]> git.lyx.org Git - lyx.git/blobdiff - src/Paragraph.cpp
Fix bug #7059: Display the vertical alignment of tables.
[lyx.git] / src / Paragraph.cpp
index 3965a61d3c96db9b0628dc6fffb6d5c97ae43993..d44d0473e11f7ad8384429d5e82ace4b794040db 100644 (file)
@@ -5,6 +5,7 @@
  *
  * \author Asger Alstrup
  * \author Lars Gullik Bjønnes
+ * \author Richard Heck (XHTML output)
  * \author Jean-Marc Lasgouttes
  * \author Angus Leeming
  * \author John Levon
@@ -36,7 +37,6 @@
 #include "OutputParams.h"
 #include "output_latex.h"
 #include "output_xhtml.h"
-#include "paragraph_funcs.h"
 #include "ParagraphParameters.h"
 #include "SpellChecker.h"
 #include "sgml.h"
@@ -51,6 +51,7 @@
 
 #include "insets/InsetBibitem.h"
 #include "insets/InsetLabel.h"
+#include "insets/InsetSpecialChar.h"
 
 #include "support/debug.h"
 #include "support/docstring_list.h"
@@ -58,7 +59,6 @@
 #include "support/gettext.h"
 #include "support/lassert.h"
 #include "support/lstrings.h"
-#include "support/Messages.h"
 #include "support/textutils.h"
 
 #include <sstream>
@@ -74,6 +74,172 @@ namespace {
 char_type const META_INSET = 0x200001;
 };
 
+
+/////////////////////////////////////////////////////////////////////
+//
+// SpellResultRange
+//
+/////////////////////////////////////////////////////////////////////
+
+class SpellResultRange {
+public:
+       SpellResultRange(FontSpan range, SpellChecker::Result result)
+       : range_(range), result_(result)
+       {}
+       ///
+       FontSpan const & range() const { return range_; }
+       ///
+       void range(FontSpan const & r) { range_ = r; }
+       ///
+       SpellChecker::Result result() const { return result_; }
+       ///
+       void result(SpellChecker::Result r) { result_ = r; }
+       ///
+       bool inside(pos_type pos) const { return range_.inside(pos); }
+       ///
+       bool covered(FontSpan const & r) const
+       {
+               // 1. first of new range inside current range or
+               // 2. last of new range inside current range or
+               // 3. first of current range inside new range or
+               // 4. last of current range inside new range
+               return range_.inside(r.first) || range_.inside(r.last) ||
+                       r.inside(range_.first) || r.inside(range_.last);
+       }
+       ///
+       void shift(pos_type pos, int offset)
+       {
+               if (range_.first > pos) {
+                       range_.first += offset;
+                       range_.last += offset;
+               } else if (range_.last > pos) {
+                       range_.last += offset;
+               }
+       }
+private:
+       FontSpan range_ ;
+       SpellChecker::Result result_ ;
+};
+
+
+/////////////////////////////////////////////////////////////////////
+//
+// SpellCheckerState
+//
+/////////////////////////////////////////////////////////////////////
+
+class SpellCheckerState {
+public:
+       SpellCheckerState() {
+               needs_refresh_ = true;
+               current_change_number_ = 0;
+       }
+
+       void setRange(FontSpan const fp, SpellChecker::Result state)
+       {
+               eraseCoveredRanges(fp);
+               if (state != SpellChecker::WORD_OK)
+                       ranges_.push_back(SpellResultRange(fp, state));
+       }
+
+       void increasePosAfterPos(pos_type pos)
+       {
+               correctRangesAfterPos(pos, 1);
+               needsRefresh(pos);
+       }
+
+       void decreasePosAfterPos(pos_type pos)
+       {
+               correctRangesAfterPos(pos, -1);
+               needsRefresh(pos);
+       }
+
+       void refreshLast(pos_type pos)
+       {
+               if (pos < refresh_.last)
+                       refresh_.last = pos;
+       }
+
+       SpellChecker::Result getState(pos_type pos) const
+       {
+               SpellChecker::Result result = SpellChecker::WORD_OK;
+               RangesIterator et = ranges_.end();
+               RangesIterator it = ranges_.begin();
+               for (; it != et; ++it) {
+                       if(it->inside(pos)) {
+                               return it->result();
+                       }
+               }
+               return result;
+       }
+
+       bool needsRefresh() const {
+               return needs_refresh_;
+       }
+
+       SpellChecker::ChangeNumber currentChangeNumber() const {
+               return current_change_number_;
+       }
+
+       void refreshRange(pos_type & first, pos_type & last) const {
+               first = refresh_.first;
+               last = refresh_.last;
+       }
+
+       void needsRefresh(pos_type pos) {
+               if (needs_refresh_ && pos != -1) {
+                       if (pos < refresh_.first)
+                               refresh_.first = pos;
+                       if (pos > refresh_.last)
+                               refresh_.last = pos;
+               } else if (pos != -1) {
+                       refresh_.first = pos;
+                       refresh_.last = pos;
+               }
+               needs_refresh_ = pos != -1;
+       }
+
+       void needsCompleteRefresh(SpellChecker::ChangeNumber change_number) {
+               needs_refresh_ = true;
+               refresh_.first = 0;
+               refresh_.last = -1;
+               current_change_number_ = change_number;
+       }
+
+private:
+       typedef vector<SpellResultRange> Ranges;
+       typedef Ranges::const_iterator RangesIterator;
+       Ranges ranges_;
+       /// the area of the paragraph with pending spell check
+       FontSpan refresh_;
+       bool needs_refresh_;
+       /// spell state cache version number
+       SpellChecker::ChangeNumber current_change_number_;
+
+
+       void eraseCoveredRanges(FontSpan const fp)
+       {
+               Ranges result;
+               RangesIterator et = ranges_.end();
+               RangesIterator it = ranges_.begin();
+               for (; it != et; ++it) {
+                       if (!it->covered(fp))
+                               result.push_back(SpellResultRange(it->range(), it->result()));
+               }
+               ranges_ = result;
+       }
+
+       void correctRangesAfterPos(pos_type pos, int offset)
+       {
+               RangesIterator et = ranges_.end();
+               Ranges::iterator it = ranges_.begin();
+               for (; it != et; ++it) {
+                       it->shift(pos, offset);
+               }
+       }
+
+};
+
 /////////////////////////////////////////////////////////////////////
 //
 // Paragraph::Private
@@ -96,7 +262,7 @@ 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).
@@ -165,8 +331,7 @@ public:
                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,
@@ -174,7 +339,90 @@ public:
 
        /// match a string against a particular point in the paragraph
        bool isTextAt(string const & str, pos_type pos) const;
+
+       /// a vector of speller skip positions
+       typedef vector<FontSpan> SkipPositions;
+       typedef SkipPositions::const_iterator SkipPositionsIterator;
+
+       void appendSkipPosition(SkipPositions & skips, pos_type const pos) const;
        
