]> git.lyx.org Git - lyx.git/blobdiff - src/Paragraph.cpp
Assure correct spacing of colored items in mathed
[lyx.git] / src / Paragraph.cpp
index c37220effc402ec6155e20329968bdd25d957403..f2d97fe7c97cb9dbc60eac15e39e056391b14ec8 100644 (file)
 #include "OutputParams.h"
 #include "output_latex.h"
 #include "output_xhtml.h"
+#include "output_docbook.h"
 #include "ParagraphParameters.h"
 #include "SpellChecker.h"
-#include "sgml.h"
+#include "xml.h"
 #include "texstream.h"
 #include "TextClass.h"
 #include "TexRow.h"
@@ -63,6 +64,7 @@
 #include "support/lassert.h"
 #include "support/lstrings.h"
 #include "support/textutils.h"
+#include "output_docbook.h"
 
 #include <atomic>
 #include <sstream>
@@ -82,14 +84,6 @@ using namespace lyx::support;
 
 namespace lyx {
 
-namespace {
-
-/// Inset identifier (above 0x10ffff, for ucs-4)
-char_type const META_INSET = 0x200001;
-
-} // namespace
-
-
 /////////////////////////////////////////////////////////////////////
 //
 // SpellResultRange
@@ -272,9 +266,9 @@ private:
        Ranges ranges_;
        /// the area of the paragraph with pending spell check
        FontSpan refresh_;
-       bool needs_refresh_;
        /// spell state cache version number
        SpellChecker::ChangeNumber current_change_number_;
+       bool needs_refresh_;
 
 
        void correctRangesAfterPos(pos_type pos, int offset)
@@ -314,25 +308,21 @@ public:
 
        /// Output the surrogate pair formed by \p c and \p next to \p os.
        /// \return the number of characters written.
-       int latexSurrogatePair(otexstream & os, char_type c, char_type next,
+       int latexSurrogatePair(BufferParams const &, otexstream & os,
+                              char_type c, char_type next,
                               OutputParams const &);
 
        /// Output a space in appropriate formatting (or a surrogate pair
        /// if the next character is a combining character).
        /// \return whether a surrogate pair was output.
-       bool simpleTeXBlanks(OutputParams const &,
+       bool simpleTeXBlanks(BufferParams const &,
+                            OutputParams const &,
                             otexstream &,
                             pos_type i,
                             unsigned int & column,
                             Font const & font,
                             Layout const & style);
 
-       /// Output consecutive unicode chars, belonging to the same script as
-       /// specified by the latex macro \p ltx, to \p os starting from \p i.
-       /// \return the number of characters written.
-       int writeScriptChars(otexstream & os, docstring const & ltx,
-                          Change const &, Encoding const &, pos_type & i);
-
        /// This could go to ParagraphParameters if we want to.
        int startTeXParParams(BufferParams const &, otexstream &,
                              OutputParams const &) const;
@@ -360,7 +350,7 @@ public:
                                   BufferParams const & bparams,
                                   OutputParams const & runparams,
                                   Font const & running_font,
-                                  Change const & running_change,
+                                  string & alien_script,
                                   Layout const & style,
                                   pos_type & i,
                                   pos_type end_pos,
@@ -449,10 +439,10 @@ public:
                        return;
                }
                pos_type endpos = last;
-               owner_->locateWord(first, endpos, WHOLE_WORD);
+               owner_->locateWord(first, endpos, WHOLE_WORD, true);
                if (endpos < last) {
                        endpos = last;
-                       owner_->locateWord(last, endpos, WHOLE_WORD);
+                       owner_->locateWord(last, endpos, WHOLE_WORD, true);
                }
                last = endpos;
        }
@@ -462,7 +452,7 @@ public:
        {
                int numskips = 0;
                while (it != et && it->first < start) {
-                       int skip = it->last - it->first + 1;
+                       long skip = it->last - it->first + 1;
                        start += skip;
                        numskips += skip;
                        ++it;
@@ -489,9 +479,6 @@ public:
        ///
        FontList fontlist_;
 
-       ///
-       int id_;
-
        ///
        ParagraphParameters params_;
 
@@ -516,11 +503,13 @@ public:
        Layout const * layout_;
        ///
        SpellCheckerState speller_state_;
+       ///
+       int id_;
 };
 
 
 Paragraph::Private::Private(Paragraph * owner, Layout const & layout)
-       : owner_(owner), inset_owner_(0), id_(-1), begin_of_body_(0), layout_(&layout)
+       : owner_(owner), inset_owner_(nullptr), begin_of_body_(0), layout_(&layout), id_(-1)
 {
        text_.reserve(100);
 }
@@ -533,17 +522,16 @@ int Paragraph::Private::make_id()
        // LFUN_PARAGRAPH_GOTO to switch to a different buffer, for instance in the
        // outliner.
        // (thread-safe)
-       static atomic_uint next_id(0);
+       static int next_id(0);
        return next_id++;
 }
 
 
 Paragraph::Private::Private(Private const & p, Paragraph * owner)
        : owner_(owner), inset_owner_(p.inset_owner_), fontlist_(p.fontlist_),
-         id_(make_id()),
          params_(p.params_), changes_(p.changes_), insetlist_(p.insetlist_),
          begin_of_body_(p.begin_of_body_), text_(p.text_), words_(p.words_),
-         layout_(p.layout_)
+         layout_(p.layout_), id_(make_id())
 {
        requestSpellCheck(p.text_.size());
 }
@@ -551,11 +539,11 @@ Paragraph::Private::Private(Private const & p, Paragraph * owner)
 
 Paragraph::Private::Private(Private const & p, Paragraph * owner,
        pos_type beg, pos_type end)
-       : owner_(owner), inset_owner_(p.inset_owner_), id_(make_id()),
+       : owner_(owner), inset_owner_(p.inset_owner_),
          params_(p.params_), changes_(p.changes_),
          insetlist_(p.insetlist_, beg, end),
          begin_of_body_(p.begin_of_body_), words_(p.words_),
-         layout_(p.layout_)
+         layout_(p.layout_), id_(make_id())
 {
        if (beg >= pos_type(p.text_.size()))
                return;
@@ -585,18 +573,6 @@ void Paragraph::addChangesToToc(DocIterator const & cdit, Buffer const & buf,
 }
 
 
-void Paragraph::addChangesToBuffer(Buffer const & buf) const
-{
-       d->changes_.updateBuffer(buf);
-}
-
-
-bool Paragraph::isChangeUpdateRequired() const
-{
-       return d->changes_.isUpdateRequired();
-}
-
-
 bool Paragraph::isDeleted(pos_type start, pos_type end) const
 {
        LASSERT(start >= 0 && start <= size(), return false);
@@ -614,6 +590,28 @@ bool Paragraph::isChanged(pos_type start, pos_type end) const
        return d->changes_.isChanged(start, end);
 }
 
+// FIXME: Ideally the diverse isChanged() methods should account for that!
+bool Paragraph::hasChangedInsets(pos_type start, pos_type end) const
+{
+       LASSERT(start >= 0 && start <= size(), return false);
+       LASSERT(end > start && end <= size() + 1, return false);
+
+       for (auto const & icit : d->insetlist_) {
+               if (icit.pos < start)
+                       continue;
+               if (icit.pos >= end)
+                       break;
+               if (icit.inset && icit.inset->isChanged())
+                       return true;
+       }
+       return false;
+}
+
+bool Paragraph::isChanged() const
+{
+       return d->changes_.isChanged();
+}
+
 
 bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const
 {
@@ -625,6 +623,11 @@ bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const
        return change.inserted() && change.currentAuthor();
 }
 
+Change Paragraph::parEndChange() const
+{
+       return d->changes_.lookup(size());
+}
+
 
 void Paragraph::setChange(Change const & change)
 {
@@ -659,8 +662,8 @@ void Paragraph::setChange(pos_type pos, Change const & change)
 
        // see comment in setChange(Change const &) above
        if (!change.deleted() && pos < size())
-                       if (Inset * inset = getInset(pos))
-                               inset->setChange(change);
+               if (Inset * inset = getInset(pos))
+                       inset->setChange(change);
 }
 
 
@@ -864,9 +867,10 @@ int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
        return end - i;
 }
 
-
-int Paragraph::Private::latexSurrogatePair(otexstream & os, char_type c,
-               char_type next, OutputParams const & runparams)
+// Handle combining characters
+int Paragraph::Private::latexSurrogatePair(BufferParams const & bparams,
+               otexstream & os, char_type c, char_type next,
+               OutputParams const & runparams)
 {
        // Writing next here may circumvent a possible font change between
        // c and next. Since next is only output if it forms a surrogate pair
@@ -883,23 +887,22 @@ int Paragraph::Private::latexSurrogatePair(otexstream & os, char_type c,
                        latex1 = from_ascii(tipashortcut);
                }
        }
-       docstring const latex2 = encoding.latexChar(c).first;
-       if (docstring(1, next) == latex1) {
-               // the encoding supports the combination
+       docstring latex2 = encoding.latexChar(c).first;
+
+       if (bparams.useNonTeXFonts || docstring(1, next) == latex1) {
+               // Encoding supports the combination:
+               // output as is (combining char after base char).
                os << latex2 << latex1;
                return latex1.length() + latex2.length();
-       } else if (runparams.local_font &&
-                  runparams.local_font->language()->lang() == "polutonikogreek") {
-               // polutonikogreek only works without the brackets
-               os << latex1 << latex2;
-               return latex1.length() + latex2.length();
-       } else
-               os << latex1 << '{' << latex2 << '}';
+       }
+
+       os << latex1 << "{" << latex2 << "}";
        return latex1.length() + latex2.length() + 2;
 }
 
 
-bool Paragraph::Private::simpleTeXBlanks(OutputParams const & runparams,
+bool Paragraph::Private::simpleTeXBlanks(BufferParams const & bparams,
+                                      OutputParams const & runparams,
                                       otexstream & os,
                                       pos_type i,
                                       unsigned int & column,
@@ -913,7 +916,7 @@ bool Paragraph::Private::simpleTeXBlanks(OutputParams const & runparams,
                char_type next = text_[i + 1];
                if (Encodings::isCombiningChar(next)) {
                        // This space has an accent, so we must always output it.
-                       column += latexSurrogatePair(os, ' ', next, runparams) - 1;
+                       column += latexSurrogatePair(bparams, os, ' ', next, runparams) - 1;
                        return true;
                }
        }
@@ -944,86 +947,6 @@ bool Paragraph::Private::simpleTeXBlanks(OutputParams const & runparams,
 }
 
