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                 _("Invalid position given to lookupChange()"));
 665         return d->changes_.lookup(pos);
 666 }
 667
 668
 669 void Paragraph::acceptChanges(pos_type start, pos_type end)
 670 {
 671         LASSERT(start >= 0 && start <= size(), return);
 672         LASSERT(end > start && end <= size() + 1, return);
 673
 674         for (pos_type pos = start; pos < end; ++pos) {
 675                 switch (lookupChange(pos).type) {
 676                         case Change::UNCHANGED:
 677                                 // accept changes in nested inset
 678                                 if (Inset * inset = getInset(pos))
 679                                         inset->acceptChanges();
 680                                 break;
 681
 682                         case Change::INSERTED:
 683                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 684                                 // also accept changes in nested inset
 685                                 if (Inset * inset = getInset(pos))
 686                                         inset->acceptChanges();
 687                                 break;
 688
 689                         case Change::DELETED:
 690                                 // Suppress access to non-existent
 691                                 // "end-of-paragraph char"
 692                                 if (pos < size()) {
 693                                         eraseChar(pos, false);
 694                                         --end;
 695                                         --pos;
 696                                 }
 697                                 break;
 698                 }
 699
 700         }
 701 }
 702
 703
 704 void Paragraph::rejectChanges(pos_type start, pos_type end)
 705 {
 706         LASSERT(start >= 0 && start <= size(), return);
 707         LASSERT(end > start && end <= size() + 1, return);
 708
 709         for (pos_type pos = start; pos < end; ++pos) {
 710                 switch (lookupChange(pos).type) {
 711                         case Change::UNCHANGED:
 712                                 // reject changes in nested inset
 713                                 if (Inset * inset = getInset(pos))
 714                                                 inset->rejectChanges();
 715                                 break;
 716
 717                         case Change::INSERTED:
 718                                 // Suppress access to non-existent
 719                                 // "end-of-paragraph char"
 720                                 if (pos < size()) {
 721                                         eraseChar(pos, false);
 722                                         --end;
 723                                         --pos;
 724                                 }
 725                                 break;
 726
 727                         case Change::DELETED:
 728                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 729
 730                                 // Do NOT reject changes within a deleted inset!
 731                                 // There may be insertions of a co-author inside of it!
 732
 733                                 break;
 734                 }
 735         }
 736 }
 737
 738
 739 void Paragraph::Private::insertChar(pos_type pos, char_type c,
 740                 Change const & change)
 741 {
 742         LASSERT(pos >= 0 && pos <= int(text_.size()), return);
 743
 744         // track change
 745         changes_.insert(change, pos);
 746
 747         // This is actually very common when parsing buffers (and
 748         // maybe inserting ascii text)
 749         if (pos == pos_type(text_.size())) {
 750                 // when appending characters, no need to update tables
 751                 text_.push_back(c);
 752                 // but we want spell checking
 753                 requestSpellCheck(pos);
 754                 return;
 755         }
 756
 757         text_.insert(text_.begin() + pos, c);
 758
 759         // Update the font table.
 760         fontlist_.increasePosAfterPos(pos);
 761
 762         // Update the insets
 763         insetlist_.increasePosAfterPos(pos);
 764
 765         // Update list of misspelled positions
 766         speller_state_.increasePosAfterPos(pos);
 767 }
 768
 769
 770 bool Paragraph::insertInset(pos_type pos, Inset * inset,
 771                                    Change const & change)
 772 {
 773         LASSERT(inset, return false);
 774         LASSERT(pos >= 0 && pos <= size(), return false);
 775
 776         // Paragraph::insertInset() can be used in cut/copy/paste operation where
 777         // d->inset_owner_ is not set yet.
 778         if (d->inset_owner_ && !d->inset_owner_->insetAllowed(inset->lyxCode()))
 779                 return false;
 780
 781         d->insertChar(pos, META_INSET, change);
 782         LASSERT(d->text_[pos] == META_INSET, return false);
 783
 784         // Add a new entry in the insetlist_.
 785         d->insetlist_.insert(inset, pos);
 786
 787         // Some insets require run of spell checker
 788         requestSpellCheck(pos);
 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) {
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 bool Paragraph::insertInset(pos_type pos, Inset * inset,
1787                             Font const & font, Change const & change)
1788 {
1789         bool const success = insertInset(pos, inset, change);
1790         // Set the font/language of the inset...
1791         setFont(pos, font);
1792         return success;
1793 }
1794
1795
1796 void Paragraph::resetFonts(Font const & font)
1797 {
1798         d->fontlist_.clear();
1799         d->fontlist_.set(0, font);
1800         d->fontlist_.set(d->text_.size() - 1, font);
1801 }
1802
1803 // Gets uninstantiated font setting at position.
1804 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1805                                          pos_type pos) const
1806 {
1807         if (pos > size()) {
1808                 LYXERR0("pos: " << pos << " size: " << size());
1809                 LBUFERR(false, _("Invalid position."));
1810         }
1811
1812         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1813         if (cit != d->fontlist_.end())
1814                 return cit->font();
1815
1816         if (pos == size() && !empty())
1817                 return getFontSettings(bparams, pos - 1);
1818
1819         // Optimisation: avoid a full font instantiation if there is no
1820         // language change from previous call.
1821         static Font previous_font;
1822         static Language const * previous_lang = 0;
1823         Language const * lang = getParLanguage(bparams);
1824         if (lang != previous_lang) {
1825                 previous_lang = lang;
1826                 previous_font = Font(inherit_font, lang);
1827         }
1828         return previous_font;
1829 }
1830
1831
1832 FontSpan Paragraph::fontSpan(pos_type pos) const
1833 {
1834         LBUFERR(pos <= size(), _("Invalid position"));
1835         pos_type start = 0;
1836
1837         FontList::const_iterator cit = d->fontlist_.begin();
1838         FontList::const_iterator end = d->fontlist_.end();
1839         for (; cit != end; ++cit) {
1840                 if (cit->pos() >= pos) {
1841                         if (pos >= beginOfBody())
1842                                 return FontSpan(max(start, beginOfBody()),
1843                                                 cit->pos());
1844                         else
1845                                 return FontSpan(start,
1846                                                 min(beginOfBody() - 1,
1847                                                          cit->pos()));
1848                 }
1849                 start = cit->pos() + 1;
1850         }
1851
1852         // This should not happen, but if so, we take no chances.
1853         LYXERR0("Paragraph::getEndPosOfFontSpan: This should not happen!");
1854         return FontSpan(pos, pos);
1855 }
1856
1857
1858 // Gets uninstantiated font setting at position 0
1859 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1860 {
1861         if (!empty() && !d->fontlist_.empty())
1862                 return d->fontlist_.begin()->font();
1863
1864         // Optimisation: avoid a full font instantiation if there is no
1865         // language change from previous call.
1866         static Font previous_font;
1867         static Language const * previous_lang = 0;
1868         if (bparams.language != previous_lang) {
1869                 previous_lang = bparams.language;
1870                 previous_font = Font(inherit_font, bparams.language);
1871         }
1872
1873         return previous_font;
1874 }
1875
1876
1877 // Gets the fully instantiated font at a given position in a paragraph
1878 // This is basically the same function as Text::GetFont() in text2.cpp.
1879 // The difference is that this one is used for generating the LaTeX file,
1880 // and thus cosmetic "improvements" are disallowed: This has to deliver
1881 // the true picture of the buffer. (Asger)
1882 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1883                                  Font const & outerfont) const
1884 {
1885         LBUFERR(pos >= 0, _("Invalid position."));
1886
1887         Font font = getFontSettings(bparams, pos);
1888
1889         pos_type const body_pos = beginOfBody();
1890         FontInfo & fi = font.fontInfo();
1891         if (pos < body_pos)
1892                 fi.realize(d->layout_->labelfont);
1893         else
1894                 fi.realize(d->layout_->font);
1895
1896         fi.realize(outerfont.fontInfo());
1897         fi.realize(bparams.getFont().fontInfo());
1898
1899         return font;
1900 }
1901
1902
1903 Font const Paragraph::getLabelFont
1904         (BufferParams const & bparams, Font const & outerfont) const
1905 {
1906         FontInfo tmpfont = d->layout_->labelfont;
1907         tmpfont.realize(outerfont.fontInfo());
1908         tmpfont.realize(bparams.getFont().fontInfo());
1909         return Font(tmpfont, getParLanguage(bparams));
1910 }
1911
1912
1913 Font const Paragraph::getLayoutFont
1914         (BufferParams const & bparams, Font const & outerfont) const
1915 {
1916         FontInfo tmpfont = d->layout_->font;
1917         tmpfont.realize(outerfont.fontInfo());
1918         tmpfont.realize(bparams.getFont().fontInfo());
1919         return Font(tmpfont, getParLanguage(bparams));
1920 }
1921
1922
1923 /// Returns the height of the highest font in range
1924 FontSize Paragraph::highestFontInRange
1925         (pos_type startpos, pos_type endpos, FontSize def_size) const
1926 {
1927         return d->fontlist_.highestInRange(startpos, endpos, def_size);
1928 }
1929
1930
1931 char_type Paragraph::getUChar(BufferParams const & bparams, pos_type pos) const
1932 {
1933         char_type c = d->text_[pos];
1934         if (!lyxrc.rtl_support || !getFontSettings(bparams, pos).isRightToLeft())
1935                 return c;
1936
1937         // FIXME: The arabic special casing is due to the difference of arabic
1938         // round brackets input introduced in r18599. Check if this should be
1939         // unified with Hebrew or at least if all bracket types should be
1940         // handled the same (file format change in either case).