+       Language * getSpellLanguage(pos_type const from) const;
+
+       Language * locateSpellRange(pos_type & from, pos_type & to,
+                                                               SkipPositions & skips) const;
+
+       bool hasSpellerChange() const {
+               SpellChecker::ChangeNumber speller_change_number = 0;
+               if (theSpellChecker())
+                       speller_change_number = theSpellChecker()->changeNumber();
+               return speller_change_number > speller_state_.currentChangeNumber();
+       }
+
+       void setMisspelled(pos_type from, pos_type to, SpellChecker::Result state)
+       {
+               pos_type textsize = owner_->size();
+               // check for sane arguments
+               if (to < from || from >= textsize)
+                       return;
+               FontSpan fp = FontSpan(from, to);
+               // don't mark end of paragraph
+               if (fp.last >= textsize)
+                       fp.last = textsize - 1;
+               speller_state_.setRange(fp, state);
+       }
+
+       void requestSpellCheck(pos_type pos) {
+               speller_state_.needsRefresh(pos);
+       }
+
+       void readySpellCheck() {
+               speller_state_.needsRefresh(-1);
+       }
+
+       bool needsSpellCheck() const
+       {
+               return speller_state_.needsRefresh();
+       }
+
+       void rangeOfSpellCheck(pos_type & first, pos_type & last) const
+       {
+               speller_state_.refreshRange(first, last);
+               if (last == -1) {
+                       last = owner_->size();
+                       return;
+               }
+               pos_type endpos = last;
+               owner_->locateWord(first, endpos, WHOLE_WORD);
+               if (endpos < last) {
+                       endpos = last;
+                       owner_->locateWord(last, endpos, WHOLE_WORD);
+               }
+               last = endpos;
+       }
+
+       int countSkips(SkipPositionsIterator & it, SkipPositionsIterator const et,
+                           int & start) const
+       {
+               int numskips = 0;
+               while (it != et && it->first < start) {
+                       int skip = it->last - it->first + 1;
+                       start += skip;
+                       numskips += skip;
+                       ++it;
+               }
+               return numskips;
+       }
+
+       void markMisspelledWords(pos_type const & first, pos_type const & last,
+                                                        SpellChecker::Result result,
+                                                        docstring const & word,
+                                                        SkipPositions const & skips);
+
+       InsetCode ownerCode() const
+       {
+               return inset_owner_ ? inset_owner_->lyxCode() : NO_CODE;
+       }
+
        /// Which Paragraph owns us?
        Paragraph * owner_;
 
@@ -185,9 +433,8 @@ public:
        FontList fontlist_;
 
        ///
-       unsigned int id_;
-       ///
-       static unsigned int paragraph_id;
+       int id_;
+
        ///
        ParagraphParameters params_;
 
@@ -203,18 +450,18 @@ public:
        typedef docstring TextContainer;
        ///
        TextContainer text_;
-       
-       typedef std::set<docstring> Words;
+
+       typedef set<docstring> Words;
+       typedef map<Language, Words> LangWordsMap;
        ///
-       Words words_;
+       LangWordsMap words_;
        ///
        Layout const * layout_;
+       ///
+       SpellCheckerState speller_state_;
 };
 
 
-// Initialization of the counter for the paragraph id's,
-unsigned int Paragraph::Private::paragraph_id = 0;
-
 namespace {
 
 struct special_phrase {
@@ -236,20 +483,27 @@ size_t const phrases_nr = sizeof(special_phrases)/sizeof(special_phrase);
 
 
 Paragraph::Private::Private(Paragraph * owner, Layout const & layout)
-       : owner_(owner), inset_owner_(0), begin_of_body_(0), layout_(&layout)
+       : owner_(owner), inset_owner_(0), id_(-1), begin_of_body_(0), layout_(&layout)
 {
-       id_ = paragraph_id++;
        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_), 
+       : owner_(owner), inset_owner_(p.inset_owner_), fontlist_(p.fontlist_),
          params_(p.params_), changes_(p.changes_), insetlist_(p.insetlist_),
          begin_of_body_(p.begin_of_body_), text_(p.text_), words_(p.words_),
          layout_(p.layout_)
 {
-       id_ = paragraph_id++;
+       id_ = ++paragraph_id;
+       requestSpellCheck(p.text_.size());
 }
 
 
@@ -261,7 +515,7 @@ Paragraph::Private::Private(Private const & p, Paragraph * owner,
          begin_of_body_(p.begin_of_body_), words_(p.words_),
          layout_(p.layout_)
 {
-       id_ = paragraph_id++;
+       id_ = ++paragraph_id;
        if (beg >= pos_type(p.text_.size()))
                return;
        text_ = p.text_.substr(beg, end - beg);
@@ -279,6 +533,7 @@ Paragraph::Private::Private(Private const & p, Paragraph * owner,
                // Add a new entry in the fontlist_.
                fontlist_.set(fcit->pos() - beg, fcit->font());
        }
+       requestSpellCheck(p.text_.size());
 }
 
 
@@ -289,12 +544,12 @@ void Paragraph::addChangesToToc(DocIterator const & cdit,
 }
 
 
-bool Paragraph::isFullyDeleted(pos_type start, pos_type end) const
+bool Paragraph::isDeleted(pos_type start, pos_type end) const
 {
        LASSERT(start >= 0 && start <= size(), /**/);
        LASSERT(end > start && end <= size() + 1, /**/);
 
-       return d->changes_.isFullyDeleted(start, end);
+       return d->changes_.isDeleted(start, end);
 }
 
 
@@ -314,7 +569,7 @@ bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const
                return true;
 
        Change const change = d->changes_.lookup(size());
-       return change.type == Change::INSERTED && change.author == 0;
+       return change.inserted() && change.currentAuthor();
 }
 
 
@@ -335,7 +590,7 @@ 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 (Inset * inset = getInset(pos))
                                inset->setChange(change);
@@ -350,7 +605,7 @@ void Paragraph::setChange(pos_type pos, Change const & change)
        d->changes_.set(change, pos);
 
        // see comment in setChange(Change const &) above
-       if (change.type != Change::DELETED && pos < size())
+       if (!change.deleted() && pos < size())
                        if (Inset * inset = getInset(pos))
                                inset->setChange(change);
 }
@@ -363,8 +618,7 @@ Change const & Paragraph::lookupChange(pos_type pos) const
 }
 
 
-void Paragraph::acceptChanges(BufferParams const & bparams, pos_type start,
-               pos_type end)
+void Paragraph::acceptChanges(pos_type start, pos_type end)
 {
        LASSERT(start >= 0 && start <= size(), /**/);
        LASSERT(end > start && end <= size() + 1, /**/);
@@ -374,14 +628,14 @@ void Paragraph::acceptChanges(BufferParams const & bparams, pos_type start,
                        case Change::UNCHANGED:
                                // accept changes in nested inset
                                if (Inset * inset = getInset(pos))
-                                       inset->acceptChanges(bparams);
+                                       inset->acceptChanges();
                                break;
 
                        case Change::INSERTED:
                                d->changes_.set(Change(Change::UNCHANGED), pos);
                                // also accept changes in nested inset
                                if (Inset * inset = getInset(pos))
-                                       inset->acceptChanges(bparams);
+                                       inset->acceptChanges();
                                break;
 
