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(BufferParams const &, otexstream & os,
 318                                char_type c, char_type next,
 319                                OutputParams const &);
 320
 321         /// Output a space in appropriate formatting (or a surrogate pair
 322         /// if the next character is a combining character).
 323         /// \return whether a surrogate pair was output.
 324         bool simpleTeXBlanks(BufferParams const &,
 325                              OutputParams const &,
 326                              otexstream &,
 327                              pos_type i,
 328                              unsigned int & column,
 329                              Font const & font,
 330                              Layout const & style);
 331
 332         /// Output consecutive unicode chars, belonging to the same script as
 333         /// specified by the latex macro \p ltx, to \p os starting from \p i.
 334         /// \return the number of characters written.
 335         int writeScriptChars(BufferParams const &, OutputParams const &,
 336                              otexstream & os,
 337                              docstring const & ltx,
 338                              Change const &, Encoding const &,
 339                              std::string const, pos_type & i);
 340
 341         /// This could go to ParagraphParameters if we want to.
 342         int startTeXParParams(BufferParams const &, otexstream &,
 343                               OutputParams const &) const;
 344
 345         /// This could go to ParagraphParameters if we want to.
 346         bool endTeXParParams(BufferParams const &, otexstream &,
 347                              OutputParams const &) const;
 348
 349         ///
 350         void latexInset(BufferParams const &,
 351                                    otexstream &,
 352                                    OutputParams &,
 353                                    Font & running_font,
 354                                    Font & basefont,
 355                                    Font const & outerfont,
 356                                    bool & open_font,
 357                                    Change & running_change,
 358                                    Layout const & style,
 359                                    pos_type & i,
 360                                    unsigned int & column);
 361
 362         ///
 363         void latexSpecialChar(
 364                                    otexstream & os,
 365                                    BufferParams const & bparams,
 366                                    OutputParams const & runparams,
 367                                    Font const & running_font,
 368                                    Change const & running_change,
 369                                    Layout const & style,
 370                                    pos_type & i,
 371                                    pos_type end_pos,
 372                                    unsigned int & column);
 373
 374         ///
 375         bool latexSpecialT1(
 376                 char_type const c,
 377                 otexstream & os,
 378                 pos_type i,
 379                 unsigned int & column);
 380         ///
 381         bool latexSpecialTU(
 382                 char_type const c,
 383                 otexstream & os,
 384                 pos_type i,
 385                 unsigned int & column);
 386         ///
 387         bool latexSpecialT3(
 388                 char_type const c,
 389                 otexstream & os,
 390                 pos_type i,
 391                 unsigned int & column);
 392
 393         ///
 394         void validate(LaTeXFeatures & features) const;
 395
 396         /// Checks if the paragraph contains only text and no inset or font change.
 397         bool onlyText(Buffer const & buf, Font const & outerfont,
 398                       pos_type initial) const;
 399
 400         /// a vector of speller skip positions
 401         typedef vector<FontSpan> SkipPositions;
 402         typedef SkipPositions::const_iterator SkipPositionsIterator;
 403
 404         void appendSkipPosition(SkipPositions & skips, pos_type const pos) const;
 405
 406         Language * getSpellLanguage(pos_type const from) const;
 407
 408         Language * locateSpellRange(pos_type & from, pos_type & to,
 409                                     SkipPositions & skips) const;
 410
 411         bool hasSpellerChange() const
 412         {
 413                 SpellChecker::ChangeNumber speller_change_number = 0;
 414                 if (theSpellChecker())
 415                         speller_change_number = theSpellChecker()->changeNumber();
 416                 return speller_change_number > speller_state_.currentChangeNumber();
 417         }
 418
 419         bool ignoreWord(docstring const & word) const ;
 420
 421         void setMisspelled(pos_type from, pos_type to, SpellChecker::Result state)
 422         {
 423                 pos_type textsize = owner_->size();
 424                 // check for sane arguments
 425                 if (to <= from || from >= textsize)
 426                         return;
 427                 FontSpan fp = FontSpan(from, to - 1);
