]> git.lyx.org Git - lyx.git/blobdiff - src/Paragraph.cpp
Routines for calculating numerical labels for BibTeX citations.
[lyx.git] / src / Paragraph.cpp
index e0afc387287a49338842a7c00cdeec33d0a553ae..233057e9163ad9058363f378066dbef466f92de7 100644 (file)
@@ -4,13 +4,14 @@
  * Licence details can be found in the file COPYING.
  *
  * \author Asger Alstrup
- * \author Lars Gullik Bjønnes
+ * \author Lars Gullik Bjønnes
+ * \author Richard Heck (XHTML output)
  * \author Jean-Marc Lasgouttes
  * \author Angus Leeming
  * \author John Levon
- * \author André Pönitz
+ * \author André Pönitz
  * \author Dekel Tsur
- * \author Jürgen Vigna
+ * \author Jürgen Vigna
  *
  * Full author contact details are available in file CREDITS.
  */
 
 #include "Paragraph.h"
 
+#include "LayoutFile.h"
 #include "Buffer.h"
 #include "BufferParams.h"
 #include "Changes.h"
 #include "Counters.h"
 #include "Encoding.h"
-#include "debug.h"
-#include "gettext.h"
 #include "InsetList.h"
 #include "Language.h"
 #include "LaTeXFeatures.h"
 #include "Font.h"
 #include "FontList.h"
 #include "LyXRC.h"
-#include "Messages.h"
 #include "OutputParams.h"
 #include "output_latex.h"
-#include "paragraph_funcs.h"
+#include "output_xhtml.h"
 #include "ParagraphParameters.h"
+#include "SpellChecker.h"
 #include "sgml.h"
 #include "TextClass.h"
 #include "TexRow.h"
+#include "Text.h"
 #include "VSpace.h"
+#include "WordLangTuple.h"
+#include "WordList.h"
 
 #include "frontends/alert.h"
-#include "frontends/FontMetrics.h"
 
 #include "insets/InsetBibitem.h"
 #include "insets/InsetLabel.h"
 
+#include "support/debug.h"
+#include "support/docstring_list.h"
+#include "support/ExceptionMessage.h"
+#include "support/gettext.h"
+#include "support/lassert.h"
 #include "support/lstrings.h"
 #include "support/textutils.h"
-#include "support/convert.h"
-#include "support/unicode.h"
 
 #include <sstream>
 #include <vector>
 
-using std::endl;
-using std::string;
-using std::ostream;
+using namespace std;
+using namespace lyx::support;
 
 namespace lyx {
 
-using support::contains;
-using support::lowercase;
-using support::prefixIs;
-using support::suffixIs;
-using support::rsplit;
-using support::uppercase;
-
 namespace {
 /// Inset identifier (above 0x10ffff, for ucs-4)
 char_type const META_INSET = 0x200001;
@@ -86,9 +83,11 @@ class Paragraph::Private
 {
 public:
        ///
-       Private(Paragraph * owner);
+       Private(Paragraph * owner, Layout const & layout);
        /// "Copy constructor"
        Private(Private const &, Paragraph * owner);
+       /// Copy constructor from \p beg  to \p end
+       Private(Private const &, Paragraph * owner, pos_type beg, pos_type end);
 
        ///
        void insertChar(pos_type pos, char_type c, Change const & change);
@@ -96,7 +95,7 @@ public:
        /// Output the surrogate pair formed by \p c and \p next to \p os.
        /// \return the number of characters written.
        int latexSurrogatePair(odocstream & os, char_type c, char_type next,
-                              Encoding const &);
+                              OutputParams const &);
 
        /// Output a space in appropriate formatting (or a surrogate pair
        /// if the next character is a combining character).
@@ -112,18 +111,18 @@ public:
        /// specified by the latex macro \p ltx, to \p os starting from \p i.
        /// \return the number of characters written.
        int writeScriptChars(odocstream & os, docstring const & ltx,
-                          Change &, Encoding const &, pos_type & i);
+                          Change const &, Encoding const &, pos_type & i);
 
        /// This could go to ParagraphParameters if we want to.
        int startTeXParParams(BufferParams const &, odocstream &, TexRow &,
-                             bool) const;
+                             OutputParams const &) const;
 
        /// This could go to ParagraphParameters if we want to.
        int endTeXParParams(BufferParams const &, odocstream &, TexRow &,
-                           bool) const;
+                           OutputParams const &) const;
 
        ///
-       void latexInset(Buffer const &, BufferParams const &,
+       void latexInset(BufferParams const &,
                                   odocstream &,
                                   TexRow & texrow, OutputParams &,
                                   Font & running_font,
@@ -138,9 +137,9 @@ public:
        ///
        void latexSpecialChar(
                                   odocstream & os,
-                                  OutputParams & runparams,
-                                  Font & running_font,
-                                  Change & running_change,
+                                  OutputParams const & runparams,
+                                  Font const & running_font,
+                                  Change const & running_change,
                                   Layout const & style,
                                   pos_type & i,
                                   unsigned int & column);
@@ -149,81 +148,73 @@ public:
        bool latexSpecialT1(
                char_type const c,
                odocstream & os,
-               pos_type i,
+               pos_type i,
                unsigned int & column);
        ///
        bool latexSpecialTypewriter(
                char_type const c,
                odocstream & os,
-               pos_type i,
+               pos_type i,
                unsigned int & column);
        ///
        bool latexSpecialPhrase(
                odocstream & os,
                pos_type & i,
                unsigned int & column,
-               OutputParams & runparams);
+               OutputParams const & runparams);
 
        ///
-       void validate(LaTeXFeatures & features,
-                     Layout const & layout) const;
+       void validate(LaTeXFeatures & features) const;
 
        /// Checks if the paragraph contains only text and no inset or font change.
        bool onlyText(Buffer const & buf, Font const & outerfont,
                      pos_type initial) const;
 
        /// match a string against a particular point in the paragraph
-       bool isTextAt(std::string const & str, pos_type pos) const;
+       bool isTextAt(string const & str, pos_type pos) const;
+
+
+       InsetCode ownerCode() const
+       {
+               return inset_owner_ ? inset_owner_->lyxCode() : NO_CODE;
+       }
        
        /// Which Paragraph owns us?
        Paragraph * owner_;
 
        /// In which Inset?
-       Inset * inset_owner_;
+       Inset const * inset_owner_;
 
        ///
        FontList fontlist_;
 
        ///
-       unsigned int id_;
-       ///
-       static unsigned int paragraph_id;
+       int id_;
+
        ///
        ParagraphParameters params_;
 
-       /// position of the paragraph in the buffer. Only macros from
-       /// paragraphs strictly smaller are visible in this paragraph
-       unsigned int macrocontext_position_;
-       
        /// for recording and looking up changes
        Changes changes_;
 
        ///
        InsetList insetlist_;
 
-       ///
-       LayoutPtr layout_;
-
        /// end of label
        pos_type begin_of_body_;
 
        typedef docstring TextContainer;
        ///
        TextContainer text_;
+       
+       typedef set<docstring> Words;
+       ///
+       map<Language, Words> words_;
+       ///
+       Layout const * layout_;
 };
 
 
-
-
-using std::endl;
-using std::upper_bound;
-using std::lower_bound;
-using std::string;
-
-
-// Initialization of the counter for the paragraph id's,
-unsigned int Paragraph::Private::paragraph_id = 0;
-
 namespace {
 
 struct special_phrase {
@@ -244,43 +235,92 @@ size_t const phrases_nr = sizeof(special_phrases)/sizeof(special_phrase);
 } // namespace anon
 
 
-Paragraph::Private::Private(Paragraph * owner)
-       : owner_(owner), inset_owner_(0), begin_of_body_(0)
+Paragraph::Private::Private(Paragraph * owner, Layout const & layout)
+       : owner_(owner), inset_owner_(0), id_(-1), begin_of_body_(0), layout_(&layout)
 {
-       id_ = paragraph_id++;
-       macrocontext_position_ = 0;
        text_.reserve(100);
 }
 
 
+// Initialization of the counter for the paragraph id's,
+//
+// FIXME: There should be a more intelligent way to generate and use the
+// paragraph ids per buffer instead a global static counter for all InsetText
+// in the running program.
+static int paragraph_id = -1;
+
 Paragraph::Private::Private(Private const & p, Paragraph * owner)
        : owner_(owner), inset_owner_(p.inset_owner_), fontlist_(p.fontlist_), 
          params_(p.params_), changes_(p.changes_), insetlist_(p.insetlist_),
-         layout_(p.layout_), begin_of_body_(p.begin_of_body_), text_(p.text_)
+         begin_of_body_(p.begin_of_body_), text_(p.text_), words_(p.words_),
+         layout_(p.layout_)
+{
+       id_ = ++paragraph_id;
+}
+
+
+Paragraph::Private::Private(Private const & p, Paragraph * owner,
+       pos_type beg, pos_type end)
+       : 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_)
+{
+       id_ = ++paragraph_id;
+       if (beg >= pos_type(p.text_.size()))
+               return;
+       text_ = p.text_.substr(beg, end - beg);
+
+       FontList::const_iterator fcit = fontlist_.begin();
+       FontList::const_iterator fend = fontlist_.end();
+       for (; fcit != fend; ++fcit) {
+               if (fcit->pos() < beg)
+                       continue;
+               if (fcit->pos() >= end) {
+                       // Add last entry in the fontlist_.
+                       fontlist_.set(text_.size() - 1, fcit->font());
+                       break;
+               }
+               // Add a new entry in the fontlist_.
+               fontlist_.set(fcit->pos() - beg, fcit->font());
+       }
+}
+
+
+void Paragraph::addChangesToToc(DocIterator const & cdit,
+       Buffer const & buf) const
+{
+       d->changes_.addToToc(cdit, buf);
+}
+
+
+bool Paragraph::isDeleted(pos_type start, pos_type end) const
 {
-       id_ = paragraph_id++;
+       LASSERT(start >= 0 && start <= size(), /**/);
+       LASSERT(end > start && end <= size() + 1, /**/);
+
+       return d->changes_.isDeleted(start, end);
 }
 
 
 bool Paragraph::isChanged(pos_type start, pos_type end) const
 {
-       BOOST_ASSERT(start >= 0 && start <= size());
-       BOOST_ASSERT(end > start && end <= size() + 1);
+       LASSERT(start >= 0 && start <= size(), /**/);
+       LASSERT(end > start && end <= size() + 1, /**/);
 
        return d->changes_.isChanged(start, end);
 }
 
 
-bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const {
+bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const
+{
        // keep the logic here in sync with the logic of eraseChars()
-
-       if (!trackChanges) {
+       if (!trackChanges)
                return true;
-       }
 
-       Change change = d->changes_.lookup(size());
-
-       return change.type == Change::INSERTED && change.author == 0;
+       Change const change = d->changes_.lookup(size());
+       return change.inserted() && change.currentAuthor();
 }
 
 
@@ -301,10 +341,10 @@ void Paragraph::setChange(Change const & change)
         * Conclusion: An inset's content should remain untouched if you delete it
         */
 