                        case Change::DELETED:
@@ -399,8 +653,7 @@ 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)
 {
        LASSERT(start >= 0 && start <= size(), /**/);
        LASSERT(end > start && end <= size() + 1, /**/);
@@ -410,7 +663,7 @@ void Paragraph::rejectChanges(BufferParams const & bparams,
                        case Change::UNCHANGED:
                                // reject changes in nested inset
                                if (Inset * inset = getInset(pos))
-                                               inset->rejectChanges(bparams);
+                                               inset->rejectChanges();
                                break;
 
                        case Change::INSERTED:
@@ -448,6 +701,8 @@ void Paragraph::Private::insertChar(pos_type pos, char_type c,
        if (pos == pos_type(text_.size())) {
                // when appending characters, no need to update tables
                text_.push_back(c);
+               // but we want spell checking
+               requestSpellCheck(pos);
                return;
        }
 
@@ -458,6 +713,9 @@ void Paragraph::Private::insertChar(pos_type pos, char_type c,
 
        // Update the insets
        insetlist_.increasePosAfterPos(pos);
+
+       // Update list of misspelled positions
+       speller_state_.increasePosAfterPos(pos);
 }
 
 
@@ -477,6 +735,9 @@ bool Paragraph::insertInset(pos_type pos, Inset * inset,
 
        // Add a new entry in the insetlist_.
        d->insetlist_.insert(inset, pos);
+
+       // Some insets require run of spell checker
+       requestSpellCheck(pos);
        return true;
 }
 
@@ -494,13 +755,15 @@ 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));
+                       // request run of spell checker
+                       requestSpellCheck(pos);
                        return false;
                }
 
-               if (change.type == Change::DELETED)
+               if (change.deleted())
                        return false;
        }
 
@@ -526,6 +789,10 @@ bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
        // Update the insetlist_
        d->insetlist_.decreasePosAfterPos(pos);
 
+       // Update list of misspelled positions
+       d->speller_state_.decreasePosAfterPos(pos);
+       d->speller_state_.refreshLast(size());
+
        return true;
 }
 
@@ -545,7 +812,7 @@ 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
@@ -554,12 +821,18 @@ 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);
        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;
@@ -573,21 +846,20 @@ bool Paragraph::Private::simpleTeXBlanks(OutputParams const & runparams,
                                       Font const & font,
                                       Layout const & style)
 {
-       if (style.pass_thru || runparams.verbatim)
+       if (style.pass_thru || runparams.pass_thru)
                return false;
 
        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()))
@@ -706,19 +978,18 @@ bool Paragraph::Private::isTextAt(string const & str, pos_type pos) const
 }
 
 
-void Paragraph::Private::latexInset(
-                                            BufferParams const & bparams,
-                                            odocstream & os,
-                                            TexRow & texrow,
-                                            OutputParams & runparams,
-                                            Font & running_font,
-                                            Font & basefont,
-                                            Font const & outerfont,
-                                            bool & open_font,
-                                            Change & running_change,
-                                            Layout const & style,
-                                            pos_type & i,
-                                            unsigned int & column)
+void Paragraph::Private::latexInset(BufferParams const & bparams,
+                                   odocstream & os,
+                                   TexRow & texrow,
+                                   OutputParams & runparams,
+                                   Font & running_font,
+                                   Font & basefont,
+                                   Font const & outerfont,
+                                   bool & open_font,
+                                   Change & running_change,
+                                   Layout const & style,
+                                   pos_type & i,
+                                   unsigned int & column)
 {
        Inset * inset = owner_->getInset(i);
        LASSERT(inset, /**/);
@@ -757,25 +1028,26 @@ void Paragraph::Private::latexInset(
                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) {
-               if (running_font.language()->lang() == "farsi")
+           // 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
                        os << "\\L{";
@@ -800,10 +1072,10 @@ void Paragraph::Private::latexInset(
                unsigned int count = running_font.latexWriteEndChanges(os,
                        bparams, runparams, basefont, basefont, closeLanguage);
                column += count;
-               // if any font properties were closed, update the running_font, 
+               // if any font properties were closed, update the running_font,
                // making sure, however, to leave the language as it was
                if (count > 0) {
-                       // FIXME: probably a better way to keep track of the old 
+                       // FIXME: probably a better way to keep track of the old
                        // language, than copying the entire font?
                        Font const copy_font(running_font);
                        basefont = owner_->getLayoutFont(bparams, outerfont);
@@ -836,16 +1108,14 @@ void Paragraph::Private::latexInset(
        }
 
        if (tmp) {
-               for (int j = 0; j < tmp; ++j)
-                       texrow.newline();
-
+               texrow.newlines(tmp);
                texrow.start(owner_->id(), i + 1);
                column = 0;
        } else {
-               column += os.tellp() - len;
+               column += (unsigned int)(os.tellp() - len);
        }
 
-       if (owner_->lookupChange(i).type == Change::DELETED)
+       if (owner_->isDeleted(i))
                --runparams.inDeletedInset;
 }
 
@@ -861,7 +1131,7 @@ void Paragraph::Private::latexSpecialChar(
 {
        char_type const c = text_[i];
 
-       if (style.pass_thru) {
+       if (style.pass_thru || runparams.pass_thru) {
                if (c != '\0')
                        // FIXME UNICODE: This can fail if c cannot
                        // be encoded in the current encoding.
@@ -869,13 +1139,6 @@ void Paragraph::Private::latexSpecialChar(
                return;
        }
 
-       if (runparams.verbatim) {
-               // FIXME UNICODE: This can fail if c cannot
-               // be encoded in the current encoding.
-               os.put(c);
-               return;
-       }
-
        // If T1 font encoding is used, use the special
        // characters it provides.
        // NOTE: some languages reset the font encoding
@@ -963,7 +1226,7 @@ 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;
                        }
@@ -1062,15 +1325,37 @@ 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;
+               // we have to provide all the optional arguments here, even though
+               // the last one is the only one we care about.
+               owner_->latex(bp, f, ods, tr, features.runparams(), 0, -1, true);
+               docstring const d = ods.str();
+               if (!d.empty()) {
+                       // this will have "{" at the beginning, but not at the end
+                       string const content = to_utf8(d);
+                       string const cmd = layout_->latexname();
+                       features.addPreambleSnippet("\\" + cmd + content + "}");
+               }
+       }
+
+       if (features.runparams().flavor == OutputParams::HTML
+           && layout_->htmltitle()) {
+               features.setHTMLTitle(owner_->asString(AS_STR_INSETS));
+       }
+
        // 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);
@@ -1085,7 +1370,7 @@ 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");
                }
@@ -1114,7 +1399,7 @@ namespace {
        Layout const emptyParagraphLayout;
 }
 
-Paragraph::Paragraph() 
+Paragraph::Paragraph()
        : d(new Paragraph::Private(this, emptyParagraphLayout))
 {
        itemdepth = 0;
@@ -1160,6 +1445,18 @@ Paragraph::~Paragraph()
 }
 
 
+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
 {
@@ -1187,20 +1484,26 @@ void Paragraph::write(ostream & os, BufferParams const & bparams,
 
