeAbsPath(name, path).toFilesystemEncoding()))
                                fix_relative_filename(name);
                        else
                                cerr << "Warning: Could not find graphics file '"
@@ -1425,7 +1625,7 @@ void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
                }
 
                else if (t.cs() == "footnote" ||
-                        (t.cs() == "thanks" && context.layout->intitle)) {
+                        (t.cs() == "thanks" && context.layout->intitle)) {
                        p.skip_spaces();
                        context.check_layout(os);
                        begin_inset(os, "Foot\n");
@@ -1446,9 +1646,7 @@ void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
                else if (t.cs() == "ensuremath") {
                        p.skip_spaces();
                        context.check_layout(os);
-                       Context newcontext(false, context.textclass);
-                       newcontext.font = context.font;
-                       string s = parse_text(p, FLAG_ITEM, false, newcontext);
+                       string const s = p.verbatim_item();
                        if (s == "±" || s == "³" || s == "²" || s == "µ")
                                os << s;
                        else
@@ -1509,48 +1707,48 @@ void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
 
                else if (t.cs() == "textrm")
                        parse_text_attributes(p, os, FLAG_ITEM, outer,
-                                             context, "\\family",
-                                             context.font.family, "roman");
+                                             context, "\\family",
+                                             context.font.family, "roman");
 
                else if (t.cs() == "textsf")
                        parse_text_attributes(p, os, FLAG_ITEM, outer,
-                                             context, "\\family",
-                                             context.font.family, "sans");
+                                             context, "\\family",
+                                             context.font.family, "sans");
 
                else if (t.cs() == "texttt")
                        parse_text_attributes(p, os, FLAG_ITEM, outer,
-                                             context, "\\family",
-                                             context.font.family, "typewriter");
+                                             context, "\\family",
+                                             context.font.family, "typewriter");
 
                else if (t.cs() == "textmd")
                        parse_text_attributes(p, os, FLAG_ITEM, outer,
-                                             context, "\\series",
-                                             context.font.series, "medium");
+                                             context, "\\series",
+                                             context.font.series, "medium");
 
                else if (t.cs() == "textbf")
                        parse_text_attributes(p, os, FLAG_ITEM, outer,
-                                             context, "\\series",
-                                             context.font.series, "bold");
+                                             context, "\\series",
+                                             context.font.series, "bold");
 
                else if (t.cs() == "textup")
                        parse_text_attributes(p, os, FLAG_ITEM, outer,
-                                             context, "\\shape",
-                                             context.font.shape, "up");
+                                             context, "\\shape",
+                                             context.font.shape, "up");
 
                else if (t.cs() == "textit")
                        parse_text_attributes(p, os, FLAG_ITEM, outer,
-                                             context, "\\shape",
-                                             context.font.shape, "italic");
+                                             context, "\\shape",
+                                             context.font.shape, "italic");
 
                else if (t.cs() == "textsl")
                        parse_text_attributes(p, os, FLAG_ITEM, outer,
-                                             context, "\\shape",
-                                             context.font.shape, "slanted");
+                                             context, "\\shape",
+                                             context.font.shape, "slanted");
 
                else if (t.cs() == "textsc")
                        parse_text_attributes(p, os, FLAG_ITEM, outer,
-                                             context, "\\shape",
-                                             context.font.shape, "smallcaps");
+                                             context, "\\shape",
+                                             context.font.shape, "smallcaps");
 
                else if (t.cs() == "textnormal" || t.cs() == "normalfont") {
                        context.check_layout(os);
@@ -1575,6 +1773,7 @@ void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
                        context.check_layout(os);
                        os << "\n\\bar under\n";
                        parse_text_snippet(p, os, FLAG_ITEM, outer, context);
+                       context.check_layout(os);
                        os << "\n\\bar default\n";
                }
 