 428                 speller_state_.setRange(fp, state);
 429         }
 430
 431         void requestSpellCheck(pos_type pos)
 432         {
 433                 if (pos == -1)
 434                         speller_state_.needsCompleteRefresh(speller_state_.currentChangeNumber());
 435                 else
 436                         speller_state_.needsRefresh(pos);
 437         }
 438
 439         void readySpellCheck()
 440         {
 441                 speller_state_.needsRefresh(-1);
 442         }
 443
 444         bool needsSpellCheck() const
 445         {
 446                 return speller_state_.needsRefresh();
 447         }
 448
 449         void rangeOfSpellCheck(pos_type & first, pos_type & last) const
 450         {
 451                 speller_state_.refreshRange(first, last);
 452                 if (last == -1) {
 453                         last = owner_->size();
 454                         return;
 455                 }
 456                 pos_type endpos = last;
 457                 owner_->locateWord(first, endpos, WHOLE_WORD);
 458                 if (endpos < last) {
 459                         endpos = last;
 460                         owner_->locateWord(last, endpos, WHOLE_WORD);
 461                 }
 462                 last = endpos;
 463         }
 464
 465         int countSkips(SkipPositionsIterator & it, SkipPositionsIterator const et,
 466                             int & start) const
 467         {
 468                 int numskips = 0;
 469                 while (it != et && it->first < start) {
 470                         int skip = it->last - it->first + 1;
 471                         start += skip;
 472                         numskips += skip;
 473                         ++it;
 474                 }
 475                 return numskips;
 476         }
 477
 478         void markMisspelledWords(pos_type const & first, pos_type const & last,
 479                                                          SpellChecker::Result result,
 480                                                          docstring const & word,
 481                                                          SkipPositions const & skips);
 482
 483         InsetCode ownerCode() const
 484         {
 485                 return inset_owner_ ? inset_owner_->lyxCode() : NO_CODE;
 486         }
 487
 488         /// Which Paragraph owns us?
 489         Paragraph * owner_;
 490
 491         /// In which Inset?
 492         Inset const * inset_owner_;
 493
 494         ///
 495         FontList fontlist_;
 496
 497         ///
 498         int id_;
 499
 500         ///
 501         ParagraphParameters params_;
 502
 503         /// for recording and looking up changes
 504         Changes changes_;
 505
 506         ///
 507         InsetList insetlist_;
 508
 509         /// end of label
 510         pos_type begin_of_body_;
 511
 512         typedef docstring TextContainer;
 513         ///
 514         TextContainer text_;
 515
 516         typedef set<docstring> Words;
 517         typedef map<string, Words> LangWordsMap;
 518         ///
 519         LangWordsMap words_;
 520         ///
 521         Layout const * layout_;
 522         ///
 523         SpellCheckerState speller_state_;
 524 };
 525
 526
 527 Paragraph::Private::Private(Paragraph * owner, Layout const & layout)
 528         : owner_(owner), inset_owner_(0), id_(-1), begin_of_body_(0), layout_(&layout)
 529 {
 530         text_.reserve(100);
 531 }
 532
 533
 534 //static
 535 int Paragraph::Private::make_id()
 536 {
 537         // The id is unique per session across buffers because it is used in
 538         // LFUN_PARAGRAPH_GOTO to switch to a different buffer, for instance in the
 539         // outliner.
 540         // (thread-safe)
 541         static atomic_uint next_id(0);
 542         return next_id++;
 543 }
 544
 545
 546 Paragraph::Private::Private(Private const & p, Paragraph * owner)
 547         : owner_(owner), inset_owner_(p.inset_owner_), fontlist_(p.fontlist_),
 548           id_(make_id()),
 549           params_(p.params_), changes_(p.changes_), insetlist_(p.insetlist_),
 550           begin_of_body_(p.begin_of_body_), text_(p.text_), words_(p.words_),
 551           layout_(p.layout_)
 552 {
 553         requestSpellCheck(p.text_.size());
 554 }
 555
 556
 557 Paragraph::Private::Private(Private const & p, Paragraph * owner,
 558         pos_type beg, pos_type end)
 559         : owner_(owner), inset_owner_(p.inset_owner_), id_(make_id()),
 560           params_(p.params_), changes_(p.changes_),
 561           insetlist_(p.insetlist_, beg, end),
 562           begin_of_body_(p.begin_of_body_), words_(p.words_),
 563           layout_(p.layout_)
 564 {
 565         if (beg >= pos_type(p.text_.size()))
 566                 return;
 567         text_ = p.text_.substr(beg, end - beg);
 568
 569         FontList::const_iterator fcit = fontlist_.begin();
 570         FontList::const_iterator fend = fontlist_.end();
 571         for (; fcit != fend; ++fcit) {
 572                 if (fcit->pos() < beg)
 573                         continue;
 574                 if (fcit->pos() >= end) {
 575                         // Add last entry in the fontlist_.
