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 "BufferEncodings.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 "texstream.h"
  44 #include "TextClass.h"
  45 #include "TexRow.h"
  46 #include "Text.h"
  47 #include "WordLangTuple.h"
  48 #include "WordList.h"
  49
  50 #include "frontends/alert.h"
  51
  52 #include "insets/InsetBibitem.h"
  53 #include "insets/InsetLabel.h"
  54 #include "insets/InsetSpecialChar.h"
  55 #include "insets/InsetText.h"
  56
  57 #include "mathed/InsetMathHull.h"
  58
  59 #include "support/debug.h"
  60 #include "support/docstring_list.h"
  61 #include "support/ExceptionMessage.h"
  62 #include "support/gettext.h"
  63 #include "support/lassert.h"
  64 #include "support/lstrings.h"
  65 #include "support/textutils.h"
  66
  67 #include <atomic>
  68 #include <sstream>
  69 #include <vector>
  70
  71 using namespace std;
  72 using namespace lyx::support;
  73
  74 // OSX clang, gcc < 4.8.0, and msvc < 2015 do not support C++11 thread_local
  75 #if defined(__APPLE__) || (defined(__GNUC__) && __GNUC__ == 4 && __GNUC_MINOR__ < 8)
  76 #define THREAD_LOCAL_STATIC static __thread
  77 #elif defined(_MSC_VER) && (_MSC_VER < 1900)
  78 #define THREAD_LOCAL_STATIC static __declspec(thread)
  79 #else
  80 #define THREAD_LOCAL_STATIC thread_local static
  81 #endif
  82
  83 namespace lyx {
  84
  85 namespace {
  86
  87 /// Inset identifier (above 0x10ffff, for ucs-4)
  88 char_type const META_INSET = 0x200001;
  89
  90 } // namespace
  91
  92
  93 /////////////////////////////////////////////////////////////////////
  94 //
  95 // SpellResultRange
  96 //
  97 /////////////////////////////////////////////////////////////////////
  98
  99 class SpellResultRange {
 100 public:
 101         SpellResultRange(FontSpan range, SpellChecker::Result result)
 102         : range_(range), result_(result)
 103         {}
 104         ///
 105         FontSpan const & range() const { return range_; }
 106         ///
 107         void range(FontSpan const & r) { range_ = r; }
 108         ///
 109         SpellChecker::Result result() const { return result_; }
 110         ///
 111         void result(SpellChecker::Result r) { result_ = r; }
 112         ///
 113         bool contains(pos_type pos) const { return range_.contains(pos); }
 114         ///
 115         bool covered(FontSpan const & r) const
 116         {
 117                 // 1. first of new range inside current range or
 118                 // 2. last of new range inside current range or
 119                 // 3. first of current range inside new range or
 120                 // 4. last of current range inside new range
 121                 //FIXME: is this the same as !range_.intersect(r).empty() ?
 122                 return range_.contains(r.first) || range_.contains(r.last) ||
 123                         r.contains(range_.first) || r.contains(range_.last);
 124         }
 125         ///
 126         void shift(pos_type pos, int offset)
 127         {
 128                 if (range_.first > pos) {
 129                         range_.first += offset;
 130                         range_.last += offset;
 131                 } else if (range_.last >= pos) {
 132                         range_.last += offset;
 133                 }
 134         }
 135 private:
 136         FontSpan range_ ;
 137         SpellChecker::Result result_ ;
 138 };
 139
 140
 141 /////////////////////////////////////////////////////////////////////
 142 //
 143 // SpellCheckerState
 144 //
 145 /////////////////////////////////////////////////////////////////////
 146
 147 class SpellCheckerState {
 148 public:
 149         SpellCheckerState()
 150         {
 151                 needs_refresh_ = true;
 152                 current_change_number_ = 0;
 153         }
 154
 155         void setRange(FontSpan const & fp, SpellChecker::Result state)
 156         {
 157                 Ranges result;
 158                 RangesIterator et = ranges_.end();
 159                 RangesIterator it = ranges_.begin();
 160                 for (; it != et; ++it) {
 161                         if (!it->covered(fp))
 162                                 result.push_back(SpellResultRange(it->range(), it->result()));
 163                         else if (state == SpellChecker::WORD_OK) {
 164                                 // trim or split the current misspelled range
 165                                 // store misspelled ranges only
 166                                 FontSpan range = it->range();
 167                                 if (fp.first > range.first) {
 168                                         // misspelled area in front of WORD_OK
 169                                         range.last = fp.first - 1;
 170                                         result.push_back(SpellResultRange(range, it->result()));
 171                                         range = it->range();
 172                                 }
 173                                 if (fp.last < range.last) {
 174                                         // misspelled area after WORD_OK range
 175                                         range.first = fp.last + 1;
 176                                         result.push_back(SpellResultRange(range, it->result()));
 177                                 }
 178                         }
 179                 }
 180                 ranges_ = result;
 181                 if (state != SpellChecker::WORD_OK)
 182                         ranges_.push_back(SpellResultRange(fp, state));
 183         }
 184
 185         void increasePosAfterPos(pos_type pos)
 186         {
 187                 correctRangesAfterPos(pos, 1);
 188                 needsRefresh(pos);
 189         }
 190
 191         void decreasePosAfterPos(pos_type pos)
 192         {
 193                 correctRangesAfterPos(pos, -1);
 194                 needsRefresh(pos);
 195         }
 196
 197         void refreshLast(pos_type pos)
 198         {
 199                 if (pos < refresh_.last)
 200                         refresh_.last = pos;
 201         }
 202
 203         SpellChecker::Result getState(pos_type pos) const
 204         {
 205                 SpellChecker::Result result = SpellChecker::WORD_OK;
 206                 RangesIterator et = ranges_.end();
 207                 RangesIterator it = ranges_.begin();
 208                 for (; it != et; ++it) {
 209                         if(it->contains(pos)) {
 210                                 return it->result();
 211                         }
 212                 }
 213                 return result;
 214         }
 215
 216         FontSpan const & getRange(pos_type pos) const
 217         {
 218                 /// empty span to indicate mismatch
 219                 static FontSpan empty_;
 220                 RangesIterator et = ranges_.end();
 221                 RangesIterator it = ranges_.begin();
 222                 for (; it != et; ++it) {
 223                         if(it->contains(pos)) {
 224                                 return it->range();
 225                         }
 226                 }
 227                 return empty_;
 228         }
 229
 230         bool needsRefresh() const
 231         {
 232                 return needs_refresh_;
 233         }
 234
 235         SpellChecker::ChangeNumber currentChangeNumber() const
 236         {
 237                 return current_change_number_;
 238         }
 239
 240         void refreshRange(pos_type & first, pos_type & last) const
 241         {
 242                 first = refresh_.first;
 243                 last = refresh_.last;
 244         }
 245
 246         void needsRefresh(pos_type pos)
 247         {
 248                 if (needs_refresh_ && pos != -1) {
 249                         if (pos < refresh_.first)
 250                                 refresh_.first = pos;
 251                         if (pos > refresh_.last)
 252                                 refresh_.last = pos;
 253                 } else if (pos != -1) {
 254                         // init request check for neighbour positions too
 255                         refresh_.first = pos > 0 ? pos - 1 : 0;
 256                         // no need for special end of paragraph check
 257                         refresh_.last = pos + 1;
 258                 }
 259                 needs_refresh_ = pos != -1;
 260         }
 261
 262         void needsCompleteRefresh(SpellChecker::ChangeNumber change_number)
 263         {
 264                 needs_refresh_ = true;
 265                 refresh_.first = 0;
 266                 refresh_.last = -1;
 267                 current_change_number_ = change_number;
 268         }
 269 private:
 270         typedef vector<SpellResultRange> Ranges;
 271         typedef Ranges::const_iterator RangesIterator;
 272         Ranges ranges_;
 273         /// the area of the paragraph with pending spell check
 274         FontSpan refresh_;
 275         bool needs_refresh_;
 276         /// spell state cache version number
 277         SpellChecker::ChangeNumber current_change_number_;
 278
 279
 280         void correctRangesAfterPos(pos_type pos, int offset)
 281         {
 282                 RangesIterator et = ranges_.end();
 283                 Ranges::iterator it = ranges_.begin();
 284                 for (; it != et; ++it) {
 285                         it->shift(pos, offset);
 286                 }
 287         }
 288
 289 };
 290
 291 /////////////////////////////////////////////////////////////////////
 292 //
 293 // Paragraph::Private
 294 //
 295 /////////////////////////////////////////////////////////////////////
 296
 297 class Paragraph::Private
 298 {
 299         // Enforce our own "copy" constructor
 300         Private(Private const &) = delete;
 301         Private & operator=(Private const &) = delete;
 302         // Unique ID generator
 303         static int make_id();
 304 public:
 305         ///
 306         Private(Paragraph * owner, Layout const & layout);
 307         /// "Copy constructor"
 308         Private(Private const &, Paragraph * owner);
 309         /// Copy constructor from \p beg  to \p end
 310         Private(Private const &, Paragraph * owner, pos_type beg, pos_type end);
 311
 312         ///
 313         void insertChar(pos_type pos, char_type c, Change const & change);
 314
 315         /// Output the surrogate pair formed by \p c and \p next to \p os.
 316         /// \return the number of characters written.
 317         int latexSurrogatePair(otexstream & os, char_type c, char_type next,
 318                                OutputParams const &);
 319
 320         /// Output a space in appropriate formatting (or a surrogate pair
 321         /// if the next character is a combining character).
 322         /// \return whether a surrogate pair was output.
 323         bool simpleTeXBlanks(OutputParams const &,
 324                              otexstream &,
 325                              pos_type i,
 326                              unsigned int & column,
 327                              Font const & font,
 328                              Layout const & style);
 329
 330         /// Output consecutive unicode chars, belonging to the same script as
 331         /// specified by the latex macro \p ltx, to \p os starting from \p i.
 332         /// \return the number of characters written.
 333         int writeScriptChars(OutputParams const &, otexstream & os,
 334                              docstring const & ltx,
 335                              Change const &, Encoding const &,
 336                              std::string const, pos_type & i);
 337
 338         /// This could go to ParagraphParameters if we want to.
 339         int startTeXParParams(BufferParams const &, otexstream &,
 340                               OutputParams const &) const;
 341
 342         /// This could go to ParagraphParameters if we want to.
 343         bool endTeXParParams(BufferParams const &, otexstream &,
 344                              OutputParams const &) const;
 345
 346         ///
 347         void latexInset(BufferParams const &,
 348                                    otexstream &,
 349                                    OutputParams &,
 350                                    Font & running_font,
 351                                    Font & basefont,
 352                                    Font const & outerfont,
 353                                    bool & open_font,
 354                                    Change & running_change,
 355                                    Layout const & style,
 356                                    pos_type & i,
 357                                    unsigned int & column);
 358
 359         ///
 360         void latexSpecialChar(
 361                                    otexstream & os,
 362                                    BufferParams const & bparams,
 363                                    OutputParams const & runparams,
 364                                    Font const & running_font,
 365                                    Change const & running_change,
 366                                    Layout const & style,
 367                                    pos_type & i,
 368                                    pos_type end_pos,
 369                                    unsigned int & column);
 370
 371         ///
 372         bool latexSpecialT1(
 373                 char_type const c,
 374                 otexstream & os,
 375                 pos_type i,
 376                 unsigned int & column);
 377         ///
 378         bool latexSpecialTU(
 379                 char_type const c,
 380                 otexstream & os,
 381                 pos_type i,
 382                 unsigned int & column);
 383         ///
 384         bool latexSpecialT3(
 385                 char_type const c,
 386                 otexstream & os,
 387                 pos_type i,
 388                 unsigned int & column);
 389
 390         ///
 391         void validate(LaTeXFeatures & features) const;
 392
 393         /// Checks if the paragraph contains only text and no inset or font change.
 394         bool onlyText(Buffer const & buf, Font const & outerfont,
 395                       pos_type initial) const;
 396
 397         /// a vector of speller skip positions
 398         typedef vector<FontSpan> SkipPositions;
 399         typedef SkipPositions::const_iterator SkipPositionsIterator;
 400
 401         void appendSkipPosition(SkipPositions & skips, pos_type const pos) const;
 402
 403         Language * getSpellLanguage(pos_type const from) const;
 404
 405         Language * locateSpellRange(pos_type & from, pos_type & to,
 406                                     SkipPositions & skips) const;
 407
 408         bool hasSpellerChange() const
 409         {
 410                 SpellChecker::ChangeNumber speller_change_number = 0;
 411                 if (theSpellChecker())
 412                         speller_change_number = theSpellChecker()->changeNumber();
 413                 return speller_change_number > speller_state_.currentChangeNumber();
 414         }
 415
 416         bool ignoreWord(docstring const & word) const ;
 417
 418         void setMisspelled(pos_type from, pos_type to, SpellChecker::Result state)
 419         {
 420                 pos_type textsize = owner_->size();
 421                 // check for sane arguments
 422                 if (to <= from || from >= textsize)
 423                         return;
 424                 FontSpan fp = FontSpan(from, to - 1);
 425                 speller_state_.setRange(fp, state);
 426         }
 427
 428         void requestSpellCheck(pos_type pos)
 429         {
 430                 if (pos == -1)
 431                         speller_state_.needsCompleteRefresh(speller_state_.currentChangeNumber());
 432                 else
 433                         speller_state_.needsRefresh(pos);
 434         }
 435
 436         void readySpellCheck()
 437         {
 438                 speller_state_.needsRefresh(-1);
 439         }
 440
 441         bool needsSpellCheck() const
 442         {
 443                 return speller_state_.needsRefresh();
 444         }
 445
 446         void rangeOfSpellCheck(pos_type & first, pos_type & last) const
 447         {
 448                 speller_state_.refreshRange(first, last);
 449                 if (last == -1) {
 450                         last = owner_->size();
 451                         return;
 452                 }
 453                 pos_type endpos = last;
 454                 owner_->locateWord(first, endpos, WHOLE_WORD);
 455                 if (endpos < last) {
 456                         endpos = last;
 457                         owner_->locateWord(last, endpos, WHOLE_WORD);
 458                 }
 459                 last = endpos;
 460         }
 461
 462         int countSkips(SkipPositionsIterator & it, SkipPositionsIterator const et,
 463                             int & start) const
 464         {
 465                 int numskips = 0;
 466                 while (it != et && it->first < start) {
 467                         int skip = it->last - it->first + 1;
 468                         start += skip;
 469                         numskips += skip;
 470                         ++it;
 471                 }
 472                 return numskips;
 473         }
 474
 475         void markMisspelledWords(pos_type const & first, pos_type const & last,
 476                                                          SpellChecker::Result result,
 477                                                          docstring const & word,
 478                                                          SkipPositions const & skips);
 479
 480         InsetCode ownerCode() const
 481         {
 482                 return inset_owner_ ? inset_owner_->lyxCode() : NO_CODE;
 483         }
 484
 485         /// Which Paragraph owns us?
 486         Paragraph * owner_;
 487
 488         /// In which Inset?
 489         Inset const * inset_owner_;
 490
 491         ///
 492         FontList fontlist_;
 493
 494         ///
 495         int id_;
 496
 497         ///
 498         ParagraphParameters params_;
 499
 500         /// for recording and looking up changes
 501         Changes changes_;
 502
 503         ///
 504         InsetList insetlist_;
 505
 506         /// end of label
 507         pos_type begin_of_body_;
 508
 509         typedef docstring TextContainer;
 510         ///
 511         TextContainer text_;
 512
 513         typedef set<docstring> Words;
 514         typedef map<string, Words> LangWordsMap;
 515         ///
 516         LangWordsMap words_;
 517         ///
 518         Layout const * layout_;
 519         ///
 520         SpellCheckerState speller_state_;
 521 };
 522
 523
 524 Paragraph::Private::Private(Paragraph * owner, Layout const & layout)
 525         : owner_(owner), inset_owner_(0), id_(-1), begin_of_body_(0), layout_(&layout)
 526 {
 527         text_.reserve(100);
 528 }
 529
 530
 531 //static
 532 int Paragraph::Private::make_id()
 533 {
 534         // The id is unique per session across buffers because it is used in
 535         // LFUN_PARAGRAPH_GOTO to switch to a different buffer, for instance in the
 536         // outliner.
 537         // (thread-safe)
 538         static atomic_uint next_id(0);
 539         return next_id++;
 540 }
 541
 542
 543 Paragraph::Private::Private(Private const & p, Paragraph * owner)
 544         : owner_(owner), inset_owner_(p.inset_owner_), fontlist_(p.fontlist_),
 545           id_(make_id()),
 546           params_(p.params_), changes_(p.changes_), insetlist_(p.insetlist_),
 547           begin_of_body_(p.begin_of_body_), text_(p.text_), words_(p.words_),
 548           layout_(p.layout_)
 549 {
 550         requestSpellCheck(p.text_.size());
 551 }
 552
 553
 554 Paragraph::Private::Private(Private const & p, Paragraph * owner,
 555         pos_type beg, pos_type end)
 556         : owner_(owner), inset_owner_(p.inset_owner_), id_(make_id()),
 557           params_(p.params_), changes_(p.changes_),
 558           insetlist_(p.insetlist_, beg, end),
 559           begin_of_body_(p.begin_of_body_), words_(p.words_),
 560           layout_(p.layout_)
 561 {
 562         if (beg >= pos_type(p.text_.size()))
 563                 return;
 564         text_ = p.text_.substr(beg, end - beg);
 565
 566         FontList::const_iterator fcit = fontlist_.begin();
 567         FontList::const_iterator fend = fontlist_.end();
 568         for (; fcit != fend; ++fcit) {
 569                 if (fcit->pos() < beg)
 570                         continue;
 571                 if (fcit->pos() >= end) {
 572                         // Add last entry in the fontlist_.
 573                         fontlist_.set(text_.size() - 1, fcit->font());
 574                         break;
 575                 }
 576                 // Add a new entry in the fontlist_.