 
-int Paragraph::Private::writeScriptChars(otexstream & os,
-                                        docstring const & ltx,
-                                        Change const & runningChange,
-                                        Encoding const & encoding,
-                                        pos_type & i)
-{
-       // FIXME: modifying i here is not very nice...
-
-       // We only arrive here when character text_[i] could not be translated
-       // into the current latex encoding (or its latex translation has been forced,)
-       // and it belongs to a known script.
-       // TODO: We need \textcyr and \textgreek wrappers also for characters
-       //       that can be encoded in the "LaTeX encoding" but not in the
-       //       current *font encoding*.
-       //       (See #9681 for details and test)
-       // Parameter ltx contains the latex translation of text_[i] as specified
-       // in the unicodesymbols file and is something like "\textXXX{<spec>}".
-       // The latex macro name "textXXX" specifies the script to which text_[i]
-       // belongs and we use it in order to check whether characters from the
-       // same script immediately follow, such that we can collect them in a
-       // single "\textXXX" macro. So, we have to retain "\textXXX{<spec>"
-       // for the first char but only "<spec>" for all subsequent chars.
-       docstring::size_type const brace1 = ltx.find_first_of(from_ascii("{"));
-       docstring::size_type const brace2 = ltx.find_last_of(from_ascii("}"));
-       string script = to_ascii(ltx.substr(1, brace1 - 1));
-       int pos = 0;
-       int length = brace2;
-       bool closing_brace = true;
-       if (script == "textgreek" && encoding.latexName() == "iso-8859-7") {
-               // Correct encoding is being used, so we can avoid \textgreek.
-               // TODO: wrong test: we need to check the *font encoding*
-               //       (i.e. the active language and its FontEncoding tag)
-               //       instead of the LaTeX *input encoding*!
-               //       See #9637 for details and test-cases.
-               pos = brace1 + 1;
-               length -= pos;
-               closing_brace = false;
-       }
-       os << ltx.substr(pos, length);
-       int size = text_.size();
-       while (i + 1 < size) {
-               char_type const next = text_[i + 1];
-               // Stop here if next character belongs to another script
-               // or there is a change in change tracking status.
-               if (!Encodings::isKnownScriptChar(next, script) ||
-                   runningChange != owner_->lookupChange(i + 1))
-                       break;
-               Font prev_font;
-               bool found = false;
-               FontList::const_iterator cit = fontlist_.begin();
-               FontList::const_iterator end = fontlist_.end();
-               for (; cit != end; ++cit) {
-                       if (cit->pos() >= i && !found) {
-                               prev_font = cit->font();
-                               found = true;
-                       }
-                       if (cit->pos() >= i + 1)
-                               break;
-               }
-               // Stop here if there is a font attribute or encoding change.
-               if (found && cit != end && prev_font != cit->font())
-                       break;
-               docstring const latex = encoding.latexChar(next).first;
-               docstring::size_type const b1 =
-                                       latex.find_first_of(from_ascii("{"));
-               docstring::size_type const b2 =
-                                       latex.find_last_of(from_ascii("}"));
-               int const len = b2 - b1 - 1;
-               os << latex.substr(b1 + 1, len);
-               length += len;
-               ++i;
-       }
-       if (closing_brace) {
-               os << '}';
-               ++length;
-       }
-       return length;
-}
-
-
 void Paragraph::Private::latexInset(BufferParams const & bparams,
                                    otexstream & os,
                                    OutputParams & runparams,
@@ -1089,14 +1012,16 @@ void Paragraph::Private::latexInset(BufferParams const & bparams,
        bool close = false;
        odocstream::pos_type const len = os.os().tellp();
 
-       if (inset->forceLTR()
-           && !runparams.use_polyglossia
+       if (inset->forceLTR(runparams)
            && running_font.isRightToLeft()
            // ERT is an exception, it should be output with no
            // decorations at all
            && inset->lyxCode() != ERT_CODE) {
-               if (running_font.language()->lang() == "farsi")
-                       os << "\\beginL" << termcmd;
+               if (runparams.use_polyglossia) {
+                       os << "\\LRE{";
+               } else if (running_font.language()->lang() == "farsi"
+                          || running_font.language()->lang() == "arabic_arabi")
+                       os << "\\textLR{" << termcmd;
                else
                        os << "\\L{";
                close = true;
@@ -1118,9 +1043,11 @@ void Paragraph::Private::latexInset(BufferParams const & bparams,
                bool needPar = false;
                bool closeLanguage = arabtex
                        || basefont.isRightToLeft() == running_font.isRightToLeft();
+               // We pass non_inherit_inset = true here since size switches
+               // ought not to be terminated here (#8384).
                unsigned int count = running_font.latexWriteEndChanges(os,
                                        bparams, runparams, basefont, basefont,
-                                       needPar, closeLanguage);
+                                       needPar, closeLanguage, true);
                column += count;
                // if any font properties were closed, update the running_font,
                // making sure, however, to leave the language as it was
@@ -1132,8 +1059,15 @@ void Paragraph::Private::latexInset(BufferParams const & bparams,
                        running_font = basefont;
                        if (!closeLanguage)
                                running_font.setLanguage(copy_font.language());
-                       // leave font open if language is still open
-                       open_font = (running_font.language() == basefont.language());
+                       // For these, we use switches, so no need to close
+                       basefont.fontInfo().setSize(copy_font.fontInfo().size());
+                       basefont.fontInfo().setFamily(copy_font.fontInfo().family());
+                       basefont.fontInfo().setSeries(copy_font.fontInfo().series());
+                       // leave font open if language or any of the switches is still open
+                       open_font = (running_font.language() == basefont.language()
+                                    || running_font.fontInfo().size() == basefont.fontInfo().size()
+                                    || running_font.fontInfo().family() == basefont.fontInfo().family()
+                                    || running_font.fontInfo().series() == basefont.fontInfo().series());
                        if (closeLanguage)
                                runparams.local_font = &basefont;
                }
@@ -1152,12 +1086,8 @@ void Paragraph::Private::latexInset(BufferParams const & bparams,
                throw(e);
        }
 
-       if (close) {
-               if (running_font.language()->lang() == "farsi")
-                               os << "\\endL" << termcmd;
-                       else
-                               os << '}';
-       }
+       if (close)
+               os << '}';
 
        if (os.texrow().rows() > previous_row_count) {
                os.texrow().start(owner_->id(), i + 1);
@@ -1175,16 +1105,13 @@ void Paragraph::Private::latexSpecialChar(otexstream & os,
                                          BufferParams const & bparams,
                                          OutputParams const & runparams,
                                          Font const & running_font,
-                                         Change const & running_change,
+                                         string & alien_script,
                                          Layout const & style,
                                          pos_type & i,
                                          pos_type end_pos,
                                          unsigned int & column)
 {
-       // With polyglossia, brackets and stuff need not be reversed
-       // in RTL scripts (see bug #8251)
-       char_type const c = (runparams.use_polyglossia) ?
-               owner_->getUChar(bparams, i) : text_[i];
+       char_type const c = owner_->getUChar(bparams, runparams, i);
 
        if (style.pass_thru || runparams.pass_thru
            || contains(style.pass_thru_chars, c)
@@ -1210,7 +1137,7 @@ void Paragraph::Private::latexSpecialChar(otexstream & os,
            && !runparams.isFullUnicode() && bparams.main_font_encoding() == "T1"
            && latexSpecialT1(c, os, i, column))
                return;
-       // NOTE: XeTeX and LuaTeX use EU1/2 (pre 2017) or TU (as of 2017) encoding
+       // NOTE: "fontspec" (non-TeX fonts) sets the font encoding to "TU" (untill 2017 "EU1" or "EU2")
        else if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
                 && runparams.isFullUnicode() && latexSpecialTU(c, os, i, column))
                     return;
@@ -1238,10 +1165,8 @@ void Paragraph::Private::latexSpecialChar(otexstream & os,
                if (i + 1 < static_cast<pos_type>(text_.size()) &&
                    (end_pos == -1 || i + 1 < end_pos) &&
                    text_[i+1] == '-') {
-                       // Prevent "--" becoming an endash and "---" becoming
-                       // an emdash.
-                       // Within \ttfamily, "--" is merged to "-" (no endash)
-                       // so we avoid this rather irritating ligature as well
+                       // Prevent "--" becoming an en dash and "---" an em dash.
+                       // (Within \ttfamily, "---" is merged to en dash + hyphen.)
                        os << "{}";
                        column += 2;
                }
@@ -1292,6 +1217,9 @@ void Paragraph::Private::latexSpecialChar(otexstream & os,
                // XeTeX's dash behaviour is determined via a global setting
                if (bparams.use_dash_ligatures
                    && owner_->getFontSettings(bparams, i).fontInfo().family() != TYPEWRITER_FAMILY
+                   && !runparams.inIPA
+                       // TODO #10961: && not in inset Flex Code
+                       // TODO #10961: && not in layout LyXCode
                    && (!bparams.useNonTeXFonts || runparams.flavor != OutputParams::XETEX)) {
                        if (c == 0x2013) {
                                // en-dash
@@ -1314,17 +1242,16 @@ void Paragraph::Private::latexSpecialChar(otexstream & os,
                if (i + 1 < int(text_.size())) {
                        next = text_[i + 1];
                        if (Encodings::isCombiningChar(next)) {
-                               column += latexSurrogatePair(os, c, next, runparams) - 1;
+                               column += latexSurrogatePair(bparams, os, c, next, runparams) - 1;
                                ++i;
                                break;
                        }
                }
-               string script;
                pair<docstring, bool> latex = encoding.latexChar(c);
                docstring nextlatex;
                bool nexttipas = false;
                string nexttipashortcut;
-               if (next != '\0' && next != META_INSET && encoding.encodable(next)) {
+               if (next != '\0' && next != META_INSET && !encoding.encodable(next)) {
                        nextlatex = encoding.latexChar(next).first;
                        if (runparams.inIPA) {
                                nexttipashortcut = Encodings::TIPAShortcut(next);
@@ -1340,19 +1267,22 @@ void Paragraph::Private::latexSpecialChar(otexstream & os,
                                tipas = true;
                        }
                }
-               if (Encodings::isKnownScriptChar(c, script)
-                   && prefixIs(latex.first, from_ascii("\\" + script)))
-                       column += writeScriptChars(os, latex.first,
-                                       running_change, encoding, i) - 1;
-               else if (latex.second
+               // eventually close "script wrapper" command (see `Paragraph::latex`)
+               if (!alien_script.empty()
+                       && alien_script != Encodings::isKnownScriptChar(next)) {
+                       column += latex.first.length();
+                       alien_script.clear();
+                       os << latex.first << "}";
+                       break;
+               }
+               if (latex.second
                         && ((!prefixIs(nextlatex, '\\')
                               && !prefixIs(nextlatex, '{')
                               && !prefixIs(nextlatex, '}'))
                             || (nexttipas
                                 && !prefixIs(from_ascii(nexttipashortcut), '\\')))
                         && !tipas) {
-                       // Prevent eating of a following
-                       // space or command corruption by
+                       // Prevent eating of a following space or command corruption by
                        // following characters
                        if (next == ' ' || next == '\0') {
                                column += latex.first.length() + 1;
@@ -1420,14 +1350,6 @@ bool Paragraph::Private::latexSpecialT3(char_type const c, otexstream & os,
                os << "\\textvertline" << termcmd;
                column += 14;
                return true;
-       case 0x2013:
-               os << "\\textendash" << termcmd;
-               column += 12;
-               return true;
-       case 0x2014:
-               os << "\\textemdash" << termcmd;
-               column += 12;
-               return true;
        default:
                return false;
        }
@@ -1436,44 +1358,20 @@ bool Paragraph::Private::latexSpecialT3(char_type const c, otexstream & os,
 