1941         string const & lang = getFontSettings(bparams, pos).language()->lang();
1942         bool const arabic = lang == "arabic_arabtex" || lang == "arabic_arabi"
1943                 || lang == "farsi";
1944         char_type uc = c;
1945         switch (c) {
1946         case '(':
1947                 uc = arabic ? c : ')';
1948                 break;
1949         case ')':
1950                 uc = arabic ? c : '(';
1951                 break;
1952         case '[':
1953                 uc = ']';
1954                 break;
1955         case ']':
1956                 uc = '[';
1957                 break;
1958         case '{':
1959                 uc = '}';
1960                 break;
1961         case '}':
1962                 uc = '{';
1963                 break;
1964         case '<':
1965                 uc = '>';
1966                 break;
1967         case '>':
1968                 uc = '<';
1969                 break;
1970         }
1971
1972         return uc;
1973 }
1974
1975
1976 void Paragraph::setFont(pos_type pos, Font const & font)
1977 {
1978         LASSERT(pos <= size(), return);
1979
1980         // First, reduce font against layout/label font
1981         // Update: The setCharFont() routine in text2.cpp already
1982         // reduces font, so we don't need to do that here. (Asger)
1983
1984         d->fontlist_.set(pos, font);
1985 }
1986
1987
1988 void Paragraph::makeSameLayout(Paragraph const & par)
1989 {
1990         d->layout_ = par.d->layout_;
1991         d->params_ = par.d->params_;
1992 }
1993
1994
1995 bool Paragraph::stripLeadingSpaces(bool trackChanges)
1996 {
1997         if (isFreeSpacing())
1998                 return false;
1999
2000         int pos = 0;
2001         int count = 0;
2002
2003         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
2004                 if (eraseChar(pos, trackChanges))
2005                         ++count;
2006                 else
2007                         ++pos;
2008         }
2009
2010         return count > 0 || pos > 0;
2011 }
2012
2013
2014 bool Paragraph::hasSameLayout(Paragraph const & par) const
2015 {
2016         return par.d->layout_ == d->layout_
2017                 && d->params_.sameLayout(par.d->params_);
2018 }
2019
2020
2021 depth_type Paragraph::getDepth() const
2022 {
2023         return d->params_.depth();
2024 }
2025
2026
2027 depth_type Paragraph::getMaxDepthAfter() const
2028 {
2029         if (d->layout_->isEnvironment())
2030                 return d->params_.depth() + 1;
2031         else
2032                 return d->params_.depth();
2033 }
2034
2035
2036 char Paragraph::getAlign() const
2037 {
2038         if (d->params_.align() == LYX_ALIGN_LAYOUT)
2039                 return d->layout_->align;
2040         else
2041                 return d->params_.align();
2042 }
2043
2044
2045 docstring const & Paragraph::labelString() const
2046 {
2047         return d->params_.labelString();
2048 }
2049
2050
2051 // the next two functions are for the manual labels
2052 docstring const Paragraph::getLabelWidthString() const
2053 {
2054         if (d->layout_->margintype == MARGIN_MANUAL
2055             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
2056                 return d->params_.labelWidthString();
2057         else
2058                 return _("Senseless with this layout!");
2059 }
2060
2061
2062 void Paragraph::setLabelWidthString(docstring const & s)
2063 {
2064         d->params_.labelWidthString(s);
2065 }
2066
2067
2068 docstring Paragraph::expandLabel(Layout const & layout,
2069                 BufferParams const & bparams) const
2070 {
2071         return expandParagraphLabel(layout, bparams, true);
2072 }
2073
2074
2075 docstring Paragraph::expandDocBookLabel(Layout const & layout,
2076                 BufferParams const & bparams) const
2077 {
2078         return expandParagraphLabel(layout, bparams, false);
2079 }
2080
2081
2082 docstring Paragraph::expandParagraphLabel(Layout const & layout,
2083                 BufferParams const & bparams, bool process_appendix) const
2084 {
2085         DocumentClass const & tclass = bparams.documentClass();
2086         string const & lang = getParLanguage(bparams)->code();
2087         bool const in_appendix = process_appendix && d->params_.appendix();
2088         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
2089
2090         if (fmt.empty() && !layout.counter.empty())
2091                 return tclass.counters().theCounter(layout.counter, lang);
2092
2093         // handle 'inherited level parts' in 'fmt',
2094         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
2095         size_t const i = fmt.find('@', 0);
2096         if (i != docstring::npos) {
2097                 size_t const j = fmt.find('@', i + 1);
2098                 if (j != docstring::npos) {
2099                         docstring parent(fmt, i + 1, j - i - 1);
2100                         docstring label = from_ascii("??");
2101                         if (tclass.hasLayout(parent))
2102                                 docstring label = expandParagraphLabel(tclass[parent], bparams,
2103                                                       process_appendix);
2104                         fmt = docstring(fmt, 0, i) + label
2105                                 + docstring(fmt, j + 1, docstring::npos);
2106                 }
2107         }
2108
2109         return tclass.counters().counterLabel(fmt, lang);
2110 }
2111
2112
2113 void Paragraph::applyLayout(Layout const & new_layout)
2114 {
2115         d->layout_ = &new_layout;
2116         LyXAlignment const oldAlign = d->params_.align();
2117
2118         if (!(oldAlign & d->layout_->alignpossible)) {
2119                 frontend::Alert::warning(_("Alignment not permitted"),
2120                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
2121                 d->params_.align(LYX_ALIGN_LAYOUT);
2122         }
2123 }
2124
2125
2126 pos_type Paragraph::beginOfBody() const
2127 {
2128         return d->begin_of_body_;
2129 }
2130
2131
2132 void Paragraph::setBeginOfBody()
2133 {
2134         if (d->layout_->labeltype != LABEL_MANUAL) {
2135                 d->begin_of_body_ = 0;
2136                 return;
2137         }
2138
2139         // Unroll the first two cycles of the loop
2140         // and remember the previous character to
2141         // remove unnecessary getChar() calls
2142         pos_type i = 0;
2143         pos_type end = size();
2144         if (i < end && !isNewline(i)) {
2145                 ++i;
2146                 char_type previous_char = 0;
2147                 char_type temp = 0;
2148                 if (i < end) {
2149                         previous_char = d->text_[i];
2150                         if (!isNewline(i)) {
2151                                 ++i;
2152                                 while (i < end && previous_char != ' ') {
2153                                         temp = d->text_[i];
2154                                         if (isNewline(i))
2155                                                 break;
2156                                         ++i;
2157                                         previous_char = temp;
2158                                 }
2159                         }
2160                 }
2161         }
2162
2163         d->begin_of_body_ = i;
2164 }
2165
2166
2167 bool Paragraph::allowParagraphCustomization() const
2168 {
2169         return inInset().allowParagraphCustomization();
2170 }
2171
2172
2173 bool Paragraph::usePlainLayout() const
2174 {
2175         return inInset().usePlainLayout();
2176 }
2177
2178
2179 bool Paragraph::isPassThru() const
2180 {
2181         return inInset().isPassThru() || d->layout_->pass_thru;
2182 }
2183
2184 namespace {
2185
2186 // paragraphs inside floats need different alignment tags to avoid
2187 // unwanted space
2188
2189 bool noTrivlistCentering(InsetCode code)
2190 {
2191         return code == FLOAT_CODE
2192                || code == WRAP_CODE
2193                || code == CELL_CODE;
2194 }
2195
2196
2197 string correction(string const & orig)
2198 {
2199         if (orig == "flushleft")
2200                 return "raggedright";
2201         if (orig == "flushright")
2202                 return "raggedleft";
2203         if (orig == "center")
2204                 return "centering";
2205         return orig;
2206 }
2207
2208
2209 string const corrected_env(string const & suffix, string const & env,
2210         InsetCode code, bool const lastpar)
2211 {
2212         string output = suffix + "{";
2213         if (noTrivlistCentering(code)) {
2214                 if (lastpar) {
2215                         // the last paragraph in non-trivlist-aligned
2216                         // context is special (to avoid unwanted whitespace)
2217                         if (suffix == "\\begin")
2218                                 return "\\" + correction(env) + "{}";
2219                         return string();
2220                 }
2221                 output += correction(env);
2222         } else
2223                 output += env;
2224         output += "}";
2225         if (suffix == "\\begin")
2226                 output += "\n";
2227         return output;
2228 }
2229
2230
2231 void adjust_column(string const & str, int & column)
2232 {
2233         if (!contains(str, "\n"))
2234                 column += str.size();
2235         else {
2236                 string tmp;
2237                 column = rsplit(str, tmp, '\n').size();
2238         }
2239 }
2240
2241 } // namespace anon
2242
2243
2244 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
2245                         otexstream & os, OutputParams const & runparams) const
2246 {
2247         int column = 0;
2248
2249         if (params_.noindent() && !layout_->pass_thru) {
2250                 os << "\\noindent ";
2251                 column += 10;
2252         }
2253
2254         LyXAlignment const curAlign = params_.align();
2255
2256         if (curAlign == layout_->align)
2257                 return column;
2258
2259         switch (curAlign) {
2260         case LYX_ALIGN_NONE:
2261         case LYX_ALIGN_BLOCK:
2262         case LYX_ALIGN_LAYOUT:
2263         case LYX_ALIGN_SPECIAL:
2264         case LYX_ALIGN_DECIMAL:
2265                 break;
2266         case LYX_ALIGN_LEFT:
2267         case LYX_ALIGN_RIGHT:
2268         case LYX_ALIGN_CENTER:
2269                 if (runparams.moving_arg) {
2270                         os << "\\protect";
2271                         column += 8;
2272                 }
2273                 break;
2274         }
2275
2276         string const begin_tag = "\\begin";
2277         InsetCode code = ownerCode();
2278         bool const lastpar = runparams.isLastPar;
2279
2280         switch (curAlign) {
2281         case LYX_ALIGN_NONE:
2282         case LYX_ALIGN_BLOCK:
2283         case LYX_ALIGN_LAYOUT:
2284         case LYX_ALIGN_SPECIAL:
