src/Paragraph.cpp

   1 /**
   2  * \file Paragraph.cpp
   3  * This file is part of LyX, the document processor.
   4  * Licence details can be found in the file COPYING.
   5  *
   6  * \author Asger Alstrup
   7  * \author Lars Gullik Bjønnes
   8  * \author Richard Heck (XHTML output)
   9  * \author Jean-Marc Lasgouttes
  10  * \author Angus Leeming
  11  * \author John Levon
  12  * \author André Pönitz
  13  * \author Dekel Tsur
  14  * \author Jürgen Vigna
  15  *
  16  * Full author contact details are available in file CREDITS.
  17  */
  18
  19 #include <config.h>
  20
  21 #include "Paragraph.h"
  22
  23 #include "LayoutFile.h"
  24 #include "Buffer.h"
  25 #include "BufferParams.h"
  26 #include "Changes.h"
  27 #include "Counters.h"
  28 #include "Encoding.h"
  29 #include "InsetList.h"
  30 #include "Language.h"
  31 #include "LaTeXFeatures.h"
  32 #include "Layout.h"
  33 #include "Length.h"
  34 #include "Font.h"
  35 #include "FontList.h"
  36 #include "LyXRC.h"
  37 #include "OutputParams.h"
  38 #include "output_latex.h"
  39 #include "output_xhtml.h"
  40 #include "ParagraphParameters.h"
  41 #include "SpellChecker.h"
  42 #include "sgml.h"
  43 #include "TextClass.h"
  44 #include "TexRow.h"
  45 #include "Text.h"
  46 #include "WordLangTuple.h"
  47 #include "WordList.h"
  48
  49 #include "frontends/alert.h"
  50
  51 #include "insets/InsetBibitem.h"
  52 #include "insets/InsetLabel.h"
  53 #include "insets/InsetSpecialChar.h"
  54
  55 #include "support/debug.h"
  56 #include "support/docstring_list.h"
  57 #include "support/ExceptionMessage.h"
  58 #include "support/gettext.h"
  59 #include "support/lassert.h"
  60 #include "support/lstrings.h"
  61 #include "support/textutils.h"
  62
  63 #include <sstream>
  64 #include <vector>
  65
  66 using namespace std;
  67 using namespace lyx::support;
  68
  69 namespace lyx {
  70
  71 namespace {
  72
  73 /// Inset identifier (above 0x10ffff, for ucs-4)
  74 char_type const META_INSET = 0x200001;
  75
  76 }
  77
  78
  79 /////////////////////////////////////////////////////////////////////
  80 //
  81 // SpellResultRange
  82 //
  83 /////////////////////////////////////////////////////////////////////
  84
  85 class SpellResultRange {
  86 public:
  87         SpellResultRange(FontSpan range, SpellChecker::Result result)
  88         : range_(range), result_(result)
  89         {}
  90         ///
  91         FontSpan const & range() const { return range_; }
  92         ///
  93         void range(FontSpan const & r) { range_ = r; }
  94         ///
  95         SpellChecker::Result result() const { return result_; }
  96         ///
  97         void result(SpellChecker::Result r) { result_ = r; }
  98         ///
  99         bool inside(pos_type pos) const { return range_.inside(pos); }
 100         ///
 101         bool covered(FontSpan const & r) const
 102         {
 103                 // 1. first of new range inside current range or
 104                 // 2. last of new range inside current range or
 105                 // 3. first of current range inside new range or
 106                 // 4. last of current range inside new range
 107                 return range_.inside(r.first) || range_.inside(r.last) ||
 108                         r.inside(range_.first) || r.inside(range_.last);
 109         }
 110         ///
 111         void shift(pos_type pos, int offset)
 112         {
 113                 if (range_.first > pos) {
 114                         range_.first += offset;
 115                         range_.last += offset;
 116                 } else if (range_.last >= pos) {
 117                         range_.last += offset;
 118                 }
 119         }
 120 private:
 121         FontSpan range_ ;
 122         SpellChecker::Result result_ ;
 123 };
 124
 125
 126 /////////////////////////////////////////////////////////////////////
 127 //
 128 // SpellCheckerState
 129 //
 130 /////////////////////////////////////////////////////////////////////
 131
 132 class SpellCheckerState {
 133 public:
 134         SpellCheckerState()
 135         {
 136                 needs_refresh_ = true;
 137                 current_change_number_ = 0;
 138         }
 139
 140         void setRange(FontSpan const & fp, SpellChecker::Result state)
 141         {
 142                 Ranges result;
 143                 RangesIterator et = ranges_.end();
 144                 RangesIterator it = ranges_.begin();
 145                 for (; it != et; ++it) {
 146                         if (!it->covered(fp))
 147                                 result.push_back(SpellResultRange(it->range(), it->result()));
 148                         else if (state == SpellChecker::WORD_OK) {
 149                                 // trim or split the current misspelled range
 150                                 // store misspelled ranges only
 151                                 FontSpan range = it->range();
 152                                 if (fp.first > range.first) {
 153                                         // misspelled area in front of WORD_OK
 154                                         range.last = fp.first - 1;
 155                                         result.push_back(SpellResultRange(range, it->result()));
 156                                         range = it->range();
 157                                 }
 158                                 if (fp.last < range.last) {
 159                                         // misspelled area after WORD_OK range
 160                                         range.first = fp.last + 1;
 161                                         result.push_back(SpellResultRange(range, it->result()));
 162                                 }
 163                         }
 164                 }
 165                 ranges_ = result;
 166                 if (state != SpellChecker::WORD_OK)
 167                         ranges_.push_back(SpellResultRange(fp, state));
 168         }
 169
 170         void increasePosAfterPos(pos_type pos)
 171         {
 172                 correctRangesAfterPos(pos, 1);
 173                 needsRefresh(pos);
 174         }
 175
 176         void decreasePosAfterPos(pos_type pos)
 177         {
 178                 correctRangesAfterPos(pos, -1);
 179                 needsRefresh(pos);
 180         }
 181
 182         void refreshLast(pos_type pos)
 183         {
 184                 if (pos < refresh_.last)
 185                         refresh_.last = pos;
 186         }
 187
 188         SpellChecker::Result getState(pos_type pos) const
 189         {
 190                 SpellChecker::Result result = SpellChecker::WORD_OK;
 191                 RangesIterator et = ranges_.end();
 192                 RangesIterator it = ranges_.begin();
 193                 for (; it != et; ++it) {
 194                         if(it->inside(pos)) {
 195                                 return it->result();
 196                         }
 197                 }
 198                 return result;
 199         }
 200
 201         FontSpan const & getRange(pos_type pos) const
 202         {
 203                 /// empty span to indicate mismatch
 204                 static FontSpan empty_;
 205                 RangesIterator et = ranges_.end();
 206                 RangesIterator it = ranges_.begin();
 207                 for (; it != et; ++it) {
 208                         if(it->inside(pos)) {
 209                                 return it->range();
 210                         }
 211                 }
 212                 return empty_;
 213         }
 214
 215         bool needsRefresh() const
 216         {
 217                 return needs_refresh_;
 218         }
 219
 220         SpellChecker::ChangeNumber currentChangeNumber() const
 221         {
 222                 return current_change_number_;
 223         }
 224
 225         void refreshRange(pos_type & first, pos_type & last) const
 226         {
 227                 first = refresh_.first;
 228                 last = refresh_.last;
 229         }
 230
 231         void needsRefresh(pos_type pos)
 232         {
 233                 if (needs_refresh_ && pos != -1) {
 234                         if (pos < refresh_.first)
 235                                 refresh_.first = pos;
 236                         if (pos > refresh_.last)
 237                                 refresh_.last = pos;
 238                 } else if (pos != -1) {
 239                         // init request check for neighbour positions too
 240                         refresh_.first = pos > 0 ? pos - 1 : 0;
 241                         // no need for special end of paragraph check
 242                         refresh_.last = pos + 1;
 243                 }
 244                 needs_refresh_ = pos != -1;
 245         }
 246
 247         void needsCompleteRefresh(SpellChecker::ChangeNumber change_number)
 248         {
 249                 needs_refresh_ = true;
 250                 refresh_.first = 0;
 251                 refresh_.last = -1;
 252                 current_change_number_ = change_number;
 253         }
 254 private:
 255         typedef vector<SpellResultRange> Ranges;
 256         typedef Ranges::const_iterator RangesIterator;
 257         Ranges ranges_;
 258         /// the area of the paragraph with pending spell check
 259         FontSpan refresh_;
 260         bool needs_refresh_;
 261         /// spell state cache version number
 262         SpellChecker::ChangeNumber current_change_number_;
 263
 264
 265         void correctRangesAfterPos(pos_type pos, int offset)
 266         {
 267                 RangesIterator et = ranges_.end();
 268                 Ranges::iterator it = ranges_.begin();
 269                 for (; it != et; ++it) {
 270                         it->shift(pos, offset);
 271                 }
 272         }
 273
 274 };
 275
 276 /////////////////////////////////////////////////////////////////////
 277 //
 278 // Paragraph::Private
 279 //
 280 /////////////////////////////////////////////////////////////////////
 281
 282 class Paragraph::Private
 283 {
 284 public:
 285         ///
 286         Private(Paragraph * owner, Layout const & layout);
 287         /// "Copy constructor"
 288         Private(Private const &, Paragraph * owner);
 289         /// Copy constructor from \p beg  to \p end
 290         Private(Private const &, Paragraph * owner, pos_type beg, pos_type end);
 291
 292         ///
 293         void insertChar(pos_type pos, char_type c, Change const & change);
 294
 295         /// Output the surrogate pair formed by \p c and \p next to \p os.
 296         /// \return the number of characters written.
 297         int latexSurrogatePair(otexstream & os, char_type c, char_type next,
 298                                OutputParams const &);
 299
 300         /// Output a space in appropriate formatting (or a surrogate pair
 301         /// if the next character is a combining character).
 302         /// \return whether a surrogate pair was output.
 303         bool simpleTeXBlanks(OutputParams const &,
 304                              otexstream &,
 305                              pos_type i,
 306                              unsigned int & column,
 307                              Font const & font,
 308                              Layout const & style);
 309
 310         /// Output consecutive unicode chars, belonging to the same script as
 311         /// specified by the latex macro \p ltx, to \p os starting from \p i.
 312         /// \return the number of characters written.
 313         int writeScriptChars(otexstream & os, docstring const & ltx,
 314                            Change const &, Encoding const &, pos_type & i);
 315
 316         /// This could go to ParagraphParameters if we want to.
 317         int startTeXParParams(BufferParams const &, otexstream &,
 318                               OutputParams const &) const;
 319
 320         /// This could go to ParagraphParameters if we want to.
 321         bool endTeXParParams(BufferParams const &, otexstream &,
 322                              OutputParams const &) const;
 323
 324         ///
 325         void latexInset(BufferParams const &,
 326                                    otexstream &,
 327                                    OutputParams &,
 328                                    Font & running_font,
 329                                    Font & basefont,
 330                                    Font const & outerfont,
 331                                    bool & open_font,
 332                                    Change & running_change,
 333                                    Layout const & style,
 334                                    pos_type & i,
 335                                    unsigned int & column);
 336
 337         ///
 338         void latexSpecialChar(
 339                                    otexstream & os,
 340                                    BufferParams const & bparams,
 341                                    OutputParams const & runparams,
 342                                    Font const & running_font,
 343                                    Change const & running_change,
 344                                    Layout const & style,
 345                                    pos_type & i,
 346                                    pos_type end_pos,
 347                                    unsigned int & column);
 348
 349         ///
 350         bool latexSpecialT1(
 351                 char_type const c,
 352                 otexstream & os,
 353                 pos_type i,
 354                 unsigned int & column);
 355         ///
 356         bool latexSpecialT3(
 357                 char_type const c,
 358                 otexstream & os,
 359                 pos_type i,
 360                 unsigned int & column);
 361         ///
 362         bool latexSpecialTypewriter(
 363                 char_type const c,
 364                 otexstream & os,
 365                 pos_type i,
 366                 unsigned int & column);
 367         ///
 368         bool latexSpecialPhrase(
 369                 otexstream & os,
 370                 pos_type & i,
 371                 pos_type end_pos,
 372                 unsigned int & column,
 373                 OutputParams const & runparams);
 374
 375         ///
 376         void validate(LaTeXFeatures & features) const;
 377
 378         /// Checks if the paragraph contains only text and no inset or font change.
 379         bool onlyText(Buffer const & buf, Font const & outerfont,
 380                       pos_type initial) const;
 381
 382         /// match a string against a particular point in the paragraph
 383         bool isTextAt(string const & str, pos_type pos) const;
 384
 385         /// a vector of speller skip positions
 386         typedef vector<FontSpan> SkipPositions;
 387         typedef SkipPositions::const_iterator SkipPositionsIterator;
 388
 389         void appendSkipPosition(SkipPositions & skips, pos_type const pos) const;
 390
 391         Language * getSpellLanguage(pos_type const from) const;
 392
 393         Language * locateSpellRange(pos_type & from, pos_type & to,
 394                                     SkipPositions & skips) const;
 395
 396         bool hasSpellerChange() const
 397         {
 398                 SpellChecker::ChangeNumber speller_change_number = 0;
 399                 if (theSpellChecker())
 400                         speller_change_number = theSpellChecker()->changeNumber();
 401                 return speller_change_number > speller_state_.currentChangeNumber();
 402         }
 403
 404         bool ignoreWord(docstring const & word) const ;
 405
 406         void setMisspelled(pos_type from, pos_type to, SpellChecker::Result state)
 407         {
 408                 pos_type textsize = owner_->size();
 409                 // check for sane arguments
 410                 if (to <= from || from >= textsize)
 411                         return;
 412                 FontSpan fp = FontSpan(from, to - 1);
 413                 speller_state_.setRange(fp, state);
 414         }
 415
 416         void requestSpellCheck(pos_type pos)
 417         {
 418                 if (pos == -1)
 419                         speller_state_.needsCompleteRefresh(speller_state_.currentChangeNumber());
 420                 else
 421                         speller_state_.needsRefresh(pos);
 422         }
 423
 424         void readySpellCheck()
 425         {
 426                 speller_state_.needsRefresh(-1);
 427         }
 428
 429         bool needsSpellCheck() const
 430         {
 431                 return speller_state_.needsRefresh();
 432         }
 433
 434         void rangeOfSpellCheck(pos_type & first, pos_type & last) const
 435         {
 436                 speller_state_.refreshRange(first, last);
 437                 if (last == -1) {
 438                         last = owner_->size();
 439                         return;
 440                 }
 441                 pos_type endpos = last;
 442                 owner_->locateWord(first, endpos, WHOLE_WORD);
 443                 if (endpos < last) {
 444                         endpos = last;
 445                         owner_->locateWord(last, endpos, WHOLE_WORD);
 446                 }
 447                 last = endpos;
 448         }
 449
 450         int countSkips(SkipPositionsIterator & it, SkipPositionsIterator const et,
 451                             int & start) const
 452         {
 453                 int numskips = 0;
 454                 while (it != et && it->first < start) {
 455                         int skip = it->last - it->first + 1;
 456                         start += skip;
 457                         numskips += skip;
 458                         ++it;
 459                 }
 460                 return numskips;
 461         }
 462
 463         void markMisspelledWords(pos_type const & first, pos_type const & last,
 464                                                          SpellChecker::Result result,
 465                                                          docstring const & word,
 466                                                          SkipPositions const & skips);
 467
 468         InsetCode ownerCode() const
 469         {
 470                 return inset_owner_ ? inset_owner_->lyxCode() : NO_CODE;
 471         }
 472
 473         /// Which Paragraph owns us?
 474         Paragraph * owner_;
 475
 476         /// In which Inset?
 477         Inset const * inset_owner_;
 478
 479         ///
 480         FontList fontlist_;
 481
 482         ///
 483         int id_;
 484
 485         ///
 486         ParagraphParameters params_;
 487
 488         /// for recording and looking up changes
 489         Changes changes_;
 490
 491         ///
 492         InsetList insetlist_;
 493
 494         /// end of label
 495         pos_type begin_of_body_;
 496
 497         typedef docstring TextContainer;
 498         ///
 499         TextContainer text_;
 500
 501         typedef set<docstring> Words;
 502         typedef map<Language, Words> LangWordsMap;
 503         ///
 504         LangWordsMap words_;
 505         ///
 506         Layout const * layout_;
 507         ///
 508         SpellCheckerState speller_state_;
 509 };
 510
 511
 512 namespace {
 513
 514 struct special_phrase {
 515         string phrase;
 516         docstring macro;
 517         bool builtin;
 518 };
 519
 520 special_phrase const special_phrases[] = {
 521         { "LyX", from_ascii("\\LyX{}"), false },
 522         { "TeX", from_ascii("\\TeX{}"), true },
 523         { "LaTeX2e", from_ascii("\\LaTeXe{}"), true },
 524         { "LaTeX", from_ascii("\\LaTeX{}"), true },
 525 };
 526
 527 size_t const phrases_nr = sizeof(special_phrases)/sizeof(special_phrase);
 528
 529 } // namespace anon
 530
 531
 532 Paragraph::Private::Private(Paragraph * owner, Layout const & layout)
 533         : owner_(owner), inset_owner_(0), id_(-1), begin_of_body_(0), layout_(&layout)
 534 {
 535         text_.reserve(100);
 536 }
 537
 538
 539 // Initialization of the counter for the paragraph id's,
 540 //
 541 // FIXME: There should be a more intelligent way to generate and use the
 542 // paragraph ids per buffer instead a global static counter for all InsetText
 543 // in the running program.