-       if (change.type != Change::DELETED) {
+       if (!change.deleted()) {
                for (pos_type pos = 0; pos < size(); ++pos) {
-                       if (isInset(pos))
-                               getInset(pos)->setChange(change);
+                       if (Inset * inset = getInset(pos))
+                               inset->setChange(change);
                }
        }
 }
@@ -312,48 +352,41 @@ void Paragraph::setChange(Change const & change)
 
 void Paragraph::setChange(pos_type pos, Change const & change)
 {
-       BOOST_ASSERT(pos >= 0 && pos <= size());
-
+       LASSERT(pos >= 0 && pos <= size(), /**/);
        d->changes_.set(change, pos);
 
        // see comment in setChange(Change const &) above
-
-       if (change.type != Change::DELETED &&
-           pos < size() && isInset(pos)) {
-               getInset(pos)->setChange(change);
-       }
+       if (!change.deleted() && pos < size())
+                       if (Inset * inset = getInset(pos))
+                               inset->setChange(change);
 }
 
 
 Change const & Paragraph::lookupChange(pos_type pos) const
 {
-       BOOST_ASSERT(pos >= 0 && pos <= size());
-
+       LASSERT(pos >= 0 && pos <= size(), /**/);
        return d->changes_.lookup(pos);
 }
 
 
-void Paragraph::acceptChanges(BufferParams const & bparams, pos_type start,
-               pos_type end)
+void Paragraph::acceptChanges(pos_type start, pos_type end)
 {
-       BOOST_ASSERT(start >= 0 && start <= size());
-       BOOST_ASSERT(end > start && end <= size() + 1);
+       LASSERT(start >= 0 && start <= size(), /**/);
+       LASSERT(end > start && end <= size() + 1, /**/);
 
        for (pos_type pos = start; pos < end; ++pos) {
                switch (lookupChange(pos).type) {
                        case Change::UNCHANGED:
                                // accept changes in nested inset
-                               if (pos < size() && isInset(pos))
-                                       getInset(pos)->acceptChanges(bparams);
-
+                               if (Inset * inset = getInset(pos))
+                                       inset->acceptChanges();
                                break;
 
                        case Change::INSERTED:
                                d->changes_.set(Change(Change::UNCHANGED), pos);
                                // also accept changes in nested inset
-                               if (pos < size() && isInset(pos)) {
-                                       getInset(pos)->acceptChanges(bparams);
-                               }
+                               if (Inset * inset = getInset(pos))
+                                       inset->acceptChanges();
                                break;
 
                        case Change::DELETED:
@@ -371,19 +404,17 @@ void Paragraph::acceptChanges(BufferParams const & bparams, pos_type start,
 }
 
 
-void Paragraph::rejectChanges(BufferParams const & bparams,
-               pos_type start, pos_type end)
+void Paragraph::rejectChanges(pos_type start, pos_type end)
 {
-       BOOST_ASSERT(start >= 0 && start <= size());
-       BOOST_ASSERT(end > start && end <= size() + 1);
+       LASSERT(start >= 0 && start <= size(), /**/);
+       LASSERT(end > start && end <= size() + 1, /**/);
 
        for (pos_type pos = start; pos < end; ++pos) {
                switch (lookupChange(pos).type) {
                        case Change::UNCHANGED:
                                // reject changes in nested inset
-                               if (pos < size() && isInset(pos)) {
-                                       getInset(pos)->rejectChanges(bparams);
-                               }
+                               if (Inset * inset = getInset(pos))
+                                               inset->rejectChanges();
                                break;
 
                        case Change::INSERTED:
@@ -411,7 +442,7 @@ void Paragraph::rejectChanges(BufferParams const & bparams,
 void Paragraph::Private::insertChar(pos_type pos, char_type c,
                Change const & change)
 {
-       BOOST_ASSERT(pos >= 0 && pos <= int(text_.size()));
+       LASSERT(pos >= 0 && pos <= int(text_.size()), /**/);
 
        // track change
        changes_.insert(change, pos);
@@ -434,23 +465,29 @@ void Paragraph::Private::insertChar(pos_type pos, char_type c,
 }
 
 
-void Paragraph::insertInset(pos_type pos, Inset * inset,
+bool Paragraph::insertInset(pos_type pos, Inset * inset,
                                   Change const & change)
 {
-       BOOST_ASSERT(inset);
-       BOOST_ASSERT(pos >= 0 && pos <= size());
+       LASSERT(inset, /**/);
+       LASSERT(pos >= 0 && pos <= size(), /**/);
+
+       // Paragraph::insertInset() can be used in cut/copy/paste operation where
+       // d->inset_owner_ is not set yet.
+       if (d->inset_owner_ && !d->inset_owner_->insetAllowed(inset->lyxCode()))
+               return false;
 
        d->insertChar(pos, META_INSET, change);
-       BOOST_ASSERT(d->text_[pos] == META_INSET);
+       LASSERT(d->text_[pos] == META_INSET, /**/);
 
        // Add a new entry in the insetlist_.
        d->insetlist_.insert(inset, pos);
+       return true;
 }
 
 
 bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
 {
-       BOOST_ASSERT(pos >= 0 && pos <= size());
+       LASSERT(pos >= 0 && pos <= size(), return false);
 
        // keep the logic here in sync with the logic of isMergedOnEndOfParDeletion()
 
@@ -461,13 +498,13 @@ bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
                //  a) it was previously unchanged or
                //  b) it was inserted by a co-author
 
-               if (change.type == Change::UNCHANGED ||
-                   (change.type == Change::INSERTED && change.author != 0)) {
+               if (!change.changed() ||
+                     (change.inserted() && !change.currentAuthor())) {
                        setChange(pos, Change(Change::DELETED));
                        return false;
                }
 
-               if (change.type == Change::DELETED)
+               if (change.deleted())
                        return false;
        }
 
@@ -499,8 +536,8 @@ bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
 
 int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
 {
-       BOOST_ASSERT(start >= 0 && start <= size());
-       BOOST_ASSERT(end >= start && end <= size() + 1);
+       LASSERT(start >= 0 && start <= size(), /**/);
+       LASSERT(end >= start && end <= size() + 1, /**/);
 
        pos_type i = start;
        for (pos_type count = end - start; count; --count) {
@@ -512,7 +549,7 @@ int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
 
 
 int Paragraph::Private::latexSurrogatePair(odocstream & os, char_type c,
-               char_type next, Encoding const & encoding)
+               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
@@ -521,9 +558,20 @@ int Paragraph::Private::latexSurrogatePair(odocstream & os, char_type c,
        // hopefully impossible to input.
        // FIXME: change tracking
        // Is this correct WRT change tracking?
+       Encoding const & encoding = *(runparams.encoding);
        docstring const latex1 = encoding.latexChar(next);
        docstring const latex2 = encoding.latexChar(c);
-       os << latex1 << '{' << latex2 << '}';
+       if (docstring(1, next) == latex1) {
+               // the encoding supports the combination
+               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 << '}';
        return latex1.length() + latex2.length() + 2;
 }
 
@@ -541,15 +589,14 @@ bool Paragraph::Private::simpleTeXBlanks(OutputParams const & runparams,
        if (i + 1 < int(text_.size())) {
                char_type next = text_[i + 1];
                if (Encodings::isCombiningChar(next)) {
-                       Encoding const & encoding = *(runparams.encoding);
                        // This space has an accent, so we must always output it.
-                       column += latexSurrogatePair(os, ' ', next, encoding) - 1;
+                       column += latexSurrogatePair(os, ' ', next, runparams) - 1;
                        return true;
                }
        }
 
-       if (lyxrc.plaintext_linelen > 0
-           && column > lyxrc.plaintext_linelen
+       if (runparams.linelen > 0
+           && column > runparams.linelen
            && i
            && text_[i - 1] != ' '
            && (i + 1 < int(text_.size()))
@@ -577,7 +624,7 @@ bool Paragraph::Private::simpleTeXBlanks(OutputParams const & runparams,
 
 int Paragraph::Private::writeScriptChars(odocstream & os,
                                         docstring const & ltx,
-                                        Change & runningChange,
+                                        Change const & runningChange,
                                         Encoding const & encoding,
                                         pos_type & i)
 {
@@ -585,9 +632,10 @@ int Paragraph::Private::writeScriptChars(odocstream & os,
 
        // We only arrive here when a proper language for character text_[i] has
        // not been specified (i.e., it could not be translated in the current
-       // latex encoding) and it belongs to a known script.
-       // Parameter ltx contains the latex translation of text_[i] as specified in
-       // the unicodesymbols file and is something like "\textXXX{<spec>}".
+       // latex encoding) or its latex translation has been forced, and it
+       // belongs to a known script.
+       // 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
@@ -596,8 +644,16 @@ int Paragraph::Private::writeScriptChars(odocstream & os,
        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 length = ltx.substr(0, brace2).length();
-       os << ltx.substr(0, brace2);
+       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.
+               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];
@@ -631,8 +687,10 @@ int Paragraph::Private::writeScriptChars(odocstream & os,
                length += len;
                ++i;
        }
-       os << '}';
-       ++length;
+       if (closing_brace) {
+               os << '}';
+               ++length;
+       }
        return length;
 }
 
@@ -657,7 +715,7 @@ bool Paragraph::Private::isTextAt(string const & str, pos_type pos) const
 }
 
 
-void Paragraph::Private::latexInset(Buffer const & buf,
+void Paragraph::Private::latexInset(
                                             BufferParams const & bparams,
                                             odocstream & os,
                                             TexRow & texrow,
@@ -672,10 +730,10 @@ void Paragraph::Private::latexInset(Buffer const & buf,
                                             unsigned int & column)
 {
        Inset * inset = owner_->getInset(i);
-       BOOST_ASSERT(inset);
+       LASSERT(inset, /**/);
 
        if (style.pass_thru) {
-               inset->plaintext(buf, os, runparams);
+               inset->plaintext(os, runparams);
                return;
        }
 
@@ -708,24 +766,25 @@ void Paragraph::Private::latexInset(Buffer const & buf,
                column = 0;
        }
 
-       if (owner_->lookupChange(i).type == Change::DELETED) {
+       if (owner_->isDeleted(i)) {
                if( ++runparams.inDeletedInset == 1)
                        runparams.changeOfDeletedInset = owner_->lookupChange(i);
        }
 
        if (inset->canTrackChanges()) {
                column += Changes::latexMarkChange(os, bparams, running_change,
-                       Change(Change::UNCHANGED));
+                       Change(Change::UNCHANGED), runparams);
                running_change = Change(Change::UNCHANGED);
        }
 
        bool close = false;
        odocstream::pos_type const len = os.tellp();
 
-       if (inset->forceLTR() 
+       if (inset->forceLTR()
            && running_font.isRightToLeft()
-               // ERT is an exception, it should be output with no decorations at all
-               && inset->lyxCode() != ERT_CODE) {
+           // 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{}";
                else
@@ -768,13 +827,22 @@ void Paragraph::Private::latexInset(Buffer const & buf,
                }
        }
 
-       int tmp = inset->latex(buf, os, runparams);
+       int tmp;
+
+       try {
+               tmp = inset->latex(os, runparams);
+       } catch (EncodingException & e) {
+               // add location information and throw again.
+               e.par_id = id_;
+               e.pos = i;
+               throw(e);
+       }
 
        if (close) {
-       if (running_font.language()->lang() == "farsi")
-                       os << "\\endL{}";
-               else
-                       os << '}';
+               if (running_font.language()->lang() == "farsi")
+                               os << "\\endL{}";
+                       else
+                               os << '}';
        }
 
        if (tmp) {
@@ -787,16 +855,16 @@ void Paragraph::Private::latexInset(Buffer const & buf,
                column += os.tellp() - len;
        }
 