@@ -1582,15 +1781,16 @@ void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
                        context.check_layout(os);
                        os << "\n\\" << t.cs() << " on\n";
                        parse_text_snippet(p, os, FLAG_ITEM, outer, context);
+                       context.check_layout(os);
                        os << "\n\\" << t.cs() << " default\n";
                }
 
                else if (use_natbib &&
-                        is_known(t.cs(), known_natbib_commands) &&
-                        ((t.cs() != "citefullauthor" &&
-                          t.cs() != "citeyear" &&
-                          t.cs() != "citeyearpar") ||
-                         p.next_token().asInput() != "*")) {
+                        is_known(t.cs(), known_natbib_commands) &&
+                        ((t.cs() != "citefullauthor" &&
+                          t.cs() != "citeyear" &&
+                          t.cs() != "citeyearpar") ||
+                         p.next_token().asInput() != "*")) {
                        context.check_layout(os);
                        // tex                       lyx
                        // \citet[before][after]{a}  \citet[after][before]{a}
@@ -1636,16 +1836,16 @@ void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
                }
 
                else if (use_jurabib &&
-                        is_known(t.cs(), known_jurabib_commands)) {
+                        is_known(t.cs(), known_jurabib_commands)) {
                        context.check_layout(os);
                        string const command = '\\' + t.cs();
                        char argumentOrder = '\0';
                        vector<string> const & options = used_packages["jurabib"];
                        if (std::find(options.begin(), options.end(),
-                                     "natbiborder") != options.end())
+                                     "natbiborder") != options.end())
                                argumentOrder = 'n';
                        else if (std::find(options.begin(), options.end(),
-                                          "jurabiborder") != options.end())
+                                          "jurabiborder") != options.end())
                                argumentOrder = 'j';
 
                        // text before the citation
@@ -1658,12 +1858,12 @@ void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
                        string const citation = p.verbatim_item();
                        if (!before.empty() && argumentOrder == '\0') {
                                cerr << "Warning: Assuming argument order "
-                                       "of jurabib version 0.6 for\n'"
+                                       "of jurabib version 0.6 for\n'"
                                     << command << before << after << '{'
                                     << citation << "}'.\n"
-                                       "Add 'jurabiborder' to the jurabib "
-                                       "package options if you used an\n"
-                                       "earlier jurabib version." << endl;
+                                       "Add 'jurabiborder' to the jurabib "
+                                       "package options if you used an\n"
+                                       "earlier jurabib version." << endl;
                        }
                        begin_inset(os, "LatexCommand ");
                        os << command << after << before
@@ -1699,7 +1899,8 @@ void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
                        skip_braces(p);
                }
 
-               else if (is_known(t.cs(), known_sizes)) {
+               else if (is_known(t.cs(), known_sizes) &&
+                        context.new_layout_allowed) {
                        char const * const * where = is_known(t.cs(), known_sizes);
                        context.check_layout(os);
                        Font const oldFont = context.font;
@@ -1708,7 +1909,8 @@ void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
                        eat_whitespace(p, os, context, false);
                }
 
-               else if (is_known(t.cs(), known_font_families)) {
+               else if (is_known(t.cs(), known_font_families) &&
+                        context.new_layout_allowed) {
                        char const * const * where =
                                is_known(t.cs(), known_font_families);
                        context.check_layout(os);
@@ -1719,7 +1921,8 @@ void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
                        eat_whitespace(p, os, context, false);
                }
 
-               else if (is_known(t.cs(), known_font_series)) {
+               else if (is_known(t.cs(), known_font_series) &&
+                        context.new_layout_allowed) {
                        char const * const * where =
                                is_known(t.cs(), known_font_series);
                        context.check_layout(os);
@@ -1730,7 +1933,8 @@ void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
                        eat_whitespace(p, os, context, false);
                }
 
-               else if (is_known(t.cs(), known_font_shapes)) {
+               else if (is_known(t.cs(), known_font_shapes) &&
+                        context.new_layout_allowed) {
                        char const * const * where =
                                is_known(t.cs(), known_font_shapes);
                        context.check_layout(os);
@@ -1740,7 +1944,8 @@ void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
                        output_font_change(os, oldFont, context.font);
                        eat_whitespace(p, os, context, false);
                }
-               else if (is_known(t.cs(), known_old_font_families)) {
+               else if (is_known(t.cs(), known_old_font_families) &&
+                        context.new_layout_allowed) {
                        char const * const * where =
                                is_known(t.cs(), known_old_font_families);
                        context.check_layout(os);
@@ -1753,7 +1958,8 @@ void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
                        eat_whitespace(p, os, context, false);
                }
 