 void Paragraph::Private::validate(LaTeXFeatures & features) const
 {
-       Buffer const & buf = inset_owner_->buffer();
-       BufferParams const & bp = features.runparams().is_child
-               ? buf.masterParams() : buf.params();
        if (layout_->inpreamble && inset_owner_) {
-               bool const is_command = layout_->latextype == LATEX_COMMAND;
-               Font f;
-               // Using a string stream here circumvents the encoding
+               // FIXME: Using a string stream here circumvents the encoding
                // switching machinery of odocstream. Therefore the
                // output is wrong if this paragraph contains content
                // that needs to switch encoding.
+               Buffer const & buf = inset_owner_->buffer();
                otexstringstream os;
                os << layout_->preamble();
-               if (is_command) {
-                       os << '\\' << from_ascii(layout_->latexname());
-                       // we have to provide all the optional arguments here, even though
-                       // the last one is the only one we care about.
-                       // Separate handling of optional argument inset.
-                       if (!layout_->latexargs().empty()) {
-                               OutputParams rp = features.runparams();
-                               rp.local_font = &owner_->getFirstFontSettings(bp);
-                               latexArgInsets(*owner_, os, rp, layout_->latexargs());
-                       }
-                       os << from_ascii(layout_->latexparam());
-               }
                size_t const length = os.length();
-               // this will output "{" at the beginning, but not at the end
-               owner_->latex(bp, f, os, features.runparams(), 0, -1, true);
-               if (os.length() > length) {
-                       if (is_command) {
-                               os << '}';
-                               if (!layout_->postcommandargs().empty()) {
-                                       OutputParams rp = features.runparams();
-                                       rp.local_font = &owner_->getFirstFontSettings(bp);
-                                       latexArgInsets(*owner_, os, rp, layout_->postcommandargs(), "post:");
-                               }
-                       }
+               TeXOnePar(buf, *inset_owner_->getText(int(buf.getParFromID(owner_->id()).idx())),
+                         buf.getParFromID(owner_->id()).pit(), os,
+                         features.runparams(), string(), 0, -1, true);
+               if (os.length() > length)
                        features.addPreambleSnippet(os.release(), true);
-               }
        }
 
        if (features.runparams().flavor == OutputParams::HTML
@@ -1520,19 +1418,58 @@ void Paragraph::Private::validate(LaTeXFeatures & features) const
        }
 
        // then the contents
+       BufferParams const bp = features.runparams().is_child
+               ? features.buffer().masterParams() : features.buffer().params();
        for (pos_type i = 0; i < int(text_.size()) ; ++i) {
                char_type c = text_[i];
+               CharInfo const & ci = Encodings::unicodeCharInfo(c);
                if (c == 0x0022) {
                        if (features.runparams().isFullUnicode() && bp.useNonTeXFonts)
                                features.require("textquotedblp");
                        else if (bp.main_font_encoding() != "T1"
                                 || ((&owner_->getFontSettings(bp, i))->language()->internalFontEncoding()))
                                features.require("textquotedbl");
-               }
-               if (!bp.use_dash_ligatures
-                   && (c == 0x2013 || c == 0x2014)
-                   && bp.useNonTeXFonts
-                   && features.runparams().flavor == OutputParams::XETEX)
+               } else if (ci.textfeature() && contains(ci.textpreamble(), '=')) {
+                       // features that depend on the font or input encoding
+                       string feats = ci.textpreamble();
+                       string fontenc = (&owner_->getFontSettings(bp, i))->language()->fontenc(bp);
+                       if (fontenc.empty())
+                               fontenc = features.runparams().main_fontenc;
+                       while (!feats.empty()) {
+                               string feat;
+                               feats = split(feats, feat, ',');
+                               if (contains(feat, "!=")) {
+                                       // a feature that is required except for the spcified
+                                       // font or input encodings
+                                       string realfeature;
+                                       string const contexts = ltrim(split(feat, realfeature, '!'), "=");
+                                       // multiple encodings are separated by semicolon
+                                       vector<string> context = getVectorFromString(contexts, ";");
+                                       // require feature if the context matches neither current font
+                                       // nor input encoding
+                                       if (std::find(context.begin(), context.end(), fontenc) == context.end()
+                                           && std::find(context.begin(), context.end(),
+                                                        features.runparams().encoding->name()) == context.end())
+                                               features.require(realfeature);
+                               } else if (contains(feat, '=')) {
+                                       // a feature that is required only for the spcified
+                                       // font or input encodings
+                                       string realfeature;
+                                       string const contexts = split(feat, realfeature, '=');
+                                       // multiple encodings are separated by semicolon
+                                       vector<string> context = getVectorFromString(contexts, ";");
+                                       // require feature if the context matches either current font
+                                       // or input encoding
+                                       if (std::find(context.begin(), context.end(), fontenc) != context.end()
+                                           || std::find(context.begin(), context.end(),
+                                                        features.runparams().encoding->name()) != context.end())
+                                               features.require(realfeature);
+                               }
+                       }
+               } else if (!bp.use_dash_ligatures
+                          && (c == 0x2013 || c == 0x2014)
+                          && bp.useNonTeXFonts
+                          && features.runparams().flavor == OutputParams::XETEX)
                        // XeTeX's dash behaviour is determined via a global setting
                        features.require("xetexdashbreakstate");
                BufferEncodings::validate(c, features);
@@ -1724,6 +1661,12 @@ void Paragraph::write(ostream & os, BufferParams const & bparams,
 void Paragraph::validate(LaTeXFeatures & features) const
 {
        d->validate(features);
+       bool fragile = features.runparams().moving_arg;
+       fragile |= layout().needprotect;
+       if (inInset().getLayout().isNeedProtect())
+               fragile = true;
+       if (needsCProtection(fragile))
+               features.require("cprotect");
 }
 
 
@@ -1822,7 +1765,7 @@ Font const & Paragraph::getFontSettings(BufferParams const & bparams,
        // Optimisation: avoid a full font instantiation if there is no
        // language change from previous call.
        static Font previous_font;
-       static Language const * previous_lang = 0;
+       static Language const * previous_lang = nullptr;
        Language const * lang = getParLanguage(bparams);
        if (lang != previous_lang) {
                previous_lang = lang;
@@ -1870,7 +1813,7 @@ Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
        // Optimisation: avoid a full font instantiation if there is no
        // language change from previous call.
        static Font previous_font;
-       static Language const * previous_lang = 0;
+       static Language const * previous_lang = nullptr;
        if (bparams.language != previous_lang) {
                previous_lang = bparams.language;
                previous_font = Font(inherit_font, bparams.language);
@@ -1926,32 +1869,59 @@ Font const Paragraph::getLayoutFont
 }
 
 
-char_type Paragraph::getUChar(BufferParams const & bparams, pos_type pos) const
+char_type Paragraph::getUChar(BufferParams const & bparams,
+                             OutputParams const & rp,
+                             pos_type pos) const
 {
        char_type c = d->text_[pos];
-       if (!getFontSettings(bparams, pos).isRightToLeft())
+
+       // Return unchanged character in LTR languages
+       // or if we use poylglossia/bidi (XeTeX).
+       if (rp.useBidiPackage()
+           || !getFontSettings(bparams, pos).isRightToLeft())
                return c;
 
-       // FIXME: The arabic special casing is due to the difference of arabic
-       // round brackets input introduced in r18599. Check if this should be
-       // unified with Hebrew or at least if all bracket types should be
-       // handled the same (file format change in either case).
+       // Without polyglossia/bidi, we need to account for some special cases.
+       // FIXME This needs to be audited!
+       // Check if:
+       // * The input is as expected for all delimiters
+       //   => checked for Hebrew!
+       // * The output matches the display in the LyX workarea
+       //   => checked for Hebrew!
+       // * The special cases below are really necessary
+       //   => checked for Hebrew!
+       // * In arabic_arabi, brackets are transformed to Arabic
+       //   Ornate Parentheses. Is this is really wanted?
+
        string const & lang = getFontSettings(bparams, pos).language()->lang();
-       bool const arabic = lang == "arabic_arabtex" || lang == "arabic_arabi"
-               || lang == "farsi";
        char_type uc = c;
+
+       // 1. In the following languages, parentheses need to be reversed.
+       //    Also with polyglodia/luabidi
+       bool const reverseparens = (lang == "hebrew" || rp.use_polyglossia);
+
+       // 2. In the following languages, brackets don't need to be reversed.
+       bool const reversebrackets = lang != "arabic_arabtex"
+                       && lang != "arabic_arabi"
+                       && lang != "farsi";
+
+       // Now swap delimiters if needed.
        switch (c) {
        case '(':
-               uc = arabic ? c : ')';
+               if (reverseparens)
+                       uc = ')';
                break;
        case ')':
-               uc = arabic ? c : '(';
+               if (reverseparens)
+                       uc = '(';
                break;
        case '[':
-               uc = ']';
+               if (reversebrackets)
+                       uc = ']';
                break;
        case ']':
-               uc = '[';
+               if (reversebrackets)
+                       uc = '[';
                break;
        case '{':
                uc = '}';
@@ -2031,15 +2001,29 @@ depth_type Paragraph::getMaxDepthAfter() const
 }
 
 
-LyXAlignment Paragraph::getAlign() const
+LyXAlignment Paragraph::getAlign(BufferParams const & bparams) const
 {
        if (d->params_.align() == LYX_ALIGN_LAYOUT)
-               return d->layout_->align;
+               return getDefaultAlign(bparams);
        else
                return d->params_.align();
 }
 
 
+LyXAlignment Paragraph::getDefaultAlign(BufferParams const & bparams) const
+{
+       LyXAlignment res = layout().align;
+       if (isRTL(bparams)) {
+               // Swap sides
+               if (res == LYX_ALIGN_LEFT)
+                       res = LYX_ALIGN_RIGHT;
+               else if  (res == LYX_ALIGN_RIGHT)
+                       res = LYX_ALIGN_LEFT;
+       }
+       return res;
+}
+
+
 docstring const & Paragraph::labelString() const
 {
        return d->params_.labelString();
@@ -2070,13 +2054,6 @@ docstring Paragraph::expandLabel(Layout const & layout,
 }
 
 
-docstring Paragraph::expandDocBookLabel(Layout const & layout,
-               BufferParams const & bparams) const
-{
-       return expandParagraphLabel(layout, bparams, false);
-}
-
-
 docstring Paragraph::expandParagraphLabel(Layout const & layout,
                BufferParams const & bparams, bool process_appendix) const
 {
@@ -2139,15 +2116,17 @@ void Paragraph::setBeginOfBody()
        // remove unnecessary getChar() calls
        pos_type i = 0;
        pos_type end = size();
-       if (i < end && !(isNewline(i) || isEnvSeparator(i))) {
+       bool prev_char_deleted = false;
+       if (i < end && (!(isNewline(i) || isEnvSeparator(i)) || isDeleted(i))) {
                ++i;
                if (i < end) {
                        char_type previous_char = d->text_[i];
                        if (!(isNewline(i) || isEnvSeparator(i))) {
                                ++i;
-                               while (i < end && previous_char != ' ') {
+                               while (i < end && (previous_char != ' ' || prev_char_deleted)) {
                                        char_type temp = d->text_[i];
-                                       if (isNewline(i) || isEnvSeparator(i))
+                                       prev_char_deleted = isDeleted(i);
+                                       if (!isDeleted(i) && (isNewline(i) || isEnvSeparator(i)))
                                                break;
                                        ++i;
                                        previous_char = temp;
@@ -2278,6 +2257,10 @@ int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
        string const begin_tag = "\\begin";
        InsetCode code = ownerCode();
        bool const lastpar = runparams.isLastPar;
+       // RTL in classic (PDF)LaTeX (without the Bidi package)
+       // Luabibdi (used by LuaTeX) behaves like classic
+       bool const rtl_classic = owner_->getParLanguage(bparams)->rightToLeft()
+               && !runparams.useBidiPackage();
 