 577                 fontlist_.set(fcit->pos() - beg, fcit->font());
 578         }
 579         requestSpellCheck(p.text_.size());
 580 }
 581
 582
 583 void Paragraph::addChangesToToc(DocIterator const & cdit, Buffer const & buf,
 584                                 bool output_active, TocBackend & backend) const
 585 {
 586         d->changes_.addToToc(cdit, buf, output_active, backend);
 587 }
 588
 589
 590 void Paragraph::addChangesToBuffer(Buffer const & buf) const
 591 {
 592         d->changes_.updateBuffer(buf);
 593 }
 594
 595
 596 bool Paragraph::isChangeUpdateRequired() const
 597 {
 598         return d->changes_.isUpdateRequired();
 599 }
 600
 601
 602 bool Paragraph::isDeleted(pos_type start, pos_type end) const
 603 {
 604         LASSERT(start >= 0 && start <= size(), return false);
 605         LASSERT(end > start && end <= size() + 1, return false);
 606
 607         return d->changes_.isDeleted(start, end);
 608 }
 609
 610
 611 bool Paragraph::isChanged(pos_type start, pos_type end) const
 612 {
 613         LASSERT(start >= 0 && start <= size(), return false);
 614         LASSERT(end > start && end <= size() + 1, return false);
 615
 616         return d->changes_.isChanged(start, end);
 617 }
 618
 619
 620 bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const
 621 {
 622         // keep the logic here in sync with the logic of eraseChars()
 623         if (!trackChanges)
 624                 return true;
 625
 626         Change const & change = d->changes_.lookup(size());
 627         return change.inserted() && change.currentAuthor();
 628 }
 629
 630
 631 void Paragraph::setChange(Change const & change)
 632 {
 633         // beware of the imaginary end-of-par character!
 634         d->changes_.set(change, 0, size() + 1);
 635
 636         /*
 637          * Propagate the change recursively - but not in case of DELETED!
 638          *
 639          * Imagine that your co-author makes changes in an existing inset. He
 640          * sends your document to you and you come to the conclusion that the
 641          * inset should go completely. If you erase it, LyX must not delete all
 642          * text within the inset. Otherwise, the change tracked insertions of
 643          * your co-author get lost and there is no way to restore them later.
 644          *
 645          * Conclusion: An inset's content should remain untouched if you delete it
 646          */
 647
 648         if (!change.deleted()) {
 649                 for (pos_type pos = 0; pos < size(); ++pos) {
 650                         if (Inset * inset = getInset(pos))
 651                                 inset->setChange(change);
 652                 }
 653         }
 654 }
 655
 656
 657 void Paragraph::setChange(pos_type pos, Change const & change)
 658 {
 659         LASSERT(pos >= 0 && pos <= size(), return);
 660         d->changes_.set(change, pos);
 661
 662         // see comment in setChange(Change const &) above
 663         if (!change.deleted() && pos < size())
 664                         if (Inset * inset = getInset(pos))
 665                                 inset->setChange(change);
 666 }
 667
 668
 669 Change const & Paragraph::lookupChange(pos_type pos) const
 670 {
 671         LBUFERR(pos >= 0 && pos <= size());
 672         return d->changes_.lookup(pos);
 673 }
 674
 675
 676 void Paragraph::acceptChanges(pos_type start, pos_type end)
 677 {
 678         LASSERT(start >= 0 && start <= size(), return);
 679         LASSERT(end > start && end <= size() + 1, return);
 680
 681         for (pos_type pos = start; pos < end; ++pos) {
 682                 switch (lookupChange(pos).type) {
 683                         case Change::UNCHANGED:
 684                                 // accept changes in nested inset
 685                                 if (Inset * inset = getInset(pos))
 686                                         inset->acceptChanges();
 687                                 break;
 688
 689                         case Change::INSERTED:
 690                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 691                                 // also accept changes in nested inset
 692                                 if (Inset * inset = getInset(pos))
 693                                         inset->acceptChanges();
 694                                 break;
 695
 696                         case Change::DELETED:
 697                                 // Suppress access to non-existent
 698                                 // "end-of-paragraph char"
 699                                 if (pos < size()) {
 700                                         eraseChar(pos, false);
 701                                         --end;
 702                                         --pos;
 703                                 }
 704                                 break;
 705                 }
 706
 707         }
 708 }
 709
 710
 711 void Paragraph::rejectChanges(pos_type start, pos_type end)
 712 {
 713         LASSERT(start >= 0 && start <= size(), return);
 714         LASSERT(end > start && end <= size() + 1, return);
 715
 716         for (pos_type pos = start; pos < end; ++pos) {
 717                 switch (lookupChange(pos).type) {
 718                         case Change::UNCHANGED:
 719                                 // reject changes in nested inset
 720                                 if (Inset * inset = getInset(pos))
 721                                                 inset->rejectChanges();
 722                                 break;
 723
 724                         case Change::INSERTED:
 725                                 // Suppress access to non-existent
 726                                 // "end-of-paragraph char"
 727                                 if (pos < size()) {
 728                                         eraseChar(pos, false);
 729                                         --end;
 730                                         --pos;
 731                                 }
 732                                 break;
 733
 734                         case Change::DELETED:
 735                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 736
 737                                 // Do NOT reject changes within a deleted inset!
 738                                 // There may be insertions of a co-author inside of it!
 739
 740                                 break;
 741                 }
 742         }
 743 }
 744
 745
 746 void Paragraph::Private::insertChar(pos_type pos, char_type c,
 747                 Change const & change)
 748 {
 749         LASSERT(pos >= 0 && pos <= int(text_.size()), return);
 750
 751         // track change
 752         changes_.insert(change, pos);
 753
 754         // This is actually very common when parsing buffers (and
 755         // maybe inserting ascii text)
 756         if (pos == pos_type(text_.size())) {
 757                 // when appending characters, no need to update tables
 758                 text_.push_back(c);
 759                 // but we want spell checking
 760                 requestSpellCheck(pos);
 761                 return;
 762         }
 763
 764         text_.insert(text_.begin() + pos, c);
 765
 766         // Update the font table.
 767         fontlist_.increasePosAfterPos(pos);
 768
 769         // Update the insets
 770         insetlist_.increasePosAfterPos(pos);
 771
 772         // Update list of misspelled positions
 773         speller_state_.increasePosAfterPos(pos);
 774 }
 775
 776
 777 bool Paragraph::insertInset(pos_type pos, Inset * inset,
 778                                    Font const & font, Change const & change)
 779 {
 780         LASSERT(inset, return false);
 781         LASSERT(pos >= 0 && pos <= size(), return false);
 782
 783         // Paragraph::insertInset() can be used in cut/copy/paste operation where
 784         // d->inset_owner_ is not set yet.
 785         if (d->inset_owner_ && !d->inset_owner_->insetAllowed(inset->lyxCode()))
 786                 return false;
 787
 788         d->insertChar(pos, META_INSET, change);
 789         LASSERT(d->text_[pos] == META_INSET, return false);
 790
 791         // Add a new entry in the insetlist_.
 792         d->insetlist_.insert(inset, pos);
 793
 794         // Some insets require run of spell checker
 795         requestSpellCheck(pos);
 796         setFont(pos, font);
 797         return true;
 798 }
 799
 800
 801 bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
 802 {
 803         LASSERT(pos >= 0 && pos <= size(), return false);
 804
 805         // keep the logic here in sync with the logic of isMergedOnEndOfParDeletion()
 806
 807         if (trackChanges) {
 808                 Change change = d->changes_.lookup(pos);
 809
 810                 // set the character to DELETED if
 811                 //  a) it was previously unchanged or
 812                 //  b) it was inserted by a co-author
 813
 814                 if (!change.changed() ||
 815                       (change.inserted() && !change.currentAuthor())) {
 816                         setChange(pos, Change(Change::DELETED));
 817                         // request run of spell checker
 818                         requestSpellCheck(pos);
 819                         return false;
 820                 }
 821
 822                 if (change.deleted())
 823                         return false;
 824         }
 825
 826         // Don't physically access the imaginary end-of-paragraph character.
 827         // eraseChar() can only mark it as DELETED. A physical deletion of
 828         // end-of-par must be handled externally.
 829         if (pos == size()) {
 830                 return false;
 831         }
 832
 833         // track change
 834         d->changes_.erase(pos);
 835
 836         // if it is an inset, delete the inset entry
 837         if (d->text_[pos] == META_INSET)
 838                 d->insetlist_.erase(pos);
 839
 840         d->text_.erase(d->text_.begin() + pos);
 841
 842         // Update the fontlist_
 843         d->fontlist_.erase(pos);
 844
 845         // Update the insetlist_
 846         d->insetlist_.decreasePosAfterPos(pos);
 847
 848         // Update list of misspelled positions
 849         d->speller_state_.decreasePosAfterPos(pos);
 850         d->speller_state_.refreshLast(size());
 851
 852         return true;
 853 }
 854
 855
 856 int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
 857 {
 858         LASSERT(start >= 0 && start <= size(), return 0);
 859         LASSERT(end >= start && end <= size() + 1, return 0);
 860
 861         pos_type i = start;
 862         for (pos_type count = end - start; count; --count) {
 863                 if (!eraseChar(i, trackChanges))
 864                         ++i;
 865         }
 866         return end - i;
 867 }
 868
 869
 870 int Paragraph::Private::latexSurrogatePair(otexstream & os, char_type c,
 871                 char_type next, OutputParams const & runparams)
 872 {
 873         // Writing next here may circumvent a possible font change between
 874         // c and next. Since next is only output if it forms a surrogate pair
 875         // with c we can ignore this:
 876         // A font change inside a surrogate pair does not make sense and is
 877         // hopefully impossible to input.
 878         // FIXME: change tracking
 879         // Is this correct WRT change tracking?
 880         Encoding const & encoding = *(runparams.encoding);
 881         docstring latex1 = encoding.latexChar(next).first;
 882         if (runparams.inIPA) {
 883                 string const tipashortcut = Encodings::TIPAShortcut(next);
 884                 if (!tipashortcut.empty()) {
 885                         latex1 = from_ascii(tipashortcut);
 886                 }
 887         }
 888         docstring latex2 = encoding.latexChar(c).first;
 889
 890         docstring::size_type const brace1 = latex2.find_first_of(from_ascii("{"));
 891         docstring::size_type const brace2 = latex2.find_last_of(from_ascii("}"));
 892         string script = to_ascii(latex2.substr(1, brace1 - 1));
 893         int pos = 0;
 894         int length = brace2;
 895         string fontenc;
 896         if (runparams.local_font)
 897                 fontenc = runparams.local_font->language()->fontenc();
 898         else
 899                 fontenc = runparams.main_fontenc;
 900         if ((script == "textgreek" && fontenc == "LGR")
 901             ||(script == "textcyr" && (fontenc == "T2A" || fontenc == "T2B"
 902                                        || fontenc == "T2C" || fontenc == "X2"))) {
 903                 // Correct font encoding is being used, so we can avoid \text[greek|cyr].
 904                 pos = brace1 + 1;
 905                 length -= pos;
 906                 latex2 = latex2.substr(pos, length);
 907         }
 908
 909         if (docstring(1, next) == latex1) {
 910                 // the encoding supports the combination
 911                 os << latex2 << latex1;
 912                 return latex1.length() + latex2.length();
 913         } else if (runparams.local_font &&
 914                    runparams.local_font->language()->lang() == "polutonikogreek") {
 915                 // polutonikogreek only works without the brackets
 916                 os << latex1 << latex2;
 917                 return latex1.length() + latex2.length();
 918         } else
 919                 os << latex1 << '{' << latex2 << '}';
 920         return latex1.length() + latex2.length() + 2;
 921 }
 922
 923
 924 bool Paragraph::Private::simpleTeXBlanks(OutputParams const & runparams,
 925                                        otexstream & os,
 926                                        pos_type i,
 927                                        unsigned int & column,
 928                                        Font const & font,
 929                                        Layout const & style)
 930 {
 931         if (style.pass_thru || runparams.pass_thru)
 932                 return false;
 933
 934         if (i + 1 < int(text_.size())) {
 935                 char_type next = text_[i + 1];
 936                 if (Encodings::isCombiningChar(next)) {
 937                         // This space has an accent, so we must always output it.
 938                         column += latexSurrogatePair(os, ' ', next, runparams) - 1;
 939                         return true;
 940                 }
 941         }
 942
 943         if (runparams.linelen > 0
 944             && column > runparams.linelen
 945             && i
 946             && text_[i - 1] != ' '
 947             && (i + 1 < int(text_.size()))
 948             // same in FreeSpacing mode
 949             && !owner_->isFreeSpacing()
 950             // In typewriter mode, we want to avoid
 951             // ! . ? : at the end of a line
 952             && !(font.fontInfo().family() == TYPEWRITER_FAMILY
 953                  && (text_[i - 1] == '.'
 954                      || text_[i - 1] == '?'
 955                      || text_[i - 1] == ':'
 956                      || text_[i - 1] == '!'))) {
 957                 os << '\n';
 958                 os.texrow().start(owner_->id(), i + 1);
 959                 column = 0;
 960         } else if (style.free_spacing) {
 961                 os << '~';
 962         } else {
 963                 os << ' ';
 964         }
 965         return false;
 966 }
 967
 968
 969 int Paragraph::Private::writeScriptChars(OutputParams const & runparams,
 970                                          otexstream & os,
 971                                          docstring const & ltx,
 972                                          Change const & runningChange,
 973                                          Encoding const & encoding,
 974                                          string const fontenc,
 975                                          pos_type & i)
 976 {
 977         // FIXME: modifying i here is not very nice...
 978
 979         // We only arrive here when character text_[i] could not be translated
 980         // into the current latex encoding (or its latex translation has been forced,)
 981         // and it belongs to a known script.
 982         // Parameter ltx contains the latex translation of text_[i] as specified
 983         // in the unicodesymbols file and is something like "\textXXX{<spec>}".
 984         // The latex macro name "textXXX" specifies the script to which text_[i]
 985         // belongs and we use it in order to check whether characters from the
 986         // same script immediately follow, such that we can collect them in a
 987         // single "\textXXX" macro. So, we have to retain "\textXXX{<spec>"
 988         // for the first char but only "<spec>" for all subsequent chars.
 989         docstring::size_type const brace1 = ltx.find_first_of(from_ascii("{"));
 990         docstring::size_type const brace2 = ltx.find_last_of(from_ascii("}"));
 991         string script = to_ascii(ltx.substr(1, brace1 - 1));
 992         int pos = 0;
 993         int length = brace2;
 994         bool closing_brace = true;
 995         if ((script == "textgreek" && fontenc == "LGR")
 996             ||(script == "textcyr" && (fontenc == "T2A" || fontenc == "T2B"
 997                                        || fontenc == "T2C" || fontenc == "X2"))) {
 998                 // Correct font encoding is being used, so we can avoid \text[greek|cyr].
 999                 pos = brace1 + 1;
1000                 length -= pos;
1001                 closing_brace = false;
1002         }
1003         os << ltx.substr(pos, length);
1004         int size = text_.size();
1005         while (i + 1 < size) {
1006                 char_type const next = text_[i + 1];
1007                 // Stop here if next character belongs to another script
1008                 // or there is a change in change tracking status.
1009                 if (!Encodings::isKnownScriptChar(next, script) ||
1010                     runningChange != owner_->lookupChange(i + 1))
1011                         break;
1012                 Font prev_font;
1013                 bool found = false;
1014                 FontList::const_iterator cit = fontlist_.begin();
1015                 FontList::const_iterator end = fontlist_.end();
1016                 for (; cit != end; ++cit) {
1017                         if (cit->pos() >= i && !found) {
1018                                 prev_font = cit->font();
1019                                 found = true;
1020                         }
1021                         if (cit->pos() >= i + 1)
1022                                 break;
1023                 }
1024                 // Stop here if there is a font attribute or encoding change.
1025                 if (found && cit != end && prev_font != cit->font())
1026                         break;
1027
1028                 // Check whether we have a combining pair
1029                 char_type next_next = '\0';
1030                 if (i + 2 < size) {
1031                         next_next = text_[i + 2];
1032                         if (Encodings::isCombiningChar(next_next)) {
1033                                 length += latexSurrogatePair(os, next, next_next, runparams) - 1;
1034                                 i += 2;
1035                                 continue;
1036                         }
1037                 }
1038
1039                 docstring const latex = encoding.latexChar(next).first;
1040                 docstring::size_type const b1 =
1041                                         latex.find_first_of(from_ascii("{"));
1042                 docstring::size_type const b2 =
1043                                         latex.find_last_of(from_ascii("}"));
1044                 int const len = b2 - b1 - 1;
1045                 os << latex.substr(b1 + 1, len);
1046                 length += len;
1047                 ++i;
1048         }
1049         if (closing_brace) {
1050                 os << '}';
1051                 ++length;
1052         }
1053         return length;
1054 }
1055
1056
1057 void Paragraph::Private::latexInset(BufferParams const & bparams,
1058                                     otexstream & os,
1059                                     OutputParams & runparams,
1060                                     Font & running_font,
1061                                     Font & basefont,
1062                                     Font const & outerfont,
1063                                     bool & open_font,
1064                                     Change & running_change,
1065                                     Layout const & style,
1066                                     pos_type & i,
1067                                     unsigned int & column)
1068 {
1069         Inset * inset = owner_->getInset(i);
1070         LBUFERR(inset);
1071
1072         if (style.pass_thru) {
1073                 odocstringstream ods;
1074                 inset->plaintext(ods, runparams);
1075                 os << ods.str();
1076                 return;
1077         }
1078
1079         // FIXME: move this to InsetNewline::latex
1080         if (inset->lyxCode() == NEWLINE_CODE || inset->lyxCode() == SEPARATOR_CODE) {
1081                 // newlines are handled differently here than
1082                 // the default in simpleTeXSpecialChars().