2285         case LYX_ALIGN_DECIMAL:
2286                 break;
2287         case LYX_ALIGN_LEFT: {
2288                 string output;
2289                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2290                         output = corrected_env(begin_tag, "flushleft", code, lastpar);
2291                 else
2292                         output = corrected_env(begin_tag, "flushright", code, lastpar);
2293                 os << from_ascii(output);
2294                 adjust_column(output, column);
2295                 break;
2296         } case LYX_ALIGN_RIGHT: {
2297                 string output;
2298                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2299                         output = corrected_env(begin_tag, "flushright", code, lastpar);
2300                 else
2301                         output = corrected_env(begin_tag, "flushleft", code, lastpar);
2302                 os << from_ascii(output);
2303                 adjust_column(output, column);
2304                 break;
2305         } case LYX_ALIGN_CENTER: {
2306                 string output;
2307                 output = corrected_env(begin_tag, "center", code, lastpar);
2308                 os << from_ascii(output);
2309                 adjust_column(output, column);
2310                 break;
2311         }
2312         }
2313
2314         return column;
2315 }
2316
2317
2318 bool Paragraph::Private::endTeXParParams(BufferParams const & bparams,
2319                         otexstream & os, OutputParams const & runparams) const
2320 {
2321         LyXAlignment const curAlign = params_.align();
2322
2323         if (curAlign == layout_->align)
2324                 return false;
2325
2326         switch (curAlign) {
2327         case LYX_ALIGN_NONE:
2328         case LYX_ALIGN_BLOCK:
2329         case LYX_ALIGN_LAYOUT:
2330         case LYX_ALIGN_SPECIAL:
2331         case LYX_ALIGN_DECIMAL:
2332                 break;
2333         case LYX_ALIGN_LEFT:
2334         case LYX_ALIGN_RIGHT:
2335         case LYX_ALIGN_CENTER:
2336                 if (runparams.moving_arg)
2337                         os << "\\protect";
2338                 break;
2339         }
2340
2341         string output;
2342         string const end_tag = "\n\\par\\end";
2343         InsetCode code = ownerCode();
2344         bool const lastpar = runparams.isLastPar;
2345
2346         switch (curAlign) {
2347         case LYX_ALIGN_NONE:
2348         case LYX_ALIGN_BLOCK:
2349         case LYX_ALIGN_LAYOUT:
2350         case LYX_ALIGN_SPECIAL:
2351         case LYX_ALIGN_DECIMAL:
2352                 break;
2353         case LYX_ALIGN_LEFT: {
2354                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2355                         output = corrected_env(end_tag, "flushleft", code, lastpar);
2356                 else
2357                         output = corrected_env(end_tag, "flushright", code, lastpar);
2358                 os << from_ascii(output);
2359                 break;
2360         } case LYX_ALIGN_RIGHT: {
2361                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2362                         output = corrected_env(end_tag, "flushright", code, lastpar);
2363                 else
2364                         output = corrected_env(end_tag, "flushleft", code, lastpar);
2365                 os << from_ascii(output);
2366                 break;
2367         } case LYX_ALIGN_CENTER: {
2368                 output = corrected_env(end_tag, "center", code, lastpar);
2369                 os << from_ascii(output);
2370                 break;
2371         }
2372         }
2373
2374         return !output.empty() || lastpar;
2375 }
2376
2377
2378 // This one spits out the text of the paragraph
2379 void Paragraph::latex(BufferParams const & bparams,
2380         Font const & outerfont,
2381         otexstream & os,
2382         OutputParams const & runparams,
2383         int start_pos, int end_pos, bool force) const
2384 {
2385         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
2386
2387         // FIXME This check should not be needed. Perhaps issue an
2388         // error if it triggers.
2389         Layout const & style = inInset().forcePlainLayout() ?
2390                 bparams.documentClass().plainLayout() : *d->layout_;
2391
2392         if (!force && style.inpreamble)
2393                 return;
2394
2395         bool const allowcust = allowParagraphCustomization();
2396
2397         // Current base font for all inherited font changes, without any
2398         // change caused by an individual character, except for the language:
2399         // It is set to the language of the first character.
2400         // As long as we are in the label, this font is the base font of the
2401         // label. Before the first body character it is set to the base font
2402         // of the body.
2403         Font basefont;
2404
2405         // Maybe we have to create a optional argument.
2406         pos_type body_pos = beginOfBody();
2407         unsigned int column = 0;
2408
2409         if (body_pos > 0) {
2410                 // the optional argument is kept in curly brackets in
2411                 // case it contains a ']'
2412                 // This is not strictly needed, but if this is changed it
2413                 // would be a file format change, and tex2lyx would need
2414                 // to be adjusted, since it unconditionally removes the
2415                 // braces when it parses \item.
2416                 os << "[{";
2417                 column += 2;
2418                 basefont = getLabelFont(bparams, outerfont);
2419         } else {
2420                 basefont = getLayoutFont(bparams, outerfont);
2421         }
2422
2423         // Which font is currently active?
2424         Font running_font(basefont);
2425         // Do we have an open font change?
2426         bool open_font = false;
2427
2428         Change runningChange = Change(Change::UNCHANGED);
2429
2430         Encoding const * const prev_encoding = runparams.encoding;
2431
2432         os.texrow().start(id(), 0);
2433
2434         // if the paragraph is empty, the loop will not be entered at all
2435         if (empty()) {
2436                 if (style.isCommand()) {
2437                         os << '{';
2438                         ++column;
2439                 }
2440                 if (!style.leftdelim().empty()) {
2441                         os << style.leftdelim();
2442                         column += style.leftdelim().size();
2443                 }
2444                 if (allowcust)
2445                         column += d->startTeXParParams(bparams, os, runparams);
2446         }
2447
2448         for (pos_type i = 0; i < size(); ++i) {
2449                 // First char in paragraph or after label?
2450                 if (i == body_pos) {
2451                         if (body_pos > 0) {
2452                                 if (open_font) {
2453                                         column += running_font.latexWriteEndChanges(
2454                                                 os, bparams, runparams,
2455                                                 basefont, basefont);
2456                                         open_font = false;
2457                                 }
2458                                 basefont = getLayoutFont(bparams, outerfont);
2459                                 running_font = basefont;
2460
2461                                 column += Changes::latexMarkChange(os, bparams,
2462                                                 runningChange, Change(Change::UNCHANGED),
2463                                                 runparams);
2464                                 runningChange = Change(Change::UNCHANGED);
2465
2466                                 os << "}] ";
2467                                 column +=3;
2468                         }
2469                         if (style.isCommand()) {
2470                                 os << '{';
2471                                 ++column;
2472                         }
2473
2474                         if (!style.leftdelim().empty()) {
2475                                 os << style.leftdelim();
2476                                 column += style.leftdelim().size();
2477                         }
2478
2479                         if (allowcust)
2480                                 column += d->startTeXParParams(bparams, os,
2481                                                             runparams);
2482                 }
2483
2484                 Change const & change = runparams.inDeletedInset
2485                         ? runparams.changeOfDeletedInset : lookupChange(i);
2486
2487                 if (bparams.outputChanges && runningChange != change) {
2488                         if (open_font) {
2489                                 column += running_font.latexWriteEndChanges(
2490                                                 os, bparams, runparams, basefont, basefont);
2491                                 open_font = false;
2492                         }
2493                         basefont = getLayoutFont(bparams, outerfont);
2494                         running_font = basefont;
2495
2496                         column += Changes::latexMarkChange(os, bparams, runningChange,
2497                                                            change, runparams);
2498                         runningChange = change;
2499                 }
2500
2501                 // do not output text which is marked deleted
2502                 // if change tracking output is disabled
2503                 if (!bparams.outputChanges && change.deleted()) {
2504                         continue;
2505                 }
2506
2507                 ++column;
2508
2509                 // Fully instantiated font
2510                 Font const font = getFont(bparams, i, outerfont);
2511
2512                 Font const last_font = running_font;
2513
2514                 // Do we need to close the previous font?
2515                 if (open_font &&
2516                     (font != running_font ||
2517                      font.language() != running_font.language()))
2518                 {
2519                         column += running_font.latexWriteEndChanges(
2520                                         os, bparams, runparams, basefont,
2521                                         (i == body_pos-1) ? basefont : font);
2522                         running_font = basefont;
2523                         open_font = false;
2524                 }
2525
2526                 string const running_lang = runparams.use_polyglossia ?
2527                         running_font.language()->polyglossia() : running_font.language()->babel();
2528                 // close babel's font environment before opening CJK.
2529                 string const lang_end_command = runparams.use_polyglossia ?