        switch (curAlign) {
        case LYX_ALIGN_NONE:
@@ -2287,16 +2270,18 @@ int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
        case LYX_ALIGN_DECIMAL:
                break;
        case LYX_ALIGN_LEFT: {
-               if (owner_->getParLanguage(bparams)->babel() != "hebrew")
-                       corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
-               else
+               if (rtl_classic)
+                       // Classic (PDF)LaTeX switches the left/right logic in RTL mode
                        corrected_env(os, begin_tag, "flushright", code, lastpar, column);
+               else
+                       corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
                break;
        } case LYX_ALIGN_RIGHT: {
-               if (owner_->getParLanguage(bparams)->babel() != "hebrew")
-                       corrected_env(os, begin_tag, "flushright", code, lastpar, column);
-               else
+               if (rtl_classic)
+                       // Classic (PDF)LaTeX switches the left/right logic in RTL mode
                        corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
+               else
+                       corrected_env(os, begin_tag, "flushright", code, lastpar, column);
                break;
        } case LYX_ALIGN_CENTER: {
                corrected_env(os, begin_tag, "center", code, lastpar, column);
@@ -2336,6 +2321,10 @@ bool Paragraph::Private::endTeXParParams(BufferParams const & bparams,
        string const end_tag = "\\par\\end";
        InsetCode code = ownerCode();
        bool const lastpar = runparams.isLastPar;
+       // RTL in classic (PDF)LaTeX (without the Bidi package)
+       // Luabibdi (used by LuaTeX) behaves like classic
+       bool const rtl_classic = owner_->getParLanguage(bparams)->rightToLeft()
+               && !runparams.useBidiPackage();
 
        switch (curAlign) {
        case LYX_ALIGN_NONE:
@@ -2345,16 +2334,18 @@ bool Paragraph::Private::endTeXParParams(BufferParams const & bparams,
        case LYX_ALIGN_DECIMAL:
                break;
        case LYX_ALIGN_LEFT: {
-               if (owner_->getParLanguage(bparams)->babel() != "hebrew")
-                       output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
-               else
+               if (rtl_classic)
+                       // Classic (PDF)LaTeX switches the left/right logic in RTL mode
                        output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
+               else
+                       output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
                break;
        } case LYX_ALIGN_RIGHT: {
-               if (owner_->getParLanguage(bparams)->babel() != "hebrew")
-                       output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
-               else
+               if (rtl_classic)
+                       // Classic (PDF)LaTeX switches the left/right logic in RTL mode
                        output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
+               else
+                       output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
                break;
        } case LYX_ALIGN_CENTER: {
                corrected_env(os, end_tag, "center", code, lastpar, col);
@@ -2393,10 +2384,20 @@ void Paragraph::latex(BufferParams const & bparams,
        // of the body.
        Font basefont;
 
+       // If there is an open font-encoding changing command (script wrapper),
+       // alien_script is set to its name
+       string alien_script;
+       string script;
+
        // Maybe we have to create a optional argument.
        pos_type body_pos = beginOfBody();
        unsigned int column = 0;
 
+       // If we are inside an non inheritFont() inset, the real outerfont is local_font
+       Font const real_outerfont = (!inInset().inheritFont()
+                                    && runparams.local_font != nullptr)
+                       ? Font(runparams.local_font->fontInfo()) : outerfont;
+
        if (body_pos > 0) {
                // the optional argument is kept in curly brackets in
                // case it contains a ']'
@@ -2406,9 +2407,9 @@ void Paragraph::latex(BufferParams const & bparams,
                // braces when it parses \item.
                os << "[{";
                column += 2;
-               basefont = getLabelFont(bparams, outerfont);
+               basefont = getLabelFont(bparams, real_outerfont);
        } else {
-               basefont = getLayoutFont(bparams, outerfont);
+               basefont = getLayoutFont(bparams, real_outerfont);
        }
 
        // Which font is currently active?
@@ -2426,7 +2427,7 @@ void Paragraph::latex(BufferParams const & bparams,
        if (empty()) {
                // For InTitle commands, we have already opened a group
                // in output_latex::TeXOnePar.
-               if (style.isCommand() && (!style.intitle || style.inpreamble)) {
+               if (style.isCommand() && !style.intitle) {
                        os << '{';
                        ++column;
                }
@@ -2453,7 +2454,7 @@ void Paragraph::latex(BufferParams const & bparams,
                                                basefont, basefont, needPar);
                                        open_font = false;
                                }
-                               basefont = getLayoutFont(bparams, outerfont);
+                               basefont = getLayoutFont(bparams, real_outerfont);
                                running_font = basefont;
 
                                column += Changes::latexMarkChange(os, bparams,
@@ -2466,7 +2467,7 @@ void Paragraph::latex(BufferParams const & bparams,
                        }
                        // For InTitle commands, we have already opened a group
                        // in output_latex::TeXOnePar.
-                       if (style.isCommand() && (!style.intitle || style.inpreamble)) {
+                       if (style.isCommand() && !style.intitle) {
                                os << '{';
                                ++column;
                        }
@@ -2487,9 +2488,14 @@ void Paragraph::latex(BufferParams const & bparams,
                Change const & change = runparams.inDeletedInset
                        ? runparams.changeOfDeletedInset : lookupChange(i);
 
+               char_type const c = d->text_[i];
+
                // Check whether a display math inset follows
-               if (d->text_[i] == META_INSET
+               if (c == META_INSET
                    && i >= start_pos && (end_pos == -1 || i < end_pos)) {
+                       if (isDeleted(i))
+                               runparams.ctObject = getInset(i)->CtObject(runparams);
+       
                        InsetMath const * im = getInset(i)->asInsetMath();
                        if (im && im->asHullInset()
                            && im->asHullInset()->outerDisplay()) {
@@ -2515,7 +2521,8 @@ void Paragraph::latex(BufferParams const & bparams,
                                                basefont, needPar);
                                        open_font = false;
                                }
-                               basefont = getLayoutFont(bparams, outerfont);
+                               basefont = (body_pos > i) ? getLabelFont(bparams, real_outerfont)
+                                                         : getLayoutFont(bparams, real_outerfont);
                                running_font = basefont;
                                column += Changes::latexMarkChange(os, bparams,
                                        Change(Change::INSERTED), change, rp);
@@ -2523,6 +2530,11 @@ void Paragraph::latex(BufferParams const & bparams,
                }
 
                if (bparams.output_changes && runningChange != change) {
+                       if (!alien_script.empty()) {
+                               column += 1;
+                               os << "}";
+                               alien_script.clear();
+                       }
                        if (open_font) {
                                bool needPar = false;
                                column += running_font.latexWriteEndChanges(
@@ -2530,7 +2542,8 @@ void Paragraph::latex(BufferParams const & bparams,
                                                basefont, basefont, needPar);
                                open_font = false;
                        }
-                       basefont = getLayoutFont(bparams, outerfont);
+                       basefont = (body_pos > i) ? getLabelFont(bparams, real_outerfont)
+                                                 : getLayoutFont(bparams, real_outerfont);
                        running_font = basefont;
                        column += Changes::latexMarkChange(os, bparams, runningChange,
                                                           change, runparams);
@@ -2549,38 +2562,55 @@ void Paragraph::latex(BufferParams const & bparams,
                Font const current_font = getFont(bparams, i, outerfont);
 
                Font const last_font = running_font;
+               bool const in_ct_deletion = (bparams.output_changes
+                                            && runningChange == change
+                                            && change.type == Change::DELETED
+                                            && !os.afterParbreak());
 
                // Do we need to close the previous font?
                if (open_font &&
                    (current_font != running_font ||
                     current_font.language() != running_font.language()))
                {
+                       // ensure there is no open script-wrapper
+                       if (!alien_script.empty()) {
+                               column += 1;
+                               os << "}";
+                               alien_script.clear();
+                       }
                        bool needPar = false;
+                       if (in_ct_deletion) {
+                               // We have to close and then reopen \lyxdeleted,
+                               // as strikeout needs to be on lowest level.
+                               os << '}';
+                               column += 1;
+                       }
                        column += running_font.latexWriteEndChanges(
                                    os, bparams, runparams, basefont,
                                    (i == body_pos-1) ? basefont : current_font,
                                    needPar);
+                       if (in_ct_deletion) {
+                               // We have to close and then reopen \lyxdeleted,
+                               // as strikeout needs to be on lowest level.
+                               OutputParams rp = runparams;
+                               column += Changes::latexMarkChange(os, bparams,
+                                       Change(Change::UNCHANGED), Change(Change::DELETED), rp);
+                       }
                        running_font = basefont;
                        open_font = false;
                }
 
-               string const running_lang = runparams.use_polyglossia ?
-                       running_font.language()->polyglossia() : running_font.language()->babel();
-               // close babel's font environment before opening CJK.
-               string const lang_end_command = runparams.use_polyglossia ?
-                       "\\end{$$lang}" : lyxrc.language_command_end;
-               bool const using_begin_end = runparams.use_polyglossia ||
-                                               !lang_end_command.empty();
-               if (!running_lang.empty() &&
-                   (!using_begin_end || running_lang == openLanguageName()) &&
-                   current_font.language()->encoding()->package() == Encoding::CJK) {
-                               string end_tag = subst(lang_end_command,
-                                                       "$$lang",
-                                                       running_lang);
-                               os << from_ascii(end_tag);
-                               column += end_tag.length();
-                               if (using_begin_end)
-                                       popLanguageName();
+               // if necessary, close language environment before opening CJK
+               string const running_lang = running_font.language()->babel();
+               string const lang_end_command = lyxrc.language_command_end;
+               if (!lang_end_command.empty() && !bparams.useNonTeXFonts
+                       && !running_lang.empty()
+                       && running_lang == openLanguageName()
+                       && current_font.language()->encoding()->package() == Encoding::CJK) {
+                       string end_tag = subst(lang_end_command, "$$lang", running_lang);
+                       os << from_ascii(end_tag);
+                       column += end_tag.length();
+                       popLanguageName();
                }
 
                // Switch file encoding if necessary (and allowed)
@@ -2596,8 +2626,6 @@ void Paragraph::latex(BufferParams const & bparams,
                        }
                }
 
-               char_type const c = d->text_[i];
-
                // A display math inset inside an ulem command will be output
                // as a box of width \linewidth, so we have to either disable
                // indentation if the inset starts a paragraph, or start a new
@@ -2613,13 +2641,25 @@ void Paragraph::latex(BufferParams const & bparams,
 