-       if (owner_->lookupChange(i).type == Change::DELETED)
+       if (owner_->isDeleted(i))
                --runparams.inDeletedInset;
 }
 
 
 void Paragraph::Private::latexSpecialChar(
                                             odocstream & os,
-                                            OutputParams & runparams,
-                                            Font & running_font,
-                                            Change & running_change,
+                                            OutputParams const & runparams,
+                                            Font const & running_font,
+                                            Change const & running_change,
                                             Layout const & style,
                                             pos_type & i,
                                             unsigned int & column)
@@ -812,13 +880,21 @@ void Paragraph::Private::latexSpecialChar(
        }
 
        if (runparams.verbatim) {
+               // FIXME UNICODE: This can fail if c cannot
+               // be encoded in the current encoding.
                os.put(c);
                return;
        }
 
-       if (lyxrc.fontenc == "T1" && latexSpecialT1(c, os, i, column))
+       // If T1 font encoding is used, use the special
+       // characters it provides.
+       // NOTE: some languages reset the font encoding
+       // internally
+       if (!running_font.language()->internalFontEncoding()
+           && lyxrc.fontenc == "T1" && latexSpecialT1(c, os, i, column))
                return;
 
+       // \tt font needs special treatment
        if (running_font.fontInfo().family() == TYPEWRITER_FAMILY
                && latexSpecialTypewriter(c, os, i, column))
                return;
@@ -867,7 +943,9 @@ void Paragraph::Private::latexSpecialChar(
                column += 17;
                break;
 
-       case '*': case '[':
+       case '*':
+       case '[':
+       case ']':
                // avoid being mistaken for optional arguments
                os << '{';
                os.put(c);
@@ -884,7 +962,6 @@ void Paragraph::Private::latexSpecialChar(
                break;
 
        default:
-
                // LyX, LaTeX etc.
                if (latexSpecialPhrase(os, i, column, runparams))
                        return;
@@ -896,7 +973,7 @@ void Paragraph::Private::latexSpecialChar(
                if (i + 1 < int(text_.size())) {
                        char_type next = text_[i + 1];
                        if (Encodings::isCombiningChar(next)) {
-                               column += latexSurrogatePair(os, c, next, encoding) - 1;
+                               column += latexSurrogatePair(os, c, next, runparams) - 1;
                                ++i;
                                break;
                        }
@@ -923,7 +1000,7 @@ void Paragraph::Private::latexSpecialChar(
 
 
 bool Paragraph::Private::latexSpecialT1(char_type const c, odocstream & os,
-       pos_type i, unsigned int & column)
+       pos_type i, unsigned int & column)
 {
        switch (c) {
        case '>':
@@ -931,17 +1008,19 @@ bool Paragraph::Private::latexSpecialT1(char_type const c, odocstream & os,
                os.put(c);
                // In T1 encoding, these characters exist
                // but we should avoid ligatures
-               if (i + 1 > int(text_.size()) || text_[i + 1] != c)
+               if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
                        return true;
-               os << "\\,{}";
-               column += 3;
-               // Alternative code:
-               //os << "\\textcompwordmark{}";
-               //column += 19;
+               os << "\\textcompwordmark{}";
+               column += 19;
                return true;
        case '|':
                os.put(c);
                return true;
+       case '\"':
+               // soul.sty breaks with \char`\"
+               os << "\\textquotedbl{}";
+               column += 14;
+               return true;
        default:
                return false;
        }
@@ -949,47 +1028,29 @@ bool Paragraph::Private::latexSpecialT1(char_type const c, odocstream & os,
 
 
 bool Paragraph::Private::latexSpecialTypewriter(char_type const c, odocstream & os,
-       pos_type i, unsigned int & column)
+       pos_type i, unsigned int & column)
 {
        switch (c) {
        case '-':
+               // within \ttfamily, "--" is merged to "-" (no endash)
+               // so we avoid this rather irritating ligature
                if (i + 1 < int(text_.size()) && text_[i + 1] == '-') {
-                       // "--" in Typewriter mode -> "-{}-"
                        os << "-{}";
                        column += 2;
                } else
                        os << '-';
                return true;
 
-       // I assume this is hack treating typewriter as verbatim
-       // FIXME UNICODE: This can fail if c cannot be encoded
-       // in the current encoding.
-
-       case '\0':
-               return true;
-
-       // Those characters are not directly supported.
-       case '\\':
-       case '\"':
-       case '$': case '&':
-       case '%': case '#': case '{':
-       case '}': case '_':
-       case '~':
-       case '^':
-       case '*': case '[':
-       case ' ':
-               return false;
-
+       // everything else has to be checked separately
+       // (depending on the encoding)
        default:
-               // With Typewriter font, these characters exist.
-               os.put(c);
-               return true;
+               return false;
        }
 }
 
 
 bool Paragraph::Private::latexSpecialPhrase(odocstream & os, pos_type & i,
-       unsigned int & column, OutputParams & runparams)
+       unsigned int & column, OutputParams const & runparams)
 {
        // FIXME: if we have "LaTeX" with a font
        // change in the middle (before the 'T', then
@@ -1011,15 +1072,26 @@ bool Paragraph::Private::latexSpecialPhrase(odocstream & os, pos_type & i,
 }
 
 
-void Paragraph::Private::validate(LaTeXFeatures & features,
-                               Layout const & layout) const
+void Paragraph::Private::validate(LaTeXFeatures & features) const
 {
+       if (layout_->inpreamble && inset_owner_) {
+               Buffer const & buf = inset_owner_->buffer();
+               BufferParams const & bp = buf.params();
+               Font f;
+               TexRow tr;
+               odocstringstream ods;
+               owner_->latex(bp, f, ods, tr, features.runparams());
+               docstring d = ods.str();
+               if (!d.empty())
+                       features.addPreambleSnippet(to_utf8(d));
+       }
+       
        // check the params.
        if (!params_.spacing().isDefault())
                features.require("setspace");
 
        // then the layouts
-       features.useLayout(layout.name());
+       features.useLayout(layout_->name());
 
        // then the fonts
        fontlist_.validate(features);
@@ -1034,7 +1106,7 @@ void Paragraph::Private::validate(LaTeXFeatures & features,
        for (; icit != iend; ++icit) {
                if (icit->inset) {
                        icit->inset->validate(features);
-                       if (layout.needprotect &&
+                       if (layout_->needprotect &&
                            icit->inset->lyxCode() == FOOT_CODE)
                                features.require("NeedLyXFootnoteCode");
                }
@@ -1059,8 +1131,12 @@ void Paragraph::Private::validate(LaTeXFeatures & features,
 //
 /////////////////////////////////////////////////////////////////////
 
-Paragraph::Paragraph()
-       : d(new Paragraph::Private(this))
+namespace {
+       Layout const emptyParagraphLayout;
+}
+
+Paragraph::Paragraph() 
+       : d(new Paragraph::Private(this, emptyParagraphLayout))
 {
        itemdepth = 0;
        d->params_.clear();
@@ -1071,6 +1147,15 @@ Paragraph::Paragraph(Paragraph const & par)
        : itemdepth(par.itemdepth),
        d(new Paragraph::Private(*par.d, this))
 {
+       registerWords();
+}
+
+
+Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
+       : itemdepth(par.itemdepth),
+       d(new Paragraph::Private(*par.d, this, beg, end))
+{
+       registerWords();
 }
 
 
@@ -1080,8 +1165,10 @@ Paragraph & Paragraph::operator=(Paragraph const & par)
        if (&par != this) {
                itemdepth = par.itemdepth;
 
+               deregisterWords();
                delete d;
                d = new Private(*par.d, this);
+               registerWords();
        }
        return *this;
 }
@@ -1089,23 +1176,35 @@ Paragraph & Paragraph::operator=(Paragraph const & par)
 
 Paragraph::~Paragraph()
 {
+       deregisterWords();
        delete d;
 }
 
 
-void Paragraph::write(Buffer const & buf, ostream & os,
-                         BufferParams const & bparams,
-                         depth_type & dth) const
+namespace {
+
+// this shall be called just before every "os << ..." action.
+void flushString(ostream & os, docstring & s)
+{
+       os << to_utf8(s);
+       s.erase();
+}
+
+}
+
+
+void Paragraph::write(ostream & os, BufferParams const & bparams,
+       depth_type & dth) const
 {
        // The beginning or end of a deeper (i.e. nested) area?
-       if (dth != params().depth()) {
-               if (params().depth() > dth) {
-                       while (params().depth() > dth) {
+       if (dth != d->params_.depth()) {
+               if (d->params_.depth() > dth) {
+                       while (d->params_.depth() > dth) {
                                os << "\n\\begin_deeper";
                                ++dth;
                        }
                } else {
-                       while (params().depth() < dth) {
+                       while (d->params_.depth() < dth) {
                                os << "\n\\end_deeper";
                                --dth;
                        }
@@ -1113,27 +1212,35 @@ void Paragraph::write(Buffer const & buf, ostream & os,
        }
 
        // First write the layout
-       os << "\n\\begin_layout " << to_utf8(layout()->name()) << '\n';
+       os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
 
-       params().write(os);
+       d->params_.write(os);
 
        Font font1(inherit_font, bparams.language);
 
        Change running_change = Change(Change::UNCHANGED);
 
+       // this string is used as a buffer to avoid repetitive calls
+       // to to_utf8(), which turn out to be expensive (JMarc)
+       docstring write_buffer;
+
        int column = 0;
        for (pos_type i = 0; i <= size(); ++i) {
 
-               Change change = lookupChange(i);
-               Changes::lyxMarkChange(os, column, running_change, change);
+               Change const change = lookupChange(i);
+               if (change != running_change)
+                       flushString(os, write_buffer);
+               Changes::lyxMarkChange(os, bparams, column, running_change, change);
                running_change = change;
 
                if (i == size())
                        break;
 
-               // Write font changes
+               // Write font changes (ignore spelling markers)
                Font font2 = getFontSettings(bparams, i);
+               font2.setMisspelled(false);
                if (font2 != font1) {
+                       flushString(os, write_buffer);
                        font2.lyxWriteChanges(font1, os);
                        column = 0;
                        font1 = font2;
@@ -1142,29 +1249,30 @@ void Paragraph::write(Buffer const & buf, ostream & os,
                char_type const c = d->text_[i];
                switch (c) {
                case META_INSET:
-               {
-                       Inset const * inset = getInset(i);
-                       if (inset)
+                       if (Inset const * inset = getInset(i)) {
+                               flushString(os, write_buffer);
                                if (inset->directWrite()) {
                                        // international char, let it write
                                        // code directly so it's shorter in
                                        // the file
-                                       inset->write(buf, os);
+                                       inset->write(os);
                                } else {
                                        if (i)
                                                os << '\n';
                                        os << "\\begin_inset ";
-                                       inset->write(buf, os);
+                                       inset->write(os);
                                        os << "\n\\end_inset\n\n";
                                        column = 0;
                                }
-               }
-               break;
+                       }
+                       break;
                case '\\':
+                       flushString(os, write_buffer);
                        os << "\n\\backslash\n";
                        column = 0;
                        break;
                case '.':
+                       flushString(os, write_buffer);
                        if (i + 1 < size() && d->text_[i + 1] == ' ') {
                                os << ".\n";
                                column = 0;
@@ -1174,30 +1282,29 @@ void Paragraph::write(Buffer const & buf, ostream & os,
                default:
                        if ((column > 70 && c == ' ')
                            || column > 79) {
+                               flushString(os, write_buffer);
                                os << '\n';
                                column = 0;
                        }
                        // this check is to amend a bug. LyX sometimes
                        // inserts '\0' this could cause problems.
-                       if (c != '\0') {
-                               std::vector<char> tmp = ucs4_to_utf8(c);
-                               tmp.push_back('\0');
-                               os << &tmp[0];
-                       } else
-                               lyxerr << "ERROR (Paragraph::writeFile):"
-                                       " NULL char in structure." << endl;
+                       if (c != '\0')
+                               write_buffer.push_back(c);
+                       else
+                               LYXERR0("NUL char in structure.");
                        ++column;
                        break;
                }
        }
 