2530                         "\\end{$$lang}" : lyxrc.language_command_end;
2531                 if (!running_lang.empty() &&
2532                     font.language()->encoding()->package() == Encoding::CJK) {
2533                                 string end_tag = subst(lang_end_command,
2534                                                         "$$lang",
2535                                                         running_lang);
2536                                 os << from_ascii(end_tag);
2537                                 column += end_tag.length();
2538                 }
2539
2540                 // Switch file encoding if necessary (and allowed)
2541                 if (!runparams.pass_thru && !style.pass_thru &&
2542                     runparams.encoding->package() != Encoding::none &&
2543                     font.language()->encoding()->package() != Encoding::none) {
2544                         pair<bool, int> const enc_switch =
2545                                 switchEncoding(os.os(), bparams, runparams,
2546                                         *(font.language()->encoding()));
2547                         if (enc_switch.first) {
2548                                 column += enc_switch.second;
2549                                 runparams.encoding = font.language()->encoding();
2550                         }
2551                 }
2552
2553                 char_type const c = d->text_[i];
2554
2555                 // Do we need to change font?
2556                 if ((font != running_font ||
2557                      font.language() != running_font.language()) &&
2558                         i != body_pos - 1)
2559                 {
2560                         odocstringstream ods;
2561                         column += font.latexWriteStartChanges(ods, bparams,
2562                                                               runparams, basefont,
2563                                                               last_font);
2564                         running_font = font;
2565                         open_font = true;
2566                         docstring fontchange = ods.str();
2567                         // check whether the fontchange ends with a \\textcolor
2568                         // modifier and the text starts with a space (bug 4473)
2569                         docstring const last_modifier = rsplit(fontchange, '\\');
2570                         if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2571                                 os << fontchange << from_ascii("{}");
2572                         // check if the fontchange ends with a trailing blank
2573                         // (like "\small " (see bug 3382)
2574                         else if (suffixIs(fontchange, ' ') && c == ' ')
2575                                 os << fontchange.substr(0, fontchange.size() - 1)
2576                                    << from_ascii("{}");
2577                         else
2578                                 os << fontchange;
2579                 }
2580
2581                 // FIXME: think about end_pos implementation...
2582                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2583                         // FIXME: integrate this case in latexSpecialChar
2584                         // Do not print the separation of the optional argument
2585                         // if style.pass_thru is false. This works because
2586                         // latexSpecialChar ignores spaces if
2587                         // style.pass_thru is false.
2588                         if (i != body_pos - 1) {
2589                                 if (d->simpleTeXBlanks(runparams, os,
2590                                                 i, column, font, style)) {
2591                                         // A surrogate pair was output. We
2592                                         // must not call latexSpecialChar
2593                                         // in this iteration, since it would output
2594                                         // the combining character again.
2595                                         ++i;
2596                                         continue;
2597                                 }
2598                         }
2599                 }
2600
2601                 OutputParams rp = runparams;
2602                 rp.free_spacing = style.free_spacing;
2603                 rp.local_font = &font;
2604                 rp.intitle = style.intitle;
2605
2606                 // Two major modes:  LaTeX or plain
2607                 // Handle here those cases common to both modes
2608                 // and then split to handle the two modes separately.
2609                 if (c == META_INSET) {
2610                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2611                                 d->latexInset(bparams, os, rp, running_font,
2612                                                 basefont, outerfont, open_font,
2613                                                 runningChange, style, i, column);
2614                         }
2615                 } else {
2616                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2617                                 try {
2618                                         d->latexSpecialChar(os, bparams, rp, running_font, runningChange,
2619                                                             style, i, end_pos, column);
2620                                 } catch (EncodingException & e) {
2621                                 if (runparams.dryrun) {
2622                                         os << "<" << _("LyX Warning: ")
2623                                            << _("uncodable character") << " '";
2624                                         os.put(c);
2625                                         os << "'>";
2626                                 } else {
2627                                         // add location information and throw again.
2628                                         e.par_id = id();
2629                                         e.pos = i;
2630                                         throw(e);
2631                                 }
2632                         }
2633                 }
2634                 }
2635
2636                 // Set the encoding to that returned from latexSpecialChar (see
2637                 // comment for encoding member in OutputParams.h)
2638                 runparams.encoding = rp.encoding;
2639         }
2640
2641         // If we have an open font definition, we have to close it
2642         if (open_font) {
2643 #ifdef FIXED_LANGUAGE_END_DETECTION
2644                 if (next_) {
2645                         running_font.latexWriteEndChanges(os, bparams,
2646                                         runparams, basefont,
2647                                         next_->getFont(bparams, 0, outerfont));
2648                 } else {
2649                         running_font.latexWriteEndChanges(os, bparams,
2650                                         runparams, basefont, basefont);
2651                 }
2652 #else
2653 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2654 //FIXME: there as we start another \selectlanguage with the next paragraph if
2655 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2656                 running_font.latexWriteEndChanges(os, bparams, runparams,
2657                                 basefont, basefont);
2658 #endif
2659         }
2660
2661         column += Changes::latexMarkChange(os, bparams, runningChange,
2662                                            Change(Change::UNCHANGED), runparams);
2663
2664         // Needed if there is an optional argument but no contents.
2665         if (body_pos > 0 && body_pos == size()) {
2666                 os << "}]~";
2667         }
2668
2669         if (!style.rightdelim().empty()) {
2670                 os << style.rightdelim();
2671                 column += style.rightdelim().size();
2672         }
2673
2674         if (allowcust && d->endTeXParParams(bparams, os, runparams)
2675             && runparams.encoding != prev_encoding) {
2676                 runparams.encoding = prev_encoding;
2677                 if (!runparams.isFullUnicode())
2678                         os << setEncoding(prev_encoding->iconvName());
2679         }
2680
2681         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2682 }
2683
2684
2685 bool Paragraph::emptyTag() const
2686 {
2687         for (pos_type i = 0; i < size(); ++i) {
2688                 if (Inset const * inset = getInset(i)) {
2689                         InsetCode lyx_code = inset->lyxCode();
2690                         // FIXME testing like that is wrong. What is
2691                         // the intent?
2692                         if (lyx_code != TOC_CODE &&
2693                             lyx_code != INCLUDE_CODE &&
2694                             lyx_code != GRAPHICS_CODE &&
2695                             lyx_code != ERT_CODE &&
2696                             lyx_code != LISTINGS_CODE &&
2697                             lyx_code != FLOAT_CODE &&
2698                             lyx_code != TABULAR_CODE) {
2699                                 return false;
2700                         }
2701                 } else {
2702                         char_type c = d->text_[i];
2703                         if (c != ' ' && c != '\t')
2704                                 return false;
2705                 }
2706         }
2707         return true;
2708 }
2709
2710
2711 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
2712         const
2713 {
2714         for (pos_type i = 0; i < size(); ++i) {
2715                 if (Inset const * inset = getInset(i)) {
2716                         InsetCode lyx_code = inset->lyxCode();
2717                         if (lyx_code == LABEL_CODE) {
2718                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2719                                 docstring const & id = il->getParam("name");
2720                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2721                         }
2722                 }
2723         }
2724         return string();
2725 }
2726
2727
2728 pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runparams)
2729         const
2730 {
2731         pos_type i;
2732         for (i = 0; i < size(); ++i) {
2733                 if (Inset const * inset = getInset(i)) {
2734                         inset->docbook(os, runparams);
2735                 } else {
2736                         char_type c = d->text_[i];
2737                         if (c == ' ')
2738                                 break;
2739                         os << sgml::escapeChar(c);
2740                 }
2741         }
2742         return i;
2743 }
2744
2745
2746 pos_type Paragraph::firstWordLyXHTML(XHTMLStream & xs, OutputParams const & runparams)
2747         const
2748 {
2749         pos_type i;
2750         for (i = 0; i < size(); ++i) {