 544 static int paragraph_id = -1;
 545
 546 Paragraph::Private::Private(Private const & p, Paragraph * owner)
 547         : owner_(owner), inset_owner_(p.inset_owner_), fontlist_(p.fontlist_),
 548           params_(p.params_), changes_(p.changes_), insetlist_(p.insetlist_),
 549           begin_of_body_(p.begin_of_body_), text_(p.text_), words_(p.words_),
 550           layout_(p.layout_)
 551 {
 552         id_ = ++paragraph_id;
 553         requestSpellCheck(p.text_.size());
 554 }
 555
 556
 557 Paragraph::Private::Private(Private const & p, Paragraph * owner,
 558         pos_type beg, pos_type end)
 559         : owner_(owner), inset_owner_(p.inset_owner_),
 560           params_(p.params_), changes_(p.changes_),
 561           insetlist_(p.insetlist_, beg, end),
 562           begin_of_body_(p.begin_of_body_), words_(p.words_),
 563           layout_(p.layout_)
 564 {
 565         id_ = ++paragraph_id;
 566         if (beg >= pos_type(p.text_.size()))
 567                 return;
 568         text_ = p.text_.substr(beg, end - beg);
 569
 570         FontList::const_iterator fcit = fontlist_.begin();
 571         FontList::const_iterator fend = fontlist_.end();
 572         for (; fcit != fend; ++fcit) {
 573                 if (fcit->pos() < beg)
 574                         continue;
 575                 if (fcit->pos() >= end) {
 576                         // Add last entry in the fontlist_.
 577                         fontlist_.set(text_.size() - 1, fcit->font());
 578                         break;
 579                 }
 580                 // Add a new entry in the fontlist_.
 581                 fontlist_.set(fcit->pos() - beg, fcit->font());
 582         }
 583         requestSpellCheck(p.text_.size());
 584 }
 585
 586
 587 void Paragraph::addChangesToToc(DocIterator const & cdit,
 588         Buffer const & buf, bool output_active) const
 589 {
 590         d->changes_.addToToc(cdit, buf, output_active);
 591 }
 592
 593
 594 bool Paragraph::isDeleted(pos_type start, pos_type end) const
 595 {
 596         LASSERT(start >= 0 && start <= size(), return false);
 597         LASSERT(end > start && end <= size() + 1, return false);
 598
 599         return d->changes_.isDeleted(start, end);
 600 }
 601
 602
 603 bool Paragraph::isChanged(pos_type start, pos_type end) const
 604 {
 605         LASSERT(start >= 0 && start <= size(), return false);
 606         LASSERT(end > start && end <= size() + 1, return false);
 607
 608         return d->changes_.isChanged(start, end);
 609 }
 610
 611
 612 bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const
 613 {
 614         // keep the logic here in sync with the logic of eraseChars()
 615         if (!trackChanges)
 616                 return true;
 617
 618         Change const change = d->changes_.lookup(size());
 619         return change.inserted() && change.currentAuthor();
 620 }
 621
 622
 623 void Paragraph::setChange(Change const & change)
 624 {
 625         // beware of the imaginary end-of-par character!
 626         d->changes_.set(change, 0, size() + 1);
 627
 628         /*
 629          * Propagate the change recursively - but not in case of DELETED!
 630          *
 631          * Imagine that your co-author makes changes in an existing inset. He
 632          * sends your document to you and you come to the conclusion that the
 633          * inset should go completely. If you erase it, LyX must not delete all
 634          * text within the inset. Otherwise, the change tracked insertions of
 635          * your co-author get lost and there is no way to restore them later.
 636          *
 637          * Conclusion: An inset's content should remain untouched if you delete it
 638          */
 639
 640         if (!change.deleted()) {
 641                 for (pos_type pos = 0; pos < size(); ++pos) {
 642                         if (Inset * inset = getInset(pos))
 643                                 inset->setChange(change);
 644                 }
 645         }
 646 }
 647
 648
 649 void Paragraph::setChange(pos_type pos, Change const & change)
 650 {
 651         LASSERT(pos >= 0 && pos <= size(), return);
 652         d->changes_.set(change, pos);
 653
 654         // see comment in setChange(Change const &) above
 655         if (!change.deleted() && pos < size())
 656                         if (Inset * inset = getInset(pos))
 657                                 inset->setChange(change);
 658 }
 659
 660
 661 Change const & Paragraph::lookupChange(pos_type pos) const
 662 {
 663         LBUFERR(pos >= 0 && pos <= size());
 664         return d->changes_.lookup(pos);
 665 }
 666
 667
 668 void Paragraph::acceptChanges(pos_type start, pos_type end)
 669 {
 670         LASSERT(start >= 0 && start <= size(), return);
 671         LASSERT(end > start && end <= size() + 1, return);
 672
 673         for (pos_type pos = start; pos < end; ++pos) {
 674                 switch (lookupChange(pos).type) {
 675                         case Change::UNCHANGED:
 676                                 // accept changes in nested inset
 677                                 if (Inset * inset = getInset(pos))
 678                                         inset->acceptChanges();
 679                                 break;
 680
 681                         case Change::INSERTED:
 682                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 683                                 // also accept changes in nested inset
 684                                 if (Inset * inset = getInset(pos))
 685                                         inset->acceptChanges();
 686                                 break;
 687
 688                         case Change::DELETED:
 689                                 // Suppress access to non-existent
 690                                 // "end-of-paragraph char"
 691                                 if (pos < size()) {
 692                                         eraseChar(pos, false);
 693                                         --end;
 694                                         --pos;
 695                                 }
 696                                 break;
 697                 }
 698
 699         }
 700 }
 701
 702
 703 void Paragraph::rejectChanges(pos_type start, pos_type end)
 704 {
 705         LASSERT(start >= 0 && start <= size(), return);
 706         LASSERT(end > start && end <= size() + 1, return);
 707
 708         for (pos_type pos = start; pos < end; ++pos) {
 709                 switch (lookupChange(pos).type) {
 710                         case Change::UNCHANGED:
 711                                 // reject changes in nested inset
 712                                 if (Inset * inset = getInset(pos))
 713                                                 inset->rejectChanges();
 714                                 break;
 715
 716                         case Change::INSERTED:
 717                                 // Suppress access to non-existent
 718                                 // "end-of-paragraph char"
 719                                 if (pos < size()) {
 720                                         eraseChar(pos, false);
 721                                         --end;
 722                                         --pos;
 723                                 }
 724                                 break;
 725
 726                         case Change::DELETED:
 727                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 728
 729                                 // Do NOT reject changes within a deleted inset!
 730                                 // There may be insertions of a co-author inside of it!
 731
 732                                 break;
 733                 }
 734         }
 735 }
 736
 737
 738 void Paragraph::Private::insertChar(pos_type pos, char_type c,
 739                 Change const & change)
 740 {
 741         LASSERT(pos >= 0 && pos <= int(text_.size()), return);
 742
 743         // track change
 744         changes_.insert(change, pos);
 745
 746         // This is actually very common when parsing buffers (and
 747         // maybe inserting ascii text)
 748         if (pos == pos_type(text_.size())) {
 749                 // when appending characters, no need to update tables
 750                 text_.push_back(c);
 751                 // but we want spell checking
 752                 requestSpellCheck(pos);
 753                 return;
 754         }
 755
 756         text_.insert(text_.begin() + pos, c);
 757
 758         // Update the font table.
 759         fontlist_.increasePosAfterPos(pos);
 760
 761         // Update the insets
 762         insetlist_.increasePosAfterPos(pos);
 763
 764         // Update list of misspelled positions
 765         speller_state_.increasePosAfterPos(pos);
 766 }
 767
 768
 769 bool Paragraph::insertInset(pos_type pos, Inset * inset,
 770                                    Font const & font, Change const & change)
 771 {
 772         LASSERT(inset, return false);
 773         LASSERT(pos >= 0 && pos <= size(), return false);
 774
 775         // Paragraph::insertInset() can be used in cut/copy/paste operation where
 776         // d->inset_owner_ is not set yet.
 777         if (d->inset_owner_ && !d->inset_owner_->insetAllowed(inset->lyxCode()))
 778                 return false;
 779
 780         d->insertChar(pos, META_INSET, change);
 781         LASSERT(d->text_[pos] == META_INSET, return false);
 782
 783         // Add a new entry in the insetlist_.
 784         d->insetlist_.insert(inset, pos);
 785
 786         // Some insets require run of spell checker
 787         requestSpellCheck(pos);
 788         setFont(pos, font);
 789         return true;
 790 }
 791
 792
 793 bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
 794 {
 795         LASSERT(pos >= 0 && pos <= size(), return false);
 796
 797         // keep the logic here in sync with the logic of isMergedOnEndOfParDeletion()
 798
 799         if (trackChanges) {
 800                 Change change = d->changes_.lookup(pos);
 801
 802                 // set the character to DELETED if
 803                 //  a) it was previously unchanged or
 804                 //  b) it was inserted by a co-author
 805
 806                 if (!change.changed() ||
 807                       (change.inserted() && !change.currentAuthor())) {
 808                         setChange(pos, Change(Change::DELETED));
 809                         // request run of spell checker
 810                         requestSpellCheck(pos);
 811                         return false;
 812                 }
 813
 814                 if (change.deleted())
 815                         return false;
 816         }
 817
 818         // Don't physically access the imaginary end-of-paragraph character.
 819         // eraseChar() can only mark it as DELETED. A physical deletion of
 820         // end-of-par must be handled externally.
 821         if (pos == size()) {
 822                 return false;
 823         }
 824
 825         // track change
 826         d->changes_.erase(pos);
 827
 828         // if it is an inset, delete the inset entry
 829         if (d->text_[pos] == META_INSET)
 830                 d->insetlist_.erase(pos);
 831
 832         d->text_.erase(d->text_.begin() + pos);
 833
 834         // Update the fontlist_
 835         d->fontlist_.erase(pos);
 836
 837         // Update the insetlist_
 838         d->insetlist_.decreasePosAfterPos(pos);
 839
 840         // Update list of misspelled positions
 841         d->speller_state_.decreasePosAfterPos(pos);
 842         d->speller_state_.refreshLast(size());
 843
 844         return true;
 845 }
 846
 847
 848 int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
 849 {
 850         LASSERT(start >= 0 && start <= size(), return 0);
 851         LASSERT(end >= start && end <= size() + 1, return 0);
 852
 853         pos_type i = start;
 854         for (pos_type count = end - start; count; --count) {
 855                 if (!eraseChar(i, trackChanges))
 856                         ++i;
 857         }
 858         return end - i;
 859 }
 860
 861
 862 int Paragraph::Private::latexSurrogatePair(otexstream & os, char_type c,
 863                 char_type next, OutputParams const & runparams)
 864 {
 865         // Writing next here may circumvent a possible font change between
 866         // c and next. Since next is only output if it forms a surrogate pair
 867         // with c we can ignore this:
 868         // A font change inside a surrogate pair does not make sense and is
 869         // hopefully impossible to input.
 870         // FIXME: change tracking
 871         // Is this correct WRT change tracking?
 872         Encoding const & encoding = *(runparams.encoding);
 873         docstring latex1 = encoding.latexChar(next).first;
 874         if (runparams.inIPA) {
 875                 string const tipashortcut = Encodings::TIPAShortcut(next);
 876                 if (!tipashortcut.empty()) {
 877                         latex1 = from_ascii(tipashortcut);
 878                 }
 879         }
 880         docstring const latex2 = encoding.latexChar(c).first;
 881         if (docstring(1, next) == latex1) {
 882                 // the encoding supports the combination
 883                 os << latex2 << latex1;
 884                 return latex1.length() + latex2.length();
 885         } else if (runparams.local_font &&
 886                    runparams.local_font->language()->lang() == "polutonikogreek") {
 887                 // polutonikogreek only works without the brackets
 888                 os << latex1 << latex2;
 889                 return latex1.length() + latex2.length();
 890         } else
 891                 os << latex1 << '{' << latex2 << '}';
 892         return latex1.length() + latex2.length() + 2;
 893 }
 894
 895
 896 bool Paragraph::Private::simpleTeXBlanks(OutputParams const & runparams,
 897                                        otexstream & os,
 898                                        pos_type i,
 899                                        unsigned int & column,
 900                                        Font const & font,
 901                                        Layout const & style)
 902 {
 903         if (style.pass_thru || runparams.pass_thru)
 904                 return false;
 905
 906         if (i + 1 < int(text_.size())) {
 907                 char_type next = text_[i + 1];
 908                 if (Encodings::isCombiningChar(next)) {
 909                         // This space has an accent, so we must always output it.
 910                         column += latexSurrogatePair(os, ' ', next, runparams) - 1;
 911                         return true;
 912                 }
 913         }
 914
 915         if (runparams.linelen > 0
 916             && column > runparams.linelen
 917             && i
 918             && text_[i - 1] != ' '
 919             && (i + 1 < int(text_.size()))
 920             // same in FreeSpacing mode
 921             && !owner_->isFreeSpacing()
 922             // In typewriter mode, we want to avoid
 923             // ! . ? : at the end of a line
 924             && !(font.fontInfo().family() == TYPEWRITER_FAMILY
 925                  && (text_[i - 1] == '.'
 926                      || text_[i - 1] == '?'
 927                      || text_[i - 1] == ':'
 928                      || text_[i - 1] == '!'))) {
 929                 os << '\n';
 930                 os.texrow().start(owner_->id(), i + 1);
 931                 column = 0;
 932         } else if (style.free_spacing) {
 933                 os << '~';
 934         } else {
 935                 os << ' ';
 936         }
 937         return false;
 938 }
 939
 940
 941 int Paragraph::Private::writeScriptChars(otexstream & os,
 942                                          docstring const & ltx,
 943                                          Change const & runningChange,
 944                                          Encoding const & encoding,
 945                                          pos_type & i)
 946 {
 947         // FIXME: modifying i here is not very nice...
 948
 949         // We only arrive here when a proper language for character text_[i] has
 950         // not been specified (i.e., it could not be translated in the current
 951         // latex encoding) or its latex translation has been forced, and it
 952         // belongs to a known script.
 953         // Parameter ltx contains the latex translation of text_[i] as specified
 954         // in the unicodesymbols file and is something like "\textXXX{<spec>}".
 955         // The latex macro name "textXXX" specifies the script to which text_[i]
 956         // belongs and we use it in order to check whether characters from the
 957         // same script immediately follow, such that we can collect them in a
 958         // single "\textXXX" macro. So, we have to retain "\textXXX{<spec>"
 959         // for the first char but only "<spec>" for all subsequent chars.
 960         docstring::size_type const brace1 = ltx.find_first_of(from_ascii("{"));
 961         docstring::size_type const brace2 = ltx.find_last_of(from_ascii("}"));
 962         string script = to_ascii(ltx.substr(1, brace1 - 1));
 963         int pos = 0;
 964         int length = brace2;
 965         bool closing_brace = true;
 966         if (script == "textgreek" && encoding.latexName() == "iso-8859-7") {
 967                 // Correct encoding is being used, so we can avoid \textgreek.
 968                 pos = brace1 + 1;
 969                 length -= pos;
 970                 closing_brace = false;
 971         }
 972         os << ltx.substr(pos, length);
 973         int size = text_.size();
 974         while (i + 1 < size) {
 975                 char_type const next = text_[i + 1];
 976                 // Stop here if next character belongs to another script
 977                 // or there is a change in change tracking status.
 978                 if (!Encodings::isKnownScriptChar(next, script) ||
 979                     runningChange != owner_->lookupChange(i + 1))
 980                         break;
 981                 Font prev_font;
 982                 bool found = false;
 983                 FontList::const_iterator cit = fontlist_.begin();
 984                 FontList::const_iterator end = fontlist_.end();
 985                 for (; cit != end; ++cit) {
 986                         if (cit->pos() >= i && !found) {
 987                                 prev_font = cit->font();
 988                                 found = true;
 989                         }
 990                         if (cit->pos() >= i + 1)
 991                                 break;
 992                 }
 993                 // Stop here if there is a font attribute or encoding change.
 994                 if (found && cit != end && prev_font != cit->font())
 995                         break;
 996                 docstring const latex = encoding.latexChar(next).first;
 997                 docstring::size_type const b1 =
 998                                         latex.find_first_of(from_ascii("{"));
 999                 docstring::size_type const b2 =
1000                                         latex.find_last_of(from_ascii("}"));
1001                 int const len = b2 - b1 - 1;
1002                 os << latex.substr(b1 + 1, len);
1003                 length += len;
1004                 ++i;
1005         }
1006         if (closing_brace) {
1007                 os << '}';
1008                 ++length;
1009         }
1010         return length;
1011 }
1012
1013
1014 bool Paragraph::Private::isTextAt(string const & str, pos_type pos) const
1015 {
1016         pos_type const len = str.length();
1017
1018         // is the paragraph large enough?
1019         if (pos + len > int(text_.size()))
1020                 return false;
1021
1022         // does the wanted text start at point?
1023         for (string::size_type i = 0; i < str.length(); ++i) {
1024                 // Caution: direct comparison of characters works only
1025                 // because str is pure ASCII.