 576                         fontlist_.set(text_.size() - 1, fcit->font());
 577                         break;
 578                 }
 579                 // Add a new entry in the fontlist_.
 580                 fontlist_.set(fcit->pos() - beg, fcit->font());
 581         }
 582         requestSpellCheck(p.text_.size());
 583 }
 584
 585
 586 void Paragraph::addChangesToToc(DocIterator const & cdit, Buffer const & buf,
 587                                 bool output_active, TocBackend & backend) const
 588 {
 589         d->changes_.addToToc(cdit, buf, output_active, backend);
 590 }
 591
 592
 593 void Paragraph::addChangesToBuffer(Buffer const & buf) const
 594 {
 595         d->changes_.updateBuffer(buf);
 596 }
 597
 598
 599 bool Paragraph::isChangeUpdateRequired() const
 600 {
 601         return d->changes_.isUpdateRequired();
 602 }
 603
 604
 605 bool Paragraph::isDeleted(pos_type start, pos_type end) const
 606 {
 607         LASSERT(start >= 0 && start <= size(), return false);
 608         LASSERT(end > start && end <= size() + 1, return false);
 609
 610         return d->changes_.isDeleted(start, end);
 611 }
 612
 613
 614 bool Paragraph::isChanged(pos_type start, pos_type end) const
 615 {
 616         LASSERT(start >= 0 && start <= size(), return false);
 617         LASSERT(end > start && end <= size() + 1, return false);
 618
 619         return d->changes_.isChanged(start, end);
 620 }
 621
 622
 623 bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const
 624 {
 625         // keep the logic here in sync with the logic of eraseChars()
 626         if (!trackChanges)
 627                 return true;
 628
 629         Change const & change = d->changes_.lookup(size());
 630         return change.inserted() && change.currentAuthor();
 631 }
 632
 633
 634 void Paragraph::setChange(Change const & change)
 635 {
 636         // beware of the imaginary end-of-par character!
 637         d->changes_.set(change, 0, size() + 1);
 638
 639         /*
 640          * Propagate the change recursively - but not in case of DELETED!
 641          *
 642          * Imagine that your co-author makes changes in an existing inset. He
 643          * sends your document to you and you come to the conclusion that the
 644          * inset should go completely. If you erase it, LyX must not delete all
 645          * text within the inset. Otherwise, the change tracked insertions of
 646          * your co-author get lost and there is no way to restore them later.
 647          *
 648          * Conclusion: An inset's content should remain untouched if you delete it
 649          */
 650
 651         if (!change.deleted()) {
 652                 for (pos_type pos = 0; pos < size(); ++pos) {
 653                         if (Inset * inset = getInset(pos))
 654                                 inset->setChange(change);
 655                 }
 656         }
 657 }
 658
 659
 660 void Paragraph::setChange(pos_type pos, Change const & change)
 661 {
 662         LASSERT(pos >= 0 && pos <= size(), return);
 663         d->changes_.set(change, pos);
 664
 665         // see comment in setChange(Change const &) above
 666         if (!change.deleted() && pos < size())
 667                         if (Inset * inset = getInset(pos))
 668                                 inset->setChange(change);
 669 }
 670
 671
 672 Change const & Paragraph::lookupChange(pos_type pos) const
 673 {
 674         LBUFERR(pos >= 0 && pos <= size());
 675         return d->changes_.lookup(pos);
 676 }
 677
 678
 679 void Paragraph::acceptChanges(pos_type start, pos_type end)
 680 {
 681         LASSERT(start >= 0 && start <= size(), return);
 682         LASSERT(end > start && end <= size() + 1, return);
 683
 684         for (pos_type pos = start; pos < end; ++pos) {
 685                 switch (lookupChange(pos).type) {
 686                         case Change::UNCHANGED:
 687                                 // accept changes in nested inset
 688                                 if (Inset * inset = getInset(pos))
 689                                         inset->acceptChanges();
 690                                 break;
 691
 692                         case Change::INSERTED:
 693                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 694                                 // also accept changes in nested inset
 695                                 if (Inset * inset = getInset(pos))
 696                                         inset->acceptChanges();
 697                                 break;
 698
 699                         case Change::DELETED:
 700                                 // Suppress access to non-existent
 701                                 // "end-of-paragraph char"
 702                                 if (pos < size()) {
 703                                         eraseChar(pos, false);