+       flushString(os, write_buffer);
        os << "\n\\end_layout\n";
 }
 
 
 void Paragraph::validate(LaTeXFeatures & features) const
 {
-       d->validate(features, *layout());
+       d->validate(features);
 }
 
 
@@ -1228,13 +1335,13 @@ void Paragraph::appendString(docstring const & s, Font const & font,
        size_t newsize = oldsize + end;
        size_t capacity = d->text_.capacity();
        if (newsize >= capacity)
-               d->text_.reserve(std::max(capacity + 100, newsize));
+               d->text_.reserve(max(capacity + 100, newsize));
 
        // when appending characters, no need to update tables
        d->text_.append(s);
 
        // FIXME: Optimize this!
-       for (pos_type i = 0; i != end; ++i) {
+       for (size_t i = oldsize; i != newsize; ++i) {
                // track change
                d->changes_.insert(change, i);
        }
@@ -1268,18 +1375,13 @@ void Paragraph::insertChar(pos_type pos, char_type c,
 }
 
 
-void Paragraph::insertInset(pos_type pos, Inset * inset,
+bool Paragraph::insertInset(pos_type pos, Inset * inset,
                            Font const & font, Change const & change)
 {
-       insertInset(pos, inset, change);
+       bool const success = insertInset(pos, inset, change);
        // Set the font/language of the inset...
        setFont(pos, font);
-}
-
-
-bool Paragraph::insetAllowed(InsetCode code)
-{
-       return !d->inset_owner_ || d->inset_owner_->insetAllowed(code);
+       return success;
 }
 
 
@@ -1291,12 +1393,12 @@ void Paragraph::resetFonts(Font const & font)
 }
 
 // Gets uninstantiated font setting at position.
-Font const Paragraph::getFontSettings(BufferParams const & bparams,
+Font const Paragraph::getFontSettings(BufferParams const & bparams,
                                         pos_type pos) const
 {
        if (pos > size()) {
-               lyxerr << " pos: " << pos << " size: " << size() << endl;
-               BOOST_ASSERT(pos <= size());
+               LYXERR0("pos: " << pos << " size: " << size());
+               LASSERT(pos <= size(), /**/);
        }
 
        FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
@@ -1306,13 +1408,22 @@ Font const Paragraph::getFontSettings(BufferParams const & bparams,
        if (pos == size() && !empty())
                return getFontSettings(bparams, pos - 1);
 
-       return Font(inherit_font, getParLanguage(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;
+       Language const * lang = getParLanguage(bparams);
+       if (lang != previous_lang) {
+               previous_lang = lang;
+               previous_font = Font(inherit_font, lang);
+       }
+       return previous_font;
 }
 
 
 FontSpan Paragraph::fontSpan(pos_type pos) const
 {
-       BOOST_ASSERT(pos <= size());
+       LASSERT(pos <= size(), /**/);
        pos_type start = 0;
 
        FontList::const_iterator cit = d->fontlist_.begin();
@@ -1320,30 +1431,38 @@ FontSpan Paragraph::fontSpan(pos_type pos) const
        for (; cit != end; ++cit) {
                if (cit->pos() >= pos) {
                        if (pos >= beginOfBody())
-                               return FontSpan(std::max(start, beginOfBody()),
+                               return FontSpan(max(start, beginOfBody()),
                                                cit->pos());
                        else
                                return FontSpan(start,
-                                               std::min(beginOfBody() - 1,
+                                               min(beginOfBody() - 1,
                                                         cit->pos()));
                }
                start = cit->pos() + 1;
        }
 
        // This should not happen, but if so, we take no chances.
-       //lyxerr << "Paragraph::getEndPosOfFontSpan: This should not happen!"
-       //      << endl;
+       // LYXERR0("Paragraph::getEndPosOfFontSpan: This should not happen!");
        return FontSpan(pos, pos);
 }
 
 
 // Gets uninstantiated font setting at position 0
-Font const Paragraph::getFirstFontSettings(BufferParams const & bparams) const
+Font const Paragraph::getFirstFontSettings(BufferParams const & bparams) const
 {
        if (!empty() && !d->fontlist_.empty())
                return d->fontlist_.begin()->font();
 
-       return Font(inherit_font, bparams.language);
+       // 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;
+       if (bparams.language != previous_lang) {
+               previous_lang = bparams.language;
+               previous_font = Font(inherit_font, bparams.language);
+       }
+
+       return previous_font;
 }
 
 
@@ -1355,18 +1474,19 @@ Font const Paragraph::getFirstFontSettings(BufferParams const & bparams) const
 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
                                 Font const & outerfont) const
 {
-       BOOST_ASSERT(pos >= 0);
+       LASSERT(pos >= 0, /**/);
 
        Font font = getFontSettings(bparams, pos);
 
        pos_type const body_pos = beginOfBody();
+       FontInfo & fi = font.fontInfo();
        if (pos < body_pos)
-               font.fontInfo().realize(d->layout_->labelfont);
+               fi.realize(d->layout_->labelfont);
        else
-               font.fontInfo().realize(d->layout_->font);
+               fi.realize(d->layout_->font);
 
-       font.fontInfo().realize(outerfont.fontInfo());
-       font.fontInfo().realize(bparams.getFont().fontInfo());
+       fi.realize(outerfont.fontInfo());
+       fi.realize(bparams.getFont().fontInfo());
 
        return font;
 }
@@ -1375,7 +1495,7 @@ Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
 Font const Paragraph::getLabelFont
        (BufferParams const & bparams, Font const & outerfont) const
 {
-       FontInfo tmpfont = layout()->labelfont;
+       FontInfo tmpfont = d->layout_->labelfont;
        tmpfont.realize(outerfont.fontInfo());
        tmpfont.realize(bparams.getFont().fontInfo());
        return Font(tmpfont, getParLanguage(bparams));
@@ -1385,7 +1505,7 @@ Font const Paragraph::getLabelFont
 Font const Paragraph::getLayoutFont
        (BufferParams const & bparams, Font const & outerfont) const
 {
-       FontInfo tmpfont = layout()->font;
+       FontInfo tmpfont = d->layout_->font;
        tmpfont.realize(outerfont.fontInfo());
        tmpfont.realize(bparams.getFont().fontInfo());
        return Font(tmpfont, getParLanguage(bparams));
@@ -1400,8 +1520,7 @@ FontSize Paragraph::highestFontInRange
 }
 
 
-char_type
-Paragraph::getUChar(BufferParams const & bparams, pos_type pos) const
+char_type Paragraph::getUChar(BufferParams const & bparams, pos_type pos) const
 {
        char_type c = d->text_[pos];
        if (!lyxrc.rtl_support)
@@ -1436,14 +1555,13 @@ Paragraph::getUChar(BufferParams const & bparams, pos_type pos) const
        }
        if (uc != c && getFontSettings(bparams, pos).isRightToLeft())
                return uc;
-       else
-               return c;
+       return c;
 }
 
 
 void Paragraph::setFont(pos_type pos, Font const & font)
 {
-       BOOST_ASSERT(pos <= size());
+       LASSERT(pos <= size(), /**/);
 
        // First, reduce font against layout/label font
        // Update: The setCharFont() routine in text2.cpp already
@@ -1455,9 +1573,8 @@ void Paragraph::setFont(pos_type pos, Font const & font)
 
 void Paragraph::makeSameLayout(Paragraph const & par)
 {
-       layout(par.layout());
-       // move to pimpl?
-       d->params_ = par.params();
+       d->layout_ = par.d->layout_;
+       d->params_ = par.d->params_;
 }
 
 
@@ -1482,45 +1599,47 @@ bool Paragraph::stripLeadingSpaces(bool trackChanges)
 
 bool Paragraph::hasSameLayout(Paragraph const & par) const
 {
-       return par.layout() == layout() && d->params_.sameLayout(par.params());
+       return par.d->layout_ == d->layout_
+               && d->params_.sameLayout(par.d->params_);
 }
 
 
 depth_type Paragraph::getDepth() const
 {
-       return params().depth();
+       return d->params_.depth();
 }
 
 
 depth_type Paragraph::getMaxDepthAfter() const
 {
-       if (layout()->isEnvironment())
-               return params().depth() + 1;
+       if (d->layout_->isEnvironment())
+               return d->params_.depth() + 1;
        else
-               return params().depth();
+               return d->params_.depth();
 }
 
 
 char Paragraph::getAlign() const
 {
-       if (params().align() == LYX_ALIGN_LAYOUT)
-               return layout()->align;
+       if (d->params_.align() == LYX_ALIGN_LAYOUT)
+               return d->layout_->align;
        else
-               return params().align();
+               return d->params_.align();
 }
 
 
-docstring const & Paragraph::getLabelstring() const
+docstring const & Paragraph::labelString() const
 {
-       return params().labelString();
+       return d->params_.labelString();
 }
 
 
 // the next two functions are for the manual labels
 docstring const Paragraph::getLabelWidthString() const
 {
-       if (layout()->margintype == MARGIN_MANUAL)
-               return params().labelWidthString();
+       if (d->layout_->margintype == MARGIN_MANUAL
+           || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
+               return d->params_.labelWidthString();
        else
                return _("Senseless with this layout!");
 }
@@ -1528,39 +1647,21 @@ docstring const Paragraph::getLabelWidthString() const
 
 void Paragraph::setLabelWidthString(docstring const & s)
 {
-       params().labelWidthString(s);
-}
-
-
-docstring const Paragraph::translateIfPossible(docstring const & s,
-               BufferParams const & bparams) const
-{
-       if (!support::isAscii(s) || s.empty()) {
-               // This must be a user defined layout. We cannot translate
-               // this, since gettext accepts only ascii keys.
-               return s;
-       }
-       // Probably standard layout, try to translate
-       Messages & m = getMessages(getParLanguage(bparams)->code());
-       return m.get(to_ascii(s));
+       d->params_.labelWidthString(s);
 }
 
 
-docstring Paragraph::expandLabel(LayoutPtr const & layout,
+docstring Paragraph::expandLabel(Layout const & layout,
                BufferParams const & bparams, bool process_appendix) const
 {
-       TextClass const & tclass = bparams.getTextClass();
+       DocumentClass const & tclass = bparams.documentClass();
+       string const & lang = getParLanguage(bparams)->code();
+       bool const in_appendix = process_appendix && d->params_.appendix();
+       docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
 