1026                 if (str[i] != text_[pos + i])
1027                         return false;
1028         }
1029
1030         return fontlist_.hasChangeInRange(pos, len);
1031 }
1032
1033
1034 void Paragraph::Private::latexInset(BufferParams const & bparams,
1035                                     otexstream & os,
1036                                     OutputParams & runparams,
1037                                     Font & running_font,
1038                                     Font & basefont,
1039                                     Font const & outerfont,
1040                                     bool & open_font,
1041                                     Change & running_change,
1042                                     Layout const & style,
1043                                     pos_type & i,
1044                                     unsigned int & column)
1045 {
1046         Inset * inset = owner_->getInset(i);
1047         LBUFERR(inset);
1048
1049         if (style.pass_thru) {
1050                 odocstringstream ods;
1051                 inset->plaintext(ods, runparams);
1052                 os << ods.str();
1053                 return;
1054         }
1055
1056         // FIXME: move this to InsetNewline::latex
1057         if (inset->lyxCode() == NEWLINE_CODE) {
1058                 // newlines are handled differently here than
1059                 // the default in simpleTeXSpecialChars().
1060                 if (!style.newline_allowed) {
1061                         os << '\n';
1062                 } else {
1063                         if (open_font) {
1064                                 column += running_font.latexWriteEndChanges(
1065                                         os, bparams, runparams,
1066                                         basefont, basefont);
1067                                 open_font = false;
1068                         }
1069
1070                         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY)
1071                                 os << '~';
1072
1073                         basefont = owner_->getLayoutFont(bparams, outerfont);
1074                         running_font = basefont;
1075
1076                         if (runparams.moving_arg)
1077                                 os << "\\protect ";
1078
1079                 }
1080                 os.texrow().start(owner_->id(), i + 1);
1081                 column = 0;
1082         }
1083
1084         if (owner_->isDeleted(i)) {
1085                 if( ++runparams.inDeletedInset == 1)
1086                         runparams.changeOfDeletedInset = owner_->lookupChange(i);
1087         }
1088
1089         if (inset->canTrackChanges()) {
1090                 column += Changes::latexMarkChange(os, bparams, running_change,
1091                         Change(Change::UNCHANGED), runparams);
1092                 running_change = Change(Change::UNCHANGED);
1093         }
1094
1095         bool close = false;
1096         odocstream::pos_type const len = os.os().tellp();
1097
1098         if (inset->forceLTR()
1099             && !runparams.use_polyglossia
1100             && running_font.isRightToLeft()
1101             // ERT is an exception, it should be output with no
1102             // decorations at all
1103             && inset->lyxCode() != ERT_CODE) {
1104                 if (running_font.language()->lang() == "farsi")
1105                         os << "\\beginL{}";
1106                 else
1107                         os << "\\L{";
1108                 close = true;
1109         }
1110
1111         // FIXME: Bug: we can have an empty font change here!
1112         // if there has just been a font change, we are going to close it
1113         // right now, which means stupid latex code like \textsf{}. AFAIK,
1114         // this does not harm dvi output. A minor bug, thus (JMarc)
1115
1116         // Some insets cannot be inside a font change command.
1117         // However, even such insets *can* be placed in \L or \R
1118         // or their equivalents (for RTL language switches), so we don't
1119         // close the language in those cases.
1120         // ArabTeX, though, cannot handle this special behavior, it seems.
1121         bool arabtex = basefont.language()->lang() == "arabic_arabtex"
1122                 || running_font.language()->lang() == "arabic_arabtex";
1123         if (open_font && !inset->inheritFont()) {
1124                 bool closeLanguage = arabtex
1125                         || basefont.isRightToLeft() == running_font.isRightToLeft();
1126                 unsigned int count = running_font.latexWriteEndChanges(os,
1127                         bparams, runparams, basefont, basefont, closeLanguage);
1128                 column += count;
1129                 // if any font properties were closed, update the running_font,
1130                 // making sure, however, to leave the language as it was
1131                 if (count > 0) {
1132                         // FIXME: probably a better way to keep track of the old
1133                         // language, than copying the entire font?
1134                         Font const copy_font(running_font);
1135                         basefont = owner_->getLayoutFont(bparams, outerfont);
1136                         running_font = basefont;
1137                         if (!closeLanguage)
1138                                 running_font.setLanguage(copy_font.language());
1139                         // leave font open if language is still open
1140                         open_font = (running_font.language() == basefont.language());
1141                         if (closeLanguage)
1142                                 runparams.local_font = &basefont;
1143                 }
1144         }
1145
1146         int prev_rows = os.texrow().rows();
1147
1148         try {
1149                 runparams.lastid = id_;
1150                 runparams.lastpos = i;
1151                 inset->latex(os, runparams);
1152         } catch (EncodingException & e) {
1153                 // add location information and throw again.
1154                 e.par_id = id_;
1155                 e.pos = i;
1156                 throw(e);
1157         }
1158
1159         if (close) {
1160                 if (running_font.language()->lang() == "farsi")
1161                                 os << "\\endL{}";
1162                         else
1163                                 os << '}';
1164         }
1165
1166         if (os.texrow().rows() > prev_rows) {
1167                 os.texrow().start(owner_->id(), i + 1);
1168                 column = 0;
1169         } else {
1170                 column += (unsigned int)(os.os().tellp() - len);
1171         }
1172
1173         if (owner_->isDeleted(i))
1174                 --runparams.inDeletedInset;
1175 }
1176
1177
1178 void Paragraph::Private::latexSpecialChar(otexstream & os,
1179                                           BufferParams const & bparams,
1180                                           OutputParams const & runparams,
1181                                           Font const & running_font,
1182                                           Change const & running_change,
1183                                           Layout const & style,
1184                                           pos_type & i,
1185                                           pos_type end_pos,
1186                                           unsigned int & column)
1187 {
1188         // With polyglossia, brackets and stuff need not be reversed
1189         // in RTL scripts (see bug #8251)
1190         char_type const c = (runparams.use_polyglossia) ?
1191                 owner_->getUChar(bparams, i) : text_[i];
1192
1193         if (style.pass_thru || runparams.pass_thru) {
1194                 if (c != '\0') {
1195                         Encoding const * const enc = runparams.encoding;
1196                         if (enc && !enc->encodable(c))
1197                                 throw EncodingException(c);
1198                         os.put(c);
1199                 }
1200                 return;
1201         }
1202
1203         // TIPA uses its own T3 encoding
1204         if (runparams.inIPA && latexSpecialT3(c, os, i, column))
1205                 return;
1206         // If T1 font encoding is used, use the special
1207         // characters it provides.
1208         // NOTE: some languages reset the font encoding
1209         // internally
1210         if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1211             && lyxrc.fontenc == "T1" && latexSpecialT1(c, os, i, column))
1212                 return;
1213
1214         // \tt font needs special treatment
1215         if (!runparams.inIPA
1216              && running_font.fontInfo().family() == TYPEWRITER_FAMILY
1217              && latexSpecialTypewriter(c, os, i, column))
1218                 return;
1219
1220         // Otherwise, we use what LaTeX provides us.
1221         switch (c) {
1222         case '\\':
1223                 os << "\\textbackslash{}";
1224                 column += 15;
1225                 break;
1226         case '<':
1227                 os << "\\textless{}";
1228                 column += 10;
1229                 break;
1230         case '>':
1231                 os << "\\textgreater{}";
1232                 column += 13;
1233                 break;
1234         case '|':
1235                 os << "\\textbar{}";
1236                 column += 9;
1237                 break;
1238         case '-':
1239                 os << '-';
1240                 break;
1241         case '\"':
1242                 os << "\\char`\\\"{}";
1243                 column += 9;
1244                 break;
1245
1246         case '$': case '&':
1247         case '%': case '#': case '{':
1248         case '}': case '_':
1249                 os << '\\';
1250                 os.put(c);
1251                 column += 1;
1252                 break;
1253
1254         case '~':
1255                 os << "\\textasciitilde{}";
1256                 column += 16;
1257                 break;
1258
1259         case '^':
1260                 os << "\\textasciicircum{}";
1261                 column += 17;
1262                 break;
1263
1264         case '*':
1265         case '[':
1266         case ']':
1267                 // avoid being mistaken for optional arguments
1268                 os << '{';
1269                 os.put(c);
1270                 os << '}';
1271                 column += 2;
1272                 break;
1273
1274         case ' ':
1275                 // Blanks are printed before font switching.
1276                 // Sure? I am not! (try nice-latex)
1277                 // I am sure it's correct. LyX might be smarter
1278                 // in the future, but for now, nothing wrong is
1279                 // written. (Asger)
1280                 break;
1281
1282         default:
1283                 // LyX, LaTeX etc.
1284                 if (latexSpecialPhrase(os, i, end_pos, column, runparams))
1285                         return;
1286
1287                 if (c == '\0')
1288                         return;
1289
1290                 Encoding const & encoding = *(runparams.encoding);
1291                 char_type next = '\0';
1292                 if (i + 1 < int(text_.size())) {
1293                         next = text_[i + 1];
1294                         if (Encodings::isCombiningChar(next)) {
1295                                 column += latexSurrogatePair(os, c, next, runparams) - 1;
1296                                 ++i;
1297                                 break;
1298                         }
1299                 }
1300                 string script;
1301                 pair<docstring, bool> latex = encoding.latexChar(c);
1302                 docstring nextlatex;
1303                 bool nexttipas = false;
1304                 string nexttipashortcut;
1305                 if (next != '\0' && next != META_INSET && encoding.encodable(next)) {
1306                         nextlatex = encoding.latexChar(next).first;
1307                         if (runparams.inIPA) {
1308                                 nexttipashortcut = Encodings::TIPAShortcut(next);
1309                                 nexttipas = !nexttipashortcut.empty();
1310                         }
1311                 }
1312                 bool tipas = false;
1313                 if (runparams.inIPA) {
1314                         string const tipashortcut = Encodings::TIPAShortcut(c);
1315                         if (!tipashortcut.empty()) {
1316                                 latex.first = from_ascii(tipashortcut);
1317                                 latex.second = false;
1318                                 tipas = true;
1319                         }
1320                 }
1321                 if (Encodings::isKnownScriptChar(c, script)
1322                     && prefixIs(latex.first, from_ascii("\\" + script)))
1323                         column += writeScriptChars(os, latex.first,
1324                                         running_change, encoding, i) - 1;
1325                 else if (latex.second
1326                          && ((!prefixIs(nextlatex, '\\')
1327                                && !prefixIs(nextlatex, '{')
1328                                && !prefixIs(nextlatex, '}'))
1329                              || (nexttipas
1330                                  && !prefixIs(from_ascii(nexttipashortcut), '\\')))
1331                          && !tipas) {
1332                         // Prevent eating of a following
1333                         // space or command corruption by
1334                         // following characters
1335                         if (next == ' ' || next == '\0') {
1336                                 column += latex.first.length() + 1;
1337                                 os << latex.first << "{}";
1338                         } else {
1339                                 column += latex.first.length();
1340                                 os << latex.first << " ";
1341                         }
1342                 } else {
1343                         column += latex.first.length() - 1;
1344                         os << latex.first;
1345                 }
1346                 break;
1347         }
1348 }
1349
1350
1351 bool Paragraph::Private::latexSpecialT1(char_type const c, otexstream & os,
1352         pos_type i, unsigned int & column)
1353 {
1354         switch (c) {
1355         case '>':
1356         case '<':
1357                 os.put(c);
1358                 // In T1 encoding, these characters exist
1359                 // but we should avoid ligatures
1360                 if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
1361                         return true;
1362                 os << "\\textcompwordmark{}";
1363                 column += 19;
1364                 return true;
1365         case '|':
1366                 os.put(c);
1367                 return true;
1368         case '\"':
1369                 // soul.sty breaks with \char`\"
1370                 os << "\\textquotedbl{}";
1371                 column += 14;
1372                 return true;
1373         default:
1374                 return false;
1375         }
1376 }
1377
1378
1379 bool Paragraph::Private::latexSpecialT3(char_type const c, otexstream & os,
1380         pos_type /*i*/, unsigned int & column)
1381 {
1382         switch (c) {
1383         case '*':
1384         case '[':
1385         case ']':
1386         case '\"':
1387                 os.put(c);
1388                 return true;
1389         case '|':
1390                 os << "\\textvertline{}";
1391                 column += 14;
1392                 return true;
1393         default:
1394                 return false;
1395         }
1396 }
1397
1398
1399 bool Paragraph::Private::latexSpecialTypewriter(char_type const c, otexstream & os,
1400         pos_type i, unsigned int & column)
1401 {
1402         switch (c) {
1403         case '-':
1404                 // within \ttfamily, "--" is merged to "-" (no endash)
1405                 // so we avoid this rather irritating ligature
1406                 if (i + 1 < int(text_.size()) && text_[i + 1] == '-') {
1407                         os << "-{}";
1408                         column += 2;
1409                 } else
1410                         os << '-';
1411                 return true;
1412
1413         // everything else has to be checked separately
1414         // (depending on the encoding)
1415         default:
1416                 return false;
1417         }
1418 }
1419
1420
1421 /// \param end_pos
1422 ///   If [start_pos, end_pos) does not include entirely the special phrase, then
1423 ///   do not apply the macro transformation.
1424 bool Paragraph::Private::latexSpecialPhrase(otexstream & os, pos_type & i, pos_type end_pos,
1425         unsigned int & column, OutputParams const & runparams)
1426 {
1427         // FIXME: if we have "LaTeX" with a font
1428         // change in the middle (before the 'T', then
1429         // the "TeX" part is still special cased.
1430         // Really we should only operate this on
1431         // "words" for some definition of word
1432
1433         for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
1434                 if (!isTextAt(special_phrases[pnr].phrase, i)
1435                     || (end_pos != -1 && i + int(special_phrases[pnr].phrase.size()) > end_pos))
1436                         continue;
1437                 if (runparams.moving_arg)
1438                         os << "\\protect";
1439                 os << special_phrases[pnr].macro;
1440                 i += special_phrases[pnr].phrase.length() - 1;
1441                 column += special_phrases[pnr].macro.length() - 1;
1442                 return true;
1443         }
1444         return false;
1445 }
1446
1447
1448 void Paragraph::Private::validate(LaTeXFeatures & features) const
1449 {
1450         if (layout_->inpreamble && inset_owner_) {
1451                 bool const is_command = layout_->latextype == LATEX_COMMAND;
1452                 Buffer const & buf = inset_owner_->buffer();
1453                 BufferParams const & bp = buf.params();
1454                 Font f;
1455                 TexRow texrow;
1456                 // Using a string stream here circumvents the encoding
1457                 // switching machinery of odocstream. Therefore the
1458                 // output is wrong if this paragraph contains content
1459                 // that needs to switch encoding.
1460                 odocstringstream ods;
1461                 otexstream os(ods, texrow);
1462                 if (is_command) {
1463                         os << '\\' << from_ascii(layout_->latexname());
1464                         // we have to provide all the optional arguments here, even though
1465                         // the last one is the only one we care about.
1466                         // Separate handling of optional argument inset.
1467                         if (!layout_->latexargs().empty()) {
1468                                 OutputParams rp = features.runparams();
1469                                 rp.local_font = &owner_->getFirstFontSettings(bp);
1470                                 latexArgInsets(*owner_, os, rp, layout_->latexargs());
1471                         }
1472                         os << from_ascii(layout_->latexparam());
1473                 }
1474                 docstring::size_type const length = ods.str().length();
1475                 // this will output "{" at the beginning, but not at the end
1476                 owner_->latex(bp, f, os, features.runparams(), 0, -1, true);
1477                 if (ods.str().length() > length) {
1478                         if (is_command) {
1479                                 ods << '}';
1480                                 if (!layout_->postcommandargs().empty()) {
1481                                         OutputParams rp = features.runparams();
1482                                         rp.local_font = &owner_->getFirstFontSettings(bp);
1483                                         latexArgInsets(*owner_, os, rp, layout_->postcommandargs(), "post:");
1484                                 }
1485                         }
1486                         string const snippet = to_utf8(ods.str());
1487                         features.addPreambleSnippet(snippet);
1488                 }
1489         }
1490
1491         if (features.runparams().flavor == OutputParams::HTML
1492             && layout_->htmltitle()) {
1493                 features.setHTMLTitle(owner_->asString(AS_STR_INSETS | AS_STR_SKIPDELETE));
1494         }
1495
1496         // check the params.