 704                                         --end;
 705                                         --pos;
 706                                 }
 707                                 break;
 708                 }
 709
 710         }
 711 }
 712
 713
 714 void Paragraph::rejectChanges(pos_type start, pos_type end)
 715 {
 716         LASSERT(start >= 0 && start <= size(), return);
 717         LASSERT(end > start && end <= size() + 1, return);
 718
 719         for (pos_type pos = start; pos < end; ++pos) {
 720                 switch (lookupChange(pos).type) {
 721                         case Change::UNCHANGED:
 722                                 // reject changes in nested inset
 723                                 if (Inset * inset = getInset(pos))
 724                                                 inset->rejectChanges();
 725                                 break;
 726
 727                         case Change::INSERTED:
 728                                 // Suppress access to non-existent
 729                                 // "end-of-paragraph char"
 730                                 if (pos < size()) {
 731                                         eraseChar(pos, false);
 732                                         --end;
 733                                         --pos;
 734                                 }
 735                                 break;
 736
 737                         case Change::DELETED:
 738                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 739
 740                                 // Do NOT reject changes within a deleted inset!
 741                                 // There may be insertions of a co-author inside of it!
 742
 743                                 break;
 744                 }
 745         }
 746 }
 747
 748
 749 void Paragraph::Private::insertChar(pos_type pos, char_type c,
 750                 Change const & change)
 751 {
 752         LASSERT(pos >= 0 && pos <= int(text_.size()), return);
 753
 754         // track change
 755         changes_.insert(change, pos);
 756
 757         // This is actually very common when parsing buffers (and
 758         // maybe inserting ascii text)
 759         if (pos == pos_type(text_.size())) {
 760                 // when appending characters, no need to update tables
 761                 text_.push_back(c);
 762                 // but we want spell checking
 763                 requestSpellCheck(pos);
 764                 return;
 765         }
 766
 767         text_.insert(text_.begin() + pos, c);
 768
 769         // Update the font table.
 770         fontlist_.increasePosAfterPos(pos);
 771
 772         // Update the insets
 773         insetlist_.increasePosAfterPos(pos);
 774
 775         // Update list of misspelled positions
 776         speller_state_.increasePosAfterPos(pos);
 777 }
 778
 779
 780 bool Paragraph::insertInset(pos_type pos, Inset * inset,
 781                                    Font const & font, Change const & change)
 782 {
 783         LASSERT(inset, return false);
 784         LASSERT(pos >= 0 && pos <= size(), return false);
 785
 786         // Paragraph::insertInset() can be used in cut/copy/paste operation where
 787         // d->inset_owner_ is not set yet.
 788         if (d->inset_owner_ && !d->inset_owner_->insetAllowed(inset->lyxCode()))
 789                 return false;
 790
 791         d->insertChar(pos, META_INSET, change);
 792         LASSERT(d->text_[pos] == META_INSET, return false);
 793
 794         // Add a new entry in the insetlist_.
 795         d->insetlist_.insert(inset, pos);
 796
 797         // Some insets require run of spell checker
 798         requestSpellCheck(pos);
 799         setFont(pos, font);
 800         return true;
 801 }
 802
 803
 804 bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
 805 {
 806         LASSERT(pos >= 0 && pos <= size(), return false);
 807
 808         // keep the logic here in sync with the logic of isMergedOnEndOfParDeletion()
 809
 810         if (trackChanges) {
 811                 Change change = d->changes_.lookup(pos);
 812
 813                 // set the character to DELETED if
 814                 //  a) it was previously unchanged or
 815                 //  b) it was inserted by a co-author
 816
 817                 if (!change.changed() ||
 818                       (change.inserted() && !change.currentAuthor())) {
 819                         setChange(pos, Change(Change::DELETED));
 820                         // request run of spell checker
 821                         requestSpellCheck(pos);
 822                         return false;
 823                 }
 824
 825                 if (change.deleted())
 826                         return false;
 827         }
 828
 829         // Don't physically access the imaginary end-of-paragraph character.