2751                 if (Inset const * inset = getInset(i)) {
2752                         inset->xhtml(xs, runparams);
2753                 } else {
2754                         char_type c = d->text_[i];
2755                         if (c == ' ')
2756                                 break;
2757                         xs << c;
2758                 }
2759         }
2760         return i;
2761 }
2762
2763
2764 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2765 {
2766         Font font_old;
2767         pos_type size = text_.size();
2768         for (pos_type i = initial; i < size; ++i) {
2769                 Font font = owner_->getFont(buf.params(), i, outerfont);
2770                 if (text_[i] == META_INSET)
2771                         return false;
2772                 if (i != initial && font != font_old)
2773                         return false;
2774                 font_old = font;
2775         }
2776
2777         return true;
2778 }
2779
2780
2781 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
2782                                     odocstream & os,
2783                                     OutputParams const & runparams,
2784                                     Font const & outerfont,
2785                                     pos_type initial) const
2786 {
2787         bool emph_flag = false;
2788
2789         Layout const & style = *d->layout_;
2790         FontInfo font_old =
2791                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2792
2793         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2794                 os << "]]>";
2795
2796         // parsing main loop
2797         for (pos_type i = initial; i < size(); ++i) {
2798                 Font font = getFont(buf.params(), i, outerfont);
2799
2800                 // handle <emphasis> tag
2801                 if (font_old.emph() != font.fontInfo().emph()) {
2802                         if (font.fontInfo().emph() == FONT_ON) {
2803                                 os << "<emphasis>";
2804                                 emph_flag = true;
2805                         } else if (i != initial) {
2806                                 os << "</emphasis>";
2807                                 emph_flag = false;
2808                         }
2809                 }
2810
2811                 if (Inset const * inset = getInset(i)) {
2812                         inset->docbook(os, runparams);
2813                 } else {
2814                         char_type c = d->text_[i];
2815
2816                         if (style.pass_thru)
2817                                 os.put(c);
2818                         else
2819                                 os << sgml::escapeChar(c);
2820                 }
2821                 font_old = font.fontInfo();
2822         }
2823
2824         if (emph_flag) {
2825                 os << "</emphasis>";
2826         }
2827
2828         if (style.free_spacing)
2829                 os << '\n';
2830         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2831                 os << "<![CDATA[";
2832 }
2833
2834
2835 void doFontSwitch(XHTMLStream & xs, bool startrange,
2836         bool & flag, FontState curstate, std::string tag, std::string attr = "")
2837 {
2838         if (curstate == FONT_ON) {
2839                 xs << html::StartTag(tag, attr);
2840                 flag = true;
2841         } else if (flag && !startrange) {
2842                 xs << html::EndTag(tag);
2843                 flag = false;
2844         }
2845 }
2846
2847
2848 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
2849                                     XHTMLStream & xs,
2850                                     OutputParams const & runparams,
2851                                     Font const & outerfont,
2852                                     pos_type initial) const
2853 {
2854         docstring retval;
2855
2856         bool emph_flag = false;
2857         bool bold_flag = false;
2858         bool noun_flag = false;
2859         bool ubar_flag = false;
2860         bool dbar_flag = false;
2861         bool sout_flag = false;
2862         bool wave_flag = false;
2863
2864         Layout const & style = *d->layout_;
2865
2866         xs.startParagraph(allowEmpty());
2867
2868         if (!runparams.for_toc && runparams.html_make_pars) {
2869                 // generate a magic label for this paragraph
2870                 string const attr = "id='" + magicLabel() + "'";
2871                 xs << html::CompTag("a", attr);
2872         }
2873
2874         FontInfo font_old =
2875                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2876
2877         // parsing main loop
2878         for (pos_type i = initial; i < size(); ++i) {
2879                 // let's not show deleted material in the output
2880                 if (isDeleted(i))
2881                         continue;
2882
2883                 Font font = getFont(buf.masterBuffer()->params(), i, outerfont);
2884                 bool const at_start = (i == initial);
2885
2886                 // emphasis
2887                 FontState curstate = font.fontInfo().emph();
2888                 if (font_old.emph() != curstate)
2889                         doFontSwitch(xs, at_start, emph_flag, curstate, "em");
2890
2891                 // noun
2892                 curstate = font.fontInfo().noun();
2893                 if (font_old.noun() != curstate)
2894                         doFontSwitch(xs, at_start, noun_flag, curstate, "dfn", "class='lyxnoun'");
2895
2896                 // underbar
2897                 curstate = font.fontInfo().underbar();
2898                 if (font_old.underbar() != curstate)
2899                         doFontSwitch(xs, at_start, ubar_flag, curstate, "u");
2900
2901                 // strikeout
2902                 curstate = font.fontInfo().strikeout();
2903                 if (font_old.strikeout() != curstate)
2904                         doFontSwitch(xs, at_start, sout_flag, curstate, "del", "class='strikeout'");
2905
2906                 // HTML does not really have an equivalent of the next two, so we will just
2907                 // output a single underscore with a class, and people can style it if they
2908                 // wish to do so
2909
2910                 // double underbar
2911                 curstate = font.fontInfo().uuline();
2912                 if (font_old.uuline() != curstate)
2913                         doFontSwitch(xs, at_start, dbar_flag, curstate, "u", "class='dline'");
2914
2915                 // wavy line
2916                 curstate = font.fontInfo().uwave();
2917                 if (font_old.uwave() != curstate)
2918                         doFontSwitch(xs, at_start, wave_flag, curstate, "u", "class='wavyline'");
2919
2920                 // bold
2921                 if (font_old.series() != font.fontInfo().series()) {
2922                         if (font.fontInfo().series() == BOLD_SERIES) {
2923                                 xs << html::StartTag("b");
2924                                 bold_flag = true;
2925                         } else if (bold_flag && i != initial) {
2926                                 xs << html::EndTag("b");
2927                                 bold_flag = false;
2928                         }
2929                 }
2930                 // FIXME XHTML
2931                 // Other such tags? What about the other text ranges?
2932
2933                 Inset const * inset = getInset(i);
2934                 if (inset) {
2935                         if (!runparams.for_toc || inset->isInToc()) {
2936                                 OutputParams np = runparams;
2937                                 np.local_font = &font;
2938                                 if (!inset->getLayout().htmlisblock())
2939                                         np.html_in_par = true;
2940                                 retval += inset->xhtml(xs, np);
2941                         }
2942                 } else {
2943                         char_type c = getUChar(buf.masterBuffer()->params(), i);
2944
2945                         if (style.pass_thru || runparams.pass_thru)
2946                                 xs << c;
2947                         else if (c == '-') {
2948                                 docstring str;
2949                                 int j = i + 1;
2950                                 if (j < size() && d->text_[j] == '-') {
2951                                         j += 1;
2952                                         if (j < size() && d->text_[j] == '-') {
2953                                                 str += from_ascii("&mdash;");
2954                                                 i += 2;
2955                                         } else {
2956                                                 str += from_ascii("&ndash;");
2957                                                 i += 1;
2958                                         }
2959                                 }
2960                                 else
2961                                         str += c;
2962                                 // We don't want to escape the entities. Note that
2963                                 // it is safe to do this, since str can otherwise
2964                                 // only be "-". E.g., it can't be "<".
2965                                 xs << XHTMLStream::ESCAPE_NONE << str;
2966                         } else
2967                                 xs << c;
2968                 }
2969                 font_old = font.fontInfo();
2970         }
2971
2972         xs.closeFontTags();
2973         xs.endParagraph();
2974         return retval;
2975 }
2976
2977
2978 bool Paragraph::isHfill(pos_type pos) const
2979 {
2980         Inset const * inset = getInset(pos);
2981         return inset && (inset->lyxCode() == SPACE_CODE &&
2982                          inset->isStretchableSpace());
2983 }
2984
2985
2986 bool Paragraph::isNewline(pos_type pos) const
2987 {
2988         Inset const * inset = getInset(pos);
2989         return inset && inset->lyxCode() == NEWLINE_CODE;
2990 }
2991
2992
2993 bool Paragraph::isLineSeparator(pos_type pos) const
2994 {
2995         char_type const c = d->text_[pos];
2996         if (isLineSeparatorChar(c))
2997                 return true;
2998         Inset const * inset = getInset(pos);
2999         return inset && inset->isLineSeparator();
3000 }
3001
3002
3003 bool Paragraph::isWordSeparator(pos_type pos) const
3004 {
3005         if (pos == size())
3006                 return true;
3007         if (Inset const * inset = getInset(pos))
3008                 return !inset->isLetter();
3009         // if we have a hard hyphen (no en- or emdash) or apostrophe
3010         // we pass this to the spell checker
3011         // FIXME: this method is subject to change, visit
3012         // https://bugzilla.mozilla.org/show_bug.cgi?id=355178
3013         // to get an impression how complex this is.