1497         if (!params_.spacing().isDefault())
1498                 features.require("setspace");
1499
1500         // then the layouts
1501         features.useLayout(layout_->name());
1502
1503         // then the fonts
1504         fontlist_.validate(features);
1505
1506         // then the indentation
1507         if (!params_.leftIndent().zero())
1508                 features.require("ParagraphLeftIndent");
1509
1510         // then the insets
1511         InsetList::const_iterator icit = insetlist_.begin();
1512         InsetList::const_iterator iend = insetlist_.end();
1513         for (; icit != iend; ++icit) {
1514                 if (icit->inset) {
1515                         icit->inset->validate(features);
1516                         if (layout_->needprotect &&
1517                             icit->inset->lyxCode() == FOOT_CODE)
1518                                 features.require("NeedLyXFootnoteCode");
1519                 }
1520         }
1521
1522         // then the contents
1523         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1524                 for (size_t pnr = 0; pnr < phrases_nr; ++pnr) {
1525                         if (!special_phrases[pnr].builtin
1526                             && isTextAt(special_phrases[pnr].phrase, i)) {
1527                                 features.require(special_phrases[pnr].phrase);
1528                                 break;
1529                         }
1530                 }
1531                 Encodings::validate(text_[i], features);
1532         }
1533 }
1534
1535 /////////////////////////////////////////////////////////////////////
1536 //
1537 // Paragraph
1538 //
1539 /////////////////////////////////////////////////////////////////////
1540
1541 namespace {
1542         Layout const emptyParagraphLayout;
1543 }
1544
1545 Paragraph::Paragraph()
1546         : d(new Paragraph::Private(this, emptyParagraphLayout))
1547 {
1548         itemdepth = 0;
1549         d->params_.clear();
1550 }
1551
1552
1553 Paragraph::Paragraph(Paragraph const & par)
1554         : itemdepth(par.itemdepth),
1555         d(new Paragraph::Private(*par.d, this))
1556 {
1557         registerWords();
1558 }
1559
1560
1561 Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
1562         : itemdepth(par.itemdepth),
1563         d(new Paragraph::Private(*par.d, this, beg, end))
1564 {
1565         registerWords();
1566 }
1567
1568
1569 Paragraph & Paragraph::operator=(Paragraph const & par)
1570 {
1571         // needed as we will destroy the private part before copying it
1572         if (&par != this) {
1573                 itemdepth = par.itemdepth;
1574
1575                 deregisterWords();
1576                 delete d;
1577                 d = new Private(*par.d, this);
1578                 registerWords();
1579         }
1580         return *this;
1581 }
1582
1583
1584 Paragraph::~Paragraph()
1585 {
1586         deregisterWords();
1587         delete d;
1588 }
1589
1590
1591 namespace {
1592
1593 // this shall be called just before every "os << ..." action.
1594 void flushString(ostream & os, docstring & s)
1595 {
1596         os << to_utf8(s);
1597         s.erase();
1598 }
1599
1600 }
1601
1602
1603 void Paragraph::write(ostream & os, BufferParams const & bparams,
1604         depth_type & dth) const
1605 {
1606         // The beginning or end of a deeper (i.e. nested) area?
1607         if (dth != d->params_.depth()) {
1608                 if (d->params_.depth() > dth) {
1609                         while (d->params_.depth() > dth) {
1610                                 os << "\n\\begin_deeper";
1611                                 ++dth;
1612                         }
1613                 } else {
1614                         while (d->params_.depth() < dth) {
1615                                 os << "\n\\end_deeper";
1616                                 --dth;
1617                         }
1618                 }
1619         }
1620
1621         // First write the layout
1622         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1623
1624         d->params_.write(os);
1625
1626         Font font1(inherit_font, bparams.language);
1627
1628         Change running_change = Change(Change::UNCHANGED);
1629
1630         // this string is used as a buffer to avoid repetitive calls
1631         // to to_utf8(), which turn out to be expensive (JMarc)
1632         docstring write_buffer;
1633
1634         int column = 0;
1635         for (pos_type i = 0; i <= size(); ++i) {
1636
1637                 Change const change = lookupChange(i);
1638                 if (change != running_change)
1639                         flushString(os, write_buffer);
1640                 Changes::lyxMarkChange(os, bparams, column, running_change, change);
1641                 running_change = change;
1642
1643                 if (i == size())
1644                         break;
1645
1646                 // Write font changes
1647                 Font font2 = getFontSettings(bparams, i);
1648                 if (font2 != font1) {
1649                         flushString(os, write_buffer);
1650                         font2.lyxWriteChanges(font1, os);
1651                         column = 0;
1652                         font1 = font2;
1653                 }
1654
1655                 char_type const c = d->text_[i];
1656                 switch (c) {
1657                 case META_INSET:
1658                         if (Inset const * inset = getInset(i)) {
1659                                 flushString(os, write_buffer);
1660                                 if (inset->directWrite()) {
1661                                         // international char, let it write
1662                                         // code directly so it's shorter in
1663                                         // the file
1664                                         inset->write(os);
1665                                 } else {
1666                                         if (i)
1667                                                 os << '\n';
1668                                         os << "\\begin_inset ";
1669                                         inset->write(os);
1670                                         os << "\n\\end_inset\n\n";
1671                                         column = 0;
1672                                 }
1673                         }
1674                         break;
1675                 case '\\':
1676                         flushString(os, write_buffer);
1677                         os << "\n\\backslash\n";
1678                         column = 0;
1679                         break;
1680                 case '.':
1681                         flushString(os, write_buffer);
1682                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1683                                 os << ".\n";
1684                                 column = 0;
1685                         } else
1686                                 os << '.';
1687                         break;
1688                 default:
1689                         if ((column > 70 && c == ' ')
1690                             || column > 79) {
1691                                 flushString(os, write_buffer);
1692                                 os << '\n';
1693                                 column = 0;
1694                         }
1695                         // this check is to amend a bug. LyX sometimes
1696                         // inserts '\0' this could cause problems.
1697                         if (c != '\0')
1698                                 write_buffer.push_back(c);
1699                         else
1700                                 LYXERR0("NUL char in structure.");
1701                         ++column;
1702                         break;
1703                 }
1704         }
1705
1706         flushString(os, write_buffer);
1707         os << "\n\\end_layout\n";
1708 }
1709
1710
1711 void Paragraph::validate(LaTeXFeatures & features) const
1712 {
1713         d->validate(features);
1714 }
1715
1716
1717 void Paragraph::insert(pos_type start, docstring const & str,
1718                        Font const & font, Change const & change)
1719 {
1720         for (size_t i = 0, n = str.size(); i != n ; ++i)
1721                 insertChar(start + i, str[i], font, change);
1722 }
1723
1724
1725 void Paragraph::appendChar(char_type c, Font const & font,
1726                 Change const & change)
1727 {
1728         // track change
1729         d->changes_.insert(change, d->text_.size());
1730         // when appending characters, no need to update tables
1731         d->text_.push_back(c);
1732         setFont(d->text_.size() - 1, font);
1733         d->requestSpellCheck(d->text_.size() - 1);
1734 }
1735
1736
1737 void Paragraph::appendString(docstring const & s, Font const & font,
1738                 Change const & change)
1739 {
1740         pos_type end = s.size();
1741         size_t oldsize = d->text_.size();
1742         size_t newsize = oldsize + end;
1743         size_t capacity = d->text_.capacity();
1744         if (newsize >= capacity)
1745                 d->text_.reserve(max(capacity + 100, newsize));
1746
1747         // when appending characters, no need to update tables
1748         d->text_.append(s);
1749
1750         // FIXME: Optimize this!
1751         for (size_t i = oldsize; i != newsize; ++i) {
1752                 // track change
1753                 d->changes_.insert(change, i);
1754                 d->requestSpellCheck(i);
1755         }
1756         d->fontlist_.set(oldsize, font);
1757         d->fontlist_.set(newsize - 1, font);
1758 }
1759
1760
1761 void Paragraph::insertChar(pos_type pos, char_type c,
1762                            bool trackChanges)
1763 {
1764         d->insertChar(pos, c, Change(trackChanges ?
1765                            Change::INSERTED : Change::UNCHANGED));
1766 }
1767
1768
1769 void Paragraph::insertChar(pos_type pos, char_type c,
1770                            Font const & font, bool trackChanges)
1771 {
1772         d->insertChar(pos, c, Change(trackChanges ?
1773                            Change::INSERTED : Change::UNCHANGED));
1774         setFont(pos, font);
1775 }
1776
1777
1778 void Paragraph::insertChar(pos_type pos, char_type c,
1779                            Font const & font, Change const & change)
1780 {
1781         d->insertChar(pos, c, change);
1782         setFont(pos, font);
1783 }
1784
1785
1786 void Paragraph::resetFonts(Font const & font)
1787 {
1788         d->fontlist_.clear();
1789         d->fontlist_.set(0, font);
1790         d->fontlist_.set(d->text_.size() - 1, font);
1791 }
1792
1793 // Gets uninstantiated font setting at position.
1794 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1795                                          pos_type pos) const
1796 {
1797         if (pos > size()) {
1798                 LYXERR0("pos: " << pos << " size: " << size());
1799                 LBUFERR(false);
1800         }
1801
1802         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1803         if (cit != d->fontlist_.end())
1804                 return cit->font();
1805
1806         if (pos == size() && !empty())
1807                 return getFontSettings(bparams, pos - 1);
1808
1809         // Optimisation: avoid a full font instantiation if there is no
1810         // language change from previous call.
1811         static Font previous_font;
1812         static Language const * previous_lang = 0;
1813         Language const * lang = getParLanguage(bparams);
1814         if (lang != previous_lang) {
1815                 previous_lang = lang;
1816                 previous_font = Font(inherit_font, lang);
1817         }
1818         return previous_font;
1819 }
1820
1821
1822 FontSpan Paragraph::fontSpan(pos_type pos) const
1823 {
1824         LBUFERR(pos < size());
1825
1826         pos_type start = 0;
1827         FontList::const_iterator cit = d->fontlist_.begin();
1828         FontList::const_iterator end = d->fontlist_.end();
1829         for (; cit != end; ++cit) {
1830                 if (cit->pos() >= pos) {
1831                         if (pos >= beginOfBody())
1832                                 return FontSpan(max(start, beginOfBody()),
1833                                                 cit->pos());
1834                         else
1835                                 return FontSpan(start,
1836                                                 min(beginOfBody() - 1,
1837                                                          cit->pos()));
1838                 }
1839                 start = cit->pos() + 1;
1840         }
1841
1842         // This should not happen, but if so, we take no chances.
1843         LYXERR0("Paragraph::fontSpan: position not found in fontinfo table!");
1844         LASSERT(false, return FontSpan(pos, pos));
1845 }
1846
1847
1848 // Gets uninstantiated font setting at position 0
1849 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1850 {
1851         if (!empty() && !d->fontlist_.empty())
1852                 return d->fontlist_.begin()->font();
1853
1854         // Optimisation: avoid a full font instantiation if there is no
1855         // language change from previous call.
1856         static Font previous_font;
1857         static Language const * previous_lang = 0;
1858         if (bparams.language != previous_lang) {
1859                 previous_lang = bparams.language;
1860                 previous_font = Font(inherit_font, bparams.language);
1861         }
1862
1863         return previous_font;
1864 }
1865
1866
1867 // Gets the fully instantiated font at a given position in a paragraph
1868 // This is basically the same function as Text::GetFont() in text2.cpp.
1869 // The difference is that this one is used for generating the LaTeX file,
1870 // and thus cosmetic "improvements" are disallowed: This has to deliver
1871 // the true picture of the buffer. (Asger)
1872 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1873                                  Font const & outerfont) const
1874 {
1875         LBUFERR(pos >= 0);
1876
1877         Font font = getFontSettings(bparams, pos);
1878
1879         pos_type const body_pos = beginOfBody();
1880         FontInfo & fi = font.fontInfo();
1881         if (pos < body_pos)
1882                 fi.realize(d->layout_->labelfont);
1883         else
1884                 fi.realize(d->layout_->font);
1885
1886         fi.realize(outerfont.fontInfo());
1887         fi.realize(bparams.getFont().fontInfo());
1888
1889         return font;
1890 }
1891
1892
1893 Font const Paragraph::getLabelFont
1894         (BufferParams const & bparams, Font const & outerfont) const
1895 {
1896         FontInfo tmpfont = d->layout_->labelfont;
1897         tmpfont.realize(outerfont.fontInfo());
1898         tmpfont.realize(bparams.getFont().fontInfo());
1899         return Font(tmpfont, getParLanguage(bparams));
1900 }
1901
1902
1903 Font const Paragraph::getLayoutFont
1904         (BufferParams const & bparams, Font const & outerfont) const
1905 {
1906         FontInfo tmpfont = d->layout_->font;
1907         tmpfont.realize(outerfont.fontInfo());
1908         tmpfont.realize(bparams.getFont().fontInfo());
1909         return Font(tmpfont, getParLanguage(bparams));
1910 }
1911
1912
1913 /// Returns the height of the highest font in range
1914 FontSize Paragraph::highestFontInRange
1915         (pos_type startpos, pos_type endpos, FontSize def_size) const
1916 {
1917         return d->fontlist_.highestInRange(startpos, endpos, def_size);
1918 }
1919
1920
1921 char_type Paragraph::getUChar(BufferParams const & bparams, pos_type pos) const
1922 {
1923         char_type c = d->text_[pos];
1924         if (!lyxrc.rtl_support || !getFontSettings(bparams, pos).isRightToLeft())
1925                 return c;
1926
1927         // FIXME: The arabic special casing is due to the difference of arabic
1928         // round brackets input introduced in r18599. Check if this should be
1929         // unified with Hebrew or at least if all bracket types should be
1930         // handled the same (file format change in either case).