1083                 if (!style.newline_allowed) {
1084                         os << '\n';
1085                 } else {
1086                         if (open_font) {
1087                                 bool needPar = false;
1088                                 column += running_font.latexWriteEndChanges(
1089                                         os, bparams, runparams,
1090                                         basefont, basefont, needPar);
1091                                 open_font = false;
1092                         }
1093
1094                         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY)
1095                                 os << '~';
1096
1097                         basefont = owner_->getLayoutFont(bparams, outerfont);
1098                         running_font = basefont;
1099
1100                         if (runparams.moving_arg)
1101                                 os << "\\protect ";
1102
1103                 }
1104                 os.texrow().start(owner_->id(), i + 1);
1105                 column = 0;
1106         }
1107
1108         if (owner_->isDeleted(i)) {
1109                 if( ++runparams.inDeletedInset == 1)
1110                         runparams.changeOfDeletedInset = owner_->lookupChange(i);
1111         }
1112
1113         if (inset->canTrackChanges()) {
1114                 column += Changes::latexMarkChange(os, bparams, running_change,
1115                         Change(Change::UNCHANGED), runparams);
1116                 running_change = Change(Change::UNCHANGED);
1117         }
1118
1119         bool close = false;
1120         odocstream::pos_type const len = os.os().tellp();
1121
1122         if (inset->forceLTR()
1123             && !runparams.use_polyglossia
1124             && running_font.isRightToLeft()
1125             // ERT is an exception, it should be output with no
1126             // decorations at all
1127             && inset->lyxCode() != ERT_CODE) {
1128                 if (running_font.language()->lang() == "farsi")
1129                         os << "\\beginL" << termcmd;
1130                 else
1131                         os << "\\L{";
1132                 close = true;
1133         }
1134
1135         // FIXME: Bug: we can have an empty font change here!
1136         // if there has just been a font change, we are going to close it
1137         // right now, which means stupid latex code like \textsf{}. AFAIK,
1138         // this does not harm dvi output. A minor bug, thus (JMarc)
1139
1140         // Some insets cannot be inside a font change command.
1141         // However, even such insets *can* be placed in \L or \R
1142         // or their equivalents (for RTL language switches), so we don't
1143         // close the language in those cases.
1144         // ArabTeX, though, cannot handle this special behavior, it seems.
1145         bool arabtex = basefont.language()->lang() == "arabic_arabtex"
1146                 || running_font.language()->lang() == "arabic_arabtex";
1147         if (open_font && !inset->inheritFont()) {
1148                 bool needPar = false;
1149                 bool closeLanguage = arabtex
1150                         || basefont.isRightToLeft() == running_font.isRightToLeft();
1151                 unsigned int count = running_font.latexWriteEndChanges(os,
1152                                         bparams, runparams, basefont, basefont,
1153                                         needPar, closeLanguage);
1154                 column += count;
1155                 // if any font properties were closed, update the running_font,
1156                 // making sure, however, to leave the language as it was
1157                 if (count > 0) {
1158                         // FIXME: probably a better way to keep track of the old
1159                         // language, than copying the entire font?
1160                         Font const copy_font(running_font);
1161                         basefont = owner_->getLayoutFont(bparams, outerfont);
1162                         running_font = basefont;
1163                         if (!closeLanguage)
1164                                 running_font.setLanguage(copy_font.language());
1165                         // leave font open if language is still open
1166                         open_font = (running_font.language() == basefont.language());
1167                         if (closeLanguage)
1168                                 runparams.local_font = &basefont;
1169                 }
1170         }
1171
1172         size_t const previous_row_count = os.texrow().rows();
1173
1174         try {
1175                 runparams.lastid = id_;
1176                 runparams.lastpos = i;
1177                 inset->latex(os, runparams);
1178         } catch (EncodingException & e) {
1179                 // add location information and throw again.
1180                 e.par_id = id_;
1181                 e.pos = i;
1182                 throw(e);
1183         }
1184
1185         if (close) {
1186                 if (running_font.language()->lang() == "farsi")
1187                                 os << "\\endL" << termcmd;
1188                         else
1189                                 os << '}';
1190         }
1191
1192         if (os.texrow().rows() > previous_row_count) {
1193                 os.texrow().start(owner_->id(), i + 1);
1194                 column = 0;
1195         } else {
1196                 column += (unsigned int)(os.os().tellp() - len);
1197         }
1198
1199         if (owner_->isDeleted(i))
1200                 --runparams.inDeletedInset;
1201 }
1202
1203
1204 void Paragraph::Private::latexSpecialChar(otexstream & os,
1205                                           BufferParams const & bparams,
1206                                           OutputParams const & runparams,
1207                                           Font const & running_font,
1208                                           Change const & running_change,
1209                                           Layout const & style,
1210                                           pos_type & i,
1211                                           pos_type end_pos,
1212                                           unsigned int & column)
1213 {
1214         // With polyglossia, brackets and stuff need not be reversed
1215         // in RTL scripts (see bug #8251)
1216         char_type const c = (runparams.use_polyglossia) ?
1217                 owner_->getUChar(bparams, i) : text_[i];
1218
1219         if (style.pass_thru || runparams.pass_thru
1220             || contains(style.pass_thru_chars, c)
1221             || contains(runparams.pass_thru_chars, c)) {
1222                 if (c != '\0') {
1223                         Encoding const * const enc = runparams.encoding;
1224                         if (enc && !enc->encodable(c))
1225                                 throw EncodingException(c);
1226                         os.put(c);
1227                 }
1228                 return;
1229         }
1230
1231         // TIPA uses its own T3 encoding
1232         if (runparams.inIPA && latexSpecialT3(c, os, i, column))
1233                 return;
1234         // If T1 font encoding is used, use the special
1235         // characters it provides.
1236         // NOTE: Some languages reset the font encoding internally to a
1237         //       non-standard font encoding. If we are using such a language,
1238         //       we do not output special T1 chars.
1239         if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1240             && !runparams.isFullUnicode() && bparams.main_font_encoding() == "T1"
1241             && latexSpecialT1(c, os, i, column))
1242                 return;
1243         // NOTE: XeTeX and LuaTeX use EU1/2 (pre 2017) or TU (as of 2017) encoding
1244         else if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1245                  && runparams.isFullUnicode() && latexSpecialTU(c, os, i, column))
1246                      return;
1247
1248         // Otherwise, we use what LaTeX provides us.
1249         switch (c) {
1250         case '\\':
1251                 os << "\\textbackslash" << termcmd;
1252                 column += 15;
1253                 break;
1254         case '<':
1255                 os << "\\textless" << termcmd;
1256                 column += 10;
1257                 break;
1258         case '>':
1259                 os << "\\textgreater" << termcmd;
1260                 column += 13;
1261                 break;
1262         case '|':
1263                 os << "\\textbar" << termcmd;
1264                 column += 9;
1265                 break;
1266         case '-':
1267                 os << '-';
1268                 if (i + 1 < static_cast<pos_type>(text_.size()) &&
1269                     (end_pos == -1 || i + 1 < end_pos) &&
1270                     text_[i+1] == '-') {
1271                         // Prevent "--" becoming an en dash and "---" an em dash.
1272                         // (Within \ttfamily, "---" is merged to en dash + hyphen.)
1273                         os << "{}";
1274                         column += 2;
1275                 }
1276                 break;
1277         case '\"':
1278                 os << "\\textquotedbl" << termcmd;
1279                 column += 14;
1280                 break;
1281
1282         case '$': case '&':
1283         case '%': case '#': case '{':
1284         case '}': case '_':
1285                 os << '\\';
1286                 os.put(c);
1287                 column += 1;
1288                 break;
1289
1290         case '~':
1291                 os << "\\textasciitilde" << termcmd;
1292                 column += 16;
1293                 break;
1294
1295         case '^':
1296                 os << "\\textasciicircum" << termcmd;
1297                 column += 17;
1298                 break;
1299
1300         case '*':
1301         case '[':
1302         case ']':
1303                 // avoid being mistaken for optional arguments
1304                 os << '{';
1305                 os.put(c);
1306                 os << '}';
1307                 column += 2;
1308                 break;
1309
1310         case ' ':
1311                 // Blanks are printed before font switching.
1312                 // Sure? I am not! (try nice-latex)
1313                 // I am sure it's correct. LyX might be smarter
1314                 // in the future, but for now, nothing wrong is
1315                 // written. (Asger)
1316                 break;
1317
1318         case 0x2013:
1319         case 0x2014:
1320                 // XeTeX's dash behaviour is determined via a global setting
1321                 if (bparams.use_dash_ligatures
1322                     && owner_->getFontSettings(bparams, i).fontInfo().family() != TYPEWRITER_FAMILY
1323                     && !runparams.inIPA
1324                         // TODO #10961: && not in inset Flex Code
1325                         // TODO #10961: && not in layout LyXCode
1326                     && (!bparams.useNonTeXFonts || runparams.flavor != OutputParams::XETEX)) {
1327                         if (c == 0x2013) {
1328                                 // en-dash
1329                                 os << "--";
1330                                 column +=2;
1331                         } else {
1332                                 // em-dash
1333                                 os << "---";
1334                                 column +=3;
1335                         }
1336                         break;
1337                 }
1338                 // fall through
1339         default:
1340                 if (c == '\0')
1341                         return;
1342
1343                 Encoding const & encoding = *(runparams.encoding);
1344                 char_type next = '\0';
1345                 if (i + 1 < int(text_.size())) {
1346                         next = text_[i + 1];
1347                         if (Encodings::isCombiningChar(next)) {
1348                                 column += latexSurrogatePair(os, c, next, runparams) - 1;
1349                                 ++i;
1350                                 break;
1351                         }
1352                 }
1353                 string script;
1354                 pair<docstring, bool> latex = encoding.latexChar(c);
1355                 docstring nextlatex;
1356                 bool nexttipas = false;
1357                 string nexttipashortcut;
1358                 if (next != '\0' && next != META_INSET && encoding.encodable(next)) {
1359                         nextlatex = encoding.latexChar(next).first;
1360                         if (runparams.inIPA) {
1361                                 nexttipashortcut = Encodings::TIPAShortcut(next);
1362                                 nexttipas = !nexttipashortcut.empty();
1363                         }
1364                 }
1365                 bool tipas = false;
1366                 if (runparams.inIPA) {
1367                         string const tipashortcut = Encodings::TIPAShortcut(c);
1368                         if (!tipashortcut.empty()) {
1369                                 latex.first = from_ascii(tipashortcut);
1370                                 latex.second = false;
1371                                 tipas = true;
1372                         }
1373                 }
1374                 string fontenc;
1375                 if (running_font.language()->lang() == runparams.document_language)
1376                         fontenc = runparams.main_fontenc;
1377                 else
1378                         fontenc = running_font.language()->fontenc();
1379                 if (Encodings::isKnownScriptChar(c, script)
1380                     && prefixIs(latex.first, from_ascii("\\" + script)))
1381                         column += writeScriptChars(runparams, os, latex.first,
1382                                                    running_change, encoding,
1383                                                    fontenc, i) - 1;
1384                 else if (latex.second
1385                          && ((!prefixIs(nextlatex, '\\')
1386                                && !prefixIs(nextlatex, '{')
1387                                && !prefixIs(nextlatex, '}'))
1388                              || (nexttipas
1389                                  && !prefixIs(from_ascii(nexttipashortcut), '\\')))
1390                          && !tipas) {
1391                         // Prevent eating of a following
1392                         // space or command corruption by
1393                         // following characters
1394                         if (next == ' ' || next == '\0') {
1395                                 column += latex.first.length() + 1;
1396                                 os << latex.first << "{}";
1397                         } else {
1398                                 column += latex.first.length();
1399                                 os << latex.first << " ";
1400                         }
1401                 } else {
1402                         column += latex.first.length() - 1;
1403                         os << latex.first;
1404                 }
1405                 break;
1406         }
1407 }
1408
1409
1410 bool Paragraph::Private::latexSpecialT1(char_type const c, otexstream & os,
1411         pos_type i, unsigned int & column)
1412 {
1413         switch (c) {
1414         case '>':
1415         case '<':
1416                 os.put(c);
1417                 // In T1 encoding, these characters exist
1418                 // but we should avoid ligatures
1419                 if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
1420                         return true;
1421                 os << "\\textcompwordmark" << termcmd;
1422                 column += 19;
1423                 return true;
1424         case '|':
1425                 os.put(c);
1426                 return true;
1427         case '\"':
1428                 // soul.sty breaks with \char`\"
1429                 os << "\\textquotedbl" << termcmd;
1430                 column += 14;
1431                 return true;
1432         default:
1433                 return false;
1434         }
1435 }
1436
1437
1438 bool Paragraph::Private::latexSpecialTU(char_type const c, otexstream & os,
1439         pos_type i, unsigned int & column)
1440 {
1441         // TU encoding is currently on par with T1.
1442         return latexSpecialT1(c, os, i, column);
1443 }
1444
1445
1446 bool Paragraph::Private::latexSpecialT3(char_type const c, otexstream & os,
1447         pos_type /*i*/, unsigned int & column)
1448 {
1449         switch (c) {
1450         case '*':
1451         case '[':
1452         case ']':
1453         case '\"':
1454                 os.put(c);
1455                 return true;
1456         case '|':
1457                 os << "\\textvertline" << termcmd;
1458                 column += 14;
1459                 return true;
1460         default:
1461                 return false;
1462         }
1463 }
1464
1465
1466 void Paragraph::Private::validate(LaTeXFeatures & features) const
1467 {
1468         if (layout_->inpreamble && inset_owner_) {
1469                 // FIXME: Using a string stream here circumvents the encoding
1470                 // switching machinery of odocstream. Therefore the
1471                 // output is wrong if this paragraph contains content
1472                 // that needs to switch encoding.
1473                 Buffer const & buf = inset_owner_->buffer();
1474                 otexstringstream os;
1475                 os << layout_->preamble();
1476                 size_t const length = os.length();
1477                 TeXOnePar(buf, buf.text(), buf.getParFromID(owner_->id()).pit(), os,
1478                           features.runparams(), string(), 0, -1, true);
1479                 if (os.length() > length)
1480                         features.addPreambleSnippet(os.release(), true);
1481         }
1482
1483         if (features.runparams().flavor == OutputParams::HTML
1484             && layout_->htmltitle()) {
1485                 features.setHTMLTitle(owner_->asString(AS_STR_INSETS | AS_STR_SKIPDELETE));
1486         }
1487
1488         // check the params.
1489         if (!params_.spacing().isDefault())
1490                 features.require("setspace");
1491
1492         // then the layouts
1493         features.useLayout(layout_->name());
1494
1495         // then the fonts
1496         fontlist_.validate(features);
1497
1498         // then the indentation
1499         if (!params_.leftIndent().zero())
1500                 features.require("ParagraphLeftIndent");
1501
1502         // then the insets
1503         InsetList::const_iterator icit = insetlist_.begin();
1504         InsetList::const_iterator iend = insetlist_.end();
1505         for (; icit != iend; ++icit) {
1506                 if (icit->inset) {
1507                         features.inDeletedInset(owner_->isDeleted(icit->pos));
1508                         if (icit->inset->lyxCode() == FOOT_CODE) {
1509                                 // FIXME: an item inset would make things much easier.
1510                                 if ((layout_->latextype == LATEX_LIST_ENVIRONMENT
1511                                      || (layout_->latextype == LATEX_ITEM_ENVIRONMENT
1512                                          && layout_->margintype == MARGIN_FIRST_DYNAMIC))
1513                                     && (icit->pos < begin_of_body_
1514                                         || (icit->pos == begin_of_body_
1515                                             && (icit->pos == 0 || text_[icit->pos - 1] != ' '))))
1516                                         features.saveNoteEnv("description");
1517                         }
1518                         icit->inset->validate(features);
1519                         features.inDeletedInset(false);
1520                         if (layout_->needprotect &&
1521                             icit->inset->lyxCode() == FOOT_CODE)
1522                                 features.require("NeedLyXFootnoteCode");
1523                 }
1524         }
1525
1526         // then the contents
1527         BufferParams const bp = features.runparams().is_child
1528                 ? features.buffer().masterParams() : features.buffer().params();
1529         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1530                 char_type c = text_[i];
1531                 if (c == 0x0022) {
1532                         if (features.runparams().isFullUnicode() && bp.useNonTeXFonts)
1533                                 features.require("textquotedblp");
1534                         else if (bp.main_font_encoding() != "T1"
1535                                  || ((&owner_->getFontSettings(bp, i))->language()->internalFontEncoding()))
1536                                 features.require("textquotedbl");
1537                 }
1538                 if (!bp.use_dash_ligatures
1539                     && (c == 0x2013 || c == 0x2014)
1540                     && bp.useNonTeXFonts
1541                     && features.runparams().flavor == OutputParams::XETEX)
1542                         // XeTeX's dash behaviour is determined via a global setting
1543                         features.require("xetexdashbreakstate");
1544                 BufferEncodings::validate(c, features);
1545         }
1546 }
1547
1548 /////////////////////////////////////////////////////////////////////
1549 //
1550 // Paragraph
1551 //
1552 /////////////////////////////////////////////////////////////////////
1553
1554 namespace {
1555         Layout const emptyParagraphLayout;
1556 }
1557
1558 Paragraph::Paragraph()
1559         : d(new Paragraph::Private(this, emptyParagraphLayout))
1560 {
1561         itemdepth = 0;
1562         d->params_.clear();
1563 }
1564
1565
1566 Paragraph::Paragraph(Paragraph const & par)
1567         : itemdepth(par.itemdepth),
1568         d(new Paragraph::Private(*par.d, this))
1569 {
1570         registerWords();
1571 }
1572
1573
1574 Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
1575         : itemdepth(par.itemdepth),
1576         d(new Paragraph::Private(*par.d, this, beg, end))
1577 {
1578         registerWords();
1579 }
1580
1581
1582 Paragraph & Paragraph::operator=(Paragraph const & par)
1583 {
1584         // needed as we will destroy the private part before copying it
1585         if (&par != this) {
1586                 itemdepth = par.itemdepth;
1587
1588                 deregisterWords();
1589                 delete d;
1590                 d = new Private(*par.d, this);
1591                 registerWords();
1592         }
1593         return *this;
1594 }
1595
1596
1597 Paragraph::~Paragraph()
1598 {
1599         deregisterWords();
1600         delete d;
1601 }
1602
1603
1604 namespace {
1605
1606 // this shall be called just before every "os << ..." action.
