]> git.lyx.org Git - lyx.git/blobdiff - src/Paragraph.cpp
Use \babelfonts only if babel >= 3.15 is available.
[lyx.git] / src / Paragraph.cpp
index a6c72c310e0b22a07b1251132133a1d02899041a..d129cb403e237662bfc81f496fa113cec6a04454 100644 (file)
@@ -314,25 +314,21 @@ 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 +356,7 @@ 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,
@@ -864,9 +860,10 @@ 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 +880,22 @@ 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 +909,7 @@ 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 +940,6 @@ 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,
@@ -1090,13 +1006,15 @@ void Paragraph::Private::latexInset(BufferParams const & bparams,
        odocstream::pos_type const len = os.os().tellp();
 
        if (inset->forceLTR()
-           && !runparams.use_polyglossia
            && 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;
@@ -1152,12 +1070,8 @@ 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 +1089,13 @@ 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 +1121,7 @@ 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 +1149,8 @@ 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 +1201,9 @@ 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 +1226,16 @@ 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 +1251,22 @@ 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 +1334,6 @@ 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 +1342,20 @@ 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 +1402,58 @@ 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 +1645,12 @@ 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");
 }
 
 
@@ -1926,32 +1853,59 @@ 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 +1985,29 @@ 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();
@@ -2278,6 +2246,10 @@ 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 +2259,18 @@ 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 +2310,10 @@ 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 +2323,18 @@ 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);
@@ -2393,6 +2373,11 @@ 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;
@@ -2424,7 +2409,9 @@ void Paragraph::latex(BufferParams const & bparams,
 