3014         if (isHardHyphenOrApostrophe(pos))
3015                 return false;
3016         char_type const c = d->text_[pos];
3017         // We want to pass the escape chars to the spellchecker
3018         docstring const escape_chars = from_utf8(lyxrc.spellchecker_esc_chars);
3019         return !isLetterChar(c) && !isDigitASCII(c) && !contains(escape_chars, c);
3020 }
3021
3022
3023 bool Paragraph::isHardHyphenOrApostrophe(pos_type pos) const
3024 {
3025         pos_type const psize = size();
3026         if (pos >= psize)
3027                 return false;
3028         char_type const c = d->text_[pos];
3029         if (c != '-' && c != '\'')
3030                 return false;
3031         int nextpos = pos + 1;
3032         int prevpos = pos > 0 ? pos - 1 : 0;
3033         if ((nextpos == psize || isSpace(nextpos))
3034                 && (pos == 0 || isSpace(prevpos)))
3035                 return false;
3036         return c == '\''
3037                 || ((nextpos == psize || d->text_[nextpos] != '-')
3038                 && (pos == 0 || d->text_[prevpos] != '-'));
3039 }
3040
3041
3042 bool Paragraph::isSameSpellRange(pos_type pos1, pos_type pos2) const
3043 {
3044         return pos1 == pos2
3045                 || d->speller_state_.getRange(pos1) == d->speller_state_.getRange(pos2);
3046 }
3047
3048
3049 bool Paragraph::isChar(pos_type pos) const
3050 {
3051         if (Inset const * inset = getInset(pos))
3052                 return inset->isChar();
3053         char_type const c = d->text_[pos];
3054         return !isLetterChar(c) && !isDigitASCII(c) && !lyx::isSpace(c);
3055 }
3056
3057
3058 bool Paragraph::isSpace(pos_type pos) const
3059 {
3060         if (Inset const * inset = getInset(pos))
3061                 return inset->isSpace();
3062         char_type const c = d->text_[pos];
3063         return lyx::isSpace(c);
3064 }
3065
3066
3067 Language const *
3068 Paragraph::getParLanguage(BufferParams const & bparams) const
3069 {
3070         if (!empty())
3071                 return getFirstFontSettings(bparams).language();
3072         // FIXME: we should check the prev par as well (Lgb)
3073         return bparams.language;
3074 }
3075
3076
3077 bool Paragraph::isRTL(BufferParams const & bparams) const
3078 {
3079         return lyxrc.rtl_support
3080                 && getParLanguage(bparams)->rightToLeft()
3081                 && !inInset().getLayout().forceLTR();
3082 }
3083
3084
3085 void Paragraph::changeLanguage(BufferParams const & bparams,
3086                                Language const * from, Language const * to)
3087 {
3088         // change language including dummy font change at the end
3089         for (pos_type i = 0; i <= size(); ++i) {
3090                 Font font = getFontSettings(bparams, i);
3091                 if (font.language() == from) {
3092                         font.setLanguage(to);
3093                         setFont(i, font);
3094                         d->requestSpellCheck(i);
3095                 }
3096         }
3097 }
3098
3099
3100 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
3101 {
3102         Language const * doc_language = bparams.language;
3103         FontList::const_iterator cit = d->fontlist_.begin();
3104         FontList::const_iterator end = d->fontlist_.end();
3105
3106         for (; cit != end; ++cit)
3107                 if (cit->font().language() != ignore_language &&
3108                     cit->font().language() != latex_language &&
3109                     cit->font().language() != doc_language)
3110                         return true;
3111         return false;
3112 }
3113
3114
3115 void Paragraph::getLanguages(std::set<Language const *> & languages) const
3116 {
3117         FontList::const_iterator cit = d->fontlist_.begin();
3118         FontList::const_iterator end = d->fontlist_.end();
3119
3120         for (; cit != end; ++cit) {
3121                 Language const * lang = cit->font().language();
3122                 if (lang != ignore_language &&
3123                     lang != latex_language)
3124                         languages.insert(lang);
3125         }
3126 }
3127
3128
3129 docstring Paragraph::asString(int options) const
3130 {
3131         return asString(0, size(), options);
3132 }
3133
3134
3135 docstring Paragraph::asString(pos_type beg, pos_type end, int options) const
3136 {
3137         odocstringstream os;
3138
3139         if (beg == 0
3140             && options & AS_STR_LABEL
3141             && !d->params_.labelString().empty())
3142                 os << d->params_.labelString() << ' ';
3143
3144         for (pos_type i = beg; i < end; ++i) {
3145                 if ((options & AS_STR_SKIPDELETE) && isDeleted(i))
3146                         continue;
3147                 char_type const c = d->text_[i];
3148                 if (isPrintable(c) || c == '\t'
3149                     || (c == '\n' && (options & AS_STR_NEWLINES)))
3150                         os.put(c);
3151                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
3152                         getInset(i)->toString(os);
3153                         if (getInset(i)->asInsetMath())
3154                                 os << " ";
3155                 }
3156         }
3157
3158         return os.str();
3159 }
3160
3161
3162 void Paragraph::forToc(docstring & os, size_t maxlen) const
3163 {
3164         if (!d->params_.labelString().empty())
3165                 os += d->params_.labelString() + ' ';
3166         for (pos_type i = 0; i < size() && os.length() < maxlen; ++i) {
3167                 if (isDeleted(i))
3168                         continue;
3169                 char_type const c = d->text_[i];
3170                 if (isPrintable(c))
3171                         os += c;
3172                 else if (c == '\t' || c == '\n')
3173                         os += ' ';
3174                 else if (c == META_INSET)
3175                         getInset(i)->forToc(os, maxlen);
3176         }
3177 }
3178
3179
3180 docstring Paragraph::stringify(pos_type beg, pos_type end, int options,
3181         OutputParams const & runparams) const
3182 {
3183         odocstringstream os;
3184
3185         if (beg == 0
3186                 && options & AS_STR_LABEL
3187                 && !d->params_.labelString().empty())
3188                 os << d->params_.labelString() << ' ';
3189
3190         OutputParams op = runparams;
3191         op.for_search = true;
3192
3193         for (pos_type i = beg; i < end; ++i) {
3194                 char_type const c = d->text_[i];
3195                 if (isPrintable(c) || c == '\t'
3196                     || (c == '\n' && (options & AS_STR_NEWLINES)))
3197                         os.put(c);
3198                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
3199                         getInset(i)->plaintext(os, op);
3200                 }
3201         }
3202
3203         return os.str();
3204 }
3205
3206
3207 void Paragraph::setInsetOwner(Inset const * inset)
3208 {
3209         d->inset_owner_ = inset;
3210 }
3211
3212
3213 int Paragraph::id() const
3214 {
3215         return d->id_;
3216 }
3217
3218
3219 void Paragraph::setId(int id)
3220 {
3221         d->id_ = id;
3222 }
3223
3224
3225 Layout const & Paragraph::layout() const
3226 {
3227         return *d->layout_;
3228 }
3229
3230
3231 void Paragraph::setLayout(Layout const & layout)
3232 {
3233         d->layout_ = &layout;
3234 }
3235
3236
3237 void Paragraph::setDefaultLayout(DocumentClass const & tc)
3238 {
3239         setLayout(tc.defaultLayout());
3240 }
3241
3242
3243 void Paragraph::setPlainLayout(DocumentClass const & tc)
3244 {
3245         setLayout(tc.plainLayout());
3246 }
3247
3248
3249 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
3250 {
3251         if (usePlainLayout())
3252                 setPlainLayout(tclass);
3253         else
3254                 setDefaultLayout(tclass);
3255 }
3256
3257
3258 Inset const & Paragraph::inInset() const
3259 {
3260         LBUFERR(d->inset_owner_, _("Paragraph not properly initialized"));
3261         return *d->inset_owner_;
3262 }
3263
3264
3265 ParagraphParameters & Paragraph::params()
3266 {
3267         return d->params_;
3268 }
3269
3270
3271 ParagraphParameters const & Paragraph::params() const
3272 {
3273         return d->params_;
3274 }
3275
3276
3277 bool Paragraph::isFreeSpacing() const
3278 {
3279         if (d->layout_->free_spacing)
3280                 return true;
3281         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
3282 }
3283
3284
3285 bool Paragraph::allowEmpty() const
3286 {
3287         if (d->layout_->keepempty)
3288                 return true;
3289         return d->inset_owner_ && d->inset_owner_->allowEmpty();
3290 }
3291
3292
3293 char_type Paragraph::transformChar(char_type c, pos_type pos) const
3294 {
3295         if (!Encodings::isArabicChar(c))
3296                 return c;
3297
3298         char_type prev_char = ' ';
3299         char_type next_char = ' ';
3300
3301         for (pos_type i = pos - 1; i >= 0; --i) {
3302                 char_type const par_char = d->text_[i];
3303                 if (!Encodings::isArabicComposeChar(par_char)) {
3304                         prev_char = par_char;
3305                         break;
3306                 }
3307         }
3308
3309         for (pos_type i = pos + 1, end = size(); i < end; ++i) {
3310                 char_type const par_char = d->text_[i];
3311                 if (!Encodings::isArabicComposeChar(par_char)) {
3312                         next_char = par_char;
3313                         break;
3314                 }
3315         }
3316
3317         if (Encodings::isArabicChar(next_char)) {
3318                 if (Encodings::isArabicChar(prev_char) &&
3319                         !Encodings::isArabicSpecialChar(prev_char))
3320                         return Encodings::transformChar(c, Encodings::FORM_MEDIAL);
3321                 else
3322                         return Encodings::transformChar(c, Encodings::FORM_INITIAL);
3323         } else {
3324                 if (Encodings::isArabicChar(prev_char) &&
3325                         !Encodings::isArabicSpecialChar(prev_char))
3326                         return Encodings::transformChar(c, Encodings::FORM_FINAL);
3327                 else
3328                         return Encodings::transformChar(c, Encodings::FORM_ISOLATED);
3329         }
3330 }
3331
3332
3333 bool Paragraph::brokenBiblio() const
3334 {
3335         // there is a problem if there is no bibitem at position 0 or
3336         // if there is another bibitem in the paragraph.
3337         return d->layout_->labeltype == LABEL_BIBLIO
3338                 && (d->insetlist_.find(BIBITEM_CODE) != 0
3339                     || d->insetlist_.find(BIBITEM_CODE, 1) > 0);
3340 }
3341
3342
3343 int Paragraph::fixBiblio(Buffer const & buffer)
3344 {
3345         // FIXME: What about the case where paragraph is not BIBLIO
3346         // but there is an InsetBibitem?
3347         // FIXME: when there was already an inset at 0, the return value is 1,
3348         // which does not tell whether another inset has been remove; the
3349         // cursor cannot be correctly updated.
3350
3351         if (d->layout_->labeltype != LABEL_BIBLIO)
3352                 return 0;
3353
3354         bool const track_changes = buffer.params().trackChanges;
3355         int bibitem_pos = d->insetlist_.find(BIBITEM_CODE);
3356         bool const hasbibitem0 = bibitem_pos == 0;
3357
3358         if (hasbibitem0) {
3359                 bibitem_pos = d->insetlist_.find(BIBITEM_CODE, 1);
3360                 // There was an InsetBibitem at pos 0, and no other one => OK
3361                 if (bibitem_pos == -1)
3362                         return 0;
3363                 // there is a bibitem at the 0 position, but since
3364                 // there is a second one, we copy the second on the
3365                 // first. We're assuming there are at most two of
3366                 // these, which there should be.