1931         string const & lang = getFontSettings(bparams, pos).language()->lang();
1932         bool const arabic = lang == "arabic_arabtex" || lang == "arabic_arabi"
1933                 || lang == "farsi";
1934         char_type uc = c;
1935         switch (c) {
1936         case '(':
1937                 uc = arabic ? c : ')';
1938                 break;
1939         case ')':
1940                 uc = arabic ? c : '(';
1941                 break;
1942         case '[':
1943                 uc = ']';
1944                 break;
1945         case ']':
1946                 uc = '[';
1947                 break;
1948         case '{':
1949                 uc = '}';
1950                 break;
1951         case '}':
1952                 uc = '{';
1953                 break;
1954         case '<':
1955                 uc = '>';
1956                 break;
1957         case '>':
1958                 uc = '<';
1959                 break;
1960         }
1961
1962         return uc;
1963 }
1964
1965
1966 void Paragraph::setFont(pos_type pos, Font const & font)
1967 {
1968         LASSERT(pos <= size(), return);
1969
1970         // First, reduce font against layout/label font
1971         // Update: The setCharFont() routine in text2.cpp already
1972         // reduces font, so we don't need to do that here. (Asger)
1973
1974         d->fontlist_.set(pos, font);
1975 }
1976
1977
1978 void Paragraph::makeSameLayout(Paragraph const & par)
1979 {
1980         d->layout_ = par.d->layout_;
1981         d->params_ = par.d->params_;
1982 }
1983
1984
1985 bool Paragraph::stripLeadingSpaces(bool trackChanges)
1986 {
1987         if (isFreeSpacing())
1988                 return false;
1989
1990         int pos = 0;
1991         int count = 0;
1992
1993         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
1994                 if (eraseChar(pos, trackChanges))
1995                         ++count;
1996                 else
1997                         ++pos;
1998         }
1999
2000         return count > 0 || pos > 0;
2001 }
2002
2003
2004 bool Paragraph::hasSameLayout(Paragraph const & par) const
2005 {
2006         return par.d->layout_ == d->layout_
2007                 && d->params_.sameLayout(par.d->params_);
2008 }
2009
2010
2011 depth_type Paragraph::getDepth() const
2012 {
2013         return d->params_.depth();
2014 }
2015
2016
2017 depth_type Paragraph::getMaxDepthAfter() const
2018 {
2019         if (d->layout_->isEnvironment())
2020                 return d->params_.depth() + 1;
2021         else
2022                 return d->params_.depth();
2023 }
2024
2025
2026 char Paragraph::getAlign() const
2027 {
2028         if (d->params_.align() == LYX_ALIGN_LAYOUT)
2029                 return d->layout_->align;
2030         else
2031                 return d->params_.align();
2032 }
2033
2034
2035 docstring const & Paragraph::labelString() const
2036 {
2037         return d->params_.labelString();
2038 }
2039
2040
2041 // the next two functions are for the manual labels
2042 docstring const Paragraph::getLabelWidthString() const
2043 {
2044         if (d->layout_->margintype == MARGIN_MANUAL
2045             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
2046                 return d->params_.labelWidthString();
2047         else
2048                 return _("Senseless with this layout!");
2049 }
2050
2051
2052 void Paragraph::setLabelWidthString(docstring const & s)
2053 {
2054         d->params_.labelWidthString(s);
2055 }
2056
2057
2058 docstring Paragraph::expandLabel(Layout const & layout,
2059                 BufferParams const & bparams) const
2060 {
2061         return expandParagraphLabel(layout, bparams, true);
2062 }
2063
2064
2065 docstring Paragraph::expandDocBookLabel(Layout const & layout,
2066                 BufferParams const & bparams) const
2067 {
2068         return expandParagraphLabel(layout, bparams, false);
2069 }
2070
2071
2072 docstring Paragraph::expandParagraphLabel(Layout const & layout,
2073                 BufferParams const & bparams, bool process_appendix) const
2074 {
2075         DocumentClass const & tclass = bparams.documentClass();
2076         string const & lang = getParLanguage(bparams)->code();
2077         bool const in_appendix = process_appendix && d->params_.appendix();
2078         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
2079
2080         if (fmt.empty() && !layout.counter.empty())
2081                 return tclass.counters().theCounter(layout.counter, lang);
2082
2083         // handle 'inherited level parts' in 'fmt',
2084         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
2085         size_t const i = fmt.find('@', 0);
2086         if (i != docstring::npos) {
2087                 size_t const j = fmt.find('@', i + 1);
2088                 if (j != docstring::npos) {
2089                         docstring parent(fmt, i + 1, j - i - 1);
2090                         docstring label = from_ascii("??");
2091                         if (tclass.hasLayout(parent))
2092                                 docstring label = expandParagraphLabel(tclass[parent], bparams,
2093                                                       process_appendix);
2094                         fmt = docstring(fmt, 0, i) + label
2095                                 + docstring(fmt, j + 1, docstring::npos);
2096                 }
2097         }
2098
2099         return tclass.counters().counterLabel(fmt, lang);
2100 }
2101
2102
2103 void Paragraph::applyLayout(Layout const & new_layout)
2104 {
2105         d->layout_ = &new_layout;
2106         LyXAlignment const oldAlign = d->params_.align();
2107
2108         if (!(oldAlign & d->layout_->alignpossible)) {
2109                 frontend::Alert::warning(_("Alignment not permitted"),
2110                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
2111                 d->params_.align(LYX_ALIGN_LAYOUT);
2112         }
2113 }
2114
2115
2116 pos_type Paragraph::beginOfBody() const
2117 {
2118         return d->begin_of_body_;
2119 }
2120
2121
2122 void Paragraph::setBeginOfBody()
2123 {
2124         if (d->layout_->labeltype != LABEL_MANUAL) {
2125                 d->begin_of_body_ = 0;
2126                 return;
2127         }
2128
2129         // Unroll the first two cycles of the loop
2130         // and remember the previous character to
2131         // remove unnecessary getChar() calls
2132         pos_type i = 0;
2133         pos_type end = size();
2134         if (i < end && !isNewline(i)) {
2135                 ++i;
2136                 char_type previous_char = 0;
2137                 char_type temp = 0;
2138                 if (i < end) {
2139                         previous_char = d->text_[i];
2140                         if (!isNewline(i)) {
2141                                 ++i;
2142                                 while (i < end && previous_char != ' ') {
2143                                         temp = d->text_[i];
2144                                         if (isNewline(i))
2145                                                 break;
2146                                         ++i;
2147                                         previous_char = temp;
2148                                 }
2149                         }
2150                 }
2151         }
2152
2153         d->begin_of_body_ = i;
2154 }
2155
2156
2157 bool Paragraph::allowParagraphCustomization() const
2158 {
2159         return inInset().allowParagraphCustomization();
2160 }
2161
2162
2163 bool Paragraph::usePlainLayout() const
2164 {
2165         return inInset().usePlainLayout();
2166 }
2167
2168
2169 bool Paragraph::isPassThru() const
2170 {
2171         return inInset().isPassThru() || d->layout_->pass_thru;
2172 }
2173
2174 namespace {
2175
2176 // paragraphs inside floats need different alignment tags to avoid
2177 // unwanted space
2178
2179 bool noTrivlistCentering(InsetCode code)
2180 {
2181         return code == FLOAT_CODE
2182                || code == WRAP_CODE
2183                || code == CELL_CODE;
2184 }
2185
2186
2187 string correction(string const & orig)
2188 {
2189         if (orig == "flushleft")
2190                 return "raggedright";
2191         if (orig == "flushright")
2192                 return "raggedleft";
2193         if (orig == "center")
2194                 return "centering";
2195         return orig;
2196 }
2197
2198
2199 string const corrected_env(string const & suffix, string const & env,
2200         InsetCode code, bool const lastpar)
2201 {
2202         string output = suffix + "{";
2203         if (noTrivlistCentering(code)) {
2204                 if (lastpar) {
2205                         // the last paragraph in non-trivlist-aligned
2206                         // context is special (to avoid unwanted whitespace)
2207                         if (suffix == "\\begin")
2208                                 return "\\" + correction(env) + "{}";
2209                         return string();
2210                 }
2211                 output += correction(env);
2212         } else
2213                 output += env;
2214         output += "}";
2215         if (suffix == "\\begin")
2216                 output += "\n";
2217         return output;
2218 }
2219
2220
2221 void adjust_column(string const & str, int & column)
2222 {
2223         if (!contains(str, "\n"))
2224                 column += str.size();
2225         else {
2226                 string tmp;
2227                 column = rsplit(str, tmp, '\n').size();
2228         }
2229 }
2230
2231 } // namespace anon
2232
2233
2234 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
2235                         otexstream & os, OutputParams const & runparams) const
2236 {
2237         int column = 0;
2238
2239         if (params_.noindent() && !layout_->pass_thru) {
2240                 os << "\\noindent ";
2241                 column += 10;
2242         }
2243
2244         LyXAlignment const curAlign = params_.align();
2245
2246         if (curAlign == layout_->align)
2247                 return column;
2248
2249         switch (curAlign) {
2250         case LYX_ALIGN_NONE:
2251         case LYX_ALIGN_BLOCK:
2252         case LYX_ALIGN_LAYOUT:
2253         case LYX_ALIGN_SPECIAL:
2254         case LYX_ALIGN_DECIMAL:
2255                 break;
2256         case LYX_ALIGN_LEFT:
2257         case LYX_ALIGN_RIGHT:
2258         case LYX_ALIGN_CENTER:
2259                 if (runparams.moving_arg) {
2260                         os << "\\protect";
2261                         column += 8;
2262                 }
2263                 break;
2264         }
2265
2266         string const begin_tag = "\\begin";
2267         InsetCode code = ownerCode();
2268         bool const lastpar = runparams.isLastPar;
2269
2270         switch (curAlign) {
2271         case LYX_ALIGN_NONE:
2272         case LYX_ALIGN_BLOCK:
2273         case LYX_ALIGN_LAYOUT:
2274         case LYX_ALIGN_SPECIAL:
2275         case LYX_ALIGN_DECIMAL:
2276                 break;
2277         case LYX_ALIGN_LEFT: {
2278                 string output;
2279                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2280                         output = corrected_env(begin_tag, "flushleft", code, lastpar);
2281                 else
2282                         output = corrected_env(begin_tag, "flushright", code, lastpar);
2283                 os << from_ascii(output);
2284                 adjust_column(output, column);
2285                 break;
2286         } case LYX_ALIGN_RIGHT: {
2287                 string output;
2288                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2289                         output = corrected_env(begin_tag, "flushright", code, lastpar);
2290                 else
2291                         output = corrected_env(begin_tag, "flushleft", code, lastpar);
2292                 os << from_ascii(output);
2293                 adjust_column(output, column);
2294                 break;
2295         } case LYX_ALIGN_CENTER: {
2296                 string output;
2297                 output = corrected_env(begin_tag, "center", code, lastpar);
2298                 os << from_ascii(output);
2299                 adjust_column(output, column);
2300                 break;
2301         }
2302         }
2303
2304         return column;
2305 }
2306
2307
2308 bool Paragraph::Private::endTeXParParams(BufferParams const & bparams,
2309                         otexstream & os, OutputParams const & runparams) const
2310 {
2311         LyXAlignment const curAlign = params_.align();
2312
2313         if (curAlign == layout_->align)
2314                 return false;
2315
2316         switch (curAlign) {
2317         case LYX_ALIGN_NONE:
2318         case LYX_ALIGN_BLOCK:
2319         case LYX_ALIGN_LAYOUT:
2320         case LYX_ALIGN_SPECIAL:
2321         case LYX_ALIGN_DECIMAL:
2322                 break;
2323         case LYX_ALIGN_LEFT:
2324         case LYX_ALIGN_RIGHT:
2325         case LYX_ALIGN_CENTER:
2326                 if (runparams.moving_arg)
2327                         os << "\\protect";
2328                 break;
2329         }
2330
2331         string output;
2332         string const end_tag = "\n\\par\\end";
2333         InsetCode code = ownerCode();
2334         bool const lastpar = runparams.isLastPar;
2335
2336         switch (curAlign) {
2337         case LYX_ALIGN_NONE:
2338         case LYX_ALIGN_BLOCK:
2339         case LYX_ALIGN_LAYOUT:
2340         case LYX_ALIGN_SPECIAL:
2341         case LYX_ALIGN_DECIMAL:
2342                 break;
2343         case LYX_ALIGN_LEFT: {
2344                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2345                         output = corrected_env(end_tag, "flushleft", code, lastpar);
2346                 else
2347                         output = corrected_env(end_tag, "flushright", code, lastpar);
2348                 os << from_ascii(output);
2349                 break;
2350         } case LYX_ALIGN_RIGHT: {
2351                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2352                         output = corrected_env(end_tag, "flushright", code, lastpar);
2353                 else
2354                         output = corrected_env(end_tag, "flushleft", code, lastpar);
2355                 os << from_ascii(output);
2356                 break;
2357         } case LYX_ALIGN_CENTER: {
2358                 output = corrected_env(end_tag, "center", code, lastpar);
2359                 os << from_ascii(output);
2360                 break;
2361         }
2362         }
2363
2364         return !output.empty() || lastpar;
2365 }
2366
2367
2368 // This one spits out the text of the paragraph
2369 void Paragraph::latex(BufferParams const & bparams,
2370         Font const & outerfont,
2371         otexstream & os,
2372         OutputParams const & runparams,
2373         int start_pos, int end_pos, bool force) const
2374 {
2375         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
2376
2377         // FIXME This check should not be needed. Perhaps issue an
2378         // error if it triggers.
2379         Layout const & style = inInset().forcePlainLayout() ?
2380                 bparams.documentClass().plainLayout() : *d->layout_;
2381
2382         if (!force && style.inpreamble)
2383                 return;
2384
2385         bool const allowcust = allowParagraphCustomization();
2386
2387         // Current base font for all inherited font changes, without any
2388         // change caused by an individual character, except for the language:
2389         // It is set to the language of the first character.
2390         // As long as we are in the label, this font is the base font of the
2391         // label. Before the first body character it is set to the base font
2392         // of the body.
2393         Font basefont;
2394
2395         // Maybe we have to create a optional argument.
2396         pos_type body_pos = beginOfBody();
2397         unsigned int column = 0;
2398
2399         if (body_pos > 0) {
2400                 // the optional argument is kept in curly brackets in
2401                 // case it contains a ']'
2402                 // This is not strictly needed, but if this is changed it
2403                 // would be a file format change, and tex2lyx would need
2404                 // to be adjusted, since it unconditionally removes the
2405                 // braces when it parses \item.
2406                 os << "[{";
2407                 column += 2;
2408                 basefont = getLabelFont(bparams, outerfont);
2409         } else {
2410                 basefont = getLayoutFont(bparams, outerfont);
2411         }
2412
2413         // Which font is currently active?
2414         Font running_font(basefont);
2415         // Do we have an open font change?
2416         bool open_font = false;
2417
2418         Change runningChange = Change(Change::UNCHANGED);
2419
2420         Encoding const * const prev_encoding = runparams.encoding;
2421
2422         os.texrow().start(id(), 0);
2423
2424         // if the paragraph is empty, the loop will not be entered at all
2425         if (empty()) {
2426                 if (style.isCommand()) {
2427                         os << '{';
2428                         ++column;
2429                 }
2430                 if (!style.leftdelim().empty()) {
2431                         os << style.leftdelim();
2432                         column += style.leftdelim().size();
2433                 }
2434                 if (allowcust)
2435                         column += d->startTeXParParams(bparams, os, runparams);
2436         }
2437
2438         for (pos_type i = 0; i < size(); ++i) {
2439                 // First char in paragraph or after label?
2440                 if (i == body_pos) {
2441                         if (body_pos > 0) {
2442                                 if (open_font) {
2443                                         column += running_font.latexWriteEndChanges(
2444                                                 os, bparams, runparams,
2445                                                 basefont, basefont);
2446                                         open_font = false;
2447                                 }
2448                                 basefont = getLayoutFont(bparams, outerfont);
2449                                 running_font = basefont;
2450
2451                                 column += Changes::latexMarkChange(os, bparams,
2452                                                 runningChange, Change(Change::UNCHANGED),
2453                                                 runparams);
2454                                 runningChange = Change(Change::UNCHANGED);
2455
2456                                 os << "}] ";
2457                                 column +=3;
2458                         }
2459                         if (style.isCommand()) {
2460                                 os << '{';
2461                                 ++column;
2462                         }
2463
2464                         if (!style.leftdelim().empty()) {
2465                                 os << style.leftdelim();
2466                                 column += style.leftdelim().size();
2467                         }
2468
2469                         if (allowcust)
2470                                 column += d->startTeXParParams(bparams, os,
2471                                                             runparams);
2472                 }
2473
2474                 Change const & change = runparams.inDeletedInset
2475                         ? runparams.changeOfDeletedInset : lookupChange(i);
2476
2477                 if (bparams.outputChanges && runningChange != change) {
2478                         if (open_font) {
2479                                 column += running_font.latexWriteEndChanges(
2480                                                 os, bparams, runparams, basefont, basefont);
2481                                 open_font = false;
2482                         }
2483                         basefont = getLayoutFont(bparams, outerfont);
2484                         running_font = basefont;
2485
2486                         column += Changes::latexMarkChange(os, bparams, runningChange,
2487                                                            change, runparams);
2488                         runningChange = change;
2489                 }
2490
2491                 // do not output text which is marked deleted
2492                 // if change tracking output is disabled
2493                 if (!bparams.outputChanges && change.deleted()) {
2494                         continue;
2495                 }
2496
2497                 ++column;
2498
2499                 // Fully instantiated font
2500                 Font const font = getFont(bparams, i, outerfont);
2501
2502                 Font const last_font = running_font;
2503
2504                 // Do we need to close the previous font?
2505                 if (open_font &&
2506                     (font != running_font ||
2507                      font.language() != running_font.language()))
2508                 {
2509                         column += running_font.latexWriteEndChanges(
2510                                         os, bparams, runparams, basefont,
2511                                         (i == body_pos-1) ? basefont : font);
2512                         running_font = basefont;
2513                         open_font = false;
2514                 }
2515
2516                 string const running_lang = runparams.use_polyglossia ?
2517                         running_font.language()->polyglossia() : running_font.language()->babel();
2518                 // close babel's font environment before opening CJK.
2519                 string const lang_end_command = runparams.use_polyglossia ?
2520                         "\\end{$$lang}" : lyxrc.language_command_end;
2521                 if (!running_lang.empty() &&
2522                     font.language()->encoding()->package() == Encoding::CJK) {
2523                                 string end_tag = subst(lang_end_command,
2524                                                         "$$lang",
2525                                                         running_lang);
2526                                 os << from_ascii(end_tag);
2527                                 column += end_tag.length();
2528                 }
2529
2530                 // Switch file encoding if necessary (and allowed)
2531                 if (!runparams.pass_thru && !style.pass_thru &&
2532                     runparams.encoding->package() != Encoding::none &&
2533                     font.language()->encoding()->package() != Encoding::none) {
2534                         pair<bool, int> const enc_switch =
2535                                 switchEncoding(os.os(), bparams, runparams,
2536                                         *(font.language()->encoding()));
2537                         if (enc_switch.first) {
2538                                 column += enc_switch.second;
2539                                 runparams.encoding = font.language()->encoding();
2540                         }
2541                 }
2542
2543                 char_type const c = d->text_[i];
2544
2545                 // Do we need to change font?
2546                 if ((font != running_font ||
2547                      font.language() != running_font.language()) &&
2548                         i != body_pos - 1)
2549                 {
2550                         odocstringstream ods;
2551                         column += font.latexWriteStartChanges(ods, bparams,
2552                                                               runparams, basefont,
2553                                                               last_font);
2554                         running_font = font;
2555                         open_font = true;
2556                         docstring fontchange = ods.str();
2557                         // check whether the fontchange ends with a \\textcolor
2558                         // modifier and the text starts with a space (bug 4473)
2559                         docstring const last_modifier = rsplit(fontchange, '\\');
2560                         if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2561                                 os << fontchange << from_ascii("{}");
2562                         // check if the fontchange ends with a trailing blank
2563                         // (like "\small " (see bug 3382)
2564                         else if (suffixIs(fontchange, ' ') && c == ' ')
2565                                 os << fontchange.substr(0, fontchange.size() - 1)
2566                                    << from_ascii("{}");
2567                         else
2568                                 os << fontchange;
2569                 }
2570
2571                 // FIXME: think about end_pos implementation...
2572                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2573                         // FIXME: integrate this case in latexSpecialChar
2574                         // Do not print the separation of the optional argument
2575                         // if style.pass_thru is false. This works because
2576                         // latexSpecialChar ignores spaces if
2577                         // style.pass_thru is false.
2578                         if (i != body_pos - 1) {
2579                                 if (d->simpleTeXBlanks(runparams, os,
2580                                                 i, column, font, style)) {
2581                                         // A surrogate pair was output. We
2582                                         // must not call latexSpecialChar
2583                                         // in this iteration, since it would output
2584                                         // the combining character again.
2585                                         ++i;
2586                                         continue;
2587                                 }
2588                         }
2589                 }
2590
2591                 OutputParams rp = runparams;
2592                 rp.free_spacing = style.free_spacing;
2593                 rp.local_font = &font;
2594                 rp.intitle = style.intitle;
2595
2596                 // Two major modes:  LaTeX or plain
2597                 // Handle here those cases common to both modes
2598                 // and then split to handle the two modes separately.
2599                 if (c == META_INSET) {
2600                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2601                                 d->latexInset(bparams, os, rp, running_font,
2602                                                 basefont, outerfont, open_font,
2603                                                 runningChange, style, i, column);
2604                         }
2605                 } else {
2606                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2607                                 try {
2608                                         d->latexSpecialChar(os, bparams, rp, running_font, runningChange,
2609                                                             style, i, end_pos, column);
2610                                 } catch (EncodingException & e) {
2611                                 if (runparams.dryrun) {
2612                                         os << "<" << _("LyX Warning: ")
2613                                            << _("uncodable character") << " '";
2614                                         os.put(c);
2615                                         os << "'>";
2616                                 } else {
2617                                         // add location information and throw again.