        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 const change = lookupChange(i);
-               Changes::lyxMarkChange(os, column, running_change, change);
+               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 (ignore spelling markers)
+               // Write font changes
                Font font2 = getFontSettings(bparams, i);
-               font2.setMisspelled(false);
                if (font2 != font1) {
+                       flushString(os, write_buffer);
                        font2.lyxWriteChanges(font1, os);
                        column = 0;
                        font1 = font2;
@@ -1210,6 +1513,7 @@ void Paragraph::write(ostream & os, BufferParams const & bparams,
                switch (c) {
                case META_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
@@ -1226,10 +1530,12 @@ void Paragraph::write(ostream & os, BufferParams const & bparams,
                        }
                        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;
@@ -1239,13 +1545,14 @@ void Paragraph::write(ostream & os, BufferParams const & bparams,
                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')
-                               os << to_utf8(docstring(1, c));
+                               write_buffer.push_back(c);
                        else
                                LYXERR0("NUL char in structure.");
                        ++column;
@@ -1253,13 +1560,14 @@ void Paragraph::write(ostream & os, BufferParams const & bparams,
                }
        }
 
+       flushString(os, write_buffer);
        os << "\n\\end_layout\n";
 }
 
 
 void Paragraph::validate(LaTeXFeatures & features) const
 {
-       d->validate(features, *d->layout_);
+       d->validate(features);
 }
 
 
@@ -1521,7 +1829,7 @@ void Paragraph::setFont(pos_type pos, Font const & font)
        // First, reduce font against layout/label font
        // Update: The setCharFont() routine in text2.cpp already
        // reduces font, so we don't need to do that here. (Asger)
-       
+
        d->fontlist_.set(pos, font);
 }
 
@@ -1606,35 +1914,31 @@ void Paragraph::setLabelWidthString(docstring const & s)
 }
 
 
-docstring const Paragraph::translateIfPossible(docstring const & s,
+docstring Paragraph::expandLabel(Layout const & layout,
                BufferParams const & bparams) const
 {
-       if (!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));
+       return expandParagraphLabel(layout, bparams, true);
 }
 
 
-docstring Paragraph::expandLabel(Layout const & layout,
+docstring Paragraph::expandDocBookLabel(Layout const & layout,
+               BufferParams const & bparams) const
+{
+       return expandParagraphLabel(layout, bparams, false);
+}
+
+
+docstring Paragraph::expandParagraphLabel(Layout const & layout,
                BufferParams const & bparams, bool process_appendix) const
 {
        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 && d->params_.appendix())
-               fmt = translateIfPossible(layout.labelstring_appendix(),
-                       bparams);
-       else
-               fmt = translateIfPossible(layout.labelstring(), bparams);
-
-       if (fmt.empty() && layout.labeltype == LABEL_COUNTER 
+       if (fmt.empty() && layout.labeltype == LABEL_COUNTER
            && !layout.counter.empty())
-               return tclass.counters().theCounter(layout.counter);
+               return tclass.counters().theCounter(layout.counter, lang);
 
        // handle 'inherited level parts' in 'fmt',
        // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
@@ -1645,14 +1949,14 @@ docstring Paragraph::expandLabel(Layout const & layout,
                        docstring parent(fmt, i + 1, j - i - 1);
                        docstring label = from_ascii("??");
                        if (tclass.hasLayout(parent))
-                               docstring label = expandLabel(tclass[parent], bparams,
+                               docstring label = expandParagraphLabel(tclass[parent], bparams,
                                                      process_appendix);
-                       fmt = docstring(fmt, 0, i) + label 
+                       fmt = docstring(fmt, 0, i) + label
                                + docstring(fmt, j + 1, docstring::npos);
                }
        }
 
-       return tclass.counters().counterLabel(fmt);
+       return tclass.counters().counterLabel(fmt, lang);
 }
 
 
@@ -1660,9 +1964,9 @@ void Paragraph::applyLayout(Layout const & new_layout)
 {
        d->layout_ = &new_layout;
        LyXAlignment const oldAlign = d->params_.align();
-       
+
        if (!(oldAlign & d->layout_->alignpossible)) {
-               frontend::Alert::warning(_("Alignment not permitted"), 
+               frontend::Alert::warning(_("Alignment not permitted"),
                        _("The new layout does not permit the alignment previously used.\nSetting to default."));
                d->params_.align(LYX_ALIGN_LAYOUT);
        }
@@ -1710,12 +2014,6 @@ void Paragraph::setBeginOfBody()
 }
 
 
-bool Paragraph::forcePlainLayout() const
-{
-       return inInset().forcePlainLayout();
-}
-
-
 bool Paragraph::allowParagraphCustomization() const
 {
        return inInset().allowParagraphCustomization();
@@ -1799,7 +2097,7 @@ int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
                os << "\\noindent ";
                column += 10;
        }
-       
+
        LyXAlignment const curAlign = params_.align();
 
        if (curAlign == layout_->align)
@@ -1810,6 +2108,7 @@ int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
        case LYX_ALIGN_BLOCK:
        case LYX_ALIGN_LAYOUT:
        case LYX_ALIGN_SPECIAL:
+       case LYX_ALIGN_DECIMAL:
                break;
        case LYX_ALIGN_LEFT:
        case LYX_ALIGN_RIGHT:
@@ -1822,7 +2121,7 @@ int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
        }
 
        string const begin_tag = "\\begin";
-       InsetCode code = owner_->ownerCode();
+       InsetCode code = ownerCode();
        bool const lastpar = runparams.isLastPar;
 
        switch (curAlign) {
@@ -1830,6 +2129,7 @@ int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
        case LYX_ALIGN_BLOCK:
        case LYX_ALIGN_LAYOUT:
        case LYX_ALIGN_SPECIAL:
+       case LYX_ALIGN_DECIMAL:
                break;
        case LYX_ALIGN_LEFT: {
                string output;
@@ -1878,6 +2178,7 @@ int Paragraph::Private::endTeXParParams(BufferParams const & bparams,
        case LYX_ALIGN_BLOCK:
        case LYX_ALIGN_LAYOUT:
        case LYX_ALIGN_SPECIAL:
+       case LYX_ALIGN_DECIMAL:
                break;
        case LYX_ALIGN_LEFT:
        case LYX_ALIGN_RIGHT:
@@ -1890,7 +2191,7 @@ int Paragraph::Private::endTeXParParams(BufferParams const & bparams,
        }
 
        string const end_tag = "\n\\par\\end";
-       InsetCode code = owner_->ownerCode();
+       InsetCode code = ownerCode();
        bool const lastpar = runparams.isLastPar;
 
        switch (curAlign) {
@@ -1898,6 +2199,7 @@ int Paragraph::Private::endTeXParParams(BufferParams const & bparams,
        case LYX_ALIGN_BLOCK:
        case LYX_ALIGN_LAYOUT:
        case LYX_ALIGN_SPECIAL:
+       case LYX_ALIGN_DECIMAL:
                break;
        case LYX_ALIGN_LEFT: {
                string output;
@@ -1931,23 +2233,24 @@ int Paragraph::Private::endTeXParParams(BufferParams const & bparams,
 
 
 // This one spits out the text of the paragraph
-bool Paragraph::latex(BufferParams const & bparams,
+void Paragraph::latex(BufferParams const & bparams,
        Font const & outerfont,
        odocstream & os, TexRow & texrow,
        OutputParams const & runparams,
-       int start_pos, int end_pos) const
+       int start_pos, int end_pos, bool force) const
 {
        LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
 