3367                 // FIXME: why does it make sense to do that rather
3368                 // than keep the first? (JMarc)
3369                 Inset * inset = d->insetlist_.release(bibitem_pos);
3370                 eraseChar(bibitem_pos, track_changes);
3371                 d->insetlist_.begin()->inset = inset;
3372                 return -bibitem_pos;
3373         }
3374
3375         // We need to create an inset at the beginning
3376         Inset * inset = 0;
3377         if (bibitem_pos > 0) {
3378                 // there was one somewhere in the paragraph, let's move it
3379                 inset = d->insetlist_.release(bibitem_pos);
3380                 eraseChar(bibitem_pos, track_changes);
3381         } else
3382                 // make a fresh one
3383                 inset = new InsetBibitem(const_cast<Buffer *>(&buffer),
3384                                          InsetCommandParams(BIBITEM_CODE));
3385
3386         insertInset(0, inset, Change(track_changes ? Change::INSERTED
3387                                                    : Change::UNCHANGED));
3388
3389         return 1;
3390 }
3391
3392
3393 void Paragraph::checkAuthors(AuthorList const & authorList)
3394 {
3395         d->changes_.checkAuthors(authorList);
3396 }
3397
3398
3399 bool Paragraph::isChanged(pos_type pos) const
3400 {
3401         return lookupChange(pos).changed();
3402 }
3403
3404
3405 bool Paragraph::isInserted(pos_type pos) const
3406 {
3407         return lookupChange(pos).inserted();
3408 }
3409
3410
3411 bool Paragraph::isDeleted(pos_type pos) const
3412 {
3413         return lookupChange(pos).deleted();
3414 }
3415
3416
3417 InsetList const & Paragraph::insetList() const
3418 {
3419         return d->insetlist_;
3420 }
3421
3422
3423 void Paragraph::setBuffer(Buffer & b)
3424 {
3425         d->insetlist_.setBuffer(b);
3426 }
3427
3428
3429 Inset * Paragraph::releaseInset(pos_type pos)
3430 {
3431         Inset * inset = d->insetlist_.release(pos);
3432         /// does not honour change tracking!
3433         eraseChar(pos, false);
3434         return inset;
3435 }
3436
3437
3438 Inset * Paragraph::getInset(pos_type pos)
3439 {
3440         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3441                  ? d->insetlist_.get(pos) : 0;
3442 }
3443
3444
3445 Inset const * Paragraph::getInset(pos_type pos) const
3446 {
3447         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3448                  ? d->insetlist_.get(pos) : 0;
3449 }
3450
3451
3452 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
3453                 pos_type & right, TextCase action)
3454 {
3455         // process sequences of modified characters; in change
3456         // tracking mode, this approach results in much better
3457         // usability than changing case on a char-by-char basis
3458         docstring changes;
3459
3460         bool const trackChanges = bparams.trackChanges;
3461
3462         bool capitalize = true;
3463
3464         for (; pos < right; ++pos) {
3465                 char_type oldChar = d->text_[pos];
3466                 char_type newChar = oldChar;
3467
3468                 // ignore insets and don't play with deleted text!
3469                 if (oldChar != META_INSET && !isDeleted(pos)) {
3470                         switch (action) {
3471                                 case text_lowercase:
3472                                         newChar = lowercase(oldChar);
3473                                         break;
3474                                 case text_capitalization:
3475                                         if (capitalize) {
3476                                                 newChar = uppercase(oldChar);
3477                                                 capitalize = false;
3478                                         }
3479                                         break;
3480                                 case text_uppercase:
3481                                         newChar = uppercase(oldChar);
3482                                         break;
3483                         }
3484                 }
3485
3486                 if (isWordSeparator(pos) || isDeleted(pos)) {
3487                         // permit capitalization again
3488                         capitalize = true;
3489                 }
3490
3491                 if (oldChar != newChar) {
3492                         changes += newChar;
3493                         if (pos != right - 1)
3494                                 continue;
3495                         // step behind the changing area
3496                         pos++;
3497                 }
3498
3499                 int erasePos = pos - changes.size();
3500                 for (size_t i = 0; i < changes.size(); i++) {
3501                         insertChar(pos, changes[i],
3502                                    getFontSettings(bparams,
3503                                                    erasePos),
3504                                    trackChanges);
3505                         if (!eraseChar(erasePos, trackChanges)) {
3506                                 ++erasePos;
3507                                 ++pos; // advance
3508                                 ++right; // expand selection
3509                         }
3510                 }
3511                 changes.clear();
3512         }
3513 }
3514
3515
3516 int Paragraph::find(docstring const & str, bool cs, bool mw,
3517                 pos_type start_pos, bool del) const
3518 {
3519         pos_type pos = start_pos;
3520         int const strsize = str.length();
3521         int i = 0;
3522         pos_type const parsize = d->text_.size();
3523         for (i = 0; i < strsize && pos < parsize; ++i, ++pos) {
3524                 // Ignore "invisible" letters such as ligature breaks
3525                 // and hyphenation chars while searching
3526                 while (pos < parsize - 1 && isInset(pos)) {
3527                         odocstringstream os;
3528                         getInset(pos)->toString(os);
3529                         if (!getInset(pos)->isLetter() || !os.str().empty())
3530                                 break;
3531                         pos++;
3532                 }
3533                 if (cs && str[i] != d->text_[pos])
3534                         break;
3535                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos]))
3536                         break;
3537                 if (!del && isDeleted(pos))
3538                         break;
3539         }
3540
3541         if (i != strsize)
3542                 return 0;
3543
3544         // if necessary, check whether string matches word
3545         if (mw) {
3546                 if (start_pos > 0 && !isWordSeparator(start_pos - 1))
3547                         return 0;
3548                 if (pos < parsize
3549                         && !isWordSeparator(pos))
3550                         return 0;
3551         }
3552
3553         return pos - start_pos;
3554 }
3555
3556
3557 char_type Paragraph::getChar(pos_type pos) const
3558 {
3559         return d->text_[pos];
3560 }
3561
3562
3563 pos_type Paragraph::size() const
3564 {
3565         return d->text_.size();
3566 }
3567
3568
3569 bool Paragraph::empty() const
3570 {
3571         return d->text_.empty();
3572 }
3573
3574
3575 bool Paragraph::isInset(pos_type pos) const
3576 {
3577         return d->text_[pos] == META_INSET;
3578 }
3579
3580
3581 bool Paragraph::isSeparator(pos_type pos) const
3582 {
3583         //FIXME: Are we sure this can be the only separator?
3584         return d->text_[pos] == ' ';
3585 }
3586
3587
3588 void Paragraph::deregisterWords()
3589 {
3590         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3591         Private::LangWordsMap::const_iterator ite = d->words_.end();
3592         for (; itl != ite; ++itl) {
3593                 WordList * wl = theWordList(itl->first);
3594                 Private::Words::const_iterator it = (itl->second).begin();
3595                 Private::Words::const_iterator et = (itl->second).end();
3596                 for (; it != et; ++it)
3597                         wl->remove(*it);
3598         }
3599         d->words_.clear();
3600 }
3601
3602
3603 void Paragraph::locateWord(pos_type & from, pos_type & to,
3604         word_location const loc) const
3605 {
3606         switch (loc) {
3607         case WHOLE_WORD_STRICT:
3608                 if (from == 0 || from == size()
3609                     || isWordSeparator(from)
3610                     || isWordSeparator(from - 1)) {
3611                         to = from;
3612                         return;
3613                 }
3614                 // no break here, we go to the next
3615
3616         case WHOLE_WORD:
3617                 // If we are already at the beginning of a word, do nothing
3618                 if (!from || isWordSeparator(from - 1))
3619                         break;
3620                 // no break here, we go to the next
3621
3622         case PREVIOUS_WORD:
3623                 // always move the cursor to the beginning of previous word
3624                 while (from && !isWordSeparator(from - 1))
3625                         --from;
3626                 break;
3627         case NEXT_WORD:
3628                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
3629                 break;
3630         case PARTIAL_WORD:
3631                 // no need to move the 'from' cursor
3632                 break;
3633         }
3634         to = from;
3635         while (to < size() && !isWordSeparator(to))
3636                 ++to;
3637 }
3638
3639
3640 void Paragraph::collectWords()
3641 {
3642         pos_type n = size();
3643         for (pos_type pos = 0; pos < n; ++pos) {
3644                 if (isWordSeparator(pos))
3645                         continue;
3646                 pos_type from = pos;
3647                 locateWord(from, pos, WHOLE_WORD);
3648                 if ((pos - from) >= (int)lyxrc.completion_minlength) {
3649                         docstring word = asString(from, pos, AS_STR_NONE);
3650                         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
3651                         if (cit == d->fontlist_.end())
3652                                 return;
3653                         Language const * lang = cit->font().language();
3654                         d->words_[*lang].insert(word);
3655                 }
3656         }
3657 }
3658
3659
3660 void Paragraph::registerWords()
3661 {
3662         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3663         Private::LangWordsMap::const_iterator ite = d->words_.end();
3664         for (; itl != ite; ++itl) {
3665                 WordList * wl = theWordList(itl->first);
3666                 Private::Words::const_iterator it = (itl->second).begin();
3667                 Private::Words::const_iterator et = (itl->second).end();
3668                 for (; it != et; ++it)
3669                         wl->insert(*it);
3670         }
3671 }
3672
3673
3674 void Paragraph::updateWords()
3675 {
3676         deregisterWords();
3677         collectWords();
3678         registerWords();
3679 }
3680
3681
3682 void Paragraph::Private::appendSkipPosition(SkipPositions & skips, pos_type const pos) const
3683 {
3684         SkipPositionsIterator begin = skips.begin();
3685         SkipPositions::iterator end = skips.end();
3686         if (pos > 0 && begin < end) {
3687                 --end;
3688                 if (end->last == pos - 1) {
3689                         end->last = pos;
3690                         return;
3691                 }
3692         }
3693         skips.insert(end, FontSpan(pos, pos));
3694 }
3695
3696