-       docstring fmt;
-       if (process_appendix && params().appendix())
-               fmt = translateIfPossible(layout->labelstring_appendix(),
-                       bparams);
-       else
-               fmt = translateIfPossible(layout->labelstring(), bparams);
-
-       if (fmt.empty() && layout->labeltype == LABEL_COUNTER 
-           && !layout->counter.empty())
-               fmt = "\\the" + layout->counter;
+       if (fmt.empty() && layout.labeltype == LABEL_COUNTER 
+           && !layout.counter.empty())
+               return tclass.counters().theCounter(layout.counter, lang);
 
        // handle 'inherited level parts' in 'fmt',
        // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
@@ -1578,19 +1679,19 @@ docstring Paragraph::expandLabel(LayoutPtr const & layout,
                }
        }
 
-       return tclass.counters().counterLabel(fmt);
+       return tclass.counters().counterLabel(fmt, lang);
 }
 
 
-void Paragraph::applyLayout(LayoutPtr const & new_layout)
+void Paragraph::applyLayout(Layout const & new_layout)
 {
-       layout(new_layout);
-       LyXAlignment const oldAlign = params().align();
+       d->layout_ = &new_layout;
+       LyXAlignment const oldAlign = d->params_.align();
        
-       if (!(oldAlign & layout()->alignpossible)) {
+       if (!(oldAlign & d->layout_->alignpossible)) {
                frontend::Alert::warning(_("Alignment not permitted"), 
                        _("The new layout does not permit the alignment previously used.\nSetting to default."));
-               params().align(LYX_ALIGN_LAYOUT);
+               d->params_.align(LYX_ALIGN_LAYOUT);
        }
 }
 
@@ -1603,7 +1704,7 @@ pos_type Paragraph::beginOfBody() const
 
 void Paragraph::setBeginOfBody()
 {
-       if (layout()->labeltype != LABEL_MANUAL) {
+       if (d->layout_->labeltype != LABEL_MANUAL) {
                d->begin_of_body_ = 0;
                return;
        }
@@ -1636,9 +1737,15 @@ void Paragraph::setBeginOfBody()
 }
 
 
-bool Paragraph::forceDefaultParagraphs() const
+bool Paragraph::allowParagraphCustomization() const
 {
-       return inInset() && inInset()->forceDefaultParagraphs(0);
+       return inInset().allowParagraphCustomization();
+}
+
+
+bool Paragraph::usePlainLayout() const
+{
+       return inInset().usePlainLayout();
 }
 
 
@@ -1649,7 +1756,9 @@ namespace {
 
 bool noTrivlistCentering(InsetCode code)
 {
-       return code == FLOAT_CODE || code == WRAP_CODE;
+       return code == FLOAT_CODE
+              || code == WRAP_CODE
+              || code == CELL_CODE;
 }
 
 
@@ -1666,12 +1775,19 @@ string correction(string const & orig)
 
 
 string const corrected_env(string const & suffix, string const & env,
-       InsetCode code)
+       InsetCode code, bool const lastpar)
 {
        string output = suffix + "{";
-       if (noTrivlistCentering(code))
+       if (noTrivlistCentering(code)) {
+               if (lastpar) {
+                       // the last paragraph in non-trivlist-aligned
+                       // context is special (to avoid unwanted whitespace)
+                       if (suffix == "\\begin")
+                               return "\\" + correction(env) + "{}";
+                       return string();
+               }
                output += correction(env);
-       else
+       else
                output += env;
        output += "}";
        if (suffix == "\\begin")
@@ -1696,7 +1812,7 @@ void adjust_row_column(string const & str, TexRow & texrow, int & column)
 
 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
                                 odocstream & os, TexRow & texrow,
-                                bool moving_arg) const
+                                OutputParams const & runparams) const
 {
        int column = 0;
 
@@ -1719,13 +1835,17 @@ int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
        case LYX_ALIGN_LEFT:
        case LYX_ALIGN_RIGHT:
        case LYX_ALIGN_CENTER:
-               if (moving_arg) {
+               if (runparams.moving_arg) {
                        os << "\\protect";
                        column += 8;
                }
                break;
        }
 
+       string const begin_tag = "\\begin";
+       InsetCode code = ownerCode();
+       bool const lastpar = runparams.isLastPar;
+
        switch (curAlign) {
        case LYX_ALIGN_NONE:
        case LYX_ALIGN_BLOCK:
@@ -1735,24 +1855,24 @@ int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
        case LYX_ALIGN_LEFT: {
                string output;
                if (owner_->getParLanguage(bparams)->babel() != "hebrew")
-                       output = corrected_env("\\begin", "flushleft", owner_->ownerCode());
+                       output = corrected_env(begin_tag, "flushleft", code, lastpar);
                else
-                       output = corrected_env("\\begin", "flushright", owner_->ownerCode());
+                       output = corrected_env(begin_tag, "flushright", code, lastpar);
                os << from_ascii(output);
                adjust_row_column(output, texrow, column);
                break;
        } case LYX_ALIGN_RIGHT: {
                string output;
                if (owner_->getParLanguage(bparams)->babel() != "hebrew")
-                       output = corrected_env("\\begin", "flushright", owner_->ownerCode());
+                       output = corrected_env(begin_tag, "flushright", code, lastpar);
                else
-                       output = corrected_env("\\begin", "flushleft", owner_->ownerCode());
+                       output = corrected_env(begin_tag, "flushleft", code, lastpar);
                os << from_ascii(output);
                adjust_row_column(output, texrow, column);
                break;
        } case LYX_ALIGN_CENTER: {
                string output;
-               output = corrected_env("\\begin", "center", owner_->ownerCode());
+               output = corrected_env(begin_tag, "center", code, lastpar);
                os << from_ascii(output);
                adjust_row_column(output, texrow, column);
                break;
@@ -1765,11 +1885,16 @@ int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
 
 int Paragraph::Private::endTeXParParams(BufferParams const & bparams,
                               odocstream & os, TexRow & texrow,
-                              bool moving_arg) const
+                              OutputParams const & runparams) const
 {
        int column = 0;
 
-       switch (params_.align()) {
+       LyXAlignment const curAlign = params_.align();
+
+       if (curAlign == layout_->align)
+               return column;
+
+       switch (curAlign) {
        case LYX_ALIGN_NONE:
        case LYX_ALIGN_BLOCK:
        case LYX_ALIGN_LAYOUT:
@@ -1778,14 +1903,18 @@ int Paragraph::Private::endTeXParParams(BufferParams const & bparams,
        case LYX_ALIGN_LEFT:
        case LYX_ALIGN_RIGHT:
        case LYX_ALIGN_CENTER:
-               if (moving_arg) {
+               if (runparams.moving_arg) {
                        os << "\\protect";
                        column = 8;
                }
                break;
        }
 
-       switch (params_.align()) {
+       string const end_tag = "\n\\par\\end";
+       InsetCode code = ownerCode();
+       bool const lastpar = runparams.isLastPar;
+
+       switch (curAlign) {
        case LYX_ALIGN_NONE:
        case LYX_ALIGN_BLOCK:
        case LYX_ALIGN_LAYOUT:
@@ -1794,24 +1923,24 @@ int Paragraph::Private::endTeXParParams(BufferParams const & bparams,
        case LYX_ALIGN_LEFT: {
                string output;
                if (owner_->getParLanguage(bparams)->babel() != "hebrew")
-                       output = corrected_env("\n\\par\\end", "flushleft", owner_->ownerCode());
+                       output = corrected_env(end_tag, "flushleft", code, lastpar);
                else
-                       output = corrected_env("\n\\par\\end", "flushright", owner_->ownerCode());
+                       output = corrected_env(end_tag, "flushright", code, lastpar);
                os << from_ascii(output);
                adjust_row_column(output, texrow, column);
                break;
        } case LYX_ALIGN_RIGHT: {
                string output;
                if (owner_->getParLanguage(bparams)->babel() != "hebrew")
-                       output = corrected_env("\n\\par\\end", "flushright", owner_->ownerCode());
+                       output = corrected_env(end_tag, "flushright", code, lastpar);
                else
-                       output = corrected_env("\n\\par\\end", "flushleft", owner_->ownerCode());
+                       output = corrected_env(end_tag, "flushleft", code, lastpar);
                os << from_ascii(output);
                adjust_row_column(output, texrow, column);
                break;
        } case LYX_ALIGN_CENTER: {
                string output;
-               output = corrected_env("\n\\par\\end", "center", owner_->ownerCode());
+               output = corrected_env(end_tag, "center", code, lastpar);
                os << from_ascii(output);
                adjust_row_column(output, texrow, column);
                break;
@@ -1823,30 +1952,25 @@ int Paragraph::Private::endTeXParParams(BufferParams const & bparams,
 
 
 // This one spits out the text of the paragraph
-bool Paragraph::latex(Buffer const & buf,
-                               BufferParams const & bparams,
-                               Font const & outerfont,
-                               odocstream & os, TexRow & texrow,
-                               OutputParams const & runparams) const
+bool Paragraph::latex(BufferParams const & bparams,
+       Font const & outerfont,
+       odocstream & os, TexRow & texrow,
+       OutputParams const & runparams,
+       int start_pos, int end_pos) const
 {
-       LYXERR(Debug::LATEX, "SimpleTeXOnePar...     " << this);
+       LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
 
-       bool return_value = false;
+       if (layout().inpreamble)
+               return true;
 
-       LayoutPtr style;
+       bool return_value = false;
 
-       // well we have to check if we are in an inset with unlimited
-       // length (all in one row) if that is true then we don't allow
-       // any special options in the paragraph and also we don't allow
-       // any environment other than the default layout of the text class
-       // to be valid!
-       bool asdefault = forceDefaultParagraphs();
+       bool const allowcust = allowParagraphCustomization();
 
-       if (asdefault) {
-               style = bparams.getTextClass().defaultLayout();
-       } else {
-               style = layout();
-       }
+       // FIXME This check should not be needed. Perhaps issue an
+       // error if it triggers.
+       Layout const & style = inInset().forcePlainLayout() ?
+               bparams.documentClass().plainLayout() : *d->layout_;
 
        // Current base font for all inherited font changes, without any
        // change caused by an individual character, except for the language:
@@ -1861,10 +1985,8 @@ bool Paragraph::latex(Buffer const & buf,
        unsigned int column = 0;
 
        if (body_pos > 0) {
-               // the optional argument is kept in curly brackets in
-               // case it contains a ']'
-               os << "[{";
-               column += 2;
+               os << '[';
+               column += 1;
                basefont = getLabelFont(bparams, outerfont);
        } else {
                basefont = getLayoutFont(bparams, outerfont);
@@ -1877,17 +1999,19 @@ bool Paragraph::latex(Buffer const & buf,
 
        Change runningChange = Change(Change::UNCHANGED);
 
+       Encoding const * const prev_encoding = runparams.encoding;
+
        texrow.start(id(), 0);
 
        // if the paragraph is empty, the loop will not be entered at all
        if (empty()) {
-               if (style->isCommand()) {
+               if (style.isCommand()) {
                        os << '{';
                        ++column;
                }
-               if (!asdefault)
+               if (allowcust)
                        column += d->startTeXParParams(bparams, os, texrow,
-                                                   runparams.moving_arg);
+                                                   runparams);
        }
 
        for (pos_type i = 0; i < size(); ++i) {
@@ -1904,21 +2028,22 @@ bool Paragraph::latex(Buffer const & buf,
                                running_font = basefont;
 