-       bool return_value = false;
-
-       bool const allowcust = allowParagraphCustomization();
-
        // FIXME This check should not be needed. Perhaps issue an
        // error if it triggers.
-       Layout const & style = forcePlainLayout() ?
+       Layout const & style = inInset().forcePlainLayout() ?
                bparams.documentClass().plainLayout() : *d->layout_;
 
+       if (!force && style.inpreamble)
+               return;
+
+       bool const allowcust = allowParagraphCustomization();
+
        // Current base font for all inherited font changes, without any
        // change caused by an individual character, except for the language:
        // It is set to the language of the first character.
@@ -1961,8 +2264,10 @@ bool Paragraph::latex(BufferParams const & bparams,
        unsigned int column = 0;
 
        if (body_pos > 0) {
-               os << '[';
-               column += 1;
+               // the optional argument is kept in curly brackets in
+               // case it contains a ']'
+               os << "[{";
+               column += 2;
                basefont = getLabelFont(bparams, outerfont);
        } else {
                basefont = getLayoutFont(bparams, outerfont);
@@ -1975,6 +2280,8 @@ bool Paragraph::latex(BufferParams const & bparams,
 
        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
@@ -2002,11 +2309,12 @@ bool Paragraph::latex(BufferParams const & bparams,
                                running_font = basefont;
 
                                column += Changes::latexMarkChange(os, bparams,
-                                               runningChange, Change(Change::UNCHANGED));
+                                               runningChange, Change(Change::UNCHANGED),
+                                               runparams);
                                runningChange = Change(Change::UNCHANGED);
 
-                               os << "] ";
-                               column +=2;
+                               os << "}] ";
+                               column +=3;
                        }
                        if (style.isCommand()) {
                                os << '{';
@@ -2020,7 +2328,7 @@ bool Paragraph::latex(BufferParams const & bparams,
                }
 
                Change const & change = runparams.inDeletedInset ? runparams.changeOfDeletedInset
-                                                                : lookupChange(i);
+                                                                : lookupChange(i);
 
                if (bparams.outputChanges && runningChange != change) {
                        if (open_font) {
@@ -2031,13 +2339,14 @@ bool Paragraph::latex(BufferParams const & bparams,
                        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;
                }
 
@@ -2060,18 +2369,22 @@ bool Paragraph::latex(BufferParams const & bparams,
                        open_font = false;
                }
 
+               string const running_lang = runparams.use_polyglossia ?
+                       running_font.language()->polyglossia() : running_font.language()->babel();
                // close babel's font environment before opening CJK.
-               if (!running_font.language()->babel().empty() &&
+               string const lang_end_command = runparams.use_polyglossia ?
+                       "\\end{$$lang}" : lyxrc.language_command_end;
+               if (!running_lang.empty() &&
                    font.language()->encoding()->package() == Encoding::CJK) {
-                               string end_tag = subst(lyxrc.language_command_end,
+                               string end_tag = subst(lang_end_command,
                                                        "$$lang",
-                                                       running_font.language()->babel());
+                                                       running_lang);
                                os << from_ascii(end_tag);
                                column += end_tag.length();
                }
 
                // Switch file encoding if necessary (and allowed)
-               if (!runparams.verbatim && 
+               if (!runparams.pass_thru && !style.pass_thru &&
                    runparams.encoding->package() != Encoding::none &&
                    font.language()->encoding()->package() != Encoding::none) {
                        pair<bool, int> const enc_switch = switchEncoding(os, bparams,
@@ -2096,10 +2409,15 @@ bool Paragraph::latex(BufferParams const & bparams,
                        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 == ' ')
-                               os << fontchange.substr(0, fontchange.size() - 1) 
+                       else if (suffixIs(fontchange, ' ') && c == ' ')
+                               os << fontchange.substr(0, fontchange.size() - 1)
                                   << from_ascii("{}");
                        else
                                os << fontchange;
@@ -2171,9 +2489,8 @@ bool Paragraph::latex(BufferParams const & bparams,
        if (open_font) {
 #ifdef FIXED_LANGUAGE_END_DETECTION
                if (next_) {
-                       running_font
-                               .latexWriteEndChanges(os, bparams, runparams,
-                                       basefont,
+                       running_font.latexWriteEndChanges(os, bparams,
+                                       runparams, basefont,
                                        next_->getFont(bparams, 0, outerfont));
                } else {
                        running_font.latexWriteEndChanges(os, bparams,
@@ -2188,21 +2505,22 @@ bool Paragraph::latex(BufferParams const & bparams,
 #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 << "]~";
-               return_value = false;
+               os << "}]~";
        }
 
-       if (allowcust) {
-               column += d->endTeXParParams(bparams, os, texrow,
-                                         runparams);
+       if (allowcust && d->endTeXParParams(bparams, os, texrow, runparams)
+           && runparams.encoding != prev_encoding) {
+               runparams.encoding = prev_encoding;
+               if (!runparams.isFullUnicode())
+                       os << setEncoding(prev_encoding->iconvName());
        }
 
        LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
-       return return_value;
 }
 
 
@@ -2211,6 +2529,8 @@ bool Paragraph::emptyTag() const
        for (pos_type i = 0; i < size(); ++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 &&
@@ -2265,18 +2585,18 @@ pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runpa
 }
 
 
-pos_type Paragraph::firstWordLyXHTML(odocstream & os, OutputParams const & runparams)
+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(os, runparams);
+                       inset->xhtml(xs, runparams);
                } else {
                        char_type c = d->text_[i];
                        if (c == ' ')
                                break;
-                       os << html::escapeChar(c);
+                       xs << c;
                }
        }
        return i;
@@ -2355,59 +2675,72 @@ void Paragraph::simpleDocBookOnePar(Buffer const & buf,
 
 
 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
-                                   odocstream & os,
+                                   XHTMLStream & xs,
                                    OutputParams const & runparams,
                                    Font const & outerfont,
                                    pos_type initial) const
 {
        docstring retval;
 
-       // FIXME We really need to manage the tag nesting here.
-       // Probably in the same sort of way as in output_xhtml.
        bool emph_flag = false;
        bool bold_flag = false;
-       std::string closing_tag;
+       string closing_tag;
 
        Layout const & style = *d->layout_;
+
+       if (!runparams.for_toc && runparams.html_make_pars) {
+               // generate a magic label for this paragraph
+               string const attr = "id='" + magicLabel() + "'";
+               xs << html::CompTag("a", attr);
+       }
+
        FontInfo font_old =
                style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
 
-       //if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
-       //      os << "]]>";
-
        // parsing main loop
        for (pos_type i = initial; i < size(); ++i) {
+               // let's not show deleted material in the output
+               if (isDeleted(i))
+                       continue;
+
                Font font = getFont(buf.params(), i, outerfont);
 