                // Do we need to change font?
                if ((current_font != running_font ||
-                    current_font.language() != running_font.language()) &&
-                       i != body_pos - 1)
+                    current_font.language() != running_font.language())
+                   && i != body_pos - 1)
                {
-                       odocstringstream ods;
-                       column += current_font.latexWriteStartChanges(ods, bparams,
+                       if (in_ct_deletion) {
+                               // We have to close and then reopen \lyxdeleted,
+                               // as strikeout needs to be on lowest level.
+                               bool needPar = false;
+                               OutputParams rp = runparams;
+                               column += running_font.latexWriteEndChanges(
+                                       os, bparams, rp, basefont,
+                                       basefont, needPar);
+                               os << '}';
+                               column += 1;
+                       }
+                       otexstringstream ots;
+                       bool const non_inherit_inset = (c == META_INSET && getInset(i) && !getInset(i)->inheritFont());
+                       column += current_font.latexWriteStartChanges(ots, bparams,
                                                              runparams, basefont,
-                                                             last_font);
+                                                             last_font, non_inherit_inset);
                        // Check again for display math in ulem commands as a
                        // font change may also occur just before a math inset.
                        if (runparams.inDisplayMath && !deleted_display_math
@@ -2631,19 +2671,24 @@ void Paragraph::latex(BufferParams const & bparams,
                        }
                        running_font = current_font;
                        open_font = true;
-                       docstring fontchange = ods.str();
+                       docstring fontchange = ots.str();
+                       os << fontchange;
                        // check whether the fontchange ends with a \\textcolor
-                       // modifier and the text starts with a space (bug 4473)
+                       // modifier and the text starts with a space. If so we
+                       // need to add } in order to prevent \\textcolor from gobbling
+                       // the space (bug 4473).
                        docstring const last_modifier = rsplit(fontchange, '\\');
                        if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
-                               os << fontchange << from_ascii("{}");
-                       // check if the fontchange ends with a trailing blank
-                       // (like "\small " (see bug 3382)
-                       else if (suffixIs(fontchange, ' ') && c == ' ')
-                               os << fontchange.substr(0, fontchange.size() - 1)
-                                  << from_ascii("{}");
-                       else
-                               os << fontchange;
+                               os << from_ascii("{}");
+                       else if (ots.terminateCommand())
+                               os << termcmd;
+                       if (in_ct_deletion) {
+                               // We have to close and then reopen \lyxdeleted,
+                               // as strikeout needs to be on lowest level.
+                               OutputParams rp = runparams;
+                               column += Changes::latexMarkChange(os, bparams,
+                                       Change(Change::UNCHANGED), change, rp);
+                       }
                }
 
                // FIXME: think about end_pos implementation...
@@ -2654,7 +2699,7 @@ void Paragraph::latex(BufferParams const & bparams,
                        // latexSpecialChar ignores spaces if
                        // style.pass_thru is false.
                        if (i != body_pos - 1) {
-                               if (d->simpleTeXBlanks(runparams, os,
+                               if (d->simpleTeXBlanks(bparams, runparams, os,
                                                i, column, current_font, style)) {
                                        // A surrogate pair was output. We
                                        // must not call latexSpecialChar
@@ -2685,37 +2730,61 @@ void Paragraph::latex(BufferParams const & bparams,
                                Inset const * inset = getInset(i);
                                InsetText const * textinset = inset
                                                        ? inset->asInsetText()
-                                                       : 0;
+                                                       : nullptr;
                                if (i + 1 == size() && textinset
                                    && !inset->getLayout().isDisplay()) {
                                        ParagraphList const & pars =
                                                textinset->text().paragraphs();
                                        pit_type const pit = pars.size() - 1;
-                                       Font const last_font =
+                                       Font const lastfont =
                                                pit < 0 || pars[pit].empty()
                                                ? pars[pit].getLayoutFont(
                                                                bparams,
-                                                               outerfont)
+                                                               real_outerfont)
                                                : pars[pit].getFont(bparams,
                                                        pars[pit].size() - 1,
-                                                       outerfont);
-                                       if (last_font.fontInfo().size() !=
+                                                       real_outerfont);
+                                       if (lastfont.fontInfo().size() !=
                                            basefont.fontInfo().size()) {
                                                ++parInline;
                                                incremented = true;
                                        }
                                }
                                d->latexInset(bparams, os, rp, running_font,
-                                               basefont, outerfont, open_font,
+                                               basefont, real_outerfont, open_font,
                                                runningChange, style, i, column);
                                if (incremented)
                                        --parInline;
+
+                               if (runparams.ctObject == OutputParams::CT_DISPLAYOBJECT
+                                   || runparams.ctObject == OutputParams::CT_UDISPLAYOBJECT) {
+                                       // Close \lyx*deleted and force its
+                                       // reopening (if needed)
+                                       os << '}';
+                                       column++;
+                                       runningChange = Change(Change::UNCHANGED);
+                                       runparams.ctObject = OutputParams::CT_NORMAL;
+                               }
                        }
                } else if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
+                       if (!bparams.useNonTeXFonts)
+                         script = Encodings::isKnownScriptChar(c);
+                       if (script != alien_script) {
+                               if (!alien_script.empty()) {
+                                       os << "}";
+                                       alien_script.clear();
+                               }
+                               string fontenc = running_font.language()->fontenc(bparams);
+                               if (!script.empty()
+                                       && !Encodings::fontencSupportsScript(fontenc, script)) {
+                                       column += script.length() + 2;
+                                       os << "\\" << script << "{";
+                                       alien_script = script;
+                               }
+                       }
                        try {
-                               d->latexSpecialChar(os, bparams, rp,
-                                                   running_font, runningChange,
-                                                   style, i, end_pos, column);
+                               d->latexSpecialChar(os, bparams, rp, running_font,
+                                                                       alien_script, style, i, end_pos, column);
                        } catch (EncodingException & e) {
                                if (runparams.dryrun) {
                                        os << "<" << _("LyX Warning: ")
@@ -2739,6 +2808,19 @@ void Paragraph::latex(BufferParams const & bparams,
                // such as Note that do not produce any output, so that no
                // command is ever executed but its opening was recorded.
                runparams.inulemcmd = rp.inulemcmd;
+
+               // These need to be passed upstream as well
+               runparams.need_maketitle = rp.need_maketitle;
+               runparams.have_maketitle = rp.have_maketitle;
+
+               // And finally, pass the post_macros upstream
+               runparams.post_macro = rp.post_macro;
+       }
+
+       // Close wrapper for alien script
+       if (!alien_script.empty()) {
+               os << "}";
+               alien_script.clear();
        }
 
        // If we have an open font definition, we have to close it
@@ -2754,8 +2836,8 @@ void Paragraph::latex(BufferParams const & bparams,
                // since this produces unwanted whitespace.
 
                Font const font = empty()
-                       ? getLayoutFont(bparams, outerfont)
-                       : getFont(bparams, size() - 1, outerfont);
+                       ? getLayoutFont(bparams, real_outerfont)
+                       : getFont(bparams, size() - 1, real_outerfont);
 
                InsetText const * textinset = inInset().asInsetText();
 
@@ -2850,7 +2932,7 @@ bool Paragraph::emptyTag() const
 }
 
 
-string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
+string Paragraph::getID(Buffer const &, OutputParams const &)
        const
 {
        for (pos_type i = 0; i < size(); ++i) {
@@ -2859,7 +2941,7 @@ string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
                        if (lyx_code == LABEL_CODE) {
                                InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
                                docstring const & id = il->getParam("name");
-                               return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
+                               return "id='" + to_utf8(xml::cleanID(id)) + "'";
                        }
                }
        }
@@ -2867,25 +2949,24 @@ string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
 }
 
 
-pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runparams)
-       const
+pos_type Paragraph::firstWordDocBook(XMLStream & xs, OutputParams const & runparams) const
 {
        pos_type i;
        for (i = 0; i < size(); ++i) {
                if (Inset const * inset = getInset(i)) {
-                       inset->docbook(os, runparams);
+                       inset->docbook(xs, runparams);
                } else {
                        char_type c = d->text_[i];
                        if (c == ' ')
                                break;
-                       os << sgml::escapeChar(c);
+                       xs << c;
                }
        }
        return i;
 }
 
 
-pos_type Paragraph::firstWordLyXHTML(XHTMLStream & xs, OutputParams const & runparams)
+pos_type Paragraph::firstWordLyXHTML(XMLStream & xs, OutputParams const & runparams)
        const
 {
        pos_type i;
@@ -2920,78 +3001,354 @@ bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, po
 }
 