-               else if (is_known(t.cs(), known_old_font_series)) {
+               else if (is_known(t.cs(), known_old_font_series) &&
+                        context.new_layout_allowed) {
                        char const * const * where =
                                is_known(t.cs(), known_old_font_series);
                        context.check_layout(os);
@@ -1766,7 +1972,8 @@ void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
                        eat_whitespace(p, os, context, false);
                }
 
-               else if (is_known(t.cs(), known_old_font_shapes)) {
+               else if (is_known(t.cs(), known_old_font_shapes) &&
+                        context.new_layout_allowed) {
                        char const * const * where =
                                is_known(t.cs(), known_old_font_shapes);
                        context.check_layout(os);
@@ -1862,6 +2069,15 @@ void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
                        }
                }
 
+               else if (t.cs() == "verb") {
+                       context.check_layout(os);
+                       char const delimiter = p.next_token().character();
+                       string const arg = p.getArg(delimiter, delimiter);
+                       ostringstream oss;
+                       oss << "\\verb" << delimiter << arg << delimiter;
+                       handle_ert(os, oss.str(), context);
+               }
+
                else if (t.cs() == "\"") {
                        context.check_layout(os);
                        string const name = p.verbatim_item();
@@ -1880,12 +2096,21 @@ void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
                else if (t.cs() == "=" && (flags & FLAG_TABBING))
                        handle_ert(os, t.asInput(), context);
 
-               else if (t.cs() == "H" || t.cs() == "c" || t.cs() == "^" || t.cs() == "'"
-                     || t.cs() == "~" || t.cs() == "." || t.cs() == "=") {
+               else if (t.cs() == "H" || t.cs() == "c" || t.cs() == "^"
+                        || t.cs() == "'" || t.cs() == "`"
+                        || t.cs() == "~" || t.cs() == "." || t.cs() == "=") {
                        // we need the trim as the LyX parser chokes on such spaces
+                       // The argument of InsetLatexAccent is parsed as a
+                       // subset of LaTeX, so don't parse anything here,
+                       // but use the raw argument.
+                       // Otherwise we would convert \~{\i} wrongly.
+                       // This will of course not translate \~{\ss} to \~{ß},
+                       // but that does at least compile and does only look
+                       // strange on screen.
                        context.check_layout(os);
-                       os << "\n\\i \\" << t.cs() << "{"
-                          << trim(parse_text(p, FLAG_ITEM, outer, context), " ") << "}\n";
+                       os << "\\i \\" << t.cs() << "{"
+                          << trim(p.verbatim_item(), " ")
+                          << "}\n";
                }
 
                else if (t.cs() == "ss") {
@@ -1894,9 +2119,10 @@ void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
                        skip_braces(p); // eat {}
                }
 
-               else if (t.cs() == "i" || t.cs() == "j") {
+               else if (t.cs() == "i" || t.cs() == "j" || t.cs() == "l" ||
+                        t.cs() == "L") {
                        context.check_layout(os);
-                       os << "\\" << t.cs() << ' ';
+                       os << "\\i \\" << t.cs() << "{}\n";
                        skip_braces(p); // eat {}
                }
 