1607 void flushString(ostream & os, docstring & s)
1608 {
1609         os << to_utf8(s);
1610         s.erase();
1611 }
1612
1613 } // namespace
1614
1615
1616 void Paragraph::write(ostream & os, BufferParams const & bparams,
1617         depth_type & dth) const
1618 {
1619         // The beginning or end of a deeper (i.e. nested) area?
1620         if (dth != d->params_.depth()) {
1621                 if (d->params_.depth() > dth) {
1622                         while (d->params_.depth() > dth) {
1623                                 os << "\n\\begin_deeper";
1624                                 ++dth;
1625                         }
1626                 } else {
1627                         while (d->params_.depth() < dth) {
1628                                 os << "\n\\end_deeper";
1629                                 --dth;
1630                         }
1631                 }
1632         }
1633
1634         // First write the layout
1635         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1636
1637         d->params_.write(os);
1638
1639         Font font1(inherit_font, bparams.language);
1640
1641         Change running_change = Change(Change::UNCHANGED);
1642
1643         // this string is used as a buffer to avoid repetitive calls
1644         // to to_utf8(), which turn out to be expensive (JMarc)
1645         docstring write_buffer;
1646
1647         int column = 0;
1648         for (pos_type i = 0; i <= size(); ++i) {
1649
1650                 Change const & change = lookupChange(i);
1651                 if (change != running_change)
1652                         flushString(os, write_buffer);
1653                 Changes::lyxMarkChange(os, bparams, column, running_change, change);
1654                 running_change = change;
1655
1656                 if (i == size())
1657                         break;
1658
1659                 // Write font changes
1660                 Font font2 = getFontSettings(bparams, i);
1661                 if (font2 != font1) {
1662                         flushString(os, write_buffer);
1663                         font2.lyxWriteChanges(font1, os);
1664                         column = 0;
1665                         font1 = font2;
1666                 }
1667
1668                 char_type const c = d->text_[i];
1669                 switch (c) {
1670                 case META_INSET:
1671                         if (Inset const * inset = getInset(i)) {
1672                                 flushString(os, write_buffer);
1673                                 if (inset->directWrite()) {
1674                                         // international char, let it write
1675                                         // code directly so it's shorter in
1676                                         // the file
1677                                         inset->write(os);
1678                                 } else {
1679                                         if (i)
1680                                                 os << '\n';
1681                                         os << "\\begin_inset ";
1682                                         inset->write(os);
1683                                         os << "\n\\end_inset\n\n";
1684                                         column = 0;
1685                                 }
1686                                 // FIXME This can be removed again once the mystery
1687                                 // crash has been resolved.
1688                                 os << flush;
1689                         }
1690                         break;
1691                 case '\\':
1692                         flushString(os, write_buffer);
1693                         os << "\n\\backslash\n";
1694                         column = 0;
1695                         break;
1696                 case '.':
1697                         flushString(os, write_buffer);
1698                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1699                                 os << ".\n";
1700                                 column = 0;
1701                         } else
1702                                 os << '.';
1703                         break;
1704                 default:
1705                         if ((column > 70 && c == ' ')
1706                             || column > 79) {
1707                                 flushString(os, write_buffer);
1708                                 os << '\n';
1709                                 column = 0;
1710                         }
1711                         // this check is to amend a bug. LyX sometimes
1712                         // inserts '\0' this could cause problems.
1713                         if (c != '\0')
1714                                 write_buffer.push_back(c);
1715                         else
1716                                 LYXERR0("NUL char in structure.");
1717                         ++column;
1718                         break;
1719                 }
1720         }
1721
1722         flushString(os, write_buffer);
1723         os << "\n\\end_layout\n";
1724         // FIXME This can be removed again once the mystery
1725         // crash has been resolved.
1726         os << flush;
1727 }
1728
1729
1730 void Paragraph::validate(LaTeXFeatures & features) const
1731 {
1732         d->validate(features);
1733         if (needsCProtection())
1734                 features.require("cprotect");
1735 }
1736
1737
1738 void Paragraph::insert(pos_type start, docstring const & str,
1739                        Font const & font, Change const & change)
1740 {
1741         for (size_t i = 0, n = str.size(); i != n ; ++i)
1742                 insertChar(start + i, str[i], font, change);
1743 }
1744
1745
1746 void Paragraph::appendChar(char_type c, Font const & font,
1747                 Change const & change)
1748 {
1749         // track change
1750         d->changes_.insert(change, d->text_.size());
1751         // when appending characters, no need to update tables
1752         d->text_.push_back(c);
1753         setFont(d->text_.size() - 1, font);
1754         d->requestSpellCheck(d->text_.size() - 1);
1755 }
1756
1757
1758 void Paragraph::appendString(docstring const & s, Font const & font,
1759                 Change const & change)
1760 {
1761         pos_type end = s.size();
1762         size_t oldsize = d->text_.size();
1763         size_t newsize = oldsize + end;
1764         size_t capacity = d->text_.capacity();
1765         if (newsize >= capacity)
1766                 d->text_.reserve(max(capacity + 100, newsize));
1767
1768         // when appending characters, no need to update tables
1769         d->text_.append(s);
1770
1771         // FIXME: Optimize this!
1772         for (size_t i = oldsize; i != newsize; ++i) {
1773                 // track change
1774                 d->changes_.insert(change, i);
1775                 d->requestSpellCheck(i);
1776         }
1777         d->fontlist_.set(oldsize, font);
1778         d->fontlist_.set(newsize - 1, font);
1779 }
1780
1781
1782 void Paragraph::insertChar(pos_type pos, char_type c,
1783                            bool trackChanges)
1784 {
1785         d->insertChar(pos, c, Change(trackChanges ?
1786                            Change::INSERTED : Change::UNCHANGED));
1787 }
1788
1789
1790 void Paragraph::insertChar(pos_type pos, char_type c,
1791                            Font const & font, bool trackChanges)
1792 {
1793         d->insertChar(pos, c, Change(trackChanges ?
1794                            Change::INSERTED : Change::UNCHANGED));
1795         setFont(pos, font);
1796 }
1797
1798
1799 void Paragraph::insertChar(pos_type pos, char_type c,
1800                            Font const & font, Change const & change)
1801 {
1802         d->insertChar(pos, c, change);
1803         setFont(pos, font);
1804 }
1805
1806
1807 void Paragraph::resetFonts(Font const & font)
1808 {
1809         d->fontlist_.clear();
1810         d->fontlist_.set(0, font);
1811         d->fontlist_.set(d->text_.size() - 1, font);
1812 }
1813
1814 // Gets uninstantiated font setting at position.
1815 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1816                                          pos_type pos) const
1817 {
1818         if (pos > size()) {
1819                 LYXERR0("pos: " << pos << " size: " << size());
1820                 LBUFERR(false);
1821         }
1822
1823         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1824         if (cit != d->fontlist_.end())
1825                 return cit->font();
1826
1827         if (pos == size() && !empty())
1828                 return getFontSettings(bparams, pos - 1);
1829
1830         // Optimisation: avoid a full font instantiation if there is no
1831         // language change from previous call.
1832         static Font previous_font;
1833         static Language const * previous_lang = 0;
1834         Language const * lang = getParLanguage(bparams);
1835         if (lang != previous_lang) {
1836                 previous_lang = lang;
1837                 previous_font = Font(inherit_font, lang);
1838         }
1839         return previous_font;
1840 }
1841
1842
1843 FontSpan Paragraph::fontSpan(pos_type pos) const
1844 {
1845         LBUFERR(pos <= size());
1846
1847         if (pos == size())
1848                 return FontSpan(pos, pos);
1849
1850         pos_type start = 0;
1851         FontList::const_iterator cit = d->fontlist_.begin();
1852         FontList::const_iterator end = d->fontlist_.end();
1853         for (; cit != end; ++cit) {
1854                 if (cit->pos() >= pos) {
1855                         if (pos >= beginOfBody())
1856                                 return FontSpan(max(start, beginOfBody()),
1857                                                 cit->pos());
1858                         else
1859                                 return FontSpan(start,
1860                                                 min(beginOfBody() - 1,
1861                                                          cit->pos()));
1862                 }
1863                 start = cit->pos() + 1;
1864         }
1865
1866         // This should not happen, but if so, we take no chances.
1867         LYXERR0("Paragraph::fontSpan: position not found in fontinfo table!");
1868         LASSERT(false, return FontSpan(pos, pos));
1869 }
1870
1871
1872 // Gets uninstantiated font setting at position 0
1873 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1874 {
1875         if (!empty() && !d->fontlist_.empty())
1876                 return d->fontlist_.begin()->font();
1877
1878         // Optimisation: avoid a full font instantiation if there is no
1879         // language change from previous call.
1880         static Font previous_font;
1881         static Language const * previous_lang = 0;
1882         if (bparams.language != previous_lang) {
1883                 previous_lang = bparams.language;
1884                 previous_font = Font(inherit_font, bparams.language);
1885         }
1886
1887         return previous_font;
1888 }
1889
1890
1891 // Gets the fully instantiated font at a given position in a paragraph
1892 // This is basically the same function as Text::GetFont() in text2.cpp.
1893 // The difference is that this one is used for generating the LaTeX file,
1894 // and thus cosmetic "improvements" are disallowed: This has to deliver
1895 // the true picture of the buffer. (Asger)
1896 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1897                                  Font const & outerfont) const
1898 {
1899         LBUFERR(pos >= 0);
1900
1901         Font font = getFontSettings(bparams, pos);
1902
1903         pos_type const body_pos = beginOfBody();
1904         FontInfo & fi = font.fontInfo();
1905         if (pos < body_pos)
1906                 fi.realize(d->layout_->labelfont);
1907         else
1908                 fi.realize(d->layout_->font);
1909
1910         fi.realize(outerfont.fontInfo());
1911         fi.realize(bparams.getFont().fontInfo());
1912
1913         return font;
1914 }
1915
1916
1917 Font const Paragraph::getLabelFont
1918         (BufferParams const & bparams, Font const & outerfont) const
1919 {
1920         FontInfo tmpfont = d->layout_->labelfont;
1921         tmpfont.realize(outerfont.fontInfo());
1922         tmpfont.realize(bparams.getFont().fontInfo());
1923         return Font(tmpfont, getParLanguage(bparams));
1924 }
1925
1926
1927 Font const Paragraph::getLayoutFont
1928         (BufferParams const & bparams, Font const & outerfont) const
1929 {
1930         FontInfo tmpfont = d->layout_->font;
1931         tmpfont.realize(outerfont.fontInfo());
1932         tmpfont.realize(bparams.getFont().fontInfo());
1933         return Font(tmpfont, getParLanguage(bparams));
1934 }
1935
1936
1937 char_type Paragraph::getUChar(BufferParams const & bparams, pos_type pos) const
1938 {
1939         char_type c = d->text_[pos];
1940         if (!getFontSettings(bparams, pos).isRightToLeft())
1941                 return c;
1942
1943         // FIXME: The arabic special casing is due to the difference of arabic
1944         // round brackets input introduced in r18599. Check if this should be
1945         // unified with Hebrew or at least if all bracket types should be
1946         // handled the same (file format change in either case).
1947         string const & lang = getFontSettings(bparams, pos).language()->lang();
1948         bool const arabic = lang == "arabic_arabtex" || lang == "arabic_arabi"
1949                 || lang == "farsi";
1950         char_type uc = c;
1951         switch (c) {
1952         case '(':
1953                 uc = arabic ? c : ')';
1954                 break;
1955         case ')':
1956                 uc = arabic ? c : '(';
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         case '<':
1971                 uc = '>';
1972                 break;
1973         case '>':
1974                 uc = '<';
1975                 break;
1976         }
1977
1978         return uc;
1979 }
1980
1981
1982 void Paragraph::setFont(pos_type pos, Font const & font)
1983 {
1984         LASSERT(pos <= size(), return);
1985
1986         // First, reduce font against layout/label font
1987         // Update: The setCharFont() routine in text2.cpp already
1988         // reduces font, so we don't need to do that here. (Asger)
1989
1990         d->fontlist_.set(pos, font);
1991 }
1992
1993
1994 void Paragraph::makeSameLayout(Paragraph const & par)
1995 {
1996         d->layout_ = par.d->layout_;
1997         d->params_ = par.d->params_;
1998 }
1999
2000
2001 bool Paragraph::stripLeadingSpaces(bool trackChanges)
2002 {
2003         if (isFreeSpacing())
2004                 return false;
2005
2006         int pos = 0;
2007         int count = 0;
2008
2009         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
2010                 if (eraseChar(pos, trackChanges))
2011                         ++count;
2012                 else
2013                         ++pos;
2014         }
2015
2016         return count > 0 || pos > 0;
2017 }
2018
2019
2020 bool Paragraph::hasSameLayout(Paragraph const & par) const
2021 {
2022         return par.d->layout_ == d->layout_
2023                 && d->params_.sameLayout(par.d->params_);
2024 }
2025
2026
2027 depth_type Paragraph::getDepth() const
2028 {
2029         return d->params_.depth();
2030 }
2031
2032
2033 depth_type Paragraph::getMaxDepthAfter() const
2034 {
2035         if (d->layout_->isEnvironment())
2036                 return d->params_.depth() + 1;
2037         else
2038                 return d->params_.depth();
2039 }
2040
2041
2042 LyXAlignment Paragraph::getAlign() const
2043 {
2044         if (d->params_.align() == LYX_ALIGN_LAYOUT)
2045                 return d->layout_->align;
2046         else
2047                 return d->params_.align();
2048 }
2049
2050
2051 docstring const & Paragraph::labelString() const
2052 {
2053         return d->params_.labelString();
2054 }
2055
2056
2057 // the next two functions are for the manual labels
2058 docstring const Paragraph::getLabelWidthString() const
2059 {
2060         if (d->layout_->margintype == MARGIN_MANUAL
2061             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
2062                 return d->params_.labelWidthString();
2063         else
2064                 return _("Senseless with this layout!");
2065 }
2066
2067
2068 void Paragraph::setLabelWidthString(docstring const & s)
2069 {
2070         d->params_.labelWidthString(s);
2071 }
2072
2073
2074 docstring Paragraph::expandLabel(Layout const & layout,
2075                 BufferParams const & bparams) const
2076 {
2077         return expandParagraphLabel(layout, bparams, true);
2078 }
2079
2080
2081 docstring Paragraph::expandDocBookLabel(Layout const & layout,
2082                 BufferParams const & bparams) const
2083 {
2084         return expandParagraphLabel(layout, bparams, false);
2085 }
2086
2087
2088 docstring Paragraph::expandParagraphLabel(Layout const & layout,
2089                 BufferParams const & bparams, bool process_appendix) const
2090 {
2091         DocumentClass const & tclass = bparams.documentClass();
2092         string const & lang = getParLanguage(bparams)->code();
2093         bool const in_appendix = process_appendix && d->params_.appendix();
2094         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
2095
2096         if (fmt.empty() && !layout.counter.empty())
2097                 return tclass.counters().theCounter(layout.counter, lang);
2098
2099         // handle 'inherited level parts' in 'fmt',
2100         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
2101         size_t const i = fmt.find('@', 0);
2102         if (i != docstring::npos) {
2103                 size_t const j = fmt.find('@', i + 1);
2104                 if (j != docstring::npos) {
2105                         docstring parent(fmt, i + 1, j - i - 1);
2106                         docstring label = from_ascii("??");
2107                         if (tclass.hasLayout(parent))
2108                                 label = expandParagraphLabel(tclass[parent], bparams,
2109                                                       process_appendix);
2110                         fmt = docstring(fmt, 0, i) + label
2111                                 + docstring(fmt, j + 1, docstring::npos);
2112                 }
2113         }
2114
2115         return tclass.counters().counterLabel(fmt, lang);
2116 }
2117
2118
2119 void Paragraph::applyLayout(Layout const & new_layout)
2120 {
2121         d->layout_ = &new_layout;
2122         LyXAlignment const oldAlign = d->params_.align();
2123
2124         if (!(oldAlign & d->layout_->alignpossible)) {
2125                 frontend::Alert::warning(_("Alignment not permitted"),
2126                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
2127                 d->params_.align(LYX_ALIGN_LAYOUT);
2128         }
2129 }
2130
2131
2132 pos_type Paragraph::beginOfBody() const
2133 {
2134         return d->begin_of_body_;
2135 }
2136
2137
2138 void Paragraph::setBeginOfBody()
2139 {
2140         if (d->layout_->labeltype != LABEL_MANUAL) {
2141                 d->begin_of_body_ = 0;
2142                 return;
2143         }
2144
2145         // Unroll the first two cycles of the loop
2146         // and remember the previous character to
2147         // remove unnecessary getChar() calls
2148         pos_type i = 0;
2149         pos_type end = size();
2150         if (i < end && !(isNewline(i) || isEnvSeparator(i))) {
2151                 ++i;
2152                 if (i < end) {
2153                         char_type previous_char = d->text_[i];
2154                         if (!(isNewline(i) || isEnvSeparator(i))) {
2155                                 ++i;
2156                                 while (i < end && previous_char != ' ') {
2157                                         char_type temp = d->text_[i];
2158                                         if (isNewline(i) || isEnvSeparator(i))
2159                                                 break;
2160                                         ++i;
2161                                         previous_char = temp;
2162                                 }
2163                         }
2164                 }
2165         }
2166
2167         d->begin_of_body_ = i;
2168 }
2169
2170
2171 bool Paragraph::allowParagraphCustomization() const
2172 {
2173         return inInset().allowParagraphCustomization();
2174 }
2175
2176
2177 bool Paragraph::usePlainLayout() const
2178 {
2179         return inInset().usePlainLayout();
2180 }
2181
2182
2183 bool Paragraph::isPassThru() const
2184 {
2185         return inInset().isPassThru() || d->layout_->pass_thru;
2186 }
2187
2188 namespace {
2189
2190 // paragraphs inside floats need different alignment tags to avoid
2191 // unwanted space
2192
2193 bool noTrivlistCentering(InsetCode code)
2194 {
2195         return code == FLOAT_CODE
2196                || code == WRAP_CODE
2197                || code == CELL_CODE;
2198 }
2199
2200
2201 string correction(string const & orig)
2202 {
2203         if (orig == "flushleft")
2204                 return "raggedright";
2205         if (orig == "flushright")
2206                 return "raggedleft";
2207         if (orig == "center")
2208                 return "centering";
2209         return orig;
2210 }
2211
2212
2213 bool corrected_env(otexstream & os, string const & suffix, string const & env,
2214         InsetCode code, bool const lastpar, int & col)
2215 {
2216         string macro = suffix + "{";
2217         if (noTrivlistCentering(code)) {
2218                 if (lastpar) {
2219                         // the last paragraph in non-trivlist-aligned
2220                         // context is special (to avoid unwanted whitespace)
2221                         if (suffix == "\\begin") {
2222                                 macro = "\\" + correction(env) + "{}";
2223                                 os << from_ascii(macro);
2224                                 col += macro.size();
2225                                 return true;
2226                         }
2227                         return false;
2228                 }
2229                 macro += correction(env);
2230         } else
2231                 macro += env;
2232         macro += "}";
2233         if (suffix == "\\par\\end") {
2234                 os << breakln;
2235                 col = 0;
2236         }
2237         os << from_ascii(macro);
2238         col += macro.size();
2239         if (suffix == "\\begin") {
2240                 os << breakln;
2241                 col = 0;
2242         }
2243         return true;
2244 }
2245
2246 } // namespace
2247
2248
2249 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
2250                         otexstream & os, OutputParams const & runparams) const
2251 {
2252         int column = 0;
2253
2254         bool canindent =
2255                 (bparams.paragraph_separation == BufferParams::ParagraphIndentSeparation) ?