 
-void Paragraph::simpleDocBookOnePar(Buffer const & buf,
-                                   odocstream & os,
-                                   OutputParams const & runparams,
-                                   Font const & outerfont,
-                                   pos_type initial) const
+namespace {
+
+void doFontSwitchDocBook(vector<xml::FontTag> & tagsToOpen,
+                  vector<xml::EndFontTag> & tagsToClose,
+                  bool & flag, FontState curstate, xml::FontTypes type)
 {
+       if (curstate == FONT_ON) {
+               tagsToOpen.push_back(docbookStartFontTag(type));
+               flag = true;
+       } else if (flag) {
+               tagsToClose.push_back(docbookEndFontTag(type));
+               flag = false;
+       }
+}
+
+class OptionalFontType {
+public:
+       xml::FontTypes ft;
+       bool has_value;
+
+       OptionalFontType(): ft(xml::FT_EMPH), has_value(false) {} // A possible value at random for ft.
+       OptionalFontType(xml::FontTypes ft): ft(ft), has_value(true) {}
+};
+
+OptionalFontType fontShapeToXml(FontShape fs)
+{
+       switch (fs) {
+       case ITALIC_SHAPE:
+               return {xml::FT_ITALIC};
+       case SLANTED_SHAPE:
+               return {xml::FT_SLANTED};
+       case SMALLCAPS_SHAPE:
+               return {xml::FT_SMALLCAPS};
+       case UP_SHAPE:
+       case INHERIT_SHAPE:
+               return {};
+       default:
+               // the other tags are for internal use
+               LATTEST(false);
+               return {};
+       }
+}
+
+OptionalFontType fontFamilyToXml(FontFamily fm)
+{
+       switch (fm) {
+       case ROMAN_FAMILY:
+               return {xml::FT_ROMAN};
+       case SANS_FAMILY:
+               return {xml::FT_SANS};
+       case TYPEWRITER_FAMILY:
+               return {xml::FT_TYPE};
+       case INHERIT_FAMILY:
+               return {};
+       default:
+               // the other tags are for internal use
+               LATTEST(false);
+               return {};
+       }
+}
+
+OptionalFontType fontSizeToXml(FontSize fs)
+{
+       switch (fs) {
+       case TINY_SIZE:
+               return {xml::FT_SIZE_TINY};
+       case SCRIPT_SIZE:
+               return {xml::FT_SIZE_SCRIPT};
+       case FOOTNOTE_SIZE:
+               return {xml::FT_SIZE_FOOTNOTE};
+       case SMALL_SIZE:
+               return {xml::FT_SIZE_SMALL};
+       case LARGE_SIZE:
+               return {xml::FT_SIZE_LARGE};
+       case LARGER_SIZE:
+               return {xml::FT_SIZE_LARGER};
+       case LARGEST_SIZE:
+               return {xml::FT_SIZE_LARGEST};
+       case HUGE_SIZE:
+               return {xml::FT_SIZE_HUGE};
+       case HUGER_SIZE:
+               return {xml::FT_SIZE_HUGER};
+       case INCREASE_SIZE:
+               return {xml::FT_SIZE_INCREASE};
+       case DECREASE_SIZE:
+               return {xml::FT_SIZE_DECREASE};
+       case INHERIT_SIZE:
+       case NORMAL_SIZE:
+               return {};
+       default:
+               // the other tags are for internal use
+               LATTEST(false);
+               return {};
+       }
+}
+
+struct DocBookFontState
+{
+       FontShape  curr_fs   = INHERIT_SHAPE;
+       FontFamily curr_fam  = INHERIT_FAMILY;
+       FontSize   curr_size = INHERIT_SIZE;
+
+       // track whether we have opened these tags
        bool emph_flag = false;
+       bool bold_flag = false;
+       bool noun_flag = false;
+       bool ubar_flag = false;
+       bool dbar_flag = false;
+       bool sout_flag = false;
+       bool xout_flag = false;
+       bool wave_flag = false;
+       // shape tags
+       bool shap_flag = false;
+       // family tags
+       bool faml_flag = false;
+       // size tags
+       bool size_flag = false;
+};
+
+std::tuple<vector<xml::FontTag>, vector<xml::EndFontTag>> computeDocBookFontSwitch(FontInfo const & font_old,
+                                                                                          Font const & font,
+                                                                                          std::string const & default_family,
+                                                                                          DocBookFontState & fs)
+{
+       vector<xml::FontTag> tagsToOpen;
+       vector<xml::EndFontTag> tagsToClose;
+
+       // emphasis
+       FontState curstate = font.fontInfo().emph();
+       if (font_old.emph() != curstate)
+               doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.emph_flag, curstate, xml::FT_EMPH);
+
+       // noun
+       curstate = font.fontInfo().noun();
+       if (font_old.noun() != curstate)
+               doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.noun_flag, curstate, xml::FT_NOUN);
+
+       // underbar
+       curstate = font.fontInfo().underbar();
+       if (font_old.underbar() != curstate)
+               doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.ubar_flag, curstate, xml::FT_UBAR);
+
+       // strikeout
+       curstate = font.fontInfo().strikeout();
+       if (font_old.strikeout() != curstate)
+               doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.sout_flag, curstate, xml::FT_SOUT);
+
+       // xout
+       curstate = font.fontInfo().xout();
+       if (font_old.xout() != curstate)
+               doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.xout_flag, curstate, xml::FT_XOUT);
+
+       // double underbar
+       curstate = font.fontInfo().uuline();
+       if (font_old.uuline() != curstate)
+               doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.dbar_flag, curstate, xml::FT_DBAR);
+
+       // wavy line
+       curstate = font.fontInfo().uwave();
+       if (font_old.uwave() != curstate)
+               doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.wave_flag, curstate, xml::FT_WAVE);
+
+       // bold
+       // a little hackish, but allows us to reuse what we have.
+       curstate = (font.fontInfo().series() == BOLD_SERIES ? FONT_ON : FONT_OFF);
+       if (font_old.series() != font.fontInfo().series())
+               doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.bold_flag, curstate, xml::FT_BOLD);
+
+       // Font shape
+       fs.curr_fs = font.fontInfo().shape();
+       FontShape old_fs = font_old.shape();
+       if (old_fs != fs.curr_fs) {
+               if (fs.shap_flag) {
+                       OptionalFontType tag = fontShapeToXml(old_fs);
+                       if (tag.has_value)
+                               tagsToClose.push_back(docbookEndFontTag(tag.ft));
+                       fs.shap_flag = false;
+               }
+
+               OptionalFontType tag = fontShapeToXml(fs.curr_fs);
+               if (tag.has_value)
+                       tagsToOpen.push_back(docbookStartFontTag(tag.ft));
+       }
+
+       // Font family
+       fs.curr_fam = font.fontInfo().family();
+       FontFamily old_fam = font_old.family();
+       if (old_fam != fs.curr_fam) {
+               if (fs.faml_flag) {
+                       OptionalFontType tag = fontFamilyToXml(old_fam);
+                       if (tag.has_value)
+                               tagsToClose.push_back(docbookEndFontTag(tag.ft));
+                       fs.faml_flag = false;
+               }
+               switch (fs.curr_fam) {
+                       case ROMAN_FAMILY:
+                               // we will treat a "default" font family as roman, since we have
+                               // no other idea what to do.
+                               if (default_family != "rmdefault" && default_family != "default") {
+                                       tagsToOpen.push_back(docbookStartFontTag(xml::FT_ROMAN));
+                                       fs.faml_flag = true;
+                               }
+                               break;
+                       case SANS_FAMILY:
+                               if (default_family != "sfdefault") {
+                                       tagsToOpen.push_back(docbookStartFontTag(xml::FT_SANS));
+                                       fs.faml_flag = true;
+                               }
+                               break;
+                       case TYPEWRITER_FAMILY:
+                               if (default_family != "ttdefault") {
+                                       tagsToOpen.push_back(docbookStartFontTag(xml::FT_TYPE));
+                                       fs.faml_flag = true;
+                               }
+                               break;
+                       case INHERIT_FAMILY:
+                               break;
+                       default:
+                               // the other tags are for internal use
+                               LATTEST(false);
+                               break;
+               }
+       }
+
+       // Font size
+       fs.curr_size = font.fontInfo().size();
+       FontSize old_size = font_old.size();
+       if (old_size != fs.curr_size) {
+               if (fs.size_flag) {
+                       OptionalFontType tag = fontSizeToXml(old_size);
+                       if (tag.has_value)
+                               tagsToClose.push_back(docbookEndFontTag(tag.ft));
+                       fs.size_flag = false;
+               }
+
+               OptionalFontType tag = fontSizeToXml(fs.curr_size);
+               if (tag.has_value) {
+                       tagsToOpen.push_back(docbookStartFontTag(tag.ft));
+                       fs.size_flag = true;
+               }
+       }
+
+       return std::tuple<vector<xml::FontTag>, vector<xml::EndFontTag>>(tagsToOpen, tagsToClose);
+}
+
+} // anonymous namespace
+
+
+void Paragraph::simpleDocBookOnePar(Buffer const & buf,
+                                    XMLStream & xs,
+                                    OutputParams const & runparams,
+                                    Font const & outerfont,
+                                    bool start_paragraph, bool close_paragraph,
+                                    pos_type initial) const
+{
+       // track whether we have opened these tags
+       DocBookFontState fs;
+
+       if (start_paragraph)
+               xs.startDivision(allowEmpty());
 
        Layout const & style = *d->layout_;
        FontInfo font_old =
-               style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
+                       style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
 
-       if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
-               os << "]]>";
+       string const default_family =
+                       buf.masterBuffer()->params().fonts_default_family;
+
+       vector<xml::FontTag> tagsToOpen;
+       vector<xml::EndFontTag> tagsToClose;
 
        // parsing main loop
        for (pos_type i = initial; i < size(); ++i) {
-               Font font = getFont(buf.params(), i, outerfont);
-
-               // handle <emphasis> tag
-               if (font_old.emph() != font.fontInfo().emph()) {
-                       if (font.fontInfo().emph() == FONT_ON) {
-                               os << "<emphasis>";
-                               emph_flag = true;
-                       } else if (i != initial) {
-                               os << "</emphasis>";
-                               emph_flag = false;
-                       }
-               }
+               // let's not show deleted material in the output
+               if (isDeleted(i))
+                       continue;
+
+               Font const font = getFont(buf.masterBuffer()->params(), i, outerfont);
+
+               // Determine which tags should be opened or closed.
+               tie(tagsToOpen, tagsToClose) = computeDocBookFontSwitch(font_old, font, default_family, fs);
+
+               // FIXME XHTML
+               // Other such tags? What about the other text ranges?
+
+               vector<xml::EndFontTag>::const_iterator cit = tagsToClose.begin();
+               vector<xml::EndFontTag>::const_iterator cen = tagsToClose.end();
+               for (; cit != cen; ++cit)
+                       xs << *cit;
+
+               vector<xml::FontTag>::const_iterator sit = tagsToOpen.begin();
+               vector<xml::FontTag>::const_iterator sen = tagsToOpen.end();
+               for (; sit != sen; ++sit)
+                       xs << *sit;
+
+               tagsToClose.clear();
+               tagsToOpen.clear();
 
                if (Inset const * inset = getInset(i)) {
-                       inset->docbook(os, runparams);
+                       if (!runparams.for_toc || inset->isInToc()) {
+                               OutputParams np = runparams;
+                               np.local_font = &font;
+                               // If the paragraph has size 1, then we are in the "special
+                               // case" where we do not output the containing paragraph info.
+                               // This "special case" is defined in more details in output_docbook.cpp, makeParagraphs. The results
+                               // of that brittle logic is passed to this function through open_par.
+                               if (!inset->getLayout().htmlisblock() && size() != 1) // TODO: htmlisblock here too!
+                                       np.docbook_in_par = true;
+                               inset->docbook(xs, np);
+                       }
                } else {
-                       char_type c = d->text_[i];
-
-                       if (style.pass_thru)
-                               os.put(c);
-                       else
-                               os << sgml::escapeChar(c);
+                       char_type c = getUChar(buf.masterBuffer()->params(), runparams, i);
+                       xs << c;
                }
                font_old = font.fontInfo();
        }
 
-       if (emph_flag) {
-               os << "</emphasis>";
-       }
-
-       if (style.free_spacing)
-               os << '\n';
-       if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
-               os << "<![CDATA[";
+       // FIXME, this code is just imported from XHTML
+       // I'm worried about what happens if a branch, say, is itself
+       // wrapped in some font stuff. I think that will not work.
+       xs.closeFontTags();
+       if (runparams.docbook_in_listing)
+               xs << xml::CR();
+       if (close_paragraph)
+               xs.endDivision();
 }
 
 
 namespace {
-void doFontSwitch(vector<html::FontTag> & tagsToOpen,
-                  vector<html::EndFontTag> & tagsToClose,
-                  bool & flag, FontState curstate, html::FontTypes type)
+
+void doFontSwitchXHTML(vector<xml::FontTag> & tagsToOpen,
+                  vector<xml::EndFontTag> & tagsToClose,
+                  bool & flag, FontState curstate, xml::FontTypes type)
 {
        if (curstate == FONT_ON) {
-               tagsToOpen.push_back(html::FontTag(type));
+               tagsToOpen.push_back(xhtmlStartFontTag(type));
                flag = true;
        } else if (flag) {
-               tagsToClose.push_back(html::EndFontTag(type));
+               tagsToClose.push_back(xhtmlEndFontTag(type));
                flag = false;
        }
 }
-} // namespace
+
+} // anonymous namespace
 
 
 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
-                                   XHTMLStream & xs,
+                                   XMLStream & xs,
                                    OutputParams const & runparams,
                                    Font const & outerfont,
                                    bool start_paragraph, bool close_paragraph,
@@ -3025,13 +3382,13 @@ docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
 
        FontShape  curr_fs   = INHERIT_SHAPE;
        FontFamily curr_fam  = INHERIT_FAMILY;
-       FontSize   curr_size = FONT_SIZE_INHERIT;
+       FontSize   curr_size = INHERIT_SIZE;
 
        string const default_family =
                buf.masterBuffer()->params().fonts_default_family;
 
-       vector<html::FontTag> tagsToOpen;
-       vector<html::EndFontTag> tagsToClose;
+       vector<xml::FontTag> tagsToOpen;
+       vector<xml::EndFontTag> tagsToClose;
 
        // parsing main loop
        for (pos_type i = initial; i < size(); ++i) {
@@ -3044,43 +3401,43 @@ docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
                // emphasis
                FontState curstate = font.fontInfo().emph();
                if (font_old.emph() != curstate)
-                       doFontSwitch(tagsToOpen, tagsToClose, emph_flag, curstate, html::FT_EMPH);
+                       doFontSwitchXHTML(tagsToOpen, tagsToClose, emph_flag, curstate, xml::FT_EMPH);
 
                // noun
                curstate = font.fontInfo().noun();
                if (font_old.noun() != curstate)
-                       doFontSwitch(tagsToOpen, tagsToClose, noun_flag, curstate, html::FT_NOUN);
+                       doFontSwitchXHTML(tagsToOpen, tagsToClose, noun_flag, curstate, xml::FT_NOUN);
 