                // emphasis
                if (font_old.emph() != font.fontInfo().emph()) {
                        if (font.fontInfo().emph() == FONT_ON) {
-                               os << "<em>";
+                               xs << html::StartTag("em");
                                emph_flag = true;
                        } else if (emph_flag && i != initial) {
-                               os << "</em>";
+                               xs << html::EndTag("em");
                                emph_flag = false;
                        }
                }
                // bold
                if (font_old.series() != font.fontInfo().series()) {
                        if (font.fontInfo().series() == BOLD_SERIES) {
-                               os << "<strong>";
+                               xs << html::StartTag("strong");
                                bold_flag = true;
                        } else if (bold_flag && i != initial) {
-                               os << "</strong>";
+                               xs << html::EndTag("strong");
                                bold_flag = false;
                        }
                }
-               // FIXME Other such tags? 
-
-               if (Inset const * inset = getInset(i)) {
-                       retval += inset->xhtml(os, runparams);
+               // FIXME XHTML
+               // Other such tags? What about the other text ranges?
+
+               Inset const * inset = getInset(i);
+               if (inset) {
+                       if (!runparams.for_toc || inset->isInToc()) {
+                               OutputParams np = runparams;
+                               if (!inset->getLayout().htmlisblock())
+                                       np.html_in_par = true;
+                               retval += inset->xhtml(xs, np);
+                       }
                } else {
                        char_type c = d->text_[i];
 
                        if (style.pass_thru)
-                               os.put(c);
+                               xs << c;
                        else if (c == '-') {
                                docstring str;
                                int j = i + 1;
@@ -2423,19 +2756,17 @@ docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
                                }
                                else
                                        str += c;
-                               os << str;
+                               // 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::ESCAPE_NONE << str;
                        } else
-                               os << html::escapeChar(c);
+                               xs << c;
                }
                font_old = font.fontInfo();
        }
 
-       // FIXME This could be out of order. See above.
-       if (emph_flag)
-               os << "</em>";
-       if (bold_flag)
-               os << "</strong>";
-
+       xs.closeFontTags();
        return retval;
 }
 
@@ -2509,8 +2840,7 @@ bool Paragraph::isRTL(BufferParams const & bparams) const
 {
        return lyxrc.rtl_support
                && getParLanguage(bparams)->rightToLeft()
-               && ownerCode() != ERT_CODE
-               && ownerCode() != LISTINGS_CODE;
+               && !inInset().getLayout().forceLTR();
 }
 
 
@@ -2525,12 +2855,13 @@ void Paragraph::changeLanguage(BufferParams const & bparams,
                        setFont(i, font);
                }
        }
+       d->requestSpellCheck(size());
 }
 
 
 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
 {
-       Language const * doc_language = bparams.language;
+       Language const * doc_language = bparams.language;
        FontList::const_iterator cit = d->fontlist_.begin();
        FontList::const_iterator end = d->fontlist_.end();
 
@@ -2543,6 +2874,20 @@ bool Paragraph::isMultiLingual(BufferParams const & bparams) const
 }
 
 
+void Paragraph::getLanguages(std::set<Language const *> & languages) const
+{
+       FontList::const_iterator cit = d->fontlist_.begin();
+       FontList::const_iterator end = d->fontlist_.end();
+
+       for (; cit != end; ++cit) {
+               Language const * lang = cit->font().language();
+               if (lang != ignore_language &&
+                   lang != latex_language)
+                       languages.insert(lang);
+       }
+}
+
+
 docstring Paragraph::asString(int options) const
 {
        return asString(0, size(), options);
@@ -2553,18 +2898,23 @@ 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())
+       if (beg == 0
+           && options & AS_STR_LABEL
+           && !d->params_.labelString().empty())
                os << d->params_.labelString() << ' ';
 
        for (pos_type i = beg; i < end; ++i) {
+               if ((options & AS_STR_SKIPDELETE) && isDeleted(i))
+                       continue;
                char_type const c = d->text_[i];
                if (isPrintable(c) || c == '\t'
-                   || (c == '\n' && options & AS_STR_NEWLINES))
+                   || (c == '\n' && (options & AS_STR_NEWLINES)))
                        os.put(c);
-               else if (c == META_INSET && options & AS_STR_INSETS)
+               else if (c == META_INSET && (options & AS_STR_INSETS)) {
                        getInset(i)->tocString(os);
+                       if (getInset(i)->asInsetMath())
+                               os << " ";
+               }
        }
 
        return os.str();
@@ -2575,7 +2925,7 @@ docstring Paragraph::stringify(pos_type beg, pos_type end, int options, OutputPa
 {
        odocstringstream os;
 
-       if (beg == 0 
+       if (beg == 0
                && options & AS_STR_LABEL
                && !d->params_.labelString().empty())
                os << d->params_.labelString() << ' ';
@@ -2583,9 +2933,9 @@ docstring Paragraph::stringify(pos_type beg, pos_type end, int options, OutputPa
        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))
+                   || (c == '\n' && (options & AS_STR_NEWLINES)))
                        os.put(c);
-               else if (c == META_INSET && options & AS_STR_INSETS) {
+               else if (c == META_INSET && (options & AS_STR_INSETS)) {
                        getInset(i)->plaintext(os, runparams);
                }
        }
@@ -2606,6 +2956,12 @@ int Paragraph::id() const
 }
 
 
+void Paragraph::setId(int id)
+{
+       d->id_ = id;
+}
+
+
 Layout const & Paragraph::layout() const
 {
        return *d->layout_;
@@ -2619,14 +2975,14 @@ void Paragraph::setLayout(Layout const & layout)
 
 
 void Paragraph::setDefaultLayout(DocumentClass const & tc)
-{ 
-       setLayout(tc.defaultLayout()); 
+{
+       setLayout(tc.defaultLayout());
 }
 
 
 void Paragraph::setPlainLayout(DocumentClass const & tc)
-{ 
-       setLayout(tc.plainLayout()); 
+{
+       setLayout(tc.plainLayout());
 }
 
 
@@ -2647,12 +3003,6 @@ Inset const & Paragraph::inInset() const
 }
 
 
-InsetCode Paragraph::ownerCode() const
-{
-       return d->inset_owner_ ? d->inset_owner_->lyxCode() : NO_CODE;
-}
-
-
 ParagraphParameters & Paragraph::params()
 {
        return d->params_;
@@ -2751,8 +3101,8 @@ int Paragraph::checkBiblio(Buffer const & buffer)
        InsetList::iterator end = d->insetlist_.end();
        for (; it != end; ++it)
                if (it->inset->lyxCode() == BIBITEM_CODE
-                   && it->pos > 0) {
-                       InsetBibitem * olditem = static_cast<InsetBibitem *>(it->inset);
+                     && it->pos > 0) {
+                       InsetCommand * olditem = it->inset->asInsetCommand();
                        oldkey = olditem->getParam("key");
                        oldlabel = olditem->getParam("label");
                        erasedInsetPosition = it->pos;
@@ -2768,8 +3118,7 @@ int Paragraph::checkBiblio(Buffer const & buffer)
        // 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);
+               InsetCommand * inset = d->insetlist_.begin()->inset->asInsetCommand();
                if (!oldkey.empty())
                        inset->setParam("key", oldkey);
                inset->setParam("label", oldlabel);
@@ -2778,13 +3127,13 @@ int Paragraph::checkBiblio(Buffer const & buffer)
 