2256                         (layout_->toggle_indent != ITOGGLE_NEVER) :
2257                         (layout_->toggle_indent == ITOGGLE_ALWAYS);
2258
2259         if (canindent && params_.noindent() && !layout_->pass_thru) {
2260                 os << "\\noindent ";
2261                 column += 10;
2262         }
2263
2264         LyXAlignment const curAlign = params_.align();
2265
2266         if (curAlign == layout_->align)
2267                 return column;
2268
2269         switch (curAlign) {
2270         case LYX_ALIGN_NONE:
2271         case LYX_ALIGN_BLOCK:
2272         case LYX_ALIGN_LAYOUT:
2273         case LYX_ALIGN_SPECIAL:
2274         case LYX_ALIGN_DECIMAL:
2275                 break;
2276         case LYX_ALIGN_LEFT:
2277         case LYX_ALIGN_RIGHT:
2278         case LYX_ALIGN_CENTER:
2279                 if (runparams.moving_arg) {
2280                         os << "\\protect";
2281                         column += 8;
2282                 }
2283                 break;
2284         }
2285
2286         string const begin_tag = "\\begin";
2287         InsetCode code = ownerCode();
2288         bool const lastpar = runparams.isLastPar;
2289
2290         switch (curAlign) {
2291         case LYX_ALIGN_NONE:
2292         case LYX_ALIGN_BLOCK:
2293         case LYX_ALIGN_LAYOUT:
2294         case LYX_ALIGN_SPECIAL:
2295         case LYX_ALIGN_DECIMAL:
2296                 break;
2297         case LYX_ALIGN_LEFT: {
2298                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2299                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2300                 else
2301                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2302                 break;
2303         } case LYX_ALIGN_RIGHT: {
2304                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2305                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2306                 else
2307                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2308                 break;
2309         } case LYX_ALIGN_CENTER: {
2310                 corrected_env(os, begin_tag, "center", code, lastpar, column);
2311                 break;
2312         }
2313         }
2314
2315         return column;
2316 }
2317
2318
2319 bool Paragraph::Private::endTeXParParams(BufferParams const & bparams,
2320                         otexstream & os, OutputParams const & runparams) const
2321 {
2322         LyXAlignment const curAlign = params_.align();
2323
2324         if (curAlign == layout_->align)
2325                 return false;
2326
2327         switch (curAlign) {
2328         case LYX_ALIGN_NONE:
2329         case LYX_ALIGN_BLOCK:
2330         case LYX_ALIGN_LAYOUT:
2331         case LYX_ALIGN_SPECIAL:
2332         case LYX_ALIGN_DECIMAL:
2333                 break;
2334         case LYX_ALIGN_LEFT:
2335         case LYX_ALIGN_RIGHT:
2336         case LYX_ALIGN_CENTER:
2337                 if (runparams.moving_arg)
2338                         os << "\\protect";
2339                 break;
2340         }
2341
2342         bool output = false;
2343         int col = 0;
2344         string const end_tag = "\\par\\end";
2345         InsetCode code = ownerCode();
2346         bool const lastpar = runparams.isLastPar;
2347
2348         switch (curAlign) {
2349         case LYX_ALIGN_NONE:
2350         case LYX_ALIGN_BLOCK:
2351         case LYX_ALIGN_LAYOUT:
2352         case LYX_ALIGN_SPECIAL:
2353         case LYX_ALIGN_DECIMAL:
2354                 break;
2355         case LYX_ALIGN_LEFT: {
2356                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2357                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2358                 else
2359                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2360                 break;
2361         } case LYX_ALIGN_RIGHT: {
2362                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2363                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2364                 else
2365                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2366                 break;
2367         } case LYX_ALIGN_CENTER: {
2368                 corrected_env(os, end_tag, "center", code, lastpar, col);
2369                 break;
2370         }
2371         }
2372
2373         return output || lastpar;
2374 }
2375
2376
2377 // This one spits out the text of the paragraph
2378 void Paragraph::latex(BufferParams const & bparams,
2379         Font const & outerfont,
2380         otexstream & os,
2381         OutputParams const & runparams,
2382         int start_pos, int end_pos, bool force) const
2383 {
2384         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
2385
2386         // FIXME This check should not be needed. Perhaps issue an
2387         // error if it triggers.
2388         Layout const & style = inInset().forcePlainLayout() ?
2389                 bparams.documentClass().plainLayout() : *d->layout_;
2390
2391         if (!force && style.inpreamble)
2392                 return;
2393
2394         bool const allowcust = allowParagraphCustomization();
2395
2396         // Current base font for all inherited font changes, without any
2397         // change caused by an individual character, except for the language:
2398         // It is set to the language of the first character.
2399         // As long as we are in the label, this font is the base font of the
2400         // label. Before the first body character it is set to the base font
2401         // of the body.
2402         Font basefont;
2403
2404         // Maybe we have to create a optional argument.
2405         pos_type body_pos = beginOfBody();
2406         unsigned int column = 0;
2407
2408         if (body_pos > 0) {
2409                 // the optional argument is kept in curly brackets in
2410                 // case it contains a ']'
2411                 // This is not strictly needed, but if this is changed it
2412                 // would be a file format change, and tex2lyx would need
2413                 // to be adjusted, since it unconditionally removes the
2414                 // braces when it parses \item.
2415                 os << "[{";
2416                 column += 2;
2417                 basefont = getLabelFont(bparams, outerfont);
2418         } else {
2419                 basefont = getLayoutFont(bparams, outerfont);
2420         }
2421
2422         // Which font is currently active?
2423         Font running_font(basefont);
2424         // Do we have an open font change?
2425         bool open_font = false;
2426
2427         Change runningChange = Change(Change::UNCHANGED);
2428
2429         Encoding const * const prev_encoding = runparams.encoding;
2430
2431         os.texrow().start(id(), 0);
2432
2433         // if the paragraph is empty, the loop will not be entered at all
2434         if (empty()) {
2435                 // For InTitle commands, we have already opened a group
2436                 // in output_latex::TeXOnePar.
2437                 if (style.isCommand() && !style.intitle) {
2438                         os << '{';
2439                         ++column;
2440                 }
2441                 if (!style.leftdelim().empty()) {
2442                         os << style.leftdelim();
2443                         column += style.leftdelim().size();
2444                 }
2445                 if (allowcust)
2446                         column += d->startTeXParParams(bparams, os, runparams);
2447         }
2448
2449         // Whether a \par can be issued for insets typeset inline with text.
2450         // Yes if greater than 0. This has to be static.
2451         THREAD_LOCAL_STATIC int parInline = 0;
2452
2453         for (pos_type i = 0; i < size(); ++i) {
2454                 // First char in paragraph or after label?
2455                 if (i == body_pos) {
2456                         if (body_pos > 0) {
2457                                 if (open_font) {
2458                                         bool needPar = false;
2459                                         column += running_font.latexWriteEndChanges(
2460                                                 os, bparams, runparams,
2461                                                 basefont, basefont, needPar);
2462                                         open_font = false;
2463                                 }
2464                                 basefont = getLayoutFont(bparams, outerfont);
2465                                 running_font = basefont;
2466
2467                                 column += Changes::latexMarkChange(os, bparams,
2468                                                 runningChange, Change(Change::UNCHANGED),
2469                                                 runparams);
2470                                 runningChange = Change(Change::UNCHANGED);
2471
2472                                 os << "}] ";
2473                                 column +=3;
2474                         }
2475                         // For InTitle commands, we have already opened a group
2476                         // in output_latex::TeXOnePar.
2477                         if (style.isCommand() && !style.intitle) {
2478                                 os << '{';
2479                                 ++column;
2480                         }
2481
2482                         if (!style.leftdelim().empty()) {
2483                                 os << style.leftdelim();
2484                                 column += style.leftdelim().size();
2485                         }
2486
2487                         if (allowcust)
2488                                 column += d->startTeXParParams(bparams, os,
2489                                                             runparams);
2490                 }
2491
2492                 runparams.wasDisplayMath = runparams.inDisplayMath;
2493                 runparams.inDisplayMath = false;
2494                 bool deleted_display_math = false;
2495                 Change const & change = runparams.inDeletedInset
2496                         ? runparams.changeOfDeletedInset : lookupChange(i);
2497
2498                 // Check whether a display math inset follows
2499                 if (d->text_[i] == META_INSET
2500                     && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2501                         InsetMath const * im = getInset(i)->asInsetMath();
2502                         if (im && im->asHullInset()
2503                             && im->asHullInset()->outerDisplay()) {
2504                                 runparams.inDisplayMath = true;
2505                                 // runparams.inDeletedInset will be set by
2506                                 // latexInset later, but we need this info
2507                                 // before it is called. On the other hand, we
2508                                 // cannot set it here because it is a counter.
2509                                 deleted_display_math = isDeleted(i);
2510                         }
2511                         if (bparams.output_changes && deleted_display_math
2512                             && runningChange == change
2513                             && change.type == Change::DELETED
2514                             && !os.afterParbreak()) {
2515                                 // A display math in the same paragraph follows.
2516                                 // We have to close and then reopen \lyxdeleted,
2517                                 // otherwise the math will be shifted up.
2518                                 OutputParams rp = runparams;
2519                                 if (open_font) {
2520                                         bool needPar = false;
2521                                         column += running_font.latexWriteEndChanges(
2522                                                 os, bparams, rp, basefont,
2523                                                 basefont, needPar);
2524                                         open_font = false;
2525                                 }
2526                                 basefont = getLayoutFont(bparams, outerfont);
2527                                 running_font = basefont;
2528                                 column += Changes::latexMarkChange(os, bparams,
2529                                         Change(Change::INSERTED), change, rp);
2530                         }
2531                 }
2532
2533                 if (bparams.output_changes && runningChange != change) {
2534                         if (open_font) {
2535                                 bool needPar = false;
2536                                 column += running_font.latexWriteEndChanges(
2537                                                 os, bparams, runparams,
2538                                                 basefont, basefont, needPar);
2539                                 open_font = false;
2540                         }
2541                         basefont = getLayoutFont(bparams, outerfont);
2542                         running_font = basefont;
2543                         column += Changes::latexMarkChange(os, bparams, runningChange,
2544                                                            change, runparams);
2545                         runningChange = change;
2546                 }
2547
2548                 // do not output text which is marked deleted
2549                 // if change tracking output is disabled
2550                 if (!bparams.output_changes && change.deleted()) {
2551                         continue;
2552                 }
2553
2554                 ++column;
2555
2556                 // Fully instantiated font
2557                 Font const current_font = getFont(bparams, i, outerfont);
2558
2559                 Font const last_font = running_font;
2560
2561                 // Do we need to close the previous font?
2562                 if (open_font &&
2563                     (current_font != running_font ||
2564                      current_font.language() != running_font.language()))
2565                 {
2566                         bool needPar = false;
2567                         column += running_font.latexWriteEndChanges(
2568                                     os, bparams, runparams, basefont,
2569                                     (i == body_pos-1) ? basefont : current_font,
2570                                     needPar);
2571                         running_font = basefont;
2572                         open_font = false;
2573                 }
2574
2575                 string const running_lang = runparams.use_polyglossia ?
2576                         running_font.language()->polyglossia() : running_font.language()->babel();
2577                 // close babel's font environment before opening CJK.
2578                 string const lang_end_command = runparams.use_polyglossia ?
2579                         "\\end{$$lang}" : lyxrc.language_command_end;
2580                 bool const using_begin_end = runparams.use_polyglossia ||
2581                                                 !lang_end_command.empty();
2582                 if (!running_lang.empty() &&
2583                     (!using_begin_end || running_lang == openLanguageName()) &&
2584                     current_font.language()->encoding()->package() == Encoding::CJK) {
2585                                 string end_tag = subst(lang_end_command,
2586                                                         "$$lang",
2587                                                         running_lang);
2588                                 os << from_ascii(end_tag);
2589                                 column += end_tag.length();
2590                                 if (using_begin_end)
2591                                         popLanguageName();
2592                 }
2593
2594                 // Switch file encoding if necessary (and allowed)
2595                 if (!runparams.pass_thru && !style.pass_thru &&
2596                     runparams.encoding->package() != Encoding::none &&
2597                     current_font.language()->encoding()->package() != Encoding::none) {
2598                         pair<bool, int> const enc_switch =
2599                                 switchEncoding(os.os(), bparams, runparams,
2600                                         *(current_font.language()->encoding()));
2601                         if (enc_switch.first) {
2602                                 column += enc_switch.second;
2603                                 runparams.encoding = current_font.language()->encoding();
2604                         }
2605                 }
2606
2607                 char_type const c = d->text_[i];
2608
2609                 // A display math inset inside an ulem command will be output
2610                 // as a box of width \linewidth, so we have to either disable
2611                 // indentation if the inset starts a paragraph, or start a new
2612                 // line to accommodate such box. This has to be done before
2613                 // writing any font changing commands.
2614                 if (runparams.inDisplayMath && !deleted_display_math
2615                     && runparams.inulemcmd) {
2616                         if (os.afterParbreak())
2617                                 os << "\\noindent";
2618                         else
2619                                 os << "\\\\\n";
2620                 }
2621
2622                 // Do we need to change font?
2623                 if ((current_font != running_font ||
2624                      current_font.language() != running_font.language()) &&
2625                         i != body_pos - 1)
2626                 {
2627                         odocstringstream ods;
2628                         column += current_font.latexWriteStartChanges(ods, bparams,
2629                                                               runparams, basefont,
2630                                                               last_font);
2631                         // Check again for display math in ulem commands as a
2632                         // font change may also occur just before a math inset.
2633                         if (runparams.inDisplayMath && !deleted_display_math
2634                             && runparams.inulemcmd) {
2635                                 if (os.afterParbreak())
2636                                         os << "\\noindent";
2637                                 else
2638                                         os << "\\\\\n";
2639                         }
2640                         running_font = current_font;
2641                         open_font = true;
2642                         docstring fontchange = ods.str();
2643                         // check whether the fontchange ends with a \\textcolor
2644                         // modifier and the text starts with a space (bug 4473)
2645                         docstring const last_modifier = rsplit(fontchange, '\\');
2646                         if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2647                                 os << fontchange << from_ascii("{}");
2648                         // check if the fontchange ends with a trailing blank
2649                         // (like "\small " (see bug 3382)
2650                         else if (suffixIs(fontchange, ' ') && c == ' ')
2651                                 os << fontchange.substr(0, fontchange.size() - 1)
2652                                    << from_ascii("{}");
2653                         else
2654                                 os << fontchange;
2655                 }
2656
2657                 // FIXME: think about end_pos implementation...
2658                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2659                         // FIXME: integrate this case in latexSpecialChar
2660                         // Do not print the separation of the optional argument
2661                         // if style.pass_thru is false. This works because
2662                         // latexSpecialChar ignores spaces if
2663                         // style.pass_thru is false.
2664                         if (i != body_pos - 1) {
2665                                 if (d->simpleTeXBlanks(runparams, os,
2666                                                 i, column, current_font, style)) {
2667                                         // A surrogate pair was output. We
2668                                         // must not call latexSpecialChar
2669                                         // in this iteration, since it would output
2670                                         // the combining character again.
2671                                         ++i;
2672                                         continue;
2673                                 }
2674                         }
2675                 }
2676
2677                 OutputParams rp = runparams;
2678                 rp.free_spacing = style.free_spacing;
2679                 rp.local_font = &current_font;
2680                 rp.intitle = style.intitle;
2681
2682                 // Two major modes:  LaTeX or plain
2683                 // Handle here those cases common to both modes
2684                 // and then split to handle the two modes separately.
2685                 if (c == META_INSET) {
2686                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2687                                 // Greyedout notes and, in general, all insets
2688                                 // with InsetLayout::isDisplay() == false,
2689                                 // are typeset inline with the text. So, we
2690                                 // can add a \par to the last paragraph of
2691                                 // such insets only if nothing else follows.