                                column += Changes::latexMarkChange(os, bparams,
-                                               runningChange, Change(Change::UNCHANGED));
+                                               runningChange, Change(Change::UNCHANGED),
+                                               runparams);
                                runningChange = Change(Change::UNCHANGED);
 
-                               os << "}] ";
-                               column +=3;
+                               os << "] ";
+                               column +=2;
                        }
-                       if (style->isCommand()) {
+                       if (style.isCommand()) {
                                os << '{';
                                ++column;
                        }
 
-                       if (!asdefault)
+                       if (allowcust)
                                column += d->startTeXParParams(bparams, os,
                                                            texrow,
-                                                           runparams.moving_arg);
+                                                           runparams);
                }
 
                Change const & change = runparams.inDeletedInset ? runparams.changeOfDeletedInset
@@ -1933,13 +2058,14 @@ bool Paragraph::latex(Buffer const & buf,
                        basefont = getLayoutFont(bparams, outerfont);
                        running_font = basefont;
 
-                       column += Changes::latexMarkChange(os, bparams, runningChange, change);
+                       column += Changes::latexMarkChange(os, bparams, runningChange,
+                                                          change, runparams);
                        runningChange = change;
                }
 
                // do not output text which is marked deleted
                // if change tracking output is disabled
-               if (!bparams.outputChanges && change.type == Change::DELETED) {
+               if (!bparams.outputChanges && change.deleted()) {
                        continue;
                }
 
@@ -1962,13 +2088,22 @@ bool Paragraph::latex(Buffer const & buf,
                        open_font = false;
                }
 
+               // close babel's font environment before opening CJK.
+               if (!running_font.language()->babel().empty() &&
+                   font.language()->encoding()->package() == Encoding::CJK) {
+                               string end_tag = subst(lyxrc.language_command_end,
+                                                       "$$lang",
+                                                       running_font.language()->babel());
+                               os << from_ascii(end_tag);
+                               column += end_tag.length();
+               }
+
                // Switch file encoding if necessary (and allowed)
                if (!runparams.verbatim && 
-                   runparams.encoding->package() == Encoding::inputenc &&
-                   font.language()->encoding()->package() == Encoding::inputenc) {
-                       std::pair<bool, int> const enc_switch = switchEncoding(os, bparams,
-                                       runparams.moving_arg, *(runparams.encoding),
-                                       *(font.language()->encoding()));
+                   runparams.encoding->package() != Encoding::none &&
+                   font.language()->encoding()->package() != Encoding::none) {
+                       pair<bool, int> const enc_switch = switchEncoding(os, bparams,
+                                       runparams, *(font.language()->encoding()));
                        if (enc_switch.first) {
                                column += enc_switch.second;
                                runparams.encoding = font.language()->encoding();
@@ -1989,25 +2124,31 @@ bool Paragraph::latex(Buffer const & buf,
                        running_font = font;
                        open_font = true;
                        docstring fontchange = ods.str();
+                       // check whether the fontchange ends with a \\textcolor
+                       // modifier and the text starts with a 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)
-                       if (suffixIs(fontchange, ' ') && c == ' ')
+                       else if (suffixIs(fontchange, ' ') && c == ' ')
                                os << fontchange.substr(0, fontchange.size() - 1) 
                                   << from_ascii("{}");
                        else
                                os << fontchange;
                }
 
-               if (c == ' ') {
+               // FIXME: think about end_pos implementation...
+               if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
                        // FIXME: integrate this case in latexSpecialChar
                        // Do not print the separation of the optional argument
-                       // if style->pass_thru is false. This works because
+                       // if style.pass_thru is false. This works because
                        // latexSpecialChar ignores spaces if
-                       // style->pass_thru is false.
+                       // style.pass_thru is false.
                        if (i != body_pos - 1) {
                                if (d->simpleTeXBlanks(
                                                runparams, os, texrow,
-                                               i, column, font, *style)) {
+                                               i, column, font, style)) {
                                        // A surrogate pair was output. We
                                        // must not call latexSpecialChar
                                        // in this iteration, since it would output
@@ -2019,23 +2160,42 @@ bool Paragraph::latex(Buffer const & buf,
                }
 
                OutputParams rp = runparams;
-               rp.free_spacing = style->free_spacing;
+               rp.free_spacing = style.free_spacing;
                rp.local_font = &font;
-               rp.intitle = style->intitle;
+               rp.intitle = style.intitle;
 
                // Two major modes:  LaTeX or plain
                // Handle here those cases common to both modes
                // and then split to handle the two modes separately.
-               if (c == META_INSET)
-                       d->latexInset(buf, bparams, os,
-                                       texrow, rp, running_font,
-                                       basefont, outerfont, open_font,
-                                       runningChange, *style, i, column);
-               else
-                       d->latexSpecialChar(os, rp, running_font, runningChange,
-                               *style, i, column);
+               if (c == META_INSET) {
+                       if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
+                               d->latexInset(bparams, os,
+                                               texrow, rp, running_font,
+                                               basefont, outerfont, open_font,
+                                               runningChange, style, i, column);
+                       }
+               } else {
+                       if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
+                               try {
+                                       d->latexSpecialChar(os, rp, running_font, runningChange,
+                                               style, i, column);
+                               } catch (EncodingException & e) {
+                               if (runparams.dryrun) {
+                                       os << "<" << _("LyX Warning: ")
+                                          << _("uncodable character") << " '";
+                                       os.put(c);
+                                       os << "'>";
+                               } else {
+                                       // add location information and throw again.
+                                       e.par_id = id();
+                                       e.pos = i;
+                                       throw(e);
+                               }
+                       }
+               }
+               }
 
-               // Set the encoding to that returned from simpleTeXSpecialChars (see
+               // Set the encoding to that returned from latexSpecialChar (see
                // comment for encoding member in OutputParams.h)
                runparams.encoding = rp.encoding;
        }
@@ -2061,20 +2221,23 @@ bool Paragraph::latex(Buffer const & buf,
 #endif
        }
 
-       column += Changes::latexMarkChange(os, bparams, runningChange, Change(Change::UNCHANGED));
+       column += Changes::latexMarkChange(os, bparams, runningChange,
+                                          Change(Change::UNCHANGED), runparams);
 
        // Needed if there is an optional argument but no contents.
        if (body_pos > 0 && body_pos == size()) {
-               os << "}]~";
+               os << "]~";
                return_value = false;
        }
 
-       if (!asdefault) {
-               column += d->endTeXParParams(bparams, os, texrow,
-                                         runparams.moving_arg);
+       if (allowcust && d->endTeXParParams(bparams, os, texrow, runparams)
+           && runparams.encoding != prev_encoding) {
+               runparams.encoding = prev_encoding;
+               if (!bparams.useXetex)
+                       os << setEncoding(prev_encoding->iconvName());
        }
 
-       LYXERR(Debug::LATEX, "SimpleTeXOnePar...done " << this);
+       LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
        return return_value;
 }
 
@@ -2082,9 +2245,10 @@ bool Paragraph::latex(Buffer const & buf,
 bool Paragraph::emptyTag() const
 {
        for (pos_type i = 0; i < size(); ++i) {
-               if (isInset(i)) {
-                       Inset const * inset = getInset(i);
+               if (Inset const * inset = getInset(i)) {
                        InsetCode lyx_code = inset->lyxCode();
+                       // FIXME testing like that is wrong. What is
+                       // the intent?
                        if (lyx_code != TOC_CODE &&
                            lyx_code != INCLUDE_CODE &&
                            lyx_code != GRAPHICS_CODE &&
@@ -2104,11 +2268,11 @@ bool Paragraph::emptyTag() const
 }
 
 
-string Paragraph::getID(Buffer const & buf, OutputParams const & runparams) const
+string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
+       const
 {
        for (pos_type i = 0; i < size(); ++i) {
-               if (isInset(i)) {
-                       Inset const * inset = getInset(i);
+               if (Inset const * inset = getInset(i)) {
                        InsetCode lyx_code = inset->lyxCode();
                        if (lyx_code == LABEL_CODE) {
                                InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
@@ -2116,19 +2280,18 @@ string Paragraph::getID(Buffer const & buf, OutputParams const & runparams) cons
                                return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
                        }
                }
-
        }
        return string();
 }
 
 
-pos_type Paragraph::getFirstWord(Buffer const & buf, odocstream & os, OutputParams const & runparams) const
+pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runparams)
+       const
 {
        pos_type i;
        for (i = 0; i < size(); ++i) {
-               if (isInset(i)) {
-                       Inset const * inset = getInset(i);
-                       inset->docbook(buf, os, runparams);
+               if (Inset const * inset = getInset(i)) {
+                       inset->docbook(os, runparams);
                } else {
                        char_type c = d->text_[i];
                        if (c == ' ')
@@ -2140,6 +2303,24 @@ pos_type Paragraph::getFirstWord(Buffer const & buf, odocstream & os, OutputPara
 }
 
 
+pos_type Paragraph::firstWordLyXHTML(XHTMLStream & xs, OutputParams const & runparams)
+       const
+{
+       pos_type i;
+       for (i = 0; i < size(); ++i) {
+               if (Inset const * inset = getInset(i)) {
+                       inset->xhtml(xs, runparams);
+               } else {
+                       char_type c = d->text_[i];
+                       if (c == ' ')
+                               break;
+                       xs << c;
+               }
+       }
+       return i;
+}
+
+
 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
 {
        Font font_old;
@@ -2165,11 +2346,11 @@ void Paragraph::simpleDocBookOnePar(Buffer const & buf,
 {
        bool emph_flag = false;
 
-       LayoutPtr const & style = layout();
+       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))
+       if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
                os << "]]>";
 
        // parsing main loop
@@ -2187,13 +2368,12 @@ void Paragraph::simpleDocBookOnePar(Buffer const & buf,
                        }
                }
 
-               if (isInset(i)) {
-                       Inset const * inset = getInset(i);
-                       inset->docbook(buf, os, runparams);
+               if (Inset const * inset = getInset(i)) {
+                       inset->docbook(os, runparams);
                } else {
                        char_type c = d->text_[i];
 
-                       if (style->pass_thru)
+                       if (style.pass_thru)
                                os.put(c);
                        else
                                os << sgml::escapeChar(c);
@@ -2205,45 +2385,161 @@ void Paragraph::simpleDocBookOnePar(Buffer const & buf,
                os << "</emphasis>";
        }
 
-       if (style->free_spacing)
+       if (style.free_spacing)
                os << '\n';
-       if (style->pass_thru && !d->onlyText(buf, outerfont, initial))
+       if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
                os << "<![CDATA[";
 }
 