                // underbar
                curstate = font.fontInfo().underbar();
                if (font_old.underbar() != curstate)
-                       doFontSwitch(tagsToOpen, tagsToClose, ubar_flag, curstate, html::FT_UBAR);
+                       doFontSwitchXHTML(tagsToOpen, tagsToClose, ubar_flag, curstate, xml::FT_UBAR);
 
                // strikeout
                curstate = font.fontInfo().strikeout();
                if (font_old.strikeout() != curstate)
-                       doFontSwitch(tagsToOpen, tagsToClose, sout_flag, curstate, html::FT_SOUT);
+                       doFontSwitchXHTML(tagsToOpen, tagsToClose, sout_flag, curstate, xml::FT_SOUT);
 
                // xout
                curstate = font.fontInfo().xout();
                if (font_old.xout() != curstate)
-                       doFontSwitch(tagsToOpen, tagsToClose, xout_flag, curstate, html::FT_XOUT);
+                       doFontSwitchXHTML(tagsToOpen, tagsToClose, xout_flag, curstate, xml::FT_XOUT);
 
                // double underbar
                curstate = font.fontInfo().uuline();
                if (font_old.uuline() != curstate)
-                       doFontSwitch(tagsToOpen, tagsToClose, dbar_flag, curstate, html::FT_DBAR);
+                       doFontSwitchXHTML(tagsToOpen, tagsToClose, dbar_flag, curstate, xml::FT_DBAR);
 
                // wavy line
                curstate = font.fontInfo().uwave();
                if (font_old.uwave() != curstate)
-                       doFontSwitch(tagsToOpen, tagsToClose, wave_flag, curstate, html::FT_WAVE);
+                       doFontSwitchXHTML(tagsToOpen, tagsToClose, wave_flag, curstate, xml::FT_WAVE);
 
                // bold
                // a little hackish, but allows us to reuse what we have.
                curstate = (font.fontInfo().series() == BOLD_SERIES ? FONT_ON : FONT_OFF);
                if (font_old.series() != font.fontInfo().series())
-                       doFontSwitch(tagsToOpen, tagsToClose, bold_flag, curstate, html::FT_BOLD);
+                       doFontSwitchXHTML(tagsToOpen, tagsToClose, bold_flag, curstate, xml::FT_BOLD);
 
                // Font shape
                curr_fs = font.fontInfo().shape();
@@ -3089,13 +3446,13 @@ docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
                        if (shap_flag) {
                                switch (old_fs) {
                                case ITALIC_SHAPE:
-                                       tagsToClose.push_back(html::EndFontTag(html::FT_ITALIC));
+                                       tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_ITALIC));
                                        break;
                                case SLANTED_SHAPE:
-                                       tagsToClose.push_back(html::EndFontTag(html::FT_SLANTED));
+                                       tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SLANTED));
                                        break;
                                case SMALLCAPS_SHAPE:
-                                       tagsToClose.push_back(html::EndFontTag(html::FT_SMALLCAPS));
+                                       tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SMALLCAPS));
                                        break;
                                case UP_SHAPE:
                                case INHERIT_SHAPE:
@@ -3109,15 +3466,15 @@ docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
                        }
                        switch (curr_fs) {
                        case ITALIC_SHAPE:
-                               tagsToOpen.push_back(html::FontTag(html::FT_ITALIC));
+                               tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_ITALIC));
                                shap_flag = true;
                                break;
                        case SLANTED_SHAPE:
-                               tagsToOpen.push_back(html::FontTag(html::FT_SLANTED));
+                               tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SLANTED));
                                shap_flag = true;
                                break;
                        case SMALLCAPS_SHAPE:
-                               tagsToOpen.push_back(html::FontTag(html::FT_SMALLCAPS));
+                               tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SMALLCAPS));
                                shap_flag = true;
                                break;
                        case UP_SHAPE:
@@ -3137,13 +3494,13 @@ docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
                        if (faml_flag) {
                                switch (old_fam) {
                                case ROMAN_FAMILY:
-                                       tagsToClose.push_back(html::EndFontTag(html::FT_ROMAN));
+                                       tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_ROMAN));
                                        break;
                                case SANS_FAMILY:
-                                       tagsToClose.push_back(html::EndFontTag(html::FT_SANS));
-                                       break;
+                                   tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SANS));
+                                   break;
                                case TYPEWRITER_FAMILY:
-                                       tagsToClose.push_back(html::EndFontTag(html::FT_TYPE));
+                                       tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_TYPE));
                                        break;
                                case INHERIT_FAMILY:
                                        break;
@@ -3159,19 +3516,19 @@ docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
                                // we will treat a "default" font family as roman, since we have
                                // no other idea what to do.
                                if (default_family != "rmdefault" && default_family != "default") {
-                                       tagsToOpen.push_back(html::FontTag(html::FT_ROMAN));
+                                       tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_ROMAN));
                                        faml_flag = true;
                                }
                                break;
                        case SANS_FAMILY:
                                if (default_family != "sfdefault") {
-                                       tagsToOpen.push_back(html::FontTag(html::FT_SANS));
+                                       tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SANS));
                                        faml_flag = true;
                                }
                                break;
                        case TYPEWRITER_FAMILY:
                                if (default_family != "ttdefault") {
-                                       tagsToOpen.push_back(html::FontTag(html::FT_TYPE));
+                                       tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_TYPE));
                                        faml_flag = true;
                                }
                                break;
@@ -3190,41 +3547,41 @@ docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
                if (old_size != curr_size) {
                        if (size_flag) {
                                switch (old_size) {
-                               case FONT_SIZE_TINY:
-                                       tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_TINY));
+                               case TINY_SIZE:
+                                       tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_TINY));
                                        break;
-                               case FONT_SIZE_SCRIPT:
-                                       tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SCRIPT));
+                               case SCRIPT_SIZE:
+                                       tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_SCRIPT));
                                        break;
-                               case FONT_SIZE_FOOTNOTE:
-                                       tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_FOOTNOTE));
+                               case FOOTNOTE_SIZE:
+                                       tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_FOOTNOTE));
                                        break;
-                               case FONT_SIZE_SMALL:
-                                       tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SMALL));
+                               case SMALL_SIZE:
+                                       tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_SMALL));
                                        break;
-                               case FONT_SIZE_LARGE:
-                                       tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGE));
+                               case LARGE_SIZE:
+                                       tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_LARGE));
                                        break;
-                               case FONT_SIZE_LARGER:
-                                       tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGER));
+                               case LARGER_SIZE:
+                                       tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_LARGER));
                                        break;
-                               case FONT_SIZE_LARGEST:
-                                       tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGEST));
+                               case LARGEST_SIZE:
+                                       tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_LARGEST));
                                        break;
-                               case FONT_SIZE_HUGE:
-                                       tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGE));
+                               case HUGE_SIZE:
+                                       tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_HUGE));
                                        break;
-                               case FONT_SIZE_HUGER:
-                                       tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGER));
+                               case HUGER_SIZE:
+                                       tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_HUGER));
                                        break;
-                               case FONT_SIZE_INCREASE:
-                                       tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_INCREASE));
+                               case INCREASE_SIZE:
+                                       tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_INCREASE));
                                        break;
-                               case FONT_SIZE_DECREASE:
-                                       tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_DECREASE));
+                               case DECREASE_SIZE:
+                                       tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_DECREASE));
                                        break;
-                               case FONT_SIZE_INHERIT:
-                               case FONT_SIZE_NORMAL:
+                               case INHERIT_SIZE:
+                               case NORMAL_SIZE:
                                        break;
                                default:
                                        // the other tags are for internal use
@@ -3234,52 +3591,52 @@ docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
                                size_flag = false;
                        }
                        switch (curr_size) {
-                       case FONT_SIZE_TINY:
-                               tagsToOpen.push_back(html::FontTag(html::FT_SIZE_TINY));
+                       case TINY_SIZE:
+                               tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_TINY));
                                size_flag = true;
                                break;
-                       case FONT_SIZE_SCRIPT:
-                               tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SCRIPT));
+                       case SCRIPT_SIZE:
+                               tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_SCRIPT));
                                size_flag = true;
                                break;
-                       case FONT_SIZE_FOOTNOTE:
-                               tagsToOpen.push_back(html::FontTag(html::FT_SIZE_FOOTNOTE));
+                       case FOOTNOTE_SIZE:
+                               tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_FOOTNOTE));
                                size_flag = true;
                                break;
-                       case FONT_SIZE_SMALL:
-                               tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SMALL));
+                       case SMALL_SIZE:
+                               tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_SMALL));
                                size_flag = true;
                                break;
-                       case FONT_SIZE_LARGE:
-                               tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGE));
+                       case LARGE_SIZE:
+                               tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_LARGE));
                                size_flag = true;
                                break;
-                       case FONT_SIZE_LARGER:
-                               tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGER));
+                       case LARGER_SIZE:
+                               tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_LARGER));
                                size_flag = true;
                                break;
-                       case FONT_SIZE_LARGEST:
-                               tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGEST));
+                       case LARGEST_SIZE:
+                               tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_LARGEST));
                                size_flag = true;
                                break;
-                       case FONT_SIZE_HUGE:
-                               tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGE));
+                       case HUGE_SIZE:
+                               tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_HUGE));
                                size_flag = true;
                                break;
-                       case FONT_SIZE_HUGER:
-                               tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGER));
+                       case HUGER_SIZE:
+                               tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_HUGER));
                                size_flag = true;
                                break;
-                       case FONT_SIZE_INCREASE:
-                               tagsToOpen.push_back(html::FontTag(html::FT_SIZE_INCREASE));
+                       case INCREASE_SIZE:
+                               tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_INCREASE));
                                size_flag = true;
                                break;
-                       case FONT_SIZE_DECREASE:
-                               tagsToOpen.push_back(html::FontTag(html::FT_SIZE_DECREASE));
+                       case DECREASE_SIZE:
+                               tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_DECREASE));
                                size_flag = true;
                                break;
-                       case FONT_SIZE_NORMAL:
-                       case FONT_SIZE_INHERIT:
+                       case INHERIT_SIZE:
+                       case NORMAL_SIZE:
                                break;
                        default:
                                // the other tags are for internal use
@@ -3291,13 +3648,13 @@ docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
                // FIXME XHTML
                // Other such tags? What about the other text ranges?
 