2692                                 bool incremented = false;
2693                                 Inset const * inset = getInset(i);
2694                                 InsetText const * textinset = inset
2695                                                         ? inset->asInsetText()
2696                                                         : 0;
2697                                 if (i + 1 == size() && textinset
2698                                     && !inset->getLayout().isDisplay()) {
2699                                         ParagraphList const & pars =
2700                                                 textinset->text().paragraphs();
2701                                         pit_type const pit = pars.size() - 1;
2702                                         Font const lastfont =
2703                                                 pit < 0 || pars[pit].empty()
2704                                                 ? pars[pit].getLayoutFont(
2705                                                                 bparams,
2706                                                                 outerfont)
2707                                                 : pars[pit].getFont(bparams,
2708                                                         pars[pit].size() - 1,
2709                                                         outerfont);
2710                                         if (lastfont.fontInfo().size() !=
2711                                             basefont.fontInfo().size()) {
2712                                                 ++parInline;
2713                                                 incremented = true;
2714                                         }
2715                                 }
2716                                 d->latexInset(bparams, os, rp, running_font,
2717                                                 basefont, outerfont, open_font,
2718                                                 runningChange, style, i, column);
2719                                 if (incremented)
2720                                         --parInline;
2721                         }
2722                 } else if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2723                         try {
2724                                 d->latexSpecialChar(os, bparams, rp,
2725                                                     running_font, runningChange,
2726                                                     style, i, end_pos, column);
2727                         } catch (EncodingException & e) {
2728                                 if (runparams.dryrun) {
2729                                         os << "<" << _("LyX Warning: ")
2730                                            << _("uncodable character") << " '";
2731                                         os.put(c);
2732                                         os << "'>";
2733                                 } else {
2734                                         // add location information and throw again.
2735                                         e.par_id = id();
2736                                         e.pos = i;
2737                                         throw(e);
2738                                 }
2739                         }
2740                 }
2741
2742                 // Set the encoding to that returned from latexSpecialChar (see
2743                 // comment for encoding member in OutputParams.h)
2744                 runparams.encoding = rp.encoding;
2745
2746                 // Also carry on the info on a closed ulem command for insets
2747                 // such as Note that do not produce any output, so that no
2748                 // command is ever executed but its opening was recorded.
2749                 runparams.inulemcmd = rp.inulemcmd;
2750         }
2751
2752         // If we have an open font definition, we have to close it
2753         if (open_font) {
2754                 // Make sure that \\par is done with the font of the last
2755                 // character if this has another size as the default.
2756                 // This is necessary because LaTeX (and LyX on the screen)
2757                 // calculates the space between the baselines according
2758                 // to this font. (Matthias)
2759                 //
2760                 // We must not change the font for the last paragraph
2761                 // of non-multipar insets, tabular cells or commands,
2762                 // since this produces unwanted whitespace.
2763
2764                 Font const font = empty()
2765                         ? getLayoutFont(bparams, outerfont)
2766                         : getFont(bparams, size() - 1, outerfont);
2767
2768                 InsetText const * textinset = inInset().asInsetText();
2769
2770                 bool const maintext = textinset
2771                         ? textinset->text().isMainText()
2772                         : false;
2773
2774                 size_t const numpars = textinset
2775                         ? textinset->text().paragraphs().size()
2776                         : 0;
2777
2778                 bool needPar = false;
2779
2780                 if (style.resfont.size() != font.fontInfo().size()
2781                     && (!runparams.isLastPar || maintext
2782                         || (numpars > 1 && d->ownerCode() != CELL_CODE
2783                             && (inInset().getLayout().isDisplay()
2784                                 || parInline)))
2785                     && !style.isCommand()) {
2786                         needPar = true;
2787                 }
2788 #ifdef FIXED_LANGUAGE_END_DETECTION
2789                 if (next_) {
2790                         running_font.latexWriteEndChanges(os, bparams,
2791                                         runparams, basefont,
2792                                         next_->getFont(bparams, 0, outerfont),
2793                                                        needPar);
2794                 } else {
2795                         running_font.latexWriteEndChanges(os, bparams,
2796                                         runparams, basefont, basefont, needPar);
2797                 }
2798 #else
2799 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2800 //FIXME: there as we start another \selectlanguage with the next paragraph if
2801 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2802                 running_font.latexWriteEndChanges(os, bparams, runparams,
2803                                 basefont, basefont, needPar);
2804 #endif
2805                 if (needPar) {
2806                         // The \par could not be inserted at the same nesting
2807                         // level of the font size change, so do it now.
2808                         os << "{\\" << font.latexSize() << "\\par}";
2809                 }
2810         }
2811
2812         column += Changes::latexMarkChange(os, bparams, runningChange,
2813                                            Change(Change::UNCHANGED), runparams);
2814
2815         // Needed if there is an optional argument but no contents.
2816         if (body_pos > 0 && body_pos == size()) {
2817                 os << "}]~";
2818         }
2819
2820         if (!style.rightdelim().empty()) {
2821                 os << style.rightdelim();
2822                 column += style.rightdelim().size();
2823         }
2824
2825         if (allowcust && d->endTeXParParams(bparams, os, runparams)
2826             && runparams.encoding != prev_encoding) {
2827                 runparams.encoding = prev_encoding;
2828                 os << setEncoding(prev_encoding->iconvName());
2829         }
2830
2831         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2832 }
2833
2834
2835 bool Paragraph::emptyTag() const
2836 {
2837         for (pos_type i = 0; i < size(); ++i) {
2838                 if (Inset const * inset = getInset(i)) {
2839                         InsetCode lyx_code = inset->lyxCode();
2840                         // FIXME testing like that is wrong. What is
2841                         // the intent?
2842                         if (lyx_code != TOC_CODE &&
2843                             lyx_code != INCLUDE_CODE &&
2844                             lyx_code != GRAPHICS_CODE &&
2845                             lyx_code != ERT_CODE &&
2846                             lyx_code != LISTINGS_CODE &&
2847                             lyx_code != FLOAT_CODE &&
2848                             lyx_code != TABULAR_CODE) {
2849                                 return false;
2850                         }
2851                 } else {
2852                         char_type c = d->text_[i];
2853                         if (c != ' ' && c != '\t')
2854                                 return false;
2855                 }
2856         }
2857         return true;
2858 }
2859
2860
2861 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
2862         const
2863 {
2864         for (pos_type i = 0; i < size(); ++i) {
2865                 if (Inset const * inset = getInset(i)) {
2866                         InsetCode lyx_code = inset->lyxCode();
2867                         if (lyx_code == LABEL_CODE) {
2868                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2869                                 docstring const & id = il->getParam("name");
2870                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2871                         }
2872                 }
2873         }
2874         return string();
2875 }
2876
2877
2878 pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runparams)
2879         const
2880 {
2881         pos_type i;
2882         for (i = 0; i < size(); ++i) {
2883                 if (Inset const * inset = getInset(i)) {
2884                         inset->docbook(os, runparams);
2885                 } else {
2886                         char_type c = d->text_[i];
2887                         if (c == ' ')
2888                                 break;
2889                         os << sgml::escapeChar(c);
2890                 }
2891         }
2892         return i;
2893 }
2894
2895
2896 pos_type Paragraph::firstWordLyXHTML(XHTMLStream & xs, OutputParams const & runparams)
2897         const
2898 {
2899         pos_type i;
2900         for (i = 0; i < size(); ++i) {
2901                 if (Inset const * inset = getInset(i)) {
2902                         inset->xhtml(xs, runparams);
2903                 } else {
2904                         char_type c = d->text_[i];
2905                         if (c == ' ')
2906                                 break;
2907                         xs << c;
2908                 }
2909         }
2910         return i;
2911 }
2912
2913
2914 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2915 {
2916         Font font_old;
2917         pos_type size = text_.size();
2918         for (pos_type i = initial; i < size; ++i) {
2919                 Font font = owner_->getFont(buf.params(), i, outerfont);
2920                 if (text_[i] == META_INSET)
2921                         return false;
2922                 if (i != initial && font != font_old)
2923                         return false;
2924                 font_old = font;
2925         }
2926
2927         return true;
2928 }
2929
2930
2931 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
2932                                     odocstream & os,
2933                                     OutputParams const & runparams,
2934                                     Font const & outerfont,
2935                                     pos_type initial) const
2936 {
2937         bool emph_flag = false;
2938
2939         Layout const & style = *d->layout_;
2940         FontInfo font_old =
2941                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2942
2943         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2944                 os << "]]>";
2945
2946         // parsing main loop
2947         for (pos_type i = initial; i < size(); ++i) {
2948                 Font font = getFont(buf.params(), i, outerfont);
2949
2950                 // handle <emphasis> tag
2951                 if (font_old.emph() != font.fontInfo().emph()) {
2952                         if (font.fontInfo().emph() == FONT_ON) {
2953                                 os << "<emphasis>";
2954                                 emph_flag = true;
2955                         } else if (i != initial) {
2956                                 os << "</emphasis>";
2957                                 emph_flag = false;
2958                         }
2959                 }
2960
2961                 if (Inset const * inset = getInset(i)) {
2962                         inset->docbook(os, runparams);
2963                 } else {
2964                         char_type c = d->text_[i];
2965
2966                         if (style.pass_thru)
2967                                 os.put(c);
2968                         else
2969                                 os << sgml::escapeChar(c);
2970                 }
2971                 font_old = font.fontInfo();
2972         }
2973
2974         if (emph_flag) {
2975                 os << "</emphasis>";
2976         }
2977
2978         if (style.free_spacing)
2979                 os << '\n';
2980         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2981                 os << "<![CDATA[";
2982 }
2983
2984
2985 namespace {
2986 void doFontSwitch(vector<html::FontTag> & tagsToOpen,
2987                   vector<html::EndFontTag> & tagsToClose,
2988                   bool & flag, FontState curstate, html::FontTypes type)
2989 {
2990         if (curstate == FONT_ON) {
2991                 tagsToOpen.push_back(html::FontTag(type));
2992                 flag = true;
2993         } else if (flag) {
2994                 tagsToClose.push_back(html::EndFontTag(type));
2995                 flag = false;
2996         }
2997 }
2998 } // namespace
2999
3000
3001 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
3002                                     XHTMLStream & xs,
3003                                     OutputParams const & runparams,
3004                                     Font const & outerfont,
3005                                     bool start_paragraph, bool close_paragraph,
3006                                     pos_type initial) const
3007 {
3008         docstring retval;
3009
3010         // track whether we have opened these tags
3011         bool emph_flag = false;
3012         bool bold_flag = false;
3013         bool noun_flag = false;
3014         bool ubar_flag = false;
3015         bool dbar_flag = false;
3016         bool sout_flag = false;
3017         bool xout_flag = false;
3018         bool wave_flag = false;
3019         // shape tags
3020         bool shap_flag = false;
3021         // family tags
3022         bool faml_flag = false;
3023         // size tags
3024         bool size_flag = false;
3025
3026         Layout const & style = *d->layout_;
3027
3028         if (start_paragraph)
3029                 xs.startDivision(allowEmpty());
3030
3031         FontInfo font_old =
3032                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
3033
3034         FontShape  curr_fs   = INHERIT_SHAPE;
3035         FontFamily curr_fam  = INHERIT_FAMILY;
3036         FontSize   curr_size = FONT_SIZE_INHERIT;
3037
3038         string const default_family =
3039                 buf.masterBuffer()->params().fonts_default_family;
3040
3041         vector<html::FontTag> tagsToOpen;
3042         vector<html::EndFontTag> tagsToClose;
3043
3044         // parsing main loop
3045         for (pos_type i = initial; i < size(); ++i) {
3046                 // let's not show deleted material in the output
3047                 if (isDeleted(i))
3048                         continue;
3049
3050                 Font const font = getFont(buf.masterBuffer()->params(), i, outerfont);
3051
3052                 // emphasis
3053                 FontState curstate = font.fontInfo().emph();
3054                 if (font_old.emph() != curstate)
3055                         doFontSwitch(tagsToOpen, tagsToClose, emph_flag, curstate, html::FT_EMPH);
3056
3057                 // noun
3058                 curstate = font.fontInfo().noun();
3059                 if (font_old.noun() != curstate)
3060                         doFontSwitch(tagsToOpen, tagsToClose, noun_flag, curstate, html::FT_NOUN);
3061
3062                 // underbar
3063                 curstate = font.fontInfo().underbar();
3064                 if (font_old.underbar() != curstate)
3065                         doFontSwitch(tagsToOpen, tagsToClose, ubar_flag, curstate, html::FT_UBAR);
3066
3067                 // strikeout
3068                 curstate = font.fontInfo().strikeout();
3069                 if (font_old.strikeout() != curstate)
3070                         doFontSwitch(tagsToOpen, tagsToClose, sout_flag, curstate, html::FT_SOUT);
3071
3072                 // xout
3073                 curstate = font.fontInfo().xout();
3074                 if (font_old.xout() != curstate)
3075                         doFontSwitch(tagsToOpen, tagsToClose, xout_flag, curstate, html::FT_XOUT);
3076
3077                 // double underbar
3078                 curstate = font.fontInfo().uuline();
3079                 if (font_old.uuline() != curstate)
3080                         doFontSwitch(tagsToOpen, tagsToClose, dbar_flag, curstate, html::FT_DBAR);
3081
3082                 // wavy line
3083                 curstate = font.fontInfo().uwave();
3084                 if (font_old.uwave() != curstate)
3085                         doFontSwitch(tagsToOpen, tagsToClose, wave_flag, curstate, html::FT_WAVE);
3086
3087                 // bold
3088                 // a little hackish, but allows us to reuse what we have.
3089                 curstate = (font.fontInfo().series() == BOLD_SERIES ? FONT_ON : FONT_OFF);
3090                 if (font_old.series() != font.fontInfo().series())
3091                         doFontSwitch(tagsToOpen, tagsToClose, bold_flag, curstate, html::FT_BOLD);
3092
3093                 // Font shape
3094                 curr_fs = font.fontInfo().shape();
3095                 FontShape old_fs = font_old.shape();
3096                 if (old_fs != curr_fs) {
3097                         if (shap_flag) {
3098                                 switch (old_fs) {
3099                                 case ITALIC_SHAPE:
3100                                         tagsToClose.push_back(html::EndFontTag(html::FT_ITALIC));
3101                                         break;
3102                                 case SLANTED_SHAPE:
3103                                         tagsToClose.push_back(html::EndFontTag(html::FT_SLANTED));
3104                                         break;
3105                                 case SMALLCAPS_SHAPE:
3106                                         tagsToClose.push_back(html::EndFontTag(html::FT_SMALLCAPS));
3107                                         break;
3108                                 case UP_SHAPE:
3109                                 case INHERIT_SHAPE:
3110                                         break;
3111                                 default:
3112                                         // the other tags are for internal use
3113                                         LATTEST(false);
3114                                         break;
3115                                 }
3116                                 shap_flag = false;
3117                         }
3118                         switch (curr_fs) {
3119                         case ITALIC_SHAPE:
3120                                 tagsToOpen.push_back(html::FontTag(html::FT_ITALIC));
3121                                 shap_flag = true;
3122                                 break;
3123                         case SLANTED_SHAPE:
3124                                 tagsToOpen.push_back(html::FontTag(html::FT_SLANTED));
3125                                 shap_flag = true;
3126                                 break;
3127                         case SMALLCAPS_SHAPE:
3128                                 tagsToOpen.push_back(html::FontTag(html::FT_SMALLCAPS));
3129                                 shap_flag = true;
3130                                 break;
3131                         case UP_SHAPE:
3132                         case INHERIT_SHAPE:
3133                                 break;
3134                         default:
3135                                 // the other tags are for internal use
3136                                 LATTEST(false);
3137                                 break;
3138                         }
3139                 }
3140
3141                 // Font family
3142                 curr_fam = font.fontInfo().family();
3143                 FontFamily old_fam = font_old.family();
3144                 if (old_fam != curr_fam) {
3145                         if (faml_flag) {
3146                                 switch (old_fam) {
3147                                 case ROMAN_FAMILY:
3148                                         tagsToClose.push_back(html::EndFontTag(html::FT_ROMAN));
3149                                         break;
3150                                 case SANS_FAMILY:
3151                                         tagsToClose.push_back(html::EndFontTag(html::FT_SANS));
3152                                         break;
3153                                 case TYPEWRITER_FAMILY:
3154                                         tagsToClose.push_back(html::EndFontTag(html::FT_TYPE));
3155                                         break;
3156                                 case INHERIT_FAMILY:
3157                                         break;
3158                                 default:
3159                                         // the other tags are for internal use
3160                                         LATTEST(false);
3161                                         break;
3162                                 }
3163                                 faml_flag = false;
3164                         }
3165                         switch (curr_fam) {
3166                         case ROMAN_FAMILY:
3167                                 // we will treat a "default" font family as roman, since we have
3168                                 // no other idea what to do.