        // if the paragraph is empty, the loop will not be entered at all
        if (empty()) {
-               if (style.isCommand()) {
+               // For InTitle commands, we have already opened a group
+               // in output_latex::TeXOnePar.
+               if (style.isCommand() && !style.intitle) {
                        os << '{';
                        ++column;
                }
@@ -2462,7 +2449,9 @@ void Paragraph::latex(BufferParams const & bparams,
                                os << "}] ";
                                column +=3;
                        }
-                       if (style.isCommand()) {
+                       // For InTitle commands, we have already opened a group
+                       // in output_latex::TeXOnePar.
+                       if (style.isCommand() && !style.intitle) {
                                os << '{';
                                ++column;
                        }
@@ -2519,6 +2508,11 @@ 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(
@@ -2551,6 +2545,12 @@ void Paragraph::latex(BufferParams const & bparams,
                    (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;
                        column += running_font.latexWriteEndChanges(
                                    os, bparams, runparams, basefont,
@@ -2560,23 +2560,17 @@ void Paragraph::latex(BufferParams const & bparams,
                        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)
@@ -2650,7 +2644,7 @@ 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
@@ -2687,7 +2681,7 @@ void Paragraph::latex(BufferParams const & bparams,
                                        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,
@@ -2695,7 +2689,7 @@ void Paragraph::latex(BufferParams const & bparams,
                                                : pars[pit].getFont(bparams,
                                                        pars[pit].size() - 1,
                                                        outerfont);
-                                       if (last_font.fontInfo().size() !=
+                                       if (lastfont.fontInfo().size() !=
                                            basefont.fontInfo().size()) {
                                                ++parInline;
                                                incremented = true;
@@ -2708,10 +2702,24 @@ void Paragraph::latex(BufferParams const & bparams,
                                        --parInline;
                        }
                } 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: ")
@@ -2735,6 +2743,15 @@ 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;
+
+               // 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
@@ -3021,7 +3038,7 @@ 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;
@@ -3186,41 +3203,41 @@ docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
                if (old_size != curr_size) {
                        if (size_flag) {
                                switch (old_size) {
-                               case FONT_SIZE_TINY:
+                               case TINY_SIZE:
                                        tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_TINY));
                                        break;
-                               case FONT_SIZE_SCRIPT:
+                               case SCRIPT_SIZE:
                                        tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SCRIPT));
                                        break;
-                               case FONT_SIZE_FOOTNOTE:
+                               case FOOTNOTE_SIZE:
                                        tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_FOOTNOTE));
                                        break;
-                               case FONT_SIZE_SMALL:
+                               case SMALL_SIZE:
                                        tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SMALL));
                                        break;
-                               case FONT_SIZE_LARGE:
+                               case LARGE_SIZE:
                                        tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGE));
                                        break;
-                               case FONT_SIZE_LARGER:
+                               case LARGER_SIZE:
                                        tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGER));
                                        break;
-                               case FONT_SIZE_LARGEST:
+                               case LARGEST_SIZE:
                                        tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGEST));
                                        break;
-                               case FONT_SIZE_HUGE:
+                               case HUGE_SIZE:
                                        tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGE));
                                        break;
-                               case FONT_SIZE_HUGER:
+                               case HUGER_SIZE:
                                        tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGER));
                                        break;
-                               case FONT_SIZE_INCREASE:
+                               case INCREASE_SIZE:
                                        tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_INCREASE));
                                        break;
-                               case FONT_SIZE_DECREASE:
+                               case DECREASE_SIZE:
                                        tagsToClose.push_back(html::EndFontTag(html::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
@@ -3230,52 +3247,52 @@ docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
                                size_flag = false;
                        }
                        switch (curr_size) {
-                       case FONT_SIZE_TINY:
+                       case TINY_SIZE:
                                tagsToOpen.push_back(html::FontTag(html::FT_SIZE_TINY));
                                size_flag = true;
                                break;
-                       case FONT_SIZE_SCRIPT:
+                       case SCRIPT_SIZE:
                                tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SCRIPT));
                                size_flag = true;
                                break;
-                       case FONT_SIZE_FOOTNOTE:
+                       case FOOTNOTE_SIZE:
                                tagsToOpen.push_back(html::FontTag(html::FT_SIZE_FOOTNOTE));
                                size_flag = true;
                                break;
-                       case FONT_SIZE_SMALL:
+                       case SMALL_SIZE:
                                tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SMALL));
                                size_flag = true;
                                break;
-                       case FONT_SIZE_LARGE:
+                       case LARGE_SIZE:
                                tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGE));
                                size_flag = true;
                                break;
-                       case FONT_SIZE_LARGER:
+                       case LARGER_SIZE:
                                tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGER));
                                size_flag = true;
                                break;
-                       case FONT_SIZE_LARGEST:
+                       case LARGEST_SIZE:
                                tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGEST));
                                size_flag = true;
                                break;
-                       case FONT_SIZE_HUGE:
+                       case HUGE_SIZE:
                                tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGE));
                                size_flag = true;
                                break;
-                       case FONT_SIZE_HUGER:
+                       case HUGER_SIZE:
                                tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGER));
                                size_flag = true;
                                break;
-                       case FONT_SIZE_INCREASE:
+                       case INCREASE_SIZE:
                                tagsToOpen.push_back(html::FontTag(html::FT_SIZE_INCREASE));
                                size_flag = true;
                                break;
-                       case FONT_SIZE_DECREASE:
+                       case DECREASE_SIZE:
                                tagsToOpen.push_back(html::FontTag(html::FT_SIZE_DECREASE));
                                size_flag = true;
                                break;
-                       case FONT_SIZE_NORMAL:
-                       case FONT_SIZE_INHERIT:
+                       case NORMAL_SIZE:
+                       case INHERIT_SIZE:
                                break;
                        default:
                                // the other tags are for internal use
@@ -3312,8 +3329,12 @@ 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 << XHTMLStream::ESCAPE_NONE << "&nbsp;";
+                       else
+                               xs << c;
                }
                font_old = font.fontInfo();
        }
@@ -3402,6 +3423,62 @@ 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);
@@ -3461,28 +3538,22 @@ 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);
        }
 }
 
@@ -3529,6 +3600,8 @@ 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;
@@ -3631,32 +3704,43 @@ 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
@@ -3671,7 +3755,7 @@ 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);
@@ -3685,6 +3769,8 @@ 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;
 }
 
@@ -4125,6 +4211,9 @@ SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
        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())
@@ -4174,6 +4263,15 @@ 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,