3697 Language * Paragraph::Private::locateSpellRange(
3698         pos_type & from, pos_type & to,
3699         SkipPositions & skips) const
3700 {
3701         // skip leading white space
3702         while (from < to && owner_->isWordSeparator(from))
3703                 ++from;
3704         // don't check empty range
3705         if (from >= to)
3706                 return 0;
3707         // get current language
3708         Language * lang = getSpellLanguage(from);
3709         pos_type last = from;
3710         bool samelang = true;
3711         bool sameinset = true;
3712         while (last < to && samelang && sameinset) {
3713                 // hop to end of word
3714                 while (last < to && !owner_->isWordSeparator(last)) {
3715                         if (owner_->getInset(last)) {
3716                                 appendSkipPosition(skips, last);
3717                         } else if (owner_->isDeleted(last)) {
3718                                 appendSkipPosition(skips, last);
3719                         }
3720                         ++last;
3721                 }
3722                 // hop to next word while checking for insets
3723                 while (sameinset && last < to && owner_->isWordSeparator(last)) {
3724                         if (Inset const * inset = owner_->getInset(last))
3725                                 sameinset = inset->isChar() && inset->isLetter();
3726                         if (sameinset && owner_->isDeleted(last)) {
3727                                 appendSkipPosition(skips, last);
3728                         }
3729                         if (sameinset)
3730                                 last++;
3731                 }
3732                 if (sameinset && last < to) {
3733                         // now check for language change
3734                         samelang = lang == getSpellLanguage(last);
3735                 }
3736         }
3737         // if language change detected backstep is needed
3738         if (!samelang)
3739                 --last;
3740         to = last;
3741         return lang;
3742 }
3743
3744
3745 Language * Paragraph::Private::getSpellLanguage(pos_type const from) const
3746 {
3747         Language * lang =
3748                 const_cast<Language *>(owner_->getFontSettings(
3749                         inset_owner_->buffer().params(), from).language());
3750         if (lang == inset_owner_->buffer().params().language
3751                 && !lyxrc.spellchecker_alt_lang.empty()) {
3752                 string lang_code;
3753                 string const lang_variety =
3754                         split(lyxrc.spellchecker_alt_lang, lang_code, '-');
3755                 lang->setCode(lang_code);
3756                 lang->setVariety(lang_variety);
3757         }
3758         return lang;
3759 }
3760
3761
3762 void Paragraph::requestSpellCheck(pos_type pos)
3763 {
3764         d->requestSpellCheck(pos);
3765 }
3766
3767
3768 bool Paragraph::needsSpellCheck() const
3769 {
3770         SpellChecker::ChangeNumber speller_change_number = 0;
3771         if (theSpellChecker())
3772                 speller_change_number = theSpellChecker()->changeNumber();
3773         if (speller_change_number > d->speller_state_.currentChangeNumber()) {
3774                 d->speller_state_.needsCompleteRefresh(speller_change_number);
3775         }
3776         return d->needsSpellCheck();
3777 }
3778
3779
3780 bool Paragraph::Private::ignoreWord(docstring const & word) const
3781 {
3782         // Ignore words with digits
3783         // FIXME: make this customizable
3784         // (note that some checkers ignore words with digits by default)
3785         docstring::const_iterator cit = word.begin();
3786         docstring::const_iterator const end = word.end();
3787         for (; cit != end; ++cit) {
3788                 if (isNumber((*cit)))
3789                         return true;
3790         }
3791         return false;
3792 }
3793
3794
3795 SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
3796         WordLangTuple & wl, docstring_list & suggestions,
3797         bool do_suggestion, bool check_learned) const
3798 {
3799         SpellChecker::Result result = SpellChecker::WORD_OK;
3800         SpellChecker * speller = theSpellChecker();
3801         if (!speller)
3802                 return result;
3803
3804         if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
3805                 return result;
3806
3807         locateWord(from, to, WHOLE_WORD);
3808         if (from == to || from >= size())
3809                 return result;
3810
3811         docstring word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
3812         Language * lang = d->getSpellLanguage(from);
3813
3814         wl = WordLangTuple(word, lang);
3815
3816         if (word.empty())
3817                 return result;
3818
3819         if (needsSpellCheck() || check_learned) {
3820                 pos_type end = to;
3821                 if (!d->ignoreWord(word)) {
3822                         bool const trailing_dot = to < size() && d->text_[to] == '.';
3823                         result = speller->check(wl);
3824                         if (SpellChecker::misspelled(result) && trailing_dot) {
3825                                 wl = WordLangTuple(word.append(from_ascii(".")), lang);
3826                                 result = speller->check(wl);
3827                                 if (!SpellChecker::misspelled(result)) {
3828                                         LYXERR(Debug::GUI, "misspelled word is correct with dot: \"" <<
3829                                            word << "\" [" <<
3830                                            from << ".." << to << "]");
3831                                 } else {
3832                                         // spell check with dot appended failed too
3833                                         // restore original word/lang value
3834                                         word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
3835                                         wl = WordLangTuple(word, lang);
3836                                 }
3837                         }
3838                 }
3839                 if (!SpellChecker::misspelled(result)) {
3840                         // area up to the begin of the next word is not misspelled
3841                         while (end < size() && isWordSeparator(end))
3842                                 ++end;
3843                 }
3844                 d->setMisspelled(from, end, result);
3845         } else {
3846                 result = d->speller_state_.getState(from);
3847         }
3848
3849         if (do_suggestion)
3850                 suggestions.clear();
3851
3852         if (SpellChecker::misspelled(result)) {
3853                 LYXERR(Debug::GUI, "misspelled word: \"" <<
3854                            word << "\" [" <<
3855                            from << ".." << to << "]");
3856                 if (do_suggestion)
3857                         speller->suggest(wl, suggestions);
3858         }
3859         return result;
3860 }
3861
3862
3863 void Paragraph::Private::markMisspelledWords(
3864         pos_type const & first, pos_type const & last,
3865         SpellChecker::Result result,
3866         docstring const & word,
3867         SkipPositions const & skips)
3868 {
3869         if (!SpellChecker::misspelled(result)) {
3870                 setMisspelled(first, last, SpellChecker::WORD_OK);
3871                 return;
3872         }
3873         int snext = first;
3874         SpellChecker * speller = theSpellChecker();
3875         // locate and enumerate the error positions
3876         int nerrors = speller->numMisspelledWords();
3877         int numskipped = 0;
3878         SkipPositionsIterator it = skips.begin();
3879         SkipPositionsIterator et = skips.end();
3880         for (int index = 0; index < nerrors; ++index) {
3881                 int wstart;
3882                 int wlen = 0;
3883                 speller->misspelledWord(index, wstart, wlen);
3884                 /// should not happen if speller supports range checks
3885                 if (!wlen) continue;
3886                 docstring const misspelled = word.substr(wstart, wlen);
3887                 wstart += first + numskipped;
3888                 if (snext < wstart) {
3889                         /// mark the range of correct spelling
3890                         numskipped += countSkips(it, et, wstart);
3891                         setMisspelled(snext,
3892                                 wstart - 1, SpellChecker::WORD_OK);
3893                 }
3894                 snext = wstart + wlen;
3895                 numskipped += countSkips(it, et, snext);
3896                 /// mark the range of misspelling
3897                 setMisspelled(wstart, snext, result);
3898                 LYXERR(Debug::GUI, "misspelled word: \"" <<
3899                            misspelled << "\" [" <<
3900                            wstart << ".." << (snext-1) << "]");
3901                 ++snext;
3902         }
3903         if (snext <= last) {
3904                 /// mark the range of correct spelling at end
3905                 setMisspelled(snext, last, SpellChecker::WORD_OK);
3906         }
3907 }
3908
3909
3910 void Paragraph::spellCheck() const
3911 {
3912         SpellChecker * speller = theSpellChecker();
3913         if (!speller || empty() ||!needsSpellCheck())
3914                 return;
3915         pos_type start;
3916         pos_type endpos;
3917         d->rangeOfSpellCheck(start, endpos);
3918         if (speller->canCheckParagraph()) {
3919                 // loop until we leave the range
3920                 for (pos_type first = start; first < endpos; ) {
3921                         pos_type last = endpos;
3922                         Private::SkipPositions skips;
3923                         Language * lang = d->locateSpellRange(first, last, skips);
3924                         if (first >= endpos)
3925                                 break;
3926                         // start the spell checker on the unit of meaning
3927                         docstring word = asString(first, last, AS_STR_INSETS + AS_STR_SKIPDELETE);
3928                         WordLangTuple wl = WordLangTuple(word, lang);
3929                         SpellChecker::Result result = word.size() ?
3930                                 speller->check(wl) : SpellChecker::WORD_OK;
3931                         d->markMisspelledWords(first, last, result, word, skips);
3932                         first = ++last;
3933                 }
3934         } else {
3935                 static docstring_list suggestions;
3936                 pos_type to = endpos;
3937                 while (start < endpos) {
3938                         WordLangTuple wl;
3939                         spellCheck(start, to, wl, suggestions, false);
3940                         start = to + 1;
3941                 }
3942         }
3943         d->readySpellCheck();
3944 }
3945
3946
3947 bool Paragraph::isMisspelled(pos_type pos, bool check_boundary) const
3948 {
3949         bool result = SpellChecker::misspelled(d->speller_state_.getState(pos));
3950         if (result || pos <= 0 || pos > size())
3951                 return result;
3952         if (check_boundary && (pos == size() || isWordSeparator(pos)))
3953                 result = SpellChecker::misspelled(d->speller_state_.getState(pos - 1));
3954         return result;
3955 }
3956
3957
3958 string Paragraph::magicLabel() const
3959 {
3960         stringstream ss;
3961         ss << "magicparlabel-" << id();
3962         return ss.str();
3963 }
3964
3965
3966 } // namespace lyx