2618                                         e.par_id = id();
2619                                         e.pos = i;
2620                                         throw(e);
2621                                 }
2622                         }
2623                 }
2624                 }
2625
2626                 // Set the encoding to that returned from latexSpecialChar (see
2627                 // comment for encoding member in OutputParams.h)
2628                 runparams.encoding = rp.encoding;
2629         }
2630
2631         // If we have an open font definition, we have to close it
2632         if (open_font) {
2633 #ifdef FIXED_LANGUAGE_END_DETECTION
2634                 if (next_) {
2635                         running_font.latexWriteEndChanges(os, bparams,
2636                                         runparams, basefont,
2637                                         next_->getFont(bparams, 0, outerfont));
2638                 } else {
2639                         running_font.latexWriteEndChanges(os, bparams,
2640                                         runparams, basefont, basefont);
2641                 }
2642 #else
2643 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2644 //FIXME: there as we start another \selectlanguage with the next paragraph if
2645 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2646                 running_font.latexWriteEndChanges(os, bparams, runparams,
2647                                 basefont, basefont);
2648 #endif
2649         }
2650
2651         column += Changes::latexMarkChange(os, bparams, runningChange,
2652                                            Change(Change::UNCHANGED), runparams);
2653
2654         // Needed if there is an optional argument but no contents.
2655         if (body_pos > 0 && body_pos == size()) {
2656                 os << "}]~";
2657         }
2658
2659         if (!style.rightdelim().empty()) {
2660                 os << style.rightdelim();
2661                 column += style.rightdelim().size();
2662         }
2663
2664         if (allowcust && d->endTeXParParams(bparams, os, runparams)
2665             && runparams.encoding != prev_encoding) {
2666                 runparams.encoding = prev_encoding;
2667                 if (!runparams.isFullUnicode())
2668                         os << setEncoding(prev_encoding->iconvName());
2669         }
2670
2671         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2672 }
2673
2674
2675 bool Paragraph::emptyTag() const
2676 {
2677         for (pos_type i = 0; i < size(); ++i) {
2678                 if (Inset const * inset = getInset(i)) {
2679                         InsetCode lyx_code = inset->lyxCode();
2680                         // FIXME testing like that is wrong. What is
2681                         // the intent?
2682                         if (lyx_code != TOC_CODE &&
2683                             lyx_code != INCLUDE_CODE &&
2684                             lyx_code != GRAPHICS_CODE &&
2685                             lyx_code != ERT_CODE &&
2686                             lyx_code != LISTINGS_CODE &&
2687                             lyx_code != FLOAT_CODE &&
2688                             lyx_code != TABULAR_CODE) {
2689                                 return false;
2690                         }
2691                 } else {
2692                         char_type c = d->text_[i];
2693                         if (c != ' ' && c != '\t')
2694                                 return false;
2695                 }
2696         }
2697         return true;
2698 }
2699
2700
2701 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
2702         const
2703 {
2704         for (pos_type i = 0; i < size(); ++i) {
2705                 if (Inset const * inset = getInset(i)) {
2706                         InsetCode lyx_code = inset->lyxCode();
2707                         if (lyx_code == LABEL_CODE) {
2708                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2709                                 docstring const & id = il->getParam("name");
2710                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2711                         }
2712                 }
2713         }
2714         return string();
2715 }
2716
2717
2718 pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runparams)
2719         const
2720 {
2721         pos_type i;
2722         for (i = 0; i < size(); ++i) {
2723                 if (Inset const * inset = getInset(i)) {
2724                         inset->docbook(os, runparams);
2725                 } else {
2726                         char_type c = d->text_[i];
2727                         if (c == ' ')
2728                                 break;
2729                         os << sgml::escapeChar(c);
2730                 }
2731         }
2732         return i;
2733 }
2734
2735
2736 pos_type Paragraph::firstWordLyXHTML(XHTMLStream & xs, OutputParams const & runparams)
2737         const
2738 {
2739         pos_type i;
2740         for (i = 0; i < size(); ++i) {
2741                 if (Inset const * inset = getInset(i)) {
2742                         inset->xhtml(xs, runparams);
2743                 } else {
2744                         char_type c = d->text_[i];
2745                         if (c == ' ')
2746                                 break;
2747                         xs << c;
2748                 }
2749         }
2750         return i;
2751 }
2752
2753
2754 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2755 {
2756         Font font_old;
2757         pos_type size = text_.size();
2758         for (pos_type i = initial; i < size; ++i) {
2759                 Font font = owner_->getFont(buf.params(), i, outerfont);
2760                 if (text_[i] == META_INSET)
2761                         return false;
2762                 if (i != initial && font != font_old)
2763                         return false;
2764                 font_old = font;
2765         }
2766
2767         return true;
2768 }
2769
2770
2771 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
2772                                     odocstream & os,
2773                                     OutputParams const & runparams,
2774                                     Font const & outerfont,
2775                                     pos_type initial) const
2776 {
2777         bool emph_flag = false;
2778
2779         Layout const & style = *d->layout_;
2780         FontInfo font_old =
2781                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2782
2783         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2784                 os << "]]>";
2785
2786         // parsing main loop
2787         for (pos_type i = initial; i < size(); ++i) {
2788                 Font font = getFont(buf.params(), i, outerfont);
2789
2790                 // handle <emphasis> tag
2791                 if (font_old.emph() != font.fontInfo().emph()) {
2792                         if (font.fontInfo().emph() == FONT_ON) {
2793                                 os << "<emphasis>";
2794                                 emph_flag = true;
2795                         } else if (i != initial) {
2796                                 os << "</emphasis>";
2797                                 emph_flag = false;
2798                         }
2799                 }
2800
2801                 if (Inset const * inset = getInset(i)) {
2802                         inset->docbook(os, runparams);
2803                 } else {
2804                         char_type c = d->text_[i];
2805
2806                         if (style.pass_thru)
2807                                 os.put(c);
2808                         else
2809                                 os << sgml::escapeChar(c);
2810                 }
2811                 font_old = font.fontInfo();
2812         }
2813
2814         if (emph_flag) {
2815                 os << "</emphasis>";
2816         }
2817
2818         if (style.free_spacing)
2819                 os << '\n';
2820         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2821                 os << "<![CDATA[";
2822 }
2823
2824
2825 namespace {
2826 void doFontSwitch(vector<html::FontTag> & tagsToOpen,
2827                   vector<html::EndFontTag> & tagsToClose,
2828                   bool & flag, FontState curstate, html::FontTypes type)
2829 {
2830         if (curstate == FONT_ON) {
2831                 tagsToOpen.push_back(html::FontTag(type));
2832                 flag = true;
2833         } else if (flag) {
2834                 tagsToClose.push_back(html::EndFontTag(type));
2835                 flag = false;
2836         }
2837 }
2838 }
2839
2840
2841 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
2842                                     XHTMLStream & xs,
2843                                     OutputParams const & runparams,
2844                                     Font const & outerfont,
2845                                     pos_type initial) const
2846 {
2847         docstring retval;
2848
2849         // track whether we have opened these tags
2850         bool emph_flag = false;
2851         bool bold_flag = false;
2852         bool noun_flag = false;
2853         bool ubar_flag = false;
2854         bool dbar_flag = false;
2855         bool sout_flag = false;
2856         bool wave_flag = false;
2857         // shape tags
2858         bool shap_flag = false;
2859         // family tags
2860         bool faml_flag = false;
2861         // size tags
2862         bool size_flag = false;
2863
2864         Layout const & style = *d->layout_;
2865
2866         xs.startParagraph(allowEmpty());
2867
2868         FontInfo font_old =
2869                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2870
2871         FontShape  curr_fs   = INHERIT_SHAPE;
2872         FontFamily curr_fam  = INHERIT_FAMILY;
2873         FontSize   curr_size = FONT_SIZE_INHERIT;
2874
2875         string const default_family =
2876                 buf.masterBuffer()->params().fonts_default_family;
2877
2878         vector<html::FontTag> tagsToOpen;
2879         vector<html::EndFontTag> tagsToClose;
2880
2881         // parsing main loop
2882         for (pos_type i = initial; i < size(); ++i) {
2883                 // let's not show deleted material in the output
2884                 if (isDeleted(i))
2885                         continue;
2886
2887                 Font const font = getFont(buf.masterBuffer()->params(), i, outerfont);
2888
2889                 // emphasis
2890                 FontState curstate = font.fontInfo().emph();
2891                 if (font_old.emph() != curstate)
2892                         doFontSwitch(tagsToOpen, tagsToClose, emph_flag, curstate, html::FT_EMPH);
2893
2894                 // noun
2895                 curstate = font.fontInfo().noun();
2896                 if (font_old.noun() != curstate)
2897                         doFontSwitch(tagsToOpen, tagsToClose, noun_flag, curstate, html::FT_NOUN);
2898
2899                 // underbar
2900                 curstate = font.fontInfo().underbar();
2901                 if (font_old.underbar() != curstate)
2902                         doFontSwitch(tagsToOpen, tagsToClose, ubar_flag, curstate, html::FT_UBAR);
2903
2904                 // strikeout
2905                 curstate = font.fontInfo().strikeout();
2906                 if (font_old.strikeout() != curstate)
2907                         doFontSwitch(tagsToOpen, tagsToClose, sout_flag, curstate, html::FT_SOUT);
2908
2909                 // double underbar
2910                 curstate = font.fontInfo().uuline();
2911                 if (font_old.uuline() != curstate)
2912                         doFontSwitch(tagsToOpen, tagsToClose, dbar_flag, curstate, html::FT_DBAR);
2913
2914                 // wavy line
2915                 curstate = font.fontInfo().uwave();
2916                 if (font_old.uwave() != curstate)
2917                         doFontSwitch(tagsToOpen, tagsToClose, wave_flag, curstate, html::FT_WAVE);
2918
2919                 // bold
2920                 // a little hackish, but allows us to reuse what we have.
2921                 curstate = (font.fontInfo().series() == BOLD_SERIES ? FONT_ON : FONT_OFF);
2922                 if (font_old.series() != font.fontInfo().series())
2923                         doFontSwitch(tagsToOpen, tagsToClose, bold_flag, curstate, html::FT_BOLD);
2924
2925                 // Font shape
2926                 curr_fs = font.fontInfo().shape();
2927                 FontShape old_fs = font_old.shape();
2928                 if (old_fs != curr_fs) {
2929                         if (shap_flag) {
2930                                 switch (old_fs) {
2931                                 case ITALIC_SHAPE:
2932                                         tagsToClose.push_back(html::EndFontTag(html::FT_ITALIC));
2933                                         break;
2934                                 case SLANTED_SHAPE:
2935                                         tagsToClose.push_back(html::EndFontTag(html::FT_SLANTED));
2936                                         break;
2937                                 case SMALLCAPS_SHAPE:
2938                                         tagsToClose.push_back(html::EndFontTag(html::FT_SMALLCAPS));
2939                                         break;
2940                                 case UP_SHAPE:
2941                                 case INHERIT_SHAPE:
2942                                         break;
2943                                 default:
2944                                         // the other tags are for internal use
2945                                         LATTEST(false);
2946                                         break;
2947                                 }
2948                                 shap_flag = false;
2949                         }
2950                         switch (curr_fs) {
2951                         case ITALIC_SHAPE:
2952                                 tagsToOpen.push_back(html::FontTag(html::FT_ITALIC));
2953                                 shap_flag = true;
2954                                 break;
2955                         case SLANTED_SHAPE:
2956                                 tagsToOpen.push_back(html::FontTag(html::FT_SLANTED));
2957                                 shap_flag = true;
2958                                 break;
2959                         case SMALLCAPS_SHAPE:
2960                                 tagsToOpen.push_back(html::FontTag(html::FT_SMALLCAPS));
2961                                 shap_flag = true;
2962                                 break;
2963                         case UP_SHAPE:
2964                         case INHERIT_SHAPE:
2965                                 break;
2966                         default:
2967                                 // the other tags are for internal use
2968                                 LATTEST(false);
2969                                 break;
2970                         }
2971                 }
2972
2973                 // Font family
2974                 curr_fam = font.fontInfo().family();
2975                 FontFamily old_fam = font_old.family();
2976                 if (old_fam != curr_fam) {
2977                         if (faml_flag) {
2978                                 switch (old_fam) {
2979                                 case ROMAN_FAMILY:
2980                                         tagsToClose.push_back(html::EndFontTag(html::FT_ROMAN));
2981                                         break;
2982                                 case SANS_FAMILY:
2983                                         tagsToClose.push_back(html::EndFontTag(html::FT_SANS));
2984                                         break;
2985                                 case TYPEWRITER_FAMILY:
2986                                         tagsToClose.push_back(html::EndFontTag(html::FT_TYPE));
2987                                         break;
2988                                 case INHERIT_FAMILY:
2989                                         break;
2990                                 default:
2991                                         // the other tags are for internal use
2992                                         LATTEST(false);
2993                                         break;
2994                                 }
2995                                 faml_flag = false;
2996                         }
2997                         switch (curr_fam) {
2998                         case ROMAN_FAMILY:
2999                                 // we will treat a "default" font family as roman, since we have
3000                                 // no other idea what to do.
3001                                 if (default_family != "rmdefault" && default_family != "default") {
3002                                         tagsToOpen.push_back(html::FontTag(html::FT_ROMAN));
3003                                         faml_flag = true;
3004                                 }
3005                                 break;
3006                         case SANS_FAMILY:
3007                                 if (default_family != "sfdefault") {
3008                                         tagsToOpen.push_back(html::FontTag(html::FT_SANS));
3009                                         faml_flag = true;
3010                                 }
3011                                 break;
3012                         case TYPEWRITER_FAMILY:
3013                                 if (default_family != "ttdefault") {
3014                                         tagsToOpen.push_back(html::FontTag(html::FT_TYPE));
3015                                         faml_flag = true;
3016                                 }
3017                                 break;
3018                         case INHERIT_FAMILY:
3019                                 break;
3020                         default:
3021                                 // the other tags are for internal use
3022                                 LATTEST(false);
3023                                 break;
3024                         }
3025                 }
3026
3027                 // Font size
3028                 curr_size = font.fontInfo().size();
3029                 FontSize old_size = font_old.size();
3030                 if (old_size != curr_size) {
3031                         if (size_flag) {
3032                                 switch (old_size) {
3033                                 case FONT_SIZE_TINY:
3034                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_TINY));
3035                                         break;
3036                                 case FONT_SIZE_SCRIPT:
3037                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SCRIPT));
3038                                         break;
3039                                 case FONT_SIZE_FOOTNOTE:
3040                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_FOOTNOTE));
3041                                         break;
3042                                 case FONT_SIZE_SMALL:
3043                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SMALL));
3044                                         break;
3045                                 case FONT_SIZE_LARGE:
3046                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGE));
3047                                         break;
3048                                 case FONT_SIZE_LARGER:
3049                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGER));
3050                                         break;
3051                                 case FONT_SIZE_LARGEST:
3052                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGEST));
3053                                         break;
3054                                 case FONT_SIZE_HUGE:
3055                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGE));
3056                                         break;
3057                                 case FONT_SIZE_HUGER:
3058                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGER));
3059                                         break;
3060                                 case FONT_SIZE_INCREASE:
3061                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_INCREASE));
3062                                         break;
3063                                 case FONT_SIZE_DECREASE:
3064                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_DECREASE));
3065                                         break;
3066                                 case FONT_SIZE_INHERIT:
3067                                 case FONT_SIZE_NORMAL:
3068                                         break;
3069                                 default:
3070                                         // the other tags are for internal use
3071                                         LATTEST(false);
3072                                         break;
3073                                 }
3074                                 size_flag = false;
3075                         }
3076                         switch (curr_size) {
3077                         case FONT_SIZE_TINY:
3078                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_TINY));
3079                                 size_flag = true;
3080                                 break;
3081                         case FONT_SIZE_SCRIPT:
3082                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SCRIPT));
3083                                 size_flag = true;
3084                                 break;
3085                         case FONT_SIZE_FOOTNOTE:
3086                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_FOOTNOTE));
3087                                 size_flag = true;
3088                                 break;
3089                         case FONT_SIZE_SMALL:
3090                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SMALL));
3091                                 size_flag = true;
3092                                 break;
3093                         case FONT_SIZE_LARGE:
3094                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGE));
3095                                 size_flag = true;
3096                                 break;
3097                         case FONT_SIZE_LARGER:
3098                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGER));
3099                                 size_flag = true;
3100                                 break;
3101                         case FONT_SIZE_LARGEST:
3102                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGEST));
3103                                 size_flag = true;
3104                                 break;
3105                         case FONT_SIZE_HUGE:
3106                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGE));
3107                                 size_flag = true;
3108                                 break;
3109                         case FONT_SIZE_HUGER:
3110                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGER));
3111                                 size_flag = true;
3112                                 break;
3113                         case FONT_SIZE_INCREASE:
3114                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_INCREASE));
3115                                 size_flag = true;
3116                                 break;
3117                         case FONT_SIZE_DECREASE:
3118                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_DECREASE));
3119                                 size_flag = true;
3120                                 break;
3121                         case FONT_SIZE_NORMAL:
3122                         case FONT_SIZE_INHERIT:
3123                                 break;
3124                         default:
3125                                 // the other tags are for internal use
3126                                 LATTEST(false);
3127                                 break;
3128                         }
3129                 }
3130
3131                 // FIXME XHTML
3132                 // Other such tags? What about the other text ranges?