3169                                 if (default_family != "rmdefault" && default_family != "default") {
3170                                         tagsToOpen.push_back(html::FontTag(html::FT_ROMAN));
3171                                         faml_flag = true;
3172                                 }
3173                                 break;
3174                         case SANS_FAMILY:
3175                                 if (default_family != "sfdefault") {
3176                                         tagsToOpen.push_back(html::FontTag(html::FT_SANS));
3177                                         faml_flag = true;
3178                                 }
3179                                 break;
3180                         case TYPEWRITER_FAMILY:
3181                                 if (default_family != "ttdefault") {
3182                                         tagsToOpen.push_back(html::FontTag(html::FT_TYPE));
3183                                         faml_flag = true;
3184                                 }
3185                                 break;
3186                         case INHERIT_FAMILY:
3187                                 break;
3188                         default:
3189                                 // the other tags are for internal use
3190                                 LATTEST(false);
3191                                 break;
3192                         }
3193                 }
3194
3195                 // Font size
3196                 curr_size = font.fontInfo().size();
3197                 FontSize old_size = font_old.size();
3198                 if (old_size != curr_size) {
3199                         if (size_flag) {
3200                                 switch (old_size) {
3201                                 case FONT_SIZE_TINY:
3202                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_TINY));
3203                                         break;
3204                                 case FONT_SIZE_SCRIPT:
3205                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SCRIPT));
3206                                         break;
3207                                 case FONT_SIZE_FOOTNOTE:
3208                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_FOOTNOTE));
3209                                         break;
3210                                 case FONT_SIZE_SMALL:
3211                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SMALL));
3212                                         break;
3213                                 case FONT_SIZE_LARGE:
3214                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGE));
3215                                         break;
3216                                 case FONT_SIZE_LARGER:
3217                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGER));
3218                                         break;
3219                                 case FONT_SIZE_LARGEST:
3220                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGEST));
3221                                         break;
3222                                 case FONT_SIZE_HUGE:
3223                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGE));
3224                                         break;
3225                                 case FONT_SIZE_HUGER:
3226                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGER));
3227                                         break;
3228                                 case FONT_SIZE_INCREASE:
3229                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_INCREASE));
3230                                         break;
3231                                 case FONT_SIZE_DECREASE:
3232                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_DECREASE));
3233                                         break;
3234                                 case FONT_SIZE_INHERIT:
3235                                 case FONT_SIZE_NORMAL:
3236                                         break;
3237                                 default:
3238                                         // the other tags are for internal use
3239                                         LATTEST(false);
3240                                         break;
3241                                 }
3242                                 size_flag = false;
3243                         }
3244                         switch (curr_size) {
3245                         case FONT_SIZE_TINY:
3246                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_TINY));
3247                                 size_flag = true;
3248                                 break;
3249                         case FONT_SIZE_SCRIPT:
3250                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SCRIPT));
3251                                 size_flag = true;
3252                                 break;
3253                         case FONT_SIZE_FOOTNOTE:
3254                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_FOOTNOTE));
3255                                 size_flag = true;
3256                                 break;
3257                         case FONT_SIZE_SMALL:
3258                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SMALL));
3259                                 size_flag = true;
3260                                 break;
3261                         case FONT_SIZE_LARGE:
3262                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGE));
3263                                 size_flag = true;
3264                                 break;
3265                         case FONT_SIZE_LARGER:
3266                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGER));
3267                                 size_flag = true;
3268                                 break;
3269                         case FONT_SIZE_LARGEST:
3270                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGEST));
3271                                 size_flag = true;
3272                                 break;
3273                         case FONT_SIZE_HUGE:
3274                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGE));
3275                                 size_flag = true;
3276                                 break;
3277                         case FONT_SIZE_HUGER:
3278                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGER));
3279                                 size_flag = true;
3280                                 break;
3281                         case FONT_SIZE_INCREASE:
3282                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_INCREASE));
3283                                 size_flag = true;
3284                                 break;
3285                         case FONT_SIZE_DECREASE:
3286                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_DECREASE));
3287                                 size_flag = true;
3288                                 break;
3289                         case FONT_SIZE_NORMAL:
3290                         case FONT_SIZE_INHERIT:
3291                                 break;
3292                         default:
3293                                 // the other tags are for internal use
3294                                 LATTEST(false);
3295                                 break;
3296                         }
3297                 }
3298
3299                 // FIXME XHTML
3300                 // Other such tags? What about the other text ranges?
3301
3302                 vector<html::EndFontTag>::const_iterator cit = tagsToClose.begin();
3303                 vector<html::EndFontTag>::const_iterator cen = tagsToClose.end();
3304                 for (; cit != cen; ++cit)
3305                         xs << *cit;
3306
3307                 vector<html::FontTag>::const_iterator sit = tagsToOpen.begin();
3308                 vector<html::FontTag>::const_iterator sen = tagsToOpen.end();
3309                 for (; sit != sen; ++sit)
3310                         xs << *sit;
3311
3312                 tagsToClose.clear();
3313                 tagsToOpen.clear();
3314
3315                 Inset const * inset = getInset(i);
3316                 if (inset) {
3317                         if (!runparams.for_toc || inset->isInToc()) {
3318                                 OutputParams np = runparams;
3319                                 np.local_font = &font;
3320                                 // If the paragraph has size 1, then we are in the "special
3321                                 // case" where we do not output the containing paragraph info
3322                                 if (!inset->getLayout().htmlisblock() && size() != 1)
3323                                         np.html_in_par = true;
3324                                 retval += inset->xhtml(xs, np);
3325                         }
3326                 } else {
3327                         char_type c = getUChar(buf.masterBuffer()->params(), i);
3328                         xs << c;
3329                 }
3330                 font_old = font.fontInfo();
3331         }
3332
3333         // FIXME XHTML
3334         // I'm worried about what happens if a branch, say, is itself
3335         // wrapped in some font stuff. I think that will not work.
3336         xs.closeFontTags();
3337         if (close_paragraph)
3338                 xs.endDivision();
3339
3340         return retval;
3341 }
3342
3343
3344 bool Paragraph::isHfill(pos_type pos) const
3345 {
3346         Inset const * inset = getInset(pos);
3347         return inset && inset->isHfill();
3348 }
3349
3350
3351 bool Paragraph::isNewline(pos_type pos) const
3352 {
3353         // U+2028 LINE SEPARATOR
3354         // U+2029 PARAGRAPH SEPARATOR
3355         char_type const c = d->text_[pos];
3356         if (c == 0x2028 || c == 0x2029)
3357                 return true;
3358         Inset const * inset = getInset(pos);
3359         return inset && inset->lyxCode() == NEWLINE_CODE;
3360 }
3361
3362
3363 bool Paragraph::isEnvSeparator(pos_type pos) const
3364 {
3365         Inset const * inset = getInset(pos);
3366         return inset && inset->lyxCode() == SEPARATOR_CODE;
3367 }
3368
3369
3370 bool Paragraph::isLineSeparator(pos_type pos) const
3371 {
3372         char_type const c = d->text_[pos];
3373         if (isLineSeparatorChar(c))
3374                 return true;
3375         Inset const * inset = getInset(pos);
3376         return inset && inset->isLineSeparator();
3377 }
3378
3379
3380 bool Paragraph::isWordSeparator(pos_type pos) const
3381 {
3382         if (pos == size())
3383                 return true;
3384         if (Inset const * inset = getInset(pos))
3385                 return !inset->isLetter();
3386         // if we have a hard hyphen (no en- or emdash) or apostrophe
3387         // we pass this to the spell checker
3388         // FIXME: this method is subject to change, visit
3389         // https://bugzilla.mozilla.org/show_bug.cgi?id=355178
3390         // to get an impression how complex this is.
3391         if (isHardHyphenOrApostrophe(pos))
3392                 return false;
3393         char_type const c = d->text_[pos];
3394         // We want to pass the escape chars to the spellchecker
3395         docstring const escape_chars = from_utf8(lyxrc.spellchecker_esc_chars);
3396         return !isLetterChar(c) && !isDigitASCII(c) && !contains(escape_chars, c);
3397 }
3398
3399
3400 bool Paragraph::isHardHyphenOrApostrophe(pos_type pos) const
3401 {
3402         pos_type const psize = size();
3403         if (pos >= psize)
3404                 return false;
3405         char_type const c = d->text_[pos];
3406         if (c != '-' && c != '\'')
3407                 return false;
3408         int nextpos = pos + 1;
3409         int prevpos = pos > 0 ? pos - 1 : 0;
3410         if ((nextpos == psize || isSpace(nextpos))
3411                 && (pos == 0 || isSpace(prevpos)))
3412                 return false;
3413         return true;
3414 }
3415
3416
3417 bool Paragraph::needsCProtection() const
3418 {
3419         // first check the layout of the paragraph
3420         if (layout().needcprotect) {
3421                 // Environments need cprotection regardless the content
3422                 if (layout().latextype == LATEX_ENVIRONMENT)
3423                         return true;
3424
3425                 // Commands need cprotection if they contain specific chars
3426                 int const nchars_escape = 9;
3427                 static char_type const chars_escape[nchars_escape] = {
3428                         '&', '_', '$', '%', '#', '^', '{', '}', '\\'};
3429
3430                 docstring const pars = asString();
3431                 for (int k = 0; k < nchars_escape; k++) {
3432                         if (contains(pars, chars_escape[k]))
3433                                 return true;
3434                 }
3435         }
3436
3437         // now check whether we have insets that need cprotection
3438         pos_type size = d->text_.size();
3439         for (pos_type i = 0; i < size; ++i)
3440                 if (isInset(i) && getInset(i)->needsCProtection())
3441                         return true;
3442
3443         return false;
3444 }
3445
3446
3447 FontSpan const & Paragraph::getSpellRange(pos_type pos) const
3448 {
3449         return d->speller_state_.getRange(pos);
3450 }
3451
3452
3453 bool Paragraph::isChar(pos_type pos) const
3454 {
3455         if (Inset const * inset = getInset(pos))
3456                 return inset->isChar();
3457         char_type const c = d->text_[pos];
3458         return !isLetterChar(c) && !isDigitASCII(c) && !lyx::isSpace(c);
3459 }
3460
3461
3462 bool Paragraph::isSpace(pos_type pos) const
3463 {
3464         if (Inset const * inset = getInset(pos))
3465                 return inset->isSpace();
3466         char_type const c = d->text_[pos];
3467         return lyx::isSpace(c);
3468 }
3469
3470
3471 Language const *
3472 Paragraph::getParLanguage(BufferParams const & bparams) const
3473 {
3474         if (!empty())
3475                 return getFirstFontSettings(bparams).language();
3476         // FIXME: we should check the prev par as well (Lgb)
3477         return bparams.language;
3478 }
3479
3480
3481 bool Paragraph::isRTL(BufferParams const & bparams) const
3482 {
3483         return getParLanguage(bparams)->rightToLeft()
3484                 && !inInset().getLayout().forceLTR();
3485 }
3486
3487
3488 void Paragraph::changeLanguage(BufferParams const & bparams,
3489                                Language const * from, Language const * to)
3490 {
3491         // change language including dummy font change at the end
3492         for (pos_type i = 0; i <= size(); ++i) {
3493                 Font font = getFontSettings(bparams, i);
3494                 if (font.language() == from) {
3495                         font.setLanguage(to);
3496                         setFont(i, font);
3497                         d->requestSpellCheck(i);
3498                 }
3499         }
3500 }
3501
3502
3503 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
3504 {
3505         Language const * doc_language = bparams.language;
3506         for (auto const & f : d->fontlist_)
3507                 if (f.font().language() != ignore_language &&
3508                     f.font().language() != latex_language &&
3509                     f.font().language() != doc_language)
3510                         return true;
3511         return false;
3512 }
3513
3514
3515 void Paragraph::getLanguages(std::set<Language const *> & langs) const
3516 {
3517         for (auto const & f : d->fontlist_) {
3518                 Language const * lang = f.font().language();
3519                 if (lang != ignore_language &&
3520                     lang != latex_language)
3521                         langs.insert(lang);
3522         }
3523 }
3524
3525
3526 docstring Paragraph::asString(int options) const
3527 {
3528         return asString(0, size(), options);
3529 }
3530
3531
3532 docstring Paragraph::asString(pos_type beg, pos_type end, int options, const OutputParams *runparams) const
3533 {
3534         odocstringstream os;
3535
3536         if (beg == 0
3537             && options & AS_STR_LABEL
3538             && !d->params_.labelString().empty())
3539                 os << d->params_.labelString() << ' ';
3540
3541         for (pos_type i = beg; i < end; ++i) {
3542                 if ((options & AS_STR_SKIPDELETE) && isDeleted(i))
3543                         continue;
3544                 char_type const c = d->text_[i];
3545                 if (isPrintable(c) || c == '\t'
3546                     || (c == '\n' && (options & AS_STR_NEWLINES)))
3547                         os.put(c);
3548                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
3549                         if (c == META_INSET && (options & AS_STR_PLAINTEXT)) {
3550                                 LASSERT(runparams != 0, return docstring());
3551                                 getInset(i)->plaintext(os, *runparams);
3552                         } else {
3553                                 getInset(i)->toString(os);
3554                         }
3555                 }
3556         }
3557
3558         return os.str();
3559 }
3560
3561
3562 void Paragraph::forOutliner(docstring & os, size_t const maxlen,
3563                             bool const shorten, bool const label) const
3564 {
3565         size_t tmplen = shorten ? maxlen + 1 : maxlen;
3566         if (label && !labelString().empty())
3567                 os += labelString() + ' ';
3568         if (!layout().isTocCaption())
3569                 return;
3570         for (pos_type i = 0; i < size() && os.length() < tmplen; ++i) {
3571                 if (isDeleted(i))
3572                         continue;
3573                 char_type const c = d->text_[i];
3574                 if (isPrintable(c))
3575                         os += c;
3576                 else if (c == META_INSET)
3577                         getInset(i)->forOutliner(os, tmplen, false);
3578         }
3579         if (shorten)
3580                 Text::shortenForOutliner(os, maxlen);
3581 }
3582
3583
3584 void Paragraph::setInsetOwner(Inset const * inset)
3585 {
3586         d->inset_owner_ = inset;
3587 }
3588
3589
3590 int Paragraph::id() const
3591 {
3592         return d->id_;
3593 }
3594
3595
3596 void Paragraph::setId(int id)
3597 {
3598         d->id_ = id;
3599 }
3600
3601
3602 Layout const & Paragraph::layout() const
3603 {
3604         return *d->layout_;
3605 }
3606
3607
3608 void Paragraph::setLayout(Layout const & layout)
3609 {
3610         d->layout_ = &layout;
3611 }
3612
3613
3614 void Paragraph::setDefaultLayout(DocumentClass const & tc)
3615 {
3616         setLayout(tc.defaultLayout());
3617 }
3618
3619
3620 void Paragraph::setPlainLayout(DocumentClass const & tc)
3621 {
3622         setLayout(tc.plainLayout());
3623 }
3624
3625
3626 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
3627 {
3628         if (usePlainLayout())
3629                 setPlainLayout(tclass);
3630         else
3631                 setDefaultLayout(tclass);
3632 }
3633
3634
3635 Inset const & Paragraph::inInset() const
3636 {
3637         LBUFERR(d->inset_owner_);
3638         return *d->inset_owner_;
3639 }
3640
3641
3642 ParagraphParameters & Paragraph::params()
3643 {
3644         return d->params_;
3645 }
3646
3647
3648 ParagraphParameters const & Paragraph::params() const
3649 {
3650         return d->params_;
3651 }
3652
3653
3654 bool Paragraph::isFreeSpacing() const
3655 {
3656         if (d->layout_->free_spacing)
3657                 return true;
3658         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
3659 }
3660
3661
3662 bool Paragraph::allowEmpty() const
3663 {
3664         if (d->layout_->keepempty)
3665                 return true;
3666         return d->inset_owner_ && d->inset_owner_->allowEmpty();
3667 }
3668
3669
3670 bool Paragraph::brokenBiblio() const
3671 {
3672         // there is a problem if there is no bibitem at position 0 or
3673         // if there is another bibitem in the paragraph.
3674         return d->layout_->labeltype == LABEL_BIBLIO
3675                 && (d->insetlist_.find(BIBITEM_CODE) != 0
3676                     || d->insetlist_.find(BIBITEM_CODE, 1) > 0);
3677 }
3678
3679
3680 int Paragraph::fixBiblio(Buffer const & buffer)
3681 {
3682         // FIXME: What about the case where paragraph is not BIBLIO
3683         // but there is an InsetBibitem?
3684         // FIXME: when there was already an inset at 0, the return value is 1,
3685         // which does not tell whether another inset has been remove; the
3686         // cursor cannot be correctly updated.
3687
3688         if (d->layout_->labeltype != LABEL_BIBLIO)
3689                 return 0;
3690
3691         bool const track_changes = buffer.params().track_changes;
3692         int bibitem_pos = d->insetlist_.find(BIBITEM_CODE);
3693         bool const hasbibitem0 = bibitem_pos == 0;
3694
3695         if (hasbibitem0) {
3696                 bibitem_pos = d->insetlist_.find(BIBITEM_CODE, 1);
3697                 // There was an InsetBibitem at pos 0, and no other one => OK
3698                 if (bibitem_pos == -1)
3699                         return 0;
3700                 // there is a bibitem at the 0 position, but since
3701                 // there is a second one, we copy the second on the
3702                 // first. We're assuming there are at most two of
3703                 // these, which there should be.
3704                 // FIXME: why does it make sense to do that rather
3705                 // than keep the first? (JMarc)
3706                 Inset * inset = releaseInset(bibitem_pos);
3707                 d->insetlist_.begin()->inset = inset;
3708                 return -bibitem_pos;
3709         }
3710
3711         // We need to create an inset at the beginning
3712         Inset * inset = 0;
3713         if (bibitem_pos > 0) {
3714                 // there was one somewhere in the paragraph, let's move it
3715                 inset = d->insetlist_.release(bibitem_pos);
3716                 eraseChar(bibitem_pos, track_changes);
3717         } else
3718                 // make a fresh one
3719                 inset = new InsetBibitem(const_cast<Buffer *>(&buffer),
3720                                          InsetCommandParams(BIBITEM_CODE));
3721
3722         Font font(inherit_font, buffer.params().language);
3723         insertInset(0, inset, font, Change(track_changes ? Change::INSERTED
3724                                                    : Change::UNCHANGED));
3725
3726         return 1;
3727 }
3728
3729
3730 void Paragraph::checkAuthors(AuthorList const & authorList)
3731 {
3732         d->changes_.checkAuthors(authorList);
3733 }
3734
3735
3736 bool Paragraph::isChanged(pos_type pos) const
3737 {
3738         return lookupChange(pos).changed();
3739 }
3740
3741
3742 bool Paragraph::isInserted(pos_type pos) const
3743 {
3744         return lookupChange(pos).inserted();
3745 }
3746
3747
3748 bool Paragraph::isDeleted(pos_type pos) const
3749 {
3750         return lookupChange(pos).deleted();
3751 }
3752
3753
3754 InsetList const & Paragraph::insetList() const
3755 {
3756         return d->insetlist_;
3757 }
3758
3759
3760 void Paragraph::setInsetBuffers(Buffer & b)
3761 {
3762         d->insetlist_.setBuffer(b);
3763 }
3764
3765
3766 void Paragraph::resetBuffer()
3767 {
3768         d->insetlist_.resetBuffer();
3769 }
3770
3771
3772 Inset * Paragraph::releaseInset(pos_type pos)
3773 {
3774         Inset * inset = d->insetlist_.release(pos);
3775         /// does not honour change tracking!
3776         eraseChar(pos, false);
3777         return inset;
3778 }
3779
3780
3781 Inset * Paragraph::getInset(pos_type pos)
3782 {
3783         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3784                  ? d->insetlist_.get(pos) : 0;
3785 }
3786
3787
3788 Inset const * Paragraph::getInset(pos_type pos) const
3789 {
3790         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3791                  ? d->insetlist_.get(pos) : 0;
3792 }
3793
3794
3795 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
3796                 pos_type & right, TextCase action)
3797 {
3798         // process sequences of modified characters; in change
3799         // tracking mode, this approach results in much better
3800         // usability than changing case on a char-by-char basis
3801         // We also need to track the current font, since font
3802         // changes within sequences can occur.