        // 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(buffer, InsetCommandParams(BIBITEM_CODE));
+       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);
        inset->setParam("label", oldlabel);
-       insertInset(0, static_cast<Inset *>(inset),
+       insertInset(0, inset,
                    Change(track_changes ? Change::INSERTED : Change::UNCHANGED));
 
        return 1;
@@ -2797,21 +3146,21 @@ void Paragraph::checkAuthors(AuthorList const & 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();
 }
 
 
@@ -2914,36 +3263,44 @@ void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
 }
 
 
-bool Paragraph::find(docstring const & str, bool cs, bool mw,
-               pos_type pos, bool del) const
+int Paragraph::find(docstring const & str, bool cs, bool mw,
+               pos_type start_pos, bool del) const
 {
+       pos_type pos = start_pos;
        int const strsize = str.length();
        int i = 0;
        pos_type const parsize = d->text_.size();
-       for (i = 0; pos + i < parsize; ++i) {
-               if (i >= strsize)
-                       break;
-               if (cs && str[i] != d->text_[pos + i])
+       for (i = 0; i < strsize && pos < parsize; ++i, ++pos) {
+               // Ignore ligature break and hyphenation chars while searching
+               while (pos < parsize - 1 && isInset(pos)) {
+                       const InsetSpecialChar *isc = dynamic_cast<const InsetSpecialChar*>(getInset(pos));
+                       if (isc == 0
+                           || (isc->kind() != InsetSpecialChar::HYPHENATION
+                               && isc->kind() != InsetSpecialChar::LIGATURE_BREAK))
+                               break;
+                       pos++;
+               }
+               if (cs && str[i] != d->text_[pos])
                        break;
-               if (!cs && uppercase(str[i]) != uppercase(d->text_[pos + i]))
+               if (!cs && uppercase(str[i]) != uppercase(d->text_[pos]))
                        break;
-               if (!del && isDeleted(pos + i))
+               if (!del && isDeleted(pos))
                        break;
        }
 
        if (i != strsize)
-               return false;
+               return 0;
 
        // if necessary, check whether string matches word
        if (mw) {
-               if (pos > 0 && !isWordSeparator(pos - 1))
-                       return false;
-               if (pos + strsize < parsize
-                       && !isWordSeparator(pos + strsize))
-                       return false;
+               if (start_pos > 0 && !isWordSeparator(start_pos - 1))
+                       return 0;
+               if (pos < parsize
+                       && !isWordSeparator(pos))
+                       return 0;
        }
 
-       return true;
+       return pos - start_pos;
 }
 
 
@@ -2980,10 +3337,15 @@ bool Paragraph::isSeparator(pos_type pos) const
 
 void Paragraph::deregisterWords()
 {
-       Private::Words::const_iterator it;
-       WordList & wl = theWordList();
-       for (it = d->words_.begin(); it != d->words_.end(); ++it)
-               wl.remove(*it);
+       Private::LangWordsMap::const_iterator itl = d->words_.begin();
+       Private::LangWordsMap::const_iterator ite = d->words_.end();
+       for (; itl != ite; ++itl) {
+               WordList * wl = theWordList(itl->first);
+               Private::Words::const_iterator it = (itl->second).begin();
+               Private::Words::const_iterator et = (itl->second).end();
+               for (; it != et; ++it)
+                       wl->remove(*it);
+       }
        d->words_.clear();
 }
 
@@ -3027,49 +3389,38 @@ void Paragraph::locateWord(pos_type & from, pos_type & to,
 
 void Paragraph::collectWords()
 {
-       SpellChecker * speller = theSpellChecker();
-
-       //lyxerr << "Words: ";
+       // This is the value that needs to be exposed in the preferences
+       // to resolve bug #6760.
+       static int minlength = 6;
        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 (!lyxrc.spellcheck_continuously && pos - from < 6)
-                       continue;
-
-               docstring word = asString(from, pos, false);
-               if (pos - from >= 6)
-                       d->words_.insert(word);
-
-               if (!lyxrc.spellcheck_continuously || !speller)
-                       continue;
-               
-               string const lang_code = lyxrc.spellchecker_alt_lang.empty()
-                       ? getFontSettings(d->inset_owner_->buffer().params(), from).language()->code()
-                       : lyxrc.spellchecker_alt_lang;
-               WordLangTuple wl(word, lang_code);
-               SpellChecker::Result res = speller->check(wl);
-               // ... just ignore any error that the spellchecker reports.
-               if (!speller->error().empty())
-                       continue;
-               bool const misspelled = res != SpellChecker::OK
-                       && res != SpellChecker::IGNORED_WORD;
-               d->fontlist_.setMisspelled(from, pos, misspelled);
-
-               //lyxerr << word << " ";
+               if (pos - from >= minlength) {
+                       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);
+               }
        }
-       //lyxerr << std::endl;
 }
 
 
 void Paragraph::registerWords()
 {
-       Private::Words::const_iterator it;
-       WordList & wl = theWordList();
-       for (it = d->words_.begin(); it != d->words_.end(); ++it)
-               wl.insert(*it);
+       Private::LangWordsMap::const_iterator itl = d->words_.begin();
+       Private::LangWordsMap::const_iterator ite = d->words_.end();
+       for (; itl != ite; ++itl) {
+               WordList * wl = theWordList(itl->first);
+               Private::Words::const_iterator it = (itl->second).begin();
+               Private::Words::const_iterator et = (itl->second).end();
+               for (; it != et; ++it)
+                       wl->insert(*it);
+       }
 }
 