3133
3134                 vector<html::EndFontTag>::const_iterator cit = tagsToClose.begin();
3135                 vector<html::EndFontTag>::const_iterator cen = tagsToClose.end();
3136                 for (; cit != cen; ++cit)
3137                         xs << *cit;
3138
3139                 vector<html::FontTag>::const_iterator sit = tagsToOpen.begin();
3140                 vector<html::FontTag>::const_iterator sen = tagsToOpen.end();
3141                 for (; sit != sen; ++sit)
3142                         xs << *sit;
3143
3144                 tagsToClose.clear();
3145                 tagsToOpen.clear();
3146
3147                 Inset const * inset = getInset(i);
3148                 if (inset) {
3149                         if (!runparams.for_toc || inset->isInToc()) {
3150                                 OutputParams np = runparams;
3151                                 np.local_font = &font;
3152                                 if (!inset->getLayout().htmlisblock())
3153                                         np.html_in_par = true;
3154                                 retval += inset->xhtml(xs, np);
3155                         }
3156                 } else {
3157                         char_type c = getUChar(buf.masterBuffer()->params(), i);
3158
3159                         if (style.pass_thru || runparams.pass_thru)
3160                                 xs << c;
3161                         else if (c == '-') {
3162                                 docstring str;
3163                                 int j = i + 1;
3164                                 if (j < size() && d->text_[j] == '-') {
3165                                         j += 1;
3166                                         if (j < size() && d->text_[j] == '-') {
3167                                                 str += from_ascii("&mdash;");
3168                                                 i += 2;
3169                                         } else {
3170                                                 str += from_ascii("&ndash;");
3171                                                 i += 1;
3172                                         }
3173                                 }
3174                                 else
3175                                         str += c;
3176                                 // We don't want to escape the entities. Note that
3177                                 // it is safe to do this, since str can otherwise
3178                                 // only be "-". E.g., it can't be "<".
3179                                 xs << XHTMLStream::ESCAPE_NONE << str;
3180                         } else
3181                                 xs << c;
3182                 }
3183                 font_old = font.fontInfo();
3184         }
3185
3186         xs.closeFontTags();
3187         xs.endParagraph();
3188         return retval;
3189 }
3190
3191
3192 bool Paragraph::isHfill(pos_type pos) const
3193 {
3194         Inset const * inset = getInset(pos);
3195         return inset && (inset->lyxCode() == SPACE_CODE &&
3196                          inset->isStretchableSpace());
3197 }
3198
3199
3200 bool Paragraph::isNewline(pos_type pos) const
3201 {
3202         Inset const * inset = getInset(pos);
3203         return inset && inset->lyxCode() == NEWLINE_CODE;
3204 }
3205
3206
3207 bool Paragraph::isLineSeparator(pos_type pos) const
3208 {
3209         char_type const c = d->text_[pos];
3210         if (isLineSeparatorChar(c))
3211                 return true;
3212         Inset const * inset = getInset(pos);
3213         return inset && inset->isLineSeparator();
3214 }
3215
3216
3217 bool Paragraph::isWordSeparator(pos_type pos) const
3218 {
3219         if (pos == size())
3220                 return true;
3221         if (Inset const * inset = getInset(pos))
3222                 return !inset->isLetter();
3223         // if we have a hard hyphen (no en- or emdash) or apostrophe
3224         // we pass this to the spell checker
3225         // FIXME: this method is subject to change, visit
3226         // https://bugzilla.mozilla.org/show_bug.cgi?id=355178
3227         // to get an impression how complex this is.
3228         if (isHardHyphenOrApostrophe(pos))
3229                 return false;
3230         char_type const c = d->text_[pos];
3231         // We want to pass the escape chars to the spellchecker
3232         docstring const escape_chars = from_utf8(lyxrc.spellchecker_esc_chars);
3233         return !isLetterChar(c) && !isDigitASCII(c) && !contains(escape_chars, c);
3234 }
3235
3236
3237 bool Paragraph::isHardHyphenOrApostrophe(pos_type pos) const
3238 {
3239         pos_type const psize = size();
3240         if (pos >= psize)
3241                 return false;
3242         char_type const c = d->text_[pos];
3243         if (c != '-' && c != '\'')
3244                 return false;
3245         int nextpos = pos + 1;
3246         int prevpos = pos > 0 ? pos - 1 : 0;
3247         if ((nextpos == psize || isSpace(nextpos))
3248                 && (pos == 0 || isSpace(prevpos)))
3249                 return false;
3250         return c == '\''
3251                 || ((nextpos == psize || d->text_[nextpos] != '-')
3252                 && (pos == 0 || d->text_[prevpos] != '-'));
3253 }
3254
3255
3256 bool Paragraph::isSameSpellRange(pos_type pos1, pos_type pos2) const
3257 {
3258         return pos1 == pos2
3259                 || d->speller_state_.getRange(pos1) == d->speller_state_.getRange(pos2);
3260 }
3261
3262
3263 bool Paragraph::isChar(pos_type pos) const
3264 {
3265         if (Inset const * inset = getInset(pos))
3266                 return inset->isChar();
3267         char_type const c = d->text_[pos];
3268         return !isLetterChar(c) && !isDigitASCII(c) && !lyx::isSpace(c);
3269 }
3270
3271
3272 bool Paragraph::isSpace(pos_type pos) const
3273 {
3274         if (Inset const * inset = getInset(pos))
3275                 return inset->isSpace();
3276         char_type const c = d->text_[pos];
3277         return lyx::isSpace(c);
3278 }
3279
3280
3281 Language const *
3282 Paragraph::getParLanguage(BufferParams const & bparams) const
3283 {
3284         if (!empty())
3285                 return getFirstFontSettings(bparams).language();
3286         // FIXME: we should check the prev par as well (Lgb)
3287         return bparams.language;
3288 }
3289
3290
3291 bool Paragraph::isRTL(BufferParams const & bparams) const
3292 {
3293         return lyxrc.rtl_support
3294                 && getParLanguage(bparams)->rightToLeft()
3295                 && !inInset().getLayout().forceLTR();
3296 }
3297
3298
3299 void Paragraph::changeLanguage(BufferParams const & bparams,
3300                                Language const * from, Language const * to)
3301 {
3302         // change language including dummy font change at the end
3303         for (pos_type i = 0; i <= size(); ++i) {
3304                 Font font = getFontSettings(bparams, i);
3305                 if (font.language() == from) {
3306                         font.setLanguage(to);
3307                         setFont(i, font);
3308                         d->requestSpellCheck(i);
3309                 }
3310         }
3311 }
3312
3313
3314 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
3315 {
3316         Language const * doc_language = bparams.language;
3317         FontList::const_iterator cit = d->fontlist_.begin();
3318         FontList::const_iterator end = d->fontlist_.end();
3319
3320         for (; cit != end; ++cit)
3321                 if (cit->font().language() != ignore_language &&
3322                     cit->font().language() != latex_language &&
3323                     cit->font().language() != doc_language)
3324                         return true;
3325         return false;
3326 }
3327
3328
3329 void Paragraph::getLanguages(std::set<Language const *> & languages) const
3330 {
3331         FontList::const_iterator cit = d->fontlist_.begin();
3332         FontList::const_iterator end = d->fontlist_.end();
3333
3334         for (; cit != end; ++cit) {
3335                 Language const * lang = cit->font().language();
3336                 if (lang != ignore_language &&
3337                     lang != latex_language)
3338                         languages.insert(lang);
3339         }
3340 }
3341
3342
3343 docstring Paragraph::asString(int options) const
3344 {
3345         return asString(0, size(), options);
3346 }
3347
3348
3349 docstring Paragraph::asString(pos_type beg, pos_type end, int options) const
3350 {
3351         odocstringstream os;
3352
3353         if (beg == 0
3354             && options & AS_STR_LABEL
3355             && !d->params_.labelString().empty())
3356                 os << d->params_.labelString() << ' ';
3357
3358         for (pos_type i = beg; i < end; ++i) {
3359                 if ((options & AS_STR_SKIPDELETE) && isDeleted(i))
3360                         continue;
3361                 char_type const c = d->text_[i];
3362                 if (isPrintable(c) || c == '\t'
3363                     || (c == '\n' && (options & AS_STR_NEWLINES)))
3364                         os.put(c);
3365                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
3366                         getInset(i)->toString(os);
3367                         if (getInset(i)->asInsetMath())
3368                                 os << " ";
3369                 }
3370         }
3371
3372         return os.str();
3373 }
3374
3375
3376 void Paragraph::forToc(docstring & os, size_t maxlen) const
3377 {
3378         if (!d->params_.labelString().empty())
3379                 os += d->params_.labelString() + ' ';
3380         for (pos_type i = 0; i < size() && os.length() < maxlen; ++i) {
3381                 if (isDeleted(i))
3382                         continue;
3383                 char_type const c = d->text_[i];
3384                 if (isPrintable(c))
3385                         os += c;
3386                 else if (c == '\t' || c == '\n')
3387                         os += ' ';
3388                 else if (c == META_INSET)
3389                         getInset(i)->forToc(os, maxlen);
3390         }
3391 }
3392
3393
3394 docstring Paragraph::stringify(pos_type beg, pos_type end, int options,
3395         OutputParams const & runparams) const
3396 {
3397         odocstringstream os;
3398
3399         if (beg == 0
3400                 && options & AS_STR_LABEL
3401                 && !d->params_.labelString().empty())
3402                 os << d->params_.labelString() << ' ';
3403
3404         OutputParams op = runparams;
3405         op.for_search = true;
3406
3407         for (pos_type i = beg; i < end; ++i) {
3408                 char_type const c = d->text_[i];
3409                 if (isPrintable(c) || c == '\t'
3410                     || (c == '\n' && (options & AS_STR_NEWLINES)))
3411                         os.put(c);
3412                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
3413                         getInset(i)->plaintext(os, op);
3414                 }
3415         }
3416
3417         return os.str();
3418 }
3419
3420
3421 void Paragraph::setInsetOwner(Inset const * inset)
3422 {
3423         d->inset_owner_ = inset;
3424 }
3425
3426
3427 int Paragraph::id() const
3428 {
3429         return d->id_;
3430 }
3431
3432
3433 void Paragraph::setId(int id)
3434 {
3435         d->id_ = id;
3436 }
3437
3438
3439 Layout const & Paragraph::layout() const
3440 {
3441         return *d->layout_;
3442 }
3443
3444
3445 void Paragraph::setLayout(Layout const & layout)
3446 {
3447         d->layout_ = &layout;
3448 }
3449
3450
3451 void Paragraph::setDefaultLayout(DocumentClass const & tc)
3452 {
3453         setLayout(tc.defaultLayout());
3454 }
3455
3456
3457 void Paragraph::setPlainLayout(DocumentClass const & tc)
3458 {
3459         setLayout(tc.plainLayout());
3460 }
3461
3462
3463 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
3464 {
3465         if (usePlainLayout())
3466                 setPlainLayout(tclass);
3467         else
3468                 setDefaultLayout(tclass);
3469 }
3470
3471
3472 Inset const & Paragraph::inInset() const
3473 {
3474         LBUFERR(d->inset_owner_);
3475         return *d->inset_owner_;
3476 }
3477
3478
3479 ParagraphParameters & Paragraph::params()
3480 {
3481         return d->params_;
3482 }
3483
3484
3485 ParagraphParameters const & Paragraph::params() const
3486 {
3487         return d->params_;
3488 }
3489
3490
3491 bool Paragraph::isFreeSpacing() const
3492 {
3493         if (d->layout_->free_spacing)
3494                 return true;
3495         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
3496 }
3497
3498
3499 bool Paragraph::allowEmpty() const
3500 {
3501         if (d->layout_->keepempty)
3502                 return true;
3503         return d->inset_owner_ && d->inset_owner_->allowEmpty();
3504 }
3505
3506
3507 char_type Paragraph::transformChar(char_type c, pos_type pos) const
3508 {
3509         if (!Encodings::isArabicChar(c))
3510                 return c;
3511
3512         char_type prev_char = ' ';
3513         char_type next_char = ' ';
3514
3515         for (pos_type i = pos - 1; i >= 0; --i) {
3516                 char_type const par_char = d->text_[i];
3517                 if (!Encodings::isArabicComposeChar(par_char)) {
3518                         prev_char = par_char;
3519                         break;
3520                 }
3521         }
3522
3523         for (pos_type i = pos + 1, end = size(); i < end; ++i) {
3524                 char_type const par_char = d->text_[i];
3525                 if (!Encodings::isArabicComposeChar(par_char)) {
3526                         next_char = par_char;
3527                         break;
3528                 }
3529         }
3530
3531         if (Encodings::isArabicChar(next_char)) {
3532                 if (Encodings::isArabicChar(prev_char) &&
3533                         !Encodings::isArabicSpecialChar(prev_char))
3534                         return Encodings::transformChar(c, Encodings::FORM_MEDIAL);
3535                 else
3536                         return Encodings::transformChar(c, Encodings::FORM_INITIAL);
3537         } else {
3538                 if (Encodings::isArabicChar(prev_char) &&
3539                         !Encodings::isArabicSpecialChar(prev_char))
3540                         return Encodings::transformChar(c, Encodings::FORM_FINAL);
3541                 else
3542                         return Encodings::transformChar(c, Encodings::FORM_ISOLATED);
3543         }
3544 }
3545
3546
3547 bool Paragraph::brokenBiblio() const
3548 {
3549         // there is a problem if there is no bibitem at position 0 or
3550         // if there is another bibitem in the paragraph.
3551         return d->layout_->labeltype == LABEL_BIBLIO
3552                 && (d->insetlist_.find(BIBITEM_CODE) != 0
3553                     || d->insetlist_.find(BIBITEM_CODE, 1) > 0);
3554 }
3555
3556
3557 int Paragraph::fixBiblio(Buffer const & buffer)
3558 {
3559         // FIXME: What about the case where paragraph is not BIBLIO
3560         // but there is an InsetBibitem?
3561         // FIXME: when there was already an inset at 0, the return value is 1,
3562         // which does not tell whether another inset has been remove; the
3563         // cursor cannot be correctly updated.
3564
3565         if (d->layout_->labeltype != LABEL_BIBLIO)
3566                 return 0;
3567
3568         bool const track_changes = buffer.params().trackChanges;
3569         int bibitem_pos = d->insetlist_.find(BIBITEM_CODE);
3570         bool const hasbibitem0 = bibitem_pos == 0;
3571
3572         if (hasbibitem0) {
3573                 bibitem_pos = d->insetlist_.find(BIBITEM_CODE, 1);
3574                 // There was an InsetBibitem at pos 0, and no other one => OK
3575                 if (bibitem_pos == -1)
3576                         return 0;
3577                 // there is a bibitem at the 0 position, but since
3578                 // there is a second one, we copy the second on the
3579                 // first. We're assuming there are at most two of
3580                 // these, which there should be.
3581                 // FIXME: why does it make sense to do that rather
3582                 // than keep the first? (JMarc)
3583                 Inset * inset = d->insetlist_.release(bibitem_pos);
3584                 eraseChar(bibitem_pos, track_changes);
3585                 d->insetlist_.begin()->inset = inset;
3586                 return -bibitem_pos;
3587         }
3588
3589         // We need to create an inset at the beginning
3590         Inset * inset = 0;
3591         if (bibitem_pos > 0) {
3592                 // there was one somewhere in the paragraph, let's move it
3593                 inset = d->insetlist_.release(bibitem_pos);
3594                 eraseChar(bibitem_pos, track_changes);
3595         } else
3596                 // make a fresh one
3597                 inset = new InsetBibitem(const_cast<Buffer *>(&buffer),
3598                                          InsetCommandParams(BIBITEM_CODE));
3599
3600         Font font(inherit_font, buffer.params().language);
3601         insertInset(0, inset, font, Change(track_changes ? Change::INSERTED
3602                                                    : Change::UNCHANGED));
3603
3604         return 1;
3605 }
3606
3607
3608 void Paragraph::checkAuthors(AuthorList const & authorList)
3609 {
3610         d->changes_.checkAuthors(authorList);
3611 }
3612
3613
3614 bool Paragraph::isChanged(pos_type pos) const
3615 {
3616         return lookupChange(pos).changed();
3617 }
3618
3619
3620 bool Paragraph::isInserted(pos_type pos) const
3621 {
3622         return lookupChange(pos).inserted();
3623 }
3624
3625
3626 bool Paragraph::isDeleted(pos_type pos) const
3627 {
3628         return lookupChange(pos).deleted();
3629 }
3630
3631
3632 InsetList const & Paragraph::insetList() const
3633 {
3634         return d->insetlist_;
3635 }
3636
3637
3638 void Paragraph::setBuffer(Buffer & b)
3639 {
3640         d->insetlist_.setBuffer(b);
3641 }
3642
3643
3644 Inset * Paragraph::releaseInset(pos_type pos)
3645 {
3646         Inset * inset = d->insetlist_.release(pos);
3647         /// does not honour change tracking!