 
+docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
+                                   XHTMLStream & xs,
+                                   OutputParams const & runparams,
+                                   Font const & outerfont,
+                                               bool fortoc,
+                                   pos_type initial) const
+{
+       docstring retval;
+
+       bool emph_flag = false;
+       bool bold_flag = false;
+       string closing_tag;
+
+       Layout const & style = *d->layout_;
+
+       if (!fortoc) {
+               // generate a magic label for this paragraph
+               string const attr = "id='" + magicLabel() + "'";
+               xs << CompTag("a", attr);
+       }
+
+       FontInfo font_old =
+               style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
+
+       // parsing main loop
+       for (pos_type i = initial; i < size(); ++i) {
+               Font font = getFont(buf.params(), i, outerfont);
+
+               // emphasis
+               if (font_old.emph() != font.fontInfo().emph()) {
+                       if (font.fontInfo().emph() == FONT_ON) {
+                               xs << StartTag("em");
+                               emph_flag = true;
+                       } else if (emph_flag && i != initial) {
+                               xs << EndTag("em");
+                               emph_flag = false;
+                       }
+               }
+               // bold
+               if (font_old.series() != font.fontInfo().series()) {
+                       if (font.fontInfo().series() == BOLD_SERIES) {
+                               xs << StartTag("strong");
+                               bold_flag = true;
+                       } else if (bold_flag && i != initial) {
+                               xs << EndTag("strong");
+                               bold_flag = false;
+                       }
+               }
+               // FIXME XHTML
+               // Other such tags? What about the other text ranges?
+
+               Inset const * inset = getInset(i);
+               if (inset) {
+                       InsetCommand const * ic = inset->asInsetCommand();
+                       InsetLayout const & il = inset->getLayout();
+                       if (!fortoc || il.isInToc() || (ic && ic->isInToc())) {
+                               OutputParams np = runparams;
+                               if (!il.htmlisblock())
+                                       np.html_in_par = true;
+                               retval += inset->xhtml(xs, np);
+                       }
+               } else {
+                       char_type c = d->text_[i];
+
+                       if (style.pass_thru)
+                               xs << c;
+                       else if (c == '-') {
+                               docstring str;
+                               int j = i + 1;
+                               if (j < size() && d->text_[j] == '-') {
+                                       j += 1;
+                                       if (j < size() && d->text_[j] == '-') {
+                                               str += from_ascii("&mdash;");
+                                               i += 2;
+                                       } else {
+                                               str += from_ascii("&ndash;");
+                                               i += 1;
+                                       }
+                               }
+                               else
+                                       str += c;
+                               // We don't want to escape the entities. Note that
+                               // it is safe to do this, since str can otherwise
+                               // only be "-". E.g., it can't be "<".
+                               xs << XHTMLStream::NextRaw() << str;
+                       } else
+                               xs << c;
+               }
+               font_old = font.fontInfo();
+       }
+
+       xs.closeFontTags();
+       return retval;
+}
+
+
 bool Paragraph::isHfill(pos_type pos) const
 {
-       return isInset(pos)
-               && getInset(pos)->lyxCode() == HFILL_CODE;
+       Inset const * inset = getInset(pos);
+       return inset && (inset->lyxCode() == SPACE_CODE &&
+                        inset->isStretchableSpace());
 }
 
 
 bool Paragraph::isNewline(pos_type pos) const
 {
-       return isInset(pos)
-               && getInset(pos)->lyxCode() == NEWLINE_CODE;
+       Inset const * inset = getInset(pos);
+       return inset && inset->lyxCode() == NEWLINE_CODE;
 }
 
 
 bool Paragraph::isLineSeparator(pos_type pos) const
 {
        char_type const c = d->text_[pos];
-       return isLineSeparatorChar(c)
-               || (c == META_INSET && getInset(pos) &&
-               getInset(pos)->isLineSeparator());
+       if (isLineSeparatorChar(c))
+               return true;
+       Inset const * inset = getInset(pos);
+       return inset && inset->isLineSeparator();
 }
 
 
-/// Used by the spellchecker
-bool Paragraph::isLetter(pos_type pos) const
+bool Paragraph::isWordSeparator(pos_type pos) const
 {
-       if (isInset(pos))
-               return getInset(pos)->isLetter();
-       else {
-               char_type const c = d->text_[pos];
-               return isLetterChar(c) || isDigit(c);
-       }
+       if (Inset const * inset = getInset(pos))
+               return !inset->isLetter();
+       char_type const c = d->text_[pos];
+       // We want to pass the ' and escape chars to the spellchecker
+       static docstring const quote = from_utf8(lyxrc.spellchecker_esc_chars + '\'');
+       return (!isLetterChar(c) && !isDigit(c) && !contains(quote, c))
+               || pos == size();
+}
+
+
+bool Paragraph::isChar(pos_type pos) const
+{
+       if (Inset const * inset = getInset(pos))
+               return inset->isChar();
+       char_type const c = d->text_[pos];
+       return !isLetterChar(c) && !isDigit(c) && !lyx::isSpace(c);
+}
+
+
+bool Paragraph::isSpace(pos_type pos) const
+{
+       if (Inset const * inset = getInset(pos))
+               return inset->isSpace();
+       char_type const c = d->text_[pos];
+       return lyx::isSpace(c);
 }
 
 
@@ -2261,8 +2557,7 @@ bool Paragraph::isRTL(BufferParams const & bparams) const
 {
        return lyxrc.rtl_support
                && getParLanguage(bparams)->rightToLeft()
-               && ownerCode() != ERT_CODE
-               && ownerCode() != LISTINGS_CODE;
+               && !inInset().getLayout().forceLTR();
 }
 
 
@@ -2295,36 +2590,61 @@ bool Paragraph::isMultiLingual(BufferParams const & bparams) const
 }
 
 
-// Convert the paragraph to a string.
-// Used for building the table of contents
-docstring const Paragraph::asString(Buffer const & buffer, bool label) const
+docstring Paragraph::asString(int options) const
 {
-       return asString(buffer, 0, size(), label);
+       return asString(0, size(), options);
 }
 
 
-docstring const Paragraph::asString(Buffer const & buffer,
-                                pos_type beg, pos_type end, bool label) const
+docstring Paragraph::asString(pos_type beg, pos_type end, int options) const
 {
+       odocstringstream os;
+
+       if (beg == 0 
+               && options & AS_STR_LABEL
+               && !d->params_.labelString().empty())
+               os << d->params_.labelString() << ' ';
+
+       for (pos_type i = beg; i < end; ++i) {
+               char_type const c = d->text_[i];
+               if (isPrintable(c) || c == '\t'
+                   || (c == '\n' && (options & AS_STR_NEWLINES)))
+                       os.put(c);
+               else if (c == META_INSET && (options & AS_STR_INSETS)) {
+                       getInset(i)->tocString(os);
+                       if (getInset(i)->asInsetMath())
+                               os << " ";
+               }
+       }
+
+       return os.str();
+}
 
+
+docstring Paragraph::stringify(pos_type beg, pos_type end, int options, OutputParams & runparams) const
+{
        odocstringstream os;
 
-       if (beg == 0 && label && !params().labelString().empty())
-               os << params().labelString() << ' ';
+       if (beg == 0 
+               && options & AS_STR_LABEL
+               && !d->params_.labelString().empty())
+               os << d->params_.labelString() << ' ';
 
        for (pos_type i = beg; i < end; ++i) {
                char_type const c = d->text_[i];
-               if (isPrintable(c))
+               if (isPrintable(c) || c == '\t'
+                   || (c == '\n' && (options & AS_STR_NEWLINES)))
                        os.put(c);
-               else if (c == META_INSET)
-                       getInset(i)->textString(buffer, os);
+               else if (c == META_INSET && (options & AS_STR_INSETS)) {
+                       getInset(i)->plaintext(os, runparams);
+               }
        }
 
        return os.str();
 }
 
 
-void Paragraph::setInsetOwner(Inset * inset)
+void Paragraph::setInsetOwner(Inset const * inset)
 {
        d->inset_owner_ = inset;
 }
@@ -2336,28 +2656,50 @@ int Paragraph::id() const
 }
 
 
-LayoutPtr const & Paragraph::layout() const
+void Paragraph::setId(int id)
 {
-       return d->layout_;
+       d->id_ = id;
 }
 
 
-void Paragraph::layout(LayoutPtr const & new_layout)
+Layout const & Paragraph::layout() const
 {
-       d->layout_ = new_layout;
+       return *d->layout_;
 }
 
 
-Inset * Paragraph::inInset() const
+void Paragraph::setLayout(Layout const & layout)
 {
-       return d->inset_owner_;
+       d->layout_ = &layout;
 }
 
 
-InsetCode Paragraph::ownerCode() const
+void Paragraph::setDefaultLayout(DocumentClass const & tc)
+{ 
+       setLayout(tc.defaultLayout()); 
+}
+
+
+void Paragraph::setPlainLayout(DocumentClass const & tc)
+{ 
+       setLayout(tc.plainLayout()); 
+}
+
+
+void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
 {
-       return d->inset_owner_ ?
-               d->inset_owner_->lyxCode() : NO_CODE;
+       if (usePlainLayout())
+               setPlainLayout(tclass);
+       else
+               setDefaultLayout(tclass);
+}
+
+
+Inset const & Paragraph::inInset() const
+{
+       LASSERT(d->inset_owner_, throw ExceptionMessage(BufferException,
+               _("Memory problem"), _("Paragraph not properly initialized")));
+       return *d->inset_owner_;
 }
 
 
@@ -2375,7 +2717,7 @@ ParagraphParameters const & Paragraph::params() const
 
 bool Paragraph::isFreeSpacing() const
 {
-       if (layout()->free_spacing)
+       if (d->layout_->free_spacing)
                return true;
        return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
 }
@@ -2383,7 +2725,7 @@ bool Paragraph::isFreeSpacing() const
 
 bool Paragraph::allowEmpty() const
 {
-       if (layout()->keepempty)
+       if (d->layout_->keepempty)
                return true;
        return d->inset_owner_ && d->inset_owner_->allowEmpty();
 }
@@ -2391,7 +2733,7 @@ bool Paragraph::allowEmpty() const
 
 char_type Paragraph::transformChar(char_type c, pos_type pos) const
 {
-       if (!Encodings::is_arabic(c))
+       if (!Encodings::isArabicChar(c))
                return c;
 
        char_type prev_char = ' ';
@@ -2399,7 +2741,7 @@ char_type Paragraph::transformChar(char_type c, pos_type pos) const
 
        for (pos_type i = pos - 1; i >= 0; --i) {
                char_type const par_char = d->text_[i];
-               if (!Encodings::isComposeChar_arabic(par_char)) {
+               if (!Encodings::isArabicComposeChar(par_char)) {
                        prev_char = par_char;
                        break;
                }
@@ -2407,21 +2749,21 @@ char_type Paragraph::transformChar(char_type c, pos_type pos) const
 
        for (pos_type i = pos + 1, end = size(); i < end; ++i) {
                char_type const par_char = d->text_[i];
-               if (!Encodings::isComposeChar_arabic(par_char)) {
+               if (!Encodings::isArabicComposeChar(par_char)) {
                        next_char = par_char;
                        break;
                }
        }
 
-       if (Encodings::is_arabic(next_char)) {
-               if (Encodings::is_arabic(prev_char) &&
-                       !Encodings::is_arabic_special(prev_char))
+       if (Encodings::isArabicChar(next_char)) {
+               if (Encodings::isArabicChar(prev_char) &&
+                       !Encodings::isArabicSpecialChar(prev_char))
                        return Encodings::transformChar(c, Encodings::FORM_MEDIAL);
                else
                        return Encodings::transformChar(c, Encodings::FORM_INITIAL);
        } else {
-               if (Encodings::is_arabic(prev_char) &&
-                       !Encodings::is_arabic_special(prev_char))
+               if (Encodings::isArabicChar(prev_char) &&
+                       !Encodings::isArabicSpecialChar(prev_char))
                        return Encodings::transformChar(c, Encodings::FORM_FINAL);
                else
                        return Encodings::transformChar(c, Encodings::FORM_ISOLATED);
@@ -2429,14 +2771,14 @@ char_type Paragraph::transformChar(char_type c, pos_type pos) const
 }
 
 
-int Paragraph::checkBiblio(bool track_changes)
+int Paragraph::checkBiblio(Buffer const & buffer)
 {
-       //FIXME From JS:
-       //This is getting more and more a mess. ...We really should clean
-       //up this bibitem issue for 1.6. See also bug 2743.
+       // FIXME From JS:
+       // This is getting more and more a mess. ...We really should clean
+       // up this bibitem issue for 1.6. See also bug 2743.
 