 
@@ -3081,50 +3432,259 @@ void Paragraph::updateWords()
 }
 
 
-bool Paragraph::spellCheck(pos_type & from, pos_type & to, WordLangTuple & wl,
-       docstring_list & suggestions) const
+void Paragraph::Private::appendSkipPosition(SkipPositions & skips, pos_type const pos) const
+{
+       SkipPositionsIterator begin = skips.begin();
+       SkipPositions::iterator end = skips.end();
+       if (pos > 0 && begin < end) {
+               --end;
+               if (end->last == pos - 1) {
+                       end->last = pos;
+                       return;
+               }
+       }
+       skips.insert(end, FontSpan(pos, pos));
+}
+
+
+Language * Paragraph::Private::locateSpellRange(
+       pos_type & from, pos_type & to,
+       SkipPositions & skips) const
+{
+       // skip leading white space
+       while (from < to && owner_->isWordSeparator(from))
+               ++from;
+       // don't check empty range
+       if (from >= to)
+               return 0;
+       // get current language
+       Language * lang = getSpellLanguage(from);
+       pos_type last = from;
+       bool samelang = true;
+       bool sameinset = true;
+       while (last < to && samelang && sameinset) {
+               // hop to end of word
+               while (last < to && !owner_->isWordSeparator(last)) {
+                       if (owner_->getInset(last)) {
+                               appendSkipPosition(skips, last);
+                       } else if (owner_->isDeleted(last)) {
+                               appendSkipPosition(skips, last);
+                       }
+                       ++last;
+               }
+               // hop to next word while checking for insets
+               while (sameinset && last < to && owner_->isWordSeparator(last)) {
+                       if (Inset const * inset = owner_->getInset(last))
+                               sameinset = inset->isChar() && inset->isLetter();
+                       if (sameinset && owner_->isDeleted(last)) {
+                               appendSkipPosition(skips, last);
+                       }
+                       if (sameinset)
+                               last++;
+               }
+               if (sameinset && last < to) {
+                       // now check for language change
+                       samelang = lang == getSpellLanguage(last);
+               }
+       }
+       // if language change detected backstep is needed
+       if (!samelang)
+               --last;
+       to = last;
+       return lang;
+}
+
+
+Language * Paragraph::Private::getSpellLanguage(pos_type const from) const
+{
+       Language * lang =
+               const_cast<Language *>(owner_->getFontSettings(
+                       inset_owner_->buffer().params(), from).language());
+       if (lang == inset_owner_->buffer().params().language
+               && !lyxrc.spellchecker_alt_lang.empty()) {
+               string lang_code;
+               string const lang_variety =
+                       split(lyxrc.spellchecker_alt_lang, lang_code, '-');
+               lang->setCode(lang_code);
+               lang->setVariety(lang_variety);
+       }
+       return lang;
+}
+
+
+void Paragraph::requestSpellCheck(pos_type pos)
+{
+       d->requestSpellCheck(pos == -1 ? size() : pos);
+}
+
+
+bool Paragraph::needsSpellCheck() const
+{
+       SpellChecker::ChangeNumber speller_change_number = 0;
+       if (theSpellChecker())
+               speller_change_number = theSpellChecker()->changeNumber();
+       if (speller_change_number > d->speller_state_.currentChangeNumber()) {
+               d->speller_state_.needsCompleteRefresh(speller_change_number);
+       }
+       return d->needsSpellCheck();
+}
+
+
+SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
+       WordLangTuple & wl, docstring_list & suggestions,
+       bool do_suggestion, bool check_learned) const
 {
+       SpellChecker::Result result = SpellChecker::WORD_OK;
        SpellChecker * speller = theSpellChecker();
        if (!speller)
-               return false;
+               return result;
+
+       if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
+               return result;
 
        locateWord(from, to, WHOLE_WORD);
-       if (from == to || from >= pos_type(d->text_.size()))
-               return false;
+       if (from == to || from >= size())
+               return result;
+
+       docstring word = asString(from, to, AS_STR_INSETS + AS_STR_SKIPDELETE);
+       Language * lang = d->getSpellLanguage(from);
+
+       wl = WordLangTuple(word, lang);
+
+       if (!word.size())
+               return result;
+
+       if (needsSpellCheck() || check_learned) {
+               // Ignore words with digits
+               // FIXME: make this customizable
+               // (note that some checkers ignore words with digits by default)
+               if (!hasDigit(word)) {
+                       bool const trailing_dot = to < size() && d->text_[to] == '.';
+                       result = speller->check(wl);
+                       if (SpellChecker::misspelled(result) && trailing_dot) {
+                               wl = WordLangTuple(word.append(from_ascii(".")), lang);
+                               result = speller->check(wl);
+                               if (!SpellChecker::misspelled(result)) {
+                                       LYXERR(Debug::GUI, "misspelled word is correct with dot: \"" <<
+                                          word << "\" [" <<
+                                          from << ".." << to << "]");
+                               }
+                       }
+               }
+               d->setMisspelled(from, to, result);
+       } else {
+               result = d->speller_state_.getState(from);
+       }
+
+       bool const misspelled_ = SpellChecker::misspelled(result) ;
+       if (misspelled_ && do_suggestion)
+               speller->suggest(wl, suggestions);
+       else if (misspelled_)
+               LYXERR(Debug::GUI, "misspelled word: \"" <<
+                          word << "\" [" <<
+                          from << ".." << to << "]");
+       else
+               suggestions.clear();
+
+       return result;
+}
 
-       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;
+void Paragraph::Private::markMisspelledWords(
+       pos_type const & first, pos_type const & last,
+       SpellChecker::Result result,
+       docstring const & word,
+       SkipPositions const & skips)
+{
+       if (!SpellChecker::misspelled(result)) {
+               setMisspelled(first, last, SpellChecker::WORD_OK);
+               return;
+       }
+       int snext = first;
+       SpellChecker * speller = theSpellChecker();
+       // locate and enumerate the error positions
+       int nerrors = speller->numMisspelledWords();
+       int numskipped = 0;
+       SkipPositionsIterator it = skips.begin();
+       SkipPositionsIterator et = skips.end();
+       for (int index = 0; index < nerrors; ++index) {
+               int wstart;
+               int wlen = 0;
+               speller->misspelledWord(index, wstart, wlen);
+               /// should not happen if speller supports range checks
+               if (!wlen) continue;
+               docstring const misspelled = word.substr(wstart, wlen);
+               wstart += first + numskipped;
+               if (snext < wstart) {
+                       /// mark the range of correct spelling
+                       numskipped += countSkips(it, et, wstart);
+                       setMisspelled(snext,
+                               wstart - 1, SpellChecker::WORD_OK);
+               }
+               snext = wstart + wlen;
+               numskipped += countSkips(it, et, snext);
+               /// mark the range of misspelling
+               setMisspelled(wstart, snext, result);
+               LYXERR(Debug::GUI, "misspelled word: \"" <<
+                          misspelled << "\" [" <<
+                          wstart << ".." << (snext-1) << "]");
+               ++snext;
+       }
+       if (snext <= last) {
+               /// mark the range of correct spelling at end
+               setMisspelled(snext, last, SpellChecker::WORD_OK);
+       }
+}
 
-       if (lyxrc.spellcheck_continuously)
-               d->fontlist_.setMisspelled(from, to, misspelled);
 
-       if (misspelled) {
-               while (!(word = speller->nextMiss()).empty())
-                       suggestions.push_back(word);
+void Paragraph::spellCheck() const
+{
+       SpellChecker * speller = theSpellChecker();
+       if (!speller || !size() ||!needsSpellCheck())
+               return;
+       pos_type start;
+       pos_type endpos;
+       d->rangeOfSpellCheck(start, endpos);
+       if (speller->canCheckParagraph()) {
+               // loop until we leave the range
+               for (pos_type first = start; first < endpos; ) {
+                       pos_type last = endpos;
+                       Private::SkipPositions skips;
+                       Language * lang = d->locateSpellRange(first, last, skips);
+                       if (first >= endpos)
+                               break;
+                       // start the spell checker on the unit of meaning
+                       docstring word = asString(first, last, AS_STR_INSETS + AS_STR_SKIPDELETE);
+                       WordLangTuple wl = WordLangTuple(word, lang);
+                       SpellChecker::Result result = word.size() ?
+                               speller->check(wl) : SpellChecker::WORD_OK;
+                       d->markMisspelledWords(first, last, result, word, skips);
+                       first = ++last;
+               }
+       } else {
+               static docstring_list suggestions;
+               pos_type to = endpos;
+               while (start < endpos) {
+                       WordLangTuple wl;
+                       spellCheck(start, to, wl, suggestions, false);
+                       start = to + 1;
+               }
        }
-       return misspelled;
+       d->readySpellCheck();
 }
 
 
 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);
+       return SpellChecker::misspelled(d->speller_state_.getState(pos));
+}
+
+
+string Paragraph::magicLabel() const
+{
+       stringstream ss;
+       ss << "magicparlabel-" << id();
+       return ss.str();
 }