3648         eraseChar(pos, false);
3649         return inset;
3650 }
3651
3652
3653 Inset * Paragraph::getInset(pos_type pos)
3654 {
3655         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3656                  ? d->insetlist_.get(pos) : 0;
3657 }
3658
3659
3660 Inset const * Paragraph::getInset(pos_type pos) const
3661 {
3662         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3663                  ? d->insetlist_.get(pos) : 0;
3664 }
3665
3666
3667 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
3668                 pos_type & right, TextCase action)
3669 {
3670         // process sequences of modified characters; in change
3671         // tracking mode, this approach results in much better
3672         // usability than changing case on a char-by-char basis
3673         docstring changes;
3674
3675         bool const trackChanges = bparams.trackChanges;
3676
3677         bool capitalize = true;
3678
3679         for (; pos < right; ++pos) {
3680                 char_type oldChar = d->text_[pos];
3681                 char_type newChar = oldChar;
3682
3683                 // ignore insets and don't play with deleted text!
3684                 if (oldChar != META_INSET && !isDeleted(pos)) {
3685                         switch (action) {
3686                                 case text_lowercase:
3687                                         newChar = lowercase(oldChar);
3688                                         break;
3689                                 case text_capitalization:
3690                                         if (capitalize) {
3691                                                 newChar = uppercase(oldChar);
3692                                                 capitalize = false;
3693                                         }
3694                                         break;
3695                                 case text_uppercase:
3696                                         newChar = uppercase(oldChar);
3697                                         break;
3698                         }
3699                 }
3700
3701                 if (isWordSeparator(pos) || isDeleted(pos)) {
3702                         // permit capitalization again
3703                         capitalize = true;
3704                 }
3705
3706                 if (oldChar != newChar) {
3707                         changes += newChar;
3708                         if (pos != right - 1)
3709                                 continue;
3710                         // step behind the changing area
3711                         pos++;
3712                 }
3713
3714                 int erasePos = pos - changes.size();
3715                 for (size_t i = 0; i < changes.size(); i++) {
3716                         insertChar(pos, changes[i],
3717                                    getFontSettings(bparams,
3718                                                    erasePos),
3719                                    trackChanges);
3720                         if (!eraseChar(erasePos, trackChanges)) {
3721                                 ++erasePos;
3722                                 ++pos; // advance
3723                                 ++right; // expand selection
3724                         }
3725                 }
3726                 changes.clear();
3727         }
3728 }
3729
3730
3731 int Paragraph::find(docstring const & str, bool cs, bool mw,
3732                 pos_type start_pos, bool del) const
3733 {
3734         pos_type pos = start_pos;
3735         int const strsize = str.length();
3736         int i = 0;
3737         pos_type const parsize = d->text_.size();
3738         for (i = 0; i < strsize && pos < parsize; ++i, ++pos) {
3739                 // Ignore "invisible" letters such as ligature breaks
3740                 // and hyphenation chars while searching
3741                 while (pos < parsize - 1 && isInset(pos)) {
3742                         odocstringstream os;
3743                         getInset(pos)->toString(os);
3744                         if (!getInset(pos)->isLetter() || !os.str().empty())
3745                                 break;
3746                         pos++;
3747                 }
3748                 if (cs && str[i] != d->text_[pos])
3749                         break;
3750                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos]))
3751                         break;
3752                 if (!del && isDeleted(pos))
3753                         break;
3754         }
3755
3756         if (i != strsize)
3757                 return 0;
3758
3759         // if necessary, check whether string matches word
3760         if (mw) {
3761                 if (start_pos > 0 && !isWordSeparator(start_pos - 1))
3762                         return 0;
3763                 if (pos < parsize
3764                         && !isWordSeparator(pos))
3765                         return 0;
3766         }
3767
3768         return pos - start_pos;
3769 }
3770
3771
3772 char_type Paragraph::getChar(pos_type pos) const
3773 {
3774         return d->text_[pos];
3775 }
3776
3777
3778 pos_type Paragraph::size() const
3779 {
3780         return d->text_.size();
3781 }
3782
3783
3784 bool Paragraph::empty() const
3785 {
3786         return d->text_.empty();
3787 }
3788
3789
3790 bool Paragraph::isInset(pos_type pos) const
3791 {
3792         return d->text_[pos] == META_INSET;
3793 }
3794
3795
3796 bool Paragraph::isSeparator(pos_type pos) const
3797 {
3798         //FIXME: Are we sure this can be the only separator?
3799         return d->text_[pos] == ' ';
3800 }
3801
3802
3803 void Paragraph::deregisterWords()
3804 {
3805         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3806         Private::LangWordsMap::const_iterator ite = d->words_.end();
3807         for (; itl != ite; ++itl) {
3808                 WordList * wl = theWordList(itl->first);
3809                 Private::Words::const_iterator it = (itl->second).begin();
3810                 Private::Words::const_iterator et = (itl->second).end();
3811                 for (; it != et; ++it)
3812                         wl->remove(*it);
3813         }
3814         d->words_.clear();
3815 }
3816
3817
3818 void Paragraph::locateWord(pos_type & from, pos_type & to,
3819         word_location const loc) const
3820 {
3821         switch (loc) {
3822         case WHOLE_WORD_STRICT:
3823                 if (from == 0 || from == size()
3824                     || isWordSeparator(from)
3825                     || isWordSeparator(from - 1)) {
3826                         to = from;
3827                         return;
3828                 }
3829                 // no break here, we go to the next
3830
3831         case WHOLE_WORD:
3832                 // If we are already at the beginning of a word, do nothing
3833                 if (!from || isWordSeparator(from - 1))
3834                         break;
3835                 // no break here, we go to the next
3836
3837         case PREVIOUS_WORD:
3838                 // always move the cursor to the beginning of previous word
3839                 while (from && !isWordSeparator(from - 1))
3840                         --from;
3841                 break;
3842         case NEXT_WORD:
3843                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
3844                 break;
3845         case PARTIAL_WORD:
3846                 // no need to move the 'from' cursor
3847                 break;
3848         }
3849         to = from;
3850         while (to < size() && !isWordSeparator(to))
3851                 ++to;
3852 }
3853
3854
3855 void Paragraph::collectWords()
3856 {
3857         pos_type n = size();
3858         for (pos_type pos = 0; pos < n; ++pos) {
3859                 if (isWordSeparator(pos))
3860                         continue;
3861                 pos_type from = pos;
3862                 locateWord(from, pos, WHOLE_WORD);
3863                 if ((pos - from) >= (int)lyxrc.completion_minlength) {
3864                         docstring word = asString(from, pos, AS_STR_NONE);
3865                         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
3866                         if (cit == d->fontlist_.end())
3867                                 return;
3868                         Language const * lang = cit->font().language();
3869                         d->words_[*lang].insert(word);
3870                 }
3871         }
3872 }
3873
3874
3875 void Paragraph::registerWords()
3876 {
3877         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3878         Private::LangWordsMap::const_iterator ite = d->words_.end();
3879         for (; itl != ite; ++itl) {
3880                 WordList * wl = theWordList(itl->first);
3881                 Private::Words::const_iterator it = (itl->second).begin();
3882                 Private::Words::const_iterator et = (itl->second).end();
3883                 for (; it != et; ++it)
3884                         wl->insert(*it);
3885         }
3886 }
3887
3888
3889 void Paragraph::updateWords()
3890 {
3891         deregisterWords();
3892         collectWords();
3893         registerWords();
3894 }
3895
3896
3897 void Paragraph::Private::appendSkipPosition(SkipPositions & skips, pos_type const pos) const
3898 {
3899         SkipPositionsIterator begin = skips.begin();
3900         SkipPositions::iterator end = skips.end();
3901         if (pos > 0 && begin < end) {
3902                 --end;
3903                 if (end->last == pos - 1) {
3904                         end->last = pos;
3905                         return;
3906                 }
3907         }
3908         skips.insert(end, FontSpan(pos, pos));
3909 }
3910
3911
3912 Language * Paragraph::Private::locateSpellRange(
3913         pos_type & from, pos_type & to,
3914         SkipPositions & skips) const
3915 {
3916         // skip leading white space
3917         while (from < to && owner_->isWordSeparator(from))
3918                 ++from;
3919         // don't check empty range
3920         if (from >= to)
3921                 return 0;
3922         // get current language
3923         Language * lang = getSpellLanguage(from);
3924         pos_type last = from;
3925         bool samelang = true;
3926         bool sameinset = true;
3927         while (last < to && samelang && sameinset) {
3928                 // hop to end of word
3929                 while (last < to && !owner_->isWordSeparator(last)) {
3930                         if (owner_->getInset(last)) {
3931                                 appendSkipPosition(skips, last);
3932                         } else if (owner_->isDeleted(last)) {
3933                                 appendSkipPosition(skips, last);
3934                         }
3935                         ++last;
3936                 }
3937                 // hop to next word while checking for insets
3938                 while (sameinset && last < to && owner_->isWordSeparator(last)) {
3939                         if (Inset const * inset = owner_->getInset(last))
3940                                 sameinset = inset->isChar() && inset->isLetter();
3941                         if (sameinset && owner_->isDeleted(last)) {
3942                                 appendSkipPosition(skips, last);
3943                         }
3944                         if (sameinset)
3945                                 last++;
3946                 }
3947                 if (sameinset && last < to) {
3948                         // now check for language change
3949                         samelang = lang == getSpellLanguage(last);
3950                 }
3951         }
3952         // if language change detected backstep is needed
3953         if (!samelang)
3954                 --last;
3955         to = last;
3956         return lang;
3957 }
3958
3959
3960 Language * Paragraph::Private::getSpellLanguage(pos_type const from) const
3961 {
3962         Language * lang =
3963                 const_cast<Language *>(owner_->getFontSettings(
3964                         inset_owner_->buffer().params(), from).language());
3965         if (lang == inset_owner_->buffer().params().language
3966                 && !lyxrc.spellchecker_alt_lang.empty()) {
3967                 string lang_code;
3968                 string const lang_variety =
3969                         split(lyxrc.spellchecker_alt_lang, lang_code, '-');
3970                 lang->setCode(lang_code);
3971                 lang->setVariety(lang_variety);
3972         }
3973         return lang;
3974 }
3975
3976
3977 void Paragraph::requestSpellCheck(pos_type pos)
3978 {
3979         d->requestSpellCheck(pos);
3980 }
3981
3982
3983 bool Paragraph::needsSpellCheck() const
3984 {
3985         SpellChecker::ChangeNumber speller_change_number = 0;
3986         if (theSpellChecker())
3987                 speller_change_number = theSpellChecker()->changeNumber();
3988         if (speller_change_number > d->speller_state_.currentChangeNumber()) {
3989                 d->speller_state_.needsCompleteRefresh(speller_change_number);
3990         }
3991         return d->needsSpellCheck();
3992 }
3993
3994
3995 bool Paragraph::Private::ignoreWord(docstring const & word) const
3996 {
3997         // Ignore words with digits
3998         // FIXME: make this customizable
3999         // (note that some checkers ignore words with digits by default)
4000         docstring::const_iterator cit = word.begin();
4001         docstring::const_iterator const end = word.end();
4002         for (; cit != end; ++cit) {
4003                 if (isNumber((*cit)))
4004                         return true;
4005         }
4006         return false;
4007 }
4008
4009
4010 SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
4011         WordLangTuple & wl, docstring_list & suggestions,
4012         bool do_suggestion, bool check_learned) const
4013 {
4014         SpellChecker::Result result = SpellChecker::WORD_OK;
4015         SpellChecker * speller = theSpellChecker();
4016         if (!speller)
4017                 return result;
4018
4019         if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
4020                 return result;
4021
4022         locateWord(from, to, WHOLE_WORD);
4023         if (from == to || from >= size())
4024                 return result;
4025
4026         docstring word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
4027         Language * lang = d->getSpellLanguage(from);
4028
4029         wl = WordLangTuple(word, lang);
4030
4031         if (word.empty())
4032                 return result;
4033
4034         if (needsSpellCheck() || check_learned) {
4035                 pos_type end = to;
4036                 if (!d->ignoreWord(word)) {
4037                         bool const trailing_dot = to < size() && d->text_[to] == '.';
4038                         result = speller->check(wl);
4039                         if (SpellChecker::misspelled(result) && trailing_dot) {
4040                                 wl = WordLangTuple(word.append(from_ascii(".")), lang);
4041                                 result = speller->check(wl);
4042                                 if (!SpellChecker::misspelled(result)) {
4043                                         LYXERR(Debug::GUI, "misspelled word is correct with dot: \"" <<
4044                                            word << "\" [" <<
4045                                            from << ".." << to << "]");
4046                                 } else {
4047                                         // spell check with dot appended failed too
4048                                         // restore original word/lang value
4049                                         word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
4050                                         wl = WordLangTuple(word, lang);
4051                                 }
4052                         }
4053                 }
4054                 if (!SpellChecker::misspelled(result)) {
4055                         // area up to the begin of the next word is not misspelled
4056                         while (end < size() && isWordSeparator(end))
4057                                 ++end;
4058                 }
4059                 d->setMisspelled(from, end, result);
4060         } else {
4061                 result = d->speller_state_.getState(from);
4062         }
4063
4064         if (do_suggestion)
4065                 suggestions.clear();
4066
4067         if (SpellChecker::misspelled(result)) {
4068                 LYXERR(Debug::GUI, "misspelled word: \"" <<
4069                            word << "\" [" <<
4070                            from << ".." << to << "]");
4071                 if (do_suggestion)
4072                         speller->suggest(wl, suggestions);
4073         }
4074         return result;
4075 }
4076
4077
4078 void Paragraph::Private::markMisspelledWords(
4079         pos_type const & first, pos_type const & last,
4080         SpellChecker::Result result,
4081         docstring const & word,
4082         SkipPositions const & skips)
4083 {
4084         if (!SpellChecker::misspelled(result)) {
4085                 setMisspelled(first, last, SpellChecker::WORD_OK);
4086                 return;
4087         }
4088         int snext = first;
4089         SpellChecker * speller = theSpellChecker();
4090         // locate and enumerate the error positions
4091         int nerrors = speller->numMisspelledWords();
4092         int numskipped = 0;
4093         SkipPositionsIterator it = skips.begin();
4094         SkipPositionsIterator et = skips.end();
4095         for (int index = 0; index < nerrors; ++index) {
4096                 int wstart;
4097                 int wlen = 0;
4098                 speller->misspelledWord(index, wstart, wlen);
4099                 /// should not happen if speller supports range checks
4100                 if (!wlen) continue;
4101                 docstring const misspelled = word.substr(wstart, wlen);
4102                 wstart += first + numskipped;
4103                 if (snext < wstart) {
4104                         /// mark the range of correct spelling
4105                         numskipped += countSkips(it, et, wstart);
4106                         setMisspelled(snext,
4107                                 wstart - 1, SpellChecker::WORD_OK);
4108                 }
4109                 snext = wstart + wlen;
4110                 numskipped += countSkips(it, et, snext);
4111                 /// mark the range of misspelling
4112                 setMisspelled(wstart, snext, result);
4113                 LYXERR(Debug::GUI, "misspelled word: \"" <<
4114                            misspelled << "\" [" <<
4115                            wstart << ".." << (snext-1) << "]");
4116                 ++snext;
4117         }
4118         if (snext <= last) {
4119                 /// mark the range of correct spelling at end
4120                 setMisspelled(snext, last, SpellChecker::WORD_OK);
4121         }
4122 }
4123
4124
4125 void Paragraph::spellCheck() const
4126 {
4127         SpellChecker * speller = theSpellChecker();
4128         if (!speller || empty() ||!needsSpellCheck())
4129                 return;
4130         pos_type start;
4131         pos_type endpos;
4132         d->rangeOfSpellCheck(start, endpos);
4133         if (speller->canCheckParagraph()) {
4134                 // loop until we leave the range
4135                 for (pos_type first = start; first < endpos; ) {
4136                         pos_type last = endpos;
4137                         Private::SkipPositions skips;
4138                         Language * lang = d->locateSpellRange(first, last, skips);
4139                         if (first >= endpos)
4140                                 break;
4141                         // start the spell checker on the unit of meaning
4142                         docstring word = asString(first, last, AS_STR_INSETS + AS_STR_SKIPDELETE);
4143                         WordLangTuple wl = WordLangTuple(word, lang);
4144                         SpellChecker::Result result = word.size() ?
4145                                 speller->check(wl) : SpellChecker::WORD_OK;
4146                         d->markMisspelledWords(first, last, result, word, skips);
4147                         first = ++last;
4148                 }
4149         } else {
4150                 static docstring_list suggestions;
4151                 pos_type to = endpos;
4152                 while (start < endpos) {
4153                         WordLangTuple wl;
4154                         spellCheck(start, to, wl, suggestions, false);
4155                         start = to + 1;
4156                 }
4157         }
4158         d->readySpellCheck();
4159 }
4160
4161
4162 bool Paragraph::isMisspelled(pos_type pos, bool check_boundary) const
4163 {
4164         bool result = SpellChecker::misspelled(d->speller_state_.getState(pos));
4165         if (result || pos <= 0 || pos > size())
4166                 return result;
4167         if (check_boundary && (pos == size() || isWordSeparator(pos)))
4168                 result = SpellChecker::misspelled(d->speller_state_.getState(pos - 1));
4169         return result;
4170 }
4171
4172
4173 string Paragraph::magicLabel() const
4174 {
4175         stringstream ss;
4176         ss << "magicparlabel-" << id();
4177         return ss.str();
4178 }
4179
4180
4181 } // namespace lyx