-               vector<html::EndFontTag>::const_iterator cit = tagsToClose.begin();
-               vector<html::EndFontTag>::const_iterator cen = tagsToClose.end();
+               vector<xml::EndFontTag>::const_iterator cit = tagsToClose.begin();
+               vector<xml::EndFontTag>::const_iterator cen = tagsToClose.end();
                for (; cit != cen; ++cit)
                        xs << *cit;
 
-               vector<html::FontTag>::const_iterator sit = tagsToOpen.begin();
-               vector<html::FontTag>::const_iterator sen = tagsToOpen.end();
+               vector<xml::FontTag>::const_iterator sit = tagsToOpen.begin();
+               vector<xml::FontTag>::const_iterator sen = tagsToOpen.end();
                for (; sit != sen; ++sit)
                        xs << *sit;
 
@@ -3316,8 +3673,12 @@ docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
                                retval += inset->xhtml(xs, np);
                        }
                } else {
-                       char_type c = getUChar(buf.masterBuffer()->params(), i);
-                       xs << c;
+                       char_type c = getUChar(buf.masterBuffer()->params(),
+                                              runparams, i);
+                       if (c == ' ' && (style.free_spacing || runparams.free_spacing))
+                               xs << XMLStream::ESCAPE_NONE << "&nbsp;";
+                       else
+                               xs << c;
                }
                font_old = font.fontInfo();
        }
@@ -3369,10 +3730,12 @@ bool Paragraph::isLineSeparator(pos_type pos) const
 }
 
 
-bool Paragraph::isWordSeparator(pos_type pos) const
+bool Paragraph::isWordSeparator(pos_type pos, bool const ignore_deleted) const
 {
        if (pos == size())
                return true;
+       if (ignore_deleted && isDeleted(pos))
+               return false;
        if (Inset const * inset = getInset(pos))
                return !inset->isLetter();
        // if we have a hard hyphen (no en- or emdash) or apostrophe
@@ -3406,6 +3769,62 @@ bool Paragraph::isHardHyphenOrApostrophe(pos_type pos) const
 }
 
 
+bool Paragraph::needsCProtection(bool const fragile) const
+{
+       // first check the layout of the paragraph, but only in insets
+       InsetText const * textinset = inInset().asInsetText();
+       bool const maintext = textinset
+               ? textinset->text().isMainText()
+               : false;
+
+       if (!maintext && layout().needcprotect) {
+               // Environments need cprotection regardless the content
+               if (layout().latextype == LATEX_ENVIRONMENT)
+                       return true;
+
+               // Commands need cprotection if they contain specific chars
+               int const nchars_escape = 9;
+               static char_type const chars_escape[nchars_escape] = {
+                       '&', '_', '$', '%', '#', '^', '{', '}', '\\'};
+
+               docstring const pars = asString();
+               for (int k = 0; k < nchars_escape; k++) {
+                       if (contains(pars, chars_escape[k]))
+                               return true;
+               }
+       }
+
+       // now check whether we have insets that need cprotection
+       pos_type size = pos_type(d->text_.size());
+       for (pos_type i = 0; i < size; ++i) {
+               if (!isInset(i))
+                       continue;
+               Inset const * ins = getInset(i);
+               if (ins->needsCProtection(maintext, fragile))
+                       return true;
+               if (ins->getLayout().latextype() == InsetLayout::ENVIRONMENT)
+                       // Environments need cprotection regardless the content
+                       return true;
+               // Now check math environments
+               InsetMath const * im = getInset(i)->asInsetMath();
+               if (!im || im->cell(0).empty())
+                       continue;
+               switch(im->cell(0)[0]->lyxCode()) {
+               case MATH_AMSARRAY_CODE:
+               case MATH_SUBSTACK_CODE:
+               case MATH_ENV_CODE:
+               case MATH_XYMATRIX_CODE:
+                       // these need cprotection
+                       return true;
+               default:
+                       break;
+               }
+       }
+
+       return false;
+}
+
+
 FontSpan const & Paragraph::getSpellRange(pos_type pos) const
 {
        return d->speller_state_.getRange(pos);
@@ -3465,28 +3884,22 @@ void Paragraph::changeLanguage(BufferParams const & bparams,
 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
 {
        Language const * doc_language = bparams.language;
-       FontList::const_iterator cit = d->fontlist_.begin();
-       FontList::const_iterator end = d->fontlist_.end();
-
-       for (; cit != end; ++cit)
-               if (cit->font().language() != ignore_language &&
-                   cit->font().language() != latex_language &&
-                   cit->font().language() != doc_language)
+       for (auto const & f : d->fontlist_)
+               if (f.font().language() != ignore_language &&
+                   f.font().language() != latex_language &&
+                   f.font().language() != doc_language)
                        return true;
        return false;
 }
 
 
-void Paragraph::getLanguages(std::set<Language const *> & languages) const
+void Paragraph::getLanguages(std::set<Language const *> & langs) const
 {
-       FontList::const_iterator cit = d->fontlist_.begin();
-       FontList::const_iterator end = d->fontlist_.end();
-
-       for (; cit != end; ++cit) {
-               Language const * lang = cit->font().language();
+       for (auto const & f : d->fontlist_) {
+               Language const * lang = f.font().language();
                if (lang != ignore_language &&
                    lang != latex_language)
-                       languages.insert(lang);
+                       langs.insert(lang);
        }
 }
 
@@ -3515,7 +3928,7 @@ docstring Paragraph::asString(pos_type beg, pos_type end, int options, const Out
                        os.put(c);
                else if (c == META_INSET && (options & AS_STR_INSETS)) {
                        if (c == META_INSET && (options & AS_STR_PLAINTEXT)) {
-                               LASSERT(runparams != 0, return docstring());
+                               LASSERT(runparams != nullptr, return docstring());
                                getInset(i)->plaintext(os, *runparams);
                        } else {
                                getInset(i)->toString(os);
@@ -3533,6 +3946,8 @@ void Paragraph::forOutliner(docstring & os, size_t const maxlen,
        size_t tmplen = shorten ? maxlen + 1 : maxlen;
        if (label && !labelString().empty())
                os += labelString() + ' ';
+       if (!layout().isTocCaption())
+               return;
        for (pos_type i = 0; i < size() && os.length() < tmplen; ++i) {
                if (isDeleted(i))
                        continue;
@@ -3635,32 +4050,43 @@ bool Paragraph::allowEmpty() const
 
 bool Paragraph::brokenBiblio() const
 {
-       // there is a problem if there is no bibitem at position 0 or
-       // if there is another bibitem in the paragraph.
-       return d->layout_->labeltype == LABEL_BIBLIO
+       // There is a problem if there is no bibitem at position 0 in
+       // paragraphs that need one, if there is another bibitem in the
+       // paragraph or if this paragraph is not supposed to have
+       // a bibitem inset at all.
+       return ((d->layout_->labeltype == LABEL_BIBLIO
                && (d->insetlist_.find(BIBITEM_CODE) != 0
-                   || d->insetlist_.find(BIBITEM_CODE, 1) > 0);
+                   || d->insetlist_.find(BIBITEM_CODE, 1) > 0))
+               || (d->layout_->labeltype != LABEL_BIBLIO
+                   && d->insetlist_.find(BIBITEM_CODE) != -1));
 }
 
 
 int Paragraph::fixBiblio(Buffer const & buffer)
 {
-       // FIXME: What about the case where paragraph is not BIBLIO
-       // but there is an InsetBibitem?
        // FIXME: when there was already an inset at 0, the return value is 1,
        // which does not tell whether another inset has been remove; the
        // cursor cannot be correctly updated.
 
-       if (d->layout_->labeltype != LABEL_BIBLIO)
-               return 0;
-
        bool const track_changes = buffer.params().track_changes;
        int bibitem_pos = d->insetlist_.find(BIBITEM_CODE);
-       bool const hasbibitem0 = bibitem_pos == 0;
 
+       // The case where paragraph is not BIBLIO
+       if (d->layout_->labeltype != LABEL_BIBLIO) {
+               if (bibitem_pos == -1)
+                       // No InsetBibitem => OK
+                       return 0;
+               // There is an InsetBibitem: remove it!
+               d->insetlist_.release(bibitem_pos);
+               eraseChar(bibitem_pos, track_changes);
+               return (bibitem_pos == 0) ? -1 : -bibitem_pos;
+       }
+
+       bool const hasbibitem0 = bibitem_pos == 0;
        if (hasbibitem0) {
                bibitem_pos = d->insetlist_.find(BIBITEM_CODE, 1);
-               // There was an InsetBibitem at pos 0, and no other one => OK
+               // There was an InsetBibitem at pos 0,
+               // and no other one => OK
                if (bibitem_pos == -1)
                        return 0;
                // there is a bibitem at the 0 position, but since
@@ -3675,7 +4101,7 @@ int Paragraph::fixBiblio(Buffer const & buffer)
        }
 
        // We need to create an inset at the beginning
-       Inset * inset = 0;
+       Inset * inset = nullptr;
        if (bibitem_pos > 0) {
                // there was one somewhere in the paragraph, let's move it
                inset = d->insetlist_.release(bibitem_pos);
@@ -3689,6 +4115,8 @@ int Paragraph::fixBiblio(Buffer const & buffer)
        insertInset(0, inset, font, Change(track_changes ? Change::INSERTED
                                                   : Change::UNCHANGED));
 
+       // This is needed to get the counters right
+       buffer.updateBuffer();
        return 1;
 }
 
@@ -3747,14 +4175,14 @@ Inset * Paragraph::releaseInset(pos_type pos)
 Inset * Paragraph::getInset(pos_type pos)
 {
        return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
-                ? d->insetlist_.get(pos) : 0;
+                ? d->insetlist_.get(pos) : nullptr;
 }
 
 
 Inset const * Paragraph::getInset(pos_type pos) const
 {
        return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
-                ? d->insetlist_.get(pos) : 0;
+                ? d->insetlist_.get(pos) : nullptr;
 }
 
 
@@ -3911,13 +4339,13 @@ void Paragraph::deregisterWords()
 
 
 void Paragraph::locateWord(pos_type & from, pos_type & to,
-       word_location const loc) const
+       word_location const loc, bool const ignore_deleted) const
 {
        switch (loc) {
        case WHOLE_WORD_STRICT:
                if (from == 0 || from == size()
-                   || isWordSeparator(from)
-                   || isWordSeparator(from - 1)) {
+                   || isWordSeparator(from, ignore_deleted)
+                   || isWordSeparator(from - 1, ignore_deleted)) {
                        to = from;
                        return;
                }
@@ -3925,13 +4353,13 @@ void Paragraph::locateWord(pos_type & from, pos_type & to,
 
        case WHOLE_WORD:
                // If we are already at the beginning of a word, do nothing
-               if (!from || isWordSeparator(from - 1))
+               if (!from || isWordSeparator(from - 1, ignore_deleted))
                        break;
                // fall through
 
        case PREVIOUS_WORD:
                // always move the cursor to the beginning of previous word
-               while (from && !isWordSeparator(from - 1))
+               while (from && !isWordSeparator(from - 1, ignore_deleted))
                        --from;
                break;
        case NEXT_WORD:
@@ -3942,7 +4370,7 @@ void Paragraph::locateWord(pos_type & from, pos_type & to,
                break;
        }
        to = from;
-       while (to < size() && !isWordSeparator(to))
+       while (to < size() && !isWordSeparator(to, ignore_deleted))
                ++to;
 }
 
@@ -4021,7 +4449,7 @@ Language * Paragraph::Private::locateSpellRange(
                ++from;
        // don't check empty range
        if (from >= to)
-               return 0;
+               return nullptr;
        // get current language
        Language * lang = getSpellLanguage(from);
        pos_type last = from;
@@ -4122,13 +4550,16 @@ SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
        if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
                return result;
 
-       locateWord(from, to, WHOLE_WORD);
+       locateWord(from, to, WHOLE_WORD, true);
        if (from == to || from >= size())
                return result;
 
        docstring word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
        Language * lang = d->getSpellLanguage(from);
 
+       if (getFontSettings(d->inset_owner_->buffer().params(), from).fontInfo().nospellcheck() == FONT_ON)
+               return result;
+
        wl = WordLangTuple(word, lang);
 
        if (word.empty())
@@ -4178,6 +4609,15 @@ SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
 }
 
 
+void Paragraph::anonymize()
+{
+       // This is a very crude anonymization for now
+       for (char_type & c : d->text_)
+               if (isLetterChar(c) || isNumber(c))
+                       c = 'a';
+}
+
+
 void Paragraph::Private::markMisspelledWords(
        pos_type const & first, pos_type const & last,
        SpellChecker::Result result,