3803         vector<pair<char_type, Font> > changes;
3804
3805         bool const trackChanges = bparams.track_changes;
3806
3807         bool capitalize = true;
3808
3809         for (; pos < right; ++pos) {
3810                 char_type oldChar = d->text_[pos];
3811                 char_type newChar = oldChar;
3812
3813                 // ignore insets and don't play with deleted text!
3814                 if (oldChar != META_INSET && !isDeleted(pos)) {
3815                         switch (action) {
3816                                 case text_lowercase:
3817                                         newChar = lowercase(oldChar);
3818                                         break;
3819                                 case text_capitalization:
3820                                         if (capitalize) {
3821                                                 newChar = uppercase(oldChar);
3822                                                 capitalize = false;
3823                                         }
3824                                         break;
3825                                 case text_uppercase:
3826                                         newChar = uppercase(oldChar);
3827                                         break;
3828                         }
3829                 }
3830
3831                 if (isWordSeparator(pos) || isDeleted(pos)) {
3832                         // permit capitalization again
3833                         capitalize = true;
3834                 }
3835
3836                 if (oldChar != newChar) {
3837                         changes.push_back(make_pair(newChar, getFontSettings(bparams, pos)));
3838                         if (pos != right - 1)
3839                                 continue;
3840                         // step behind the changing area
3841                         pos++;
3842                 }
3843
3844                 int erasePos = pos - changes.size();
3845                 for (size_t i = 0; i < changes.size(); i++) {
3846                         insertChar(pos, changes[i].first,
3847                                    changes[i].second,
3848                                    trackChanges);
3849                         if (!eraseChar(erasePos, trackChanges)) {
3850                                 ++erasePos;
3851                                 ++pos; // advance
3852                                 ++right; // expand selection
3853                         }
3854                 }
3855                 changes.clear();
3856         }
3857 }
3858
3859
3860 int Paragraph::find(docstring const & str, bool cs, bool mw,
3861                 pos_type start_pos, bool del) const
3862 {
3863         pos_type pos = start_pos;
3864         int const strsize = str.length();
3865         int i = 0;
3866         pos_type const parsize = d->text_.size();
3867         for (i = 0; i < strsize && pos < parsize; ++i, ++pos) {
3868                 // Ignore "invisible" letters such as ligature breaks
3869                 // and hyphenation chars while searching
3870                 while (pos < parsize - 1 && isInset(pos)) {
3871                         odocstringstream os;
3872                         getInset(pos)->toString(os);
3873                         if (!getInset(pos)->isLetter() || !os.str().empty())
3874                                 break;
3875                         pos++;
3876                 }
3877                 if (cs && str[i] != d->text_[pos])
3878                         break;
3879                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos]))
3880                         break;
3881                 if (!del && isDeleted(pos))
3882                         break;
3883         }
3884
3885         if (i != strsize)
3886                 return 0;
3887
3888         // if necessary, check whether string matches word
3889         if (mw) {
3890                 if (start_pos > 0 && !isWordSeparator(start_pos - 1))
3891                         return 0;
3892                 if (pos < parsize
3893                         && !isWordSeparator(pos))
3894                         return 0;
3895         }
3896
3897         return pos - start_pos;
3898 }
3899
3900
3901 char_type Paragraph::getChar(pos_type pos) const
3902 {
3903         return d->text_[pos];
3904 }
3905
3906
3907 pos_type Paragraph::size() const
3908 {
3909         return d->text_.size();
3910 }
3911
3912
3913 bool Paragraph::empty() const
3914 {
3915         return d->text_.empty();
3916 }
3917
3918
3919 bool Paragraph::isInset(pos_type pos) const
3920 {
3921         return d->text_[pos] == META_INSET;
3922 }
3923
3924
3925 bool Paragraph::isSeparator(pos_type pos) const
3926 {
3927         //FIXME: Are we sure this can be the only separator?
3928         return d->text_[pos] == ' ';
3929 }
3930
3931
3932 void Paragraph::deregisterWords()
3933 {
3934         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3935         Private::LangWordsMap::const_iterator ite = d->words_.end();
3936         for (; itl != ite; ++itl) {
3937                 WordList & wl = theWordList(itl->first);
3938                 Private::Words::const_iterator it = (itl->second).begin();
3939                 Private::Words::const_iterator et = (itl->second).end();
3940                 for (; it != et; ++it)
3941                         wl.remove(*it);
3942         }
3943         d->words_.clear();
3944 }
3945
3946
3947 void Paragraph::locateWord(pos_type & from, pos_type & to,
3948         word_location const loc) const
3949 {
3950         switch (loc) {
3951         case WHOLE_WORD_STRICT:
3952                 if (from == 0 || from == size()
3953                     || isWordSeparator(from)
3954                     || isWordSeparator(from - 1)) {
3955                         to = from;
3956                         return;
3957                 }
3958                 // fall through
3959
3960         case WHOLE_WORD:
3961                 // If we are already at the beginning of a word, do nothing
3962                 if (!from || isWordSeparator(from - 1))
3963                         break;
3964                 // fall through
3965
3966         case PREVIOUS_WORD:
3967                 // always move the cursor to the beginning of previous word
3968                 while (from && !isWordSeparator(from - 1))
3969                         --from;
3970                 break;
3971         case NEXT_WORD:
3972                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
3973                 break;
3974         case PARTIAL_WORD:
3975                 // no need to move the 'from' cursor
3976                 break;
3977         }
3978         to = from;
3979         while (to < size() && !isWordSeparator(to))
3980                 ++to;
3981 }
3982
3983
3984 void Paragraph::collectWords()
3985 {
3986         for (pos_type pos = 0; pos < size(); ++pos) {
3987                 if (isWordSeparator(pos))
3988                         continue;
3989                 pos_type from = pos;
3990                 locateWord(from, pos, WHOLE_WORD);
3991                 // Work around MSVC warning: The statement
3992                 // if (pos < from + lyxrc.completion_minlength)
3993                 // triggers a signed vs. unsigned warning.
3994                 // I don't know why this happens, it could be a MSVC bug, or
3995                 // related to LLP64 (windows) vs. LP64 (unix) programming
3996                 // model, or the C++ standard might be ambigous in the section
3997                 // defining the "usual arithmetic conversions". However, using
3998                 // a temporary variable is safe and works on all compilers.
3999                 pos_type const endpos = from + lyxrc.completion_minlength;
4000                 if (pos < endpos)
4001                         continue;
4002                 FontList::const_iterator cit = d->fontlist_.fontIterator(from);
4003                 if (cit == d->fontlist_.end())
4004                         return;
4005                 Language const * lang = cit->font().language();
4006                 docstring const word = asString(from, pos, AS_STR_NONE);
4007                 d->words_[lang->lang()].insert(word);
4008         }
4009 }
4010
4011
4012 void Paragraph::registerWords()
4013 {
4014         Private::LangWordsMap::const_iterator itl = d->words_.begin();
4015         Private::LangWordsMap::const_iterator ite = d->words_.end();
4016         for (; itl != ite; ++itl) {
4017                 WordList & wl = theWordList(itl->first);
4018                 Private::Words::const_iterator it = (itl->second).begin();
4019                 Private::Words::const_iterator et = (itl->second).end();
4020                 for (; it != et; ++it)
4021                         wl.insert(*it);
4022         }
4023 }
4024
4025
4026 void Paragraph::updateWords()
4027 {
4028         deregisterWords();
4029         collectWords();
4030         registerWords();
4031 }
4032
4033
4034 void Paragraph::Private::appendSkipPosition(SkipPositions & skips, pos_type const pos) const
4035 {
4036         SkipPositionsIterator begin = skips.begin();
4037         SkipPositions::iterator end = skips.end();
4038         if (pos > 0 && begin < end) {
4039                 --end;
4040                 if (end->last == pos - 1) {
4041                         end->last = pos;
4042                         return;
4043                 }
4044         }
4045         skips.insert(end, FontSpan(pos, pos));
4046 }
4047
4048
4049 Language * Paragraph::Private::locateSpellRange(
4050         pos_type & from, pos_type & to,
4051         SkipPositions & skips) const
4052 {
4053         // skip leading white space
4054         while (from < to && owner_->isWordSeparator(from))
4055                 ++from;
4056         // don't check empty range
4057         if (from >= to)
4058                 return 0;
4059         // get current language
4060         Language * lang = getSpellLanguage(from);
4061         pos_type last = from;
4062         bool samelang = true;
4063         bool sameinset = true;
4064         while (last < to && samelang && sameinset) {
4065                 // hop to end of word
4066                 while (last < to && !owner_->isWordSeparator(last)) {
4067                         if (owner_->getInset(last)) {
4068                                 appendSkipPosition(skips, last);
4069                         } else if (owner_->isDeleted(last)) {
4070                                 appendSkipPosition(skips, last);
4071                         }
4072                         ++last;
4073                 }
4074                 // hop to next word while checking for insets
4075                 while (sameinset && last < to && owner_->isWordSeparator(last)) {
4076                         if (Inset const * inset = owner_->getInset(last))
4077                                 sameinset = inset->isChar() && inset->isLetter();
4078                         if (sameinset && owner_->isDeleted(last)) {
4079                                 appendSkipPosition(skips, last);
4080                         }
4081                         if (sameinset)
4082                                 last++;
4083                 }
4084                 if (sameinset && last < to) {
4085                         // now check for language change
4086                         samelang = lang == getSpellLanguage(last);
4087                 }
4088         }
4089         // if language change detected backstep is needed
4090         if (!samelang)
4091                 --last;
4092         to = last;
4093         return lang;
4094 }
4095
4096
4097 Language * Paragraph::Private::getSpellLanguage(pos_type const from) const
4098 {
4099         Language * lang =
4100                 const_cast<Language *>(owner_->getFontSettings(
4101                         inset_owner_->buffer().params(), from).language());
4102         if (lang == inset_owner_->buffer().params().language
4103                 && !lyxrc.spellchecker_alt_lang.empty()) {
4104                 string lang_code;
4105                 string const lang_variety =
4106                         split(lyxrc.spellchecker_alt_lang, lang_code, '-');
4107                 lang->setCode(lang_code);
4108                 lang->setVariety(lang_variety);
4109         }
4110         return lang;
4111 }
4112
4113
4114 void Paragraph::requestSpellCheck(pos_type pos)
4115 {
4116         d->requestSpellCheck(pos);
4117 }
4118
4119
4120 bool Paragraph::needsSpellCheck() const
4121 {
4122         SpellChecker::ChangeNumber speller_change_number = 0;
4123         if (theSpellChecker())
4124                 speller_change_number = theSpellChecker()->changeNumber();
4125         if (speller_change_number > d->speller_state_.currentChangeNumber()) {
4126                 d->speller_state_.needsCompleteRefresh(speller_change_number);
4127         }
4128         return d->needsSpellCheck();
4129 }
4130
4131
4132 bool Paragraph::Private::ignoreWord(docstring const & word) const
4133 {
4134         // Ignore words with digits
4135         // FIXME: make this customizable
4136         // (note that some checkers ignore words with digits by default)
4137         docstring::const_iterator cit = word.begin();
4138         docstring::const_iterator const end = word.end();
4139         for (; cit != end; ++cit) {
4140                 if (isNumber((*cit)))
4141                         return true;
4142         }
4143         return false;
4144 }
4145
4146
4147 SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
4148         WordLangTuple & wl, docstring_list & suggestions,
4149         bool do_suggestion, bool check_learned) const
4150 {
4151         SpellChecker::Result result = SpellChecker::WORD_OK;
4152         SpellChecker * speller = theSpellChecker();
4153         if (!speller)
4154                 return result;
4155
4156         if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
4157                 return result;
4158
4159         locateWord(from, to, WHOLE_WORD);
4160         if (from == to || from >= size())
4161                 return result;
4162
4163         docstring word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
4164         Language * lang = d->getSpellLanguage(from);
4165
4166         wl = WordLangTuple(word, lang);
4167
4168         if (word.empty())
4169                 return result;
4170
4171         if (needsSpellCheck() || check_learned) {
4172                 pos_type end = to;
4173                 if (!d->ignoreWord(word)) {
4174                         bool const trailing_dot = to < size() && d->text_[to] == '.';
4175                         result = speller->check(wl);
4176                         if (SpellChecker::misspelled(result) && trailing_dot) {
4177                                 wl = WordLangTuple(word.append(from_ascii(".")), lang);
4178                                 result = speller->check(wl);
4179                                 if (!SpellChecker::misspelled(result)) {
4180                                         LYXERR(Debug::GUI, "misspelled word is correct with dot: \"" <<
4181                                            word << "\" [" <<
4182                                            from << ".." << to << "]");
4183                                 } else {
4184                                         // spell check with dot appended failed too
4185                                         // restore original word/lang value
4186                                         word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
4187                                         wl = WordLangTuple(word, lang);
4188                                 }
4189                         }
4190                 }
4191                 if (!SpellChecker::misspelled(result)) {
4192                         // area up to the begin of the next word is not misspelled
4193                         while (end < size() && isWordSeparator(end))
4194                                 ++end;
4195                 }
4196                 d->setMisspelled(from, end, result);
4197         } else {
4198                 result = d->speller_state_.getState(from);
4199         }
4200
4201         if (do_suggestion)
4202                 suggestions.clear();
4203
4204         if (SpellChecker::misspelled(result)) {
4205                 LYXERR(Debug::GUI, "misspelled word: \"" <<
4206                            word << "\" [" <<
4207                            from << ".." << to << "]");
4208                 if (do_suggestion)
4209                         speller->suggest(wl, suggestions);
4210         }
4211         return result;
4212 }
4213
4214
4215 void Paragraph::anonymize()
4216 {
4217         // This is a very crude anonymization for now
4218         for (char_type & c : d->text_)
4219                 if (isLetterChar(c) || isNumber(c))
4220                         c = 'a';
4221 }
4222
4223
4224 void Paragraph::Private::markMisspelledWords(
4225         pos_type const & first, pos_type const & last,
4226         SpellChecker::Result result,
4227         docstring const & word,
4228         SkipPositions const & skips)
4229 {
4230         if (!SpellChecker::misspelled(result)) {
4231                 setMisspelled(first, last, SpellChecker::WORD_OK);
4232                 return;
4233         }
4234         int snext = first;
4235         SpellChecker * speller = theSpellChecker();
4236         // locate and enumerate the error positions
4237         int nerrors = speller->numMisspelledWords();
4238         int numskipped = 0;
4239         SkipPositionsIterator it = skips.begin();
4240         SkipPositionsIterator et = skips.end();
4241         for (int index = 0; index < nerrors; ++index) {
4242                 int wstart;
4243                 int wlen = 0;
4244                 speller->misspelledWord(index, wstart, wlen);
4245                 /// should not happen if speller supports range checks
4246                 if (!wlen) continue;
4247                 docstring const misspelled = word.substr(wstart, wlen);
4248                 wstart += first + numskipped;
4249                 if (snext < wstart) {
4250                         /// mark the range of correct spelling
4251                         numskipped += countSkips(it, et, wstart);
4252                         setMisspelled(snext,
4253                                 wstart - 1, SpellChecker::WORD_OK);
4254                 }
4255                 snext = wstart + wlen;
4256                 numskipped += countSkips(it, et, snext);
4257                 /// mark the range of misspelling
4258                 setMisspelled(wstart, snext, result);
4259                 LYXERR(Debug::GUI, "misspelled word: \"" <<
4260                            misspelled << "\" [" <<
4261                            wstart << ".." << (snext-1) << "]");
4262                 ++snext;
4263         }
4264         if (snext <= last) {
4265                 /// mark the range of correct spelling at end
4266                 setMisspelled(snext, last, SpellChecker::WORD_OK);
4267         }
4268 }
4269
4270
4271 void Paragraph::spellCheck() const
4272 {
4273         SpellChecker * speller = theSpellChecker();
4274         if (!speller || empty() ||!needsSpellCheck())
4275                 return;
4276         pos_type start;
4277         pos_type endpos;
4278         d->rangeOfSpellCheck(start, endpos);
4279         if (speller->canCheckParagraph()) {
4280                 // loop until we leave the range
4281                 for (pos_type first = start; first < endpos; ) {
4282                         pos_type last = endpos;
4283                         Private::SkipPositions skips;
4284                         Language * lang = d->locateSpellRange(first, last, skips);
4285                         if (first >= endpos)
4286                                 break;
4287                         // start the spell checker on the unit of meaning
4288                         docstring word = asString(first, last, AS_STR_INSETS + AS_STR_SKIPDELETE);
4289                         WordLangTuple wl = WordLangTuple(word, lang);
4290                         SpellChecker::Result result = word.size() ?
4291                                 speller->check(wl) : SpellChecker::WORD_OK;
4292                         d->markMisspelledWords(first, last, result, word, skips);
4293                         first = ++last;
4294                 }
4295         } else {
4296                 static docstring_list suggestions;
4297                 pos_type to = endpos;
4298                 while (start < endpos) {
4299                         WordLangTuple wl;
4300                         spellCheck(start, to, wl, suggestions, false);
4301                         start = to + 1;
4302                 }
4303         }
4304         d->readySpellCheck();
4305 }
4306
4307
4308 bool Paragraph::isMisspelled(pos_type pos, bool check_boundary) const
4309 {
4310         bool result = SpellChecker::misspelled(d->speller_state_.getState(pos));
4311         if (result || pos <= 0 || pos > size())
4312                 return result;
4313         if (check_boundary && (pos == size() || isWordSeparator(pos)))
4314                 result = SpellChecker::misspelled(d->speller_state_.getState(pos - 1));
4315         return result;
4316 }
4317
4318
4319 string Paragraph::magicLabel() const
4320 {
4321         stringstream ss;
4322         ss << "magicparlabel-" << id();
4323         return ss.str();
4324 }
4325
4326
4327 } // namespace lyx