        // Add bibitem insets if necessary
-       if (layout()->labeltype != LABEL_BIBLIO)
+       if (d->layout_->labeltype != LABEL_BIBLIO)
                return 0;
 
        bool hasbibitem = !d->insetlist_.empty()
@@ -2444,6 +2786,8 @@ int Paragraph::checkBiblio(bool track_changes)
                && d->text_[0] == META_INSET
                && d->insetlist_.begin()->inset->lyxCode() == BIBITEM_CODE;
 
+       bool track_changes = buffer.params().trackChanges;
+
        docstring oldkey;
        docstring oldlabel;
 
@@ -2466,13 +2810,13 @@ int Paragraph::checkBiblio(bool track_changes)
                        break;
        }
 
-       //There was an InsetBibitem at the beginning, and we didn't
-       //have to erase one.
+       // There was an InsetBibitem at the beginning, and we didn't
+       // have to erase one.
        if (hasbibitem && erasedInsetPosition < 0)
                        return 0;
 
-       //There was an InsetBibitem at the beginning and we did have to
-       //erase one. So we give its properties to the beginning inset.
+       // There was an InsetBibitem at the beginning and we did have to
+       // erase one. So we give its properties to the beginning inset.
        if (hasbibitem) {
                InsetBibitem * inset =
                        static_cast<InsetBibitem *>(d->insetlist_.begin()->inset);
@@ -2482,9 +2826,10 @@ int Paragraph::checkBiblio(bool track_changes)
                return -erasedInsetPosition;
        }
 
-       //There was no inset at the beginning, so we need to create one with
-       //the key and label of the one we erased.
-       InsetBibitem * inset(new InsetBibitem(InsetCommandParams(BIBITEM_CODE)));
+       // There was no inset at the beginning, so we need to create one with
+       // the key and label of the one we erased.
+       InsetBibitem * inset = 
+               new InsetBibitem(const_cast<Buffer *>(&buffer), InsetCommandParams(BIBITEM_CODE));
        // restore values of previously deleted item in this par.
        if (!oldkey.empty())
                inset->setParam("key", oldkey);
@@ -2496,39 +2841,27 @@ int Paragraph::checkBiblio(bool track_changes)
 }
 
 
-unsigned int Paragraph::macrocontextPosition() const
-{
-       return d->macrocontext_position_;
-}
-
-
-void Paragraph::setMacrocontextPosition(unsigned int pos)
-{
-       d->macrocontext_position_ = pos;
-}
-
-
 void Paragraph::checkAuthors(AuthorList const & authorList)
 {
        d->changes_.checkAuthors(authorList);
 }
 
 
-bool Paragraph::isUnchanged(pos_type pos) const
+bool Paragraph::isChanged(pos_type pos) const
 {
-       return lookupChange(pos).type == Change::UNCHANGED;
+       return lookupChange(pos).changed();
 }
 
 
 bool Paragraph::isInserted(pos_type pos) const
 {
-       return lookupChange(pos).type == Change::INSERTED;
+       return lookupChange(pos).inserted();
 }
 
 
 bool Paragraph::isDeleted(pos_type pos) const
 {
-       return lookupChange(pos).type == Change::DELETED;
+       return lookupChange(pos).deleted();
 }
 
 
@@ -2538,6 +2871,12 @@ InsetList const & Paragraph::insetList() const
 }
 
 
+void Paragraph::setBuffer(Buffer & b)
+{
+       d->insetlist_.setBuffer(b);
+}
+
+
 Inset * Paragraph::releaseInset(pos_type pos)
 {
        Inset * inset = d->insetlist_.release(pos);
@@ -2549,18 +2888,20 @@ Inset * Paragraph::releaseInset(pos_type pos)
 
 Inset * Paragraph::getInset(pos_type pos)
 {
-       return d->insetlist_.get(pos);
+       return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
+                ? d->insetlist_.get(pos) : 0;
 }
 
 
 Inset const * Paragraph::getInset(pos_type pos) const
 {
-       return d->insetlist_.get(pos);
+       return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
+                ? d->insetlist_.get(pos) : 0;
 }
 
 
 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
-               pos_type right, TextCase action)
+               pos_type right, TextCase action)
 {
        // process sequences of modified characters; in change
        // tracking mode, this approach results in much better
@@ -2593,33 +2934,32 @@ void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
                        }
                }
 
-               if (!isLetter(pos) || isDeleted(pos)) {
+               if (isWordSeparator(pos) || isDeleted(pos)) {
                        // permit capitalization again
                        capitalize = true;
                }
 
-               if (oldChar != newChar)
+               if (oldChar != newChar) {
                        changes += newChar;
+                       if (pos != right - 1)
+                               continue;
+                       // step behind the changing area
+                       pos++;
+               }
 
-               if (oldChar == newChar || pos == right - 1) {
-                       if (oldChar != newChar) {
-                               // step behind the changing area
-                               pos++;
+               int erasePos = pos - changes.size();
+               for (size_t i = 0; i < changes.size(); i++) {
+                       insertChar(pos, changes[i],
+                                  getFontSettings(bparams,
+                                                  erasePos),
+                                  trackChanges);
+                       if (!eraseChar(erasePos, trackChanges)) {
+                               ++erasePos;
+                               ++pos; // advance
+                               ++right; // expand selection
                        }
-                       int erasePos = pos - changes.size();
-                       for (size_t i = 0; i < changes.size(); i++) {
-                               insertChar(pos, changes[i],
-                                       getFontSettings(bparams,
-                                       erasePos),
-                                       trackChanges);
-                               if (!eraseChar(erasePos, trackChanges)) {
-                                       ++erasePos;
-                                       ++pos; // advance
-                                       ++right; // expand selection
-                               }
-                       }
-                       changes.clear();
                }
+               changes.clear();
        }
 }
 
@@ -2646,10 +2986,10 @@ bool Paragraph::find(docstring const & str, bool cs, bool mw,
 
        // if necessary, check whether string matches word
        if (mw) {
-               if (pos > 0 && isLetter(pos - 1))
+               if (pos > 0 && !isWordSeparator(pos - 1))
                        return false;
                if (pos + strsize < parsize
-                       && isLetter(pos + strsize))
+                       && !isWordSeparator(pos + strsize))
                        return false;
        }
 
@@ -2688,4 +3028,150 @@ bool Paragraph::isSeparator(pos_type pos) const
 }
 
 
+void Paragraph::deregisterWords()
+{
+       map<Language, Private::Words>::const_iterator itl;
+       Private::Words::const_iterator it;
+       for (itl = d->words_.begin(); itl != d->words_.end(); ++itl) {
+               WordList * wl = theWordList(itl->first);
+               for (it = (itl->second).begin(); it != (itl->second).end(); ++it)
+                       wl->remove(*it);
+       }
+       d->words_.clear();
+}
+
+
+void Paragraph::locateWord(pos_type & from, pos_type & to,
+       word_location const loc) const
+{
+       switch (loc) {
+       case WHOLE_WORD_STRICT:
+               if (from == 0 || from == size()
+                   || isWordSeparator(from)
+                   || isWordSeparator(from - 1)) {
+                       to = from;
+                       return;
+               }
+               // no break here, we go to the next
+
+       case WHOLE_WORD:
+               // If we are already at the beginning of a word, do nothing
+               if (!from || isWordSeparator(from - 1))
+                       break;
+               // no break here, we go to the next
+
+       case PREVIOUS_WORD:
+               // always move the cursor to the beginning of previous word
+               while (from && !isWordSeparator(from - 1))
+                       --from;
+               break;
+       case NEXT_WORD:
+               LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
+               break;
+       case PARTIAL_WORD:
+               // no need to move the 'from' cursor
+               break;
+       }
+       to = from;
+       while (to < size() && !isWordSeparator(to))
+               ++to;
+}
+
+
+void Paragraph::collectWords()
+{
+       pos_type n = size();
+       for (pos_type pos = 0; pos < n; ++pos) {
+               if (isWordSeparator(pos))
+                       continue;
+               pos_type from = pos;
+               locateWord(from, pos, WHOLE_WORD);
+               if (pos - from >= 6) {
+                       docstring word = asString(from, pos, AS_STR_NONE);
+                       FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
+                       if (cit == d->fontlist_.end())
+                               return;
+                       Language const * lang = cit->font().language();
+                       d->words_[*lang].insert(word);
+               }
+       }
+}
+
+
+void Paragraph::registerWords()
+{
+       map<Language, Private::Words>::const_iterator itl;
+       Private::Words::const_iterator it;
+       for (itl = d->words_.begin(); itl != d->words_.end(); ++itl) {
+               WordList * wl = theWordList(itl->first);
+               for (it = (itl->second).begin(); it != (itl->second).end(); ++it)
+                       wl->insert(*it);
+       }
+}
+
+
+void Paragraph::updateWords()
+{
+       deregisterWords();
+       collectWords();
+       registerWords();
+}
+
+
+bool Paragraph::spellCheck(pos_type & from, pos_type & to, WordLangTuple & wl,
+       docstring_list & suggestions, bool do_suggestion) const
+{
+       SpellChecker * speller = theSpellChecker();
+       if (!speller)
+               return false;
+
+       locateWord(from, to, WHOLE_WORD);
+       if (from == to || from >= pos_type(d->text_.size()))
+               return false;
+
+       docstring word = asString(from, to, AS_STR_INSETS);
+       string const lang_code = lyxrc.spellchecker_alt_lang.empty()
+               ? getFontSettings(d->inset_owner_->buffer().params(), from).language()->code()
+               : lyxrc.spellchecker_alt_lang;
+       wl = WordLangTuple(word, lang_code);
+       SpellChecker::Result res = speller->check(wl);
+       // Just ignore any error that the spellchecker reports.
+       // FIXME: we should through out an exception and catch it in the GUI to
+       // display the error.
+       if (!speller->error().empty())
+               return false;
+
+       bool const misspelled = res != SpellChecker::OK
+               && res != SpellChecker::IGNORED_WORD;
+
+       if (lyxrc.spellcheck_continuously)
+               d->fontlist_.setMisspelled(from, to, misspelled);
+
+       if (misspelled && do_suggestion)
+               speller->suggest(wl, suggestions);
+       else
+               suggestions.clear();
+
+       return misspelled;
+}
+
+
+bool Paragraph::isMisspelled(pos_type pos) const
+{
+       pos_type from = pos;
+       pos_type to = pos;
+       WordLangTuple wl;
+       docstring_list suggestions;
+       return spellCheck(from, to, wl, suggestions, false);
+}
+
+
+string Paragraph::magicLabel() const
+{
+       stringstream ss;
+       ss << "magicparlabel-" << id();
+       return ss.str();
+}
+
+
 } // namespace lyx