@@ -1926,26 +2152,30 @@ void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
                        string const path = getMasterFilePath();
                        // We want to preserve relative / absolute filenames,
                        // therefore path is only used for testing
+                       // FIXME UNICODE encoding of filename and path may be
+                       // wrong (makeAbsPath expects utf8)
                        if (t.cs() == "include" &&
-                           !fs::exists(MakeAbsPath(filename, path))) {
+                           !fs::exists(makeAbsPath(filename, path).toFilesystemEncoding())) {
                                // The file extension is probably missing.
                                // Now try to find it out.
                                string const tex_name =
                                        find_file(filename, path,
-                                                 known_tex_extensions);
+                                                 known_tex_extensions);
                                if (!tex_name.empty())
                                        filename = tex_name;
                        }
-                       if (fs::exists(MakeAbsPath(filename, path))) {
+                       // FIXME UNICODE encoding of filename and path may be
+                       // wrong (makeAbsPath expects utf8)
+                       if (fs::exists(makeAbsPath(filename, path).toFilesystemEncoding())) {
                                string const abstexname =
-                                       MakeAbsPath(filename, path);
+                                       makeAbsPath(filename, path).absFilename();
                                string const abslyxname =
-                                       ChangeExtension(abstexname, ".lyx");
+                                       changeExtension(abstexname, ".lyx");
                                fix_relative_filename(filename);
                                string const lyxname =
-                                       ChangeExtension(filename, ".lyx");
+                                       changeExtension(filename, ".lyx");
                                if (t.cs() != "verbatiminput" &&
-                                   tex2lyx(abstexname, abslyxname)) {
+                                   tex2lyx(abstexname, FileName(abslyxname))) {
                                        os << name << '{' << lyxname << "}\n";
                                } else {
                                        os << name << '{' << filename << "}\n";
@@ -1959,15 +2189,6 @@ void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
                        end_inset(os);
                }
 
-               else if (t.cs() == "fancyhead") {
-                       context.check_layout(os);
-                       ostringstream ss;
-                       ss << "\\fancyhead";
-                       ss << p.getOpt();
-                       ss << '{' << p.verbatim_item() << "}\n";
-                       handle_ert(os, ss.str(), context);
-               }
-
                else if (t.cs() == "bibliographystyle") {
                        // store new bibliographystyle
                        bibliographystyle = p.verbatim_item();
@@ -1992,7 +2213,7 @@ void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
                        parse_box(p, os, FLAG_ITEM, outer, context, true);
 
                else if (t.cs() == "smallskip" ||
-                        t.cs() == "medskip" ||
+                        t.cs() == "medskip" ||
                         t.cs() == "bigskip" ||
                         t.cs() == "vfill") {
                        context.check_layout(os);
@@ -2002,9 +2223,37 @@ void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
                        skip_braces(p);
                }
 
+               else if (is_known(t.cs(), known_spaces)) {
+                       char const * const * where = is_known(t.cs(), known_spaces);
+                       context.check_layout(os);
+                       begin_inset(os, "InsetSpace ");
+                       os << '\\' << known_coded_spaces[where - known_spaces]
+                          << '\n';
+                       // LaTeX swallows whitespace after all spaces except
+                       // "\\,". We have to do that here, too, because LyX
+                       // adds "{}" which would make the spaces significant.
+                       if (t.cs() !=  ",")
+                               eat_whitespace(p, os, context, false);
+                       // LyX adds "{}" after all spaces except "\\ " and
+                       // "\\,", so we have to remove "{}".
+                       // "\\,{}" is equivalent to "\\," in LaTeX, so we
+                       // remove the braces after "\\,", too.
+                       if (t.cs() != " ")
+                               skip_braces(p);
+               }
+
+               else if (t.cs() == "newpage" ||
+                        t.cs() == "clearpage" ||
+                        t.cs() == "cleardoublepage") {
+                       context.check_layout(os);
+                       // FIXME: what about \\pagebreak?
+                       os << "\n\\" << t.cs() << "\n";
+                       skip_braces(p); // eat {}
+               }
+
                else if (t.cs() == "newcommand" ||
-                        t.cs() == "providecommand" ||
-                        t.cs() == "renewcommand") {
+                        t.cs() == "providecommand" ||
+                        t.cs() == "renewcommand") {
                        // these could be handled by parse_command(), but
                        // we need to call add_known_command() here.
                        string name = t.asInput();
@@ -2018,8 +2267,8 @@ void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
                        string const opt2 = p.getFullOpt();
                        add_known_command(command, opt1, !opt2.empty());
                        string const ert = name + '{' + command + '}' +
-                                          opt1 + opt2 +
-                                          '{' + p.verbatim_item() + '}';
+                                          opt1 + opt2 +
+                                          '{' + p.verbatim_item() + '}';
                        handle_ert(os, ert, context);
                }
 
@@ -2134,5 +2383,7 @@ void parse_text(Parser & p, ostream & os, unsigned flags, bool outer,
        }
 }
 
-
 // }])
+
+
+} // namespace lyx