 830         // eraseChar() can only mark it as DELETED. A physical deletion of
 831         // end-of-par must be handled externally.
 832         if (pos == size()) {
 833                 return false;
 834         }
 835
 836         // track change
 837         d->changes_.erase(pos);
 838
 839         // if it is an inset, delete the inset entry
 840         if (d->text_[pos] == META_INSET)
 841                 d->insetlist_.erase(pos);
 842
 843         d->text_.erase(d->text_.begin() + pos);
 844
 845         // Update the fontlist_
 846         d->fontlist_.erase(pos);
 847
 848         // Update the insetlist_
 849         d->insetlist_.decreasePosAfterPos(pos);
 850
 851         // Update list of misspelled positions
 852         d->speller_state_.decreasePosAfterPos(pos);
 853         d->speller_state_.refreshLast(size());
 854
 855         return true;
 856 }
 857
 858
 859 int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
 860 {
 861         LASSERT(start >= 0 && start <= size(), return 0);
 862         LASSERT(end >= start && end <= size() + 1, return 0);
 863
 864         pos_type i = start;
 865         for (pos_type count = end - start; count; --count) {
 866                 if (!eraseChar(i, trackChanges))
 867                         ++i;
 868         }
 869         return end - i;
 870 }
 871
 872
 873 int Paragraph::Private::latexSurrogatePair(BufferParams const & bparams,
 874                 otexstream & os, char_type c, char_type next,
 875                 OutputParams const & runparams)
 876 {
 877         // Writing next here may circumvent a possible font change between
 878         // c and next. Since next is only output if it forms a surrogate pair
 879         // with c we can ignore this:
 880         // A font change inside a surrogate pair does not make sense and is
 881         // hopefully impossible to input.
 882         // FIXME: change tracking
 883         // Is this correct WRT change tracking?
 884         Encoding const & encoding = *(runparams.encoding);
 885         docstring latex1 = encoding.latexChar(next).first;
 886         if (runparams.inIPA) {
 887                 string const tipashortcut = Encodings::TIPAShortcut(next);
 888                 if (!tipashortcut.empty()) {
 889                         latex1 = from_ascii(tipashortcut);
 890                 }
 891         }
 892         docstring latex2 = encoding.latexChar(c).first;
 893
 894         if (docstring(1, next) == latex1) {
 895                 // The encoding supports the combination:
 896                 // output as is (combining char after base char).
 897                 os << latex2 << latex1;
 898                 return latex1.length() + latex2.length();
 899         }
 900
 901         // Handle combining characters in "script" context (i.e., \textgreek and \textcyrillic)
 902         docstring::size_type const brace1 = latex2.find_first_of(from_ascii("{"));
 903         docstring::size_type const brace2 = latex2.find_last_of(from_ascii("}"));
 904         string script = to_ascii(latex2.substr(1, brace1 - 1));
 905         // "Script chars" need to embraced in \textcyrillic and \textgreek notwithstanding
 906         // whether they are encodable or not (it only depends on the font encoding)
 907         if (!runparams.isFullUnicode())
 908                 // This will get us a script value to deal with below
 909                 Encodings::isKnownScriptChar(c, script);
 910         int pos = 0;
 911         int length = brace2;
 912         string fontenc;
 913         if (runparams.local_font)
 914                 fontenc = runparams.local_font->language()->fontenc(bparams);
 915         else
 916                 fontenc = runparams.main_fontenc;
 917         docstring scriptmacro;
 918         docstring cb;
 919         if (script == "textgreek" || script == "textcyrillic") {
 920                 // We separate the script macro (\text[greek|cyr]) from the rest,
 921                 // since we need to include the combining char in it (#6463).
 922                 // This is "the rest":
 923                 pos = brace1 + 1;
 924                 length -= pos;
 925                 latex2 = latex2.substr(pos, length);
 926                 // We only need the script macro with non-native font encodings
 927                 if (Encodings::needsScriptWrapper(script, fontenc)) {
 928                         scriptmacro = from_ascii("\\" + script + "{");
 929                         cb = from_ascii("}");
 930                 }
 931         }
 932
 933         docstring lb;
 934         docstring rb;
 935         // polutonikogreek does not play nice with brackets
 936         if (!runparams.local_font
 937             || runparams.local_font->language()->lang() != "polutonikogreek") {
 938                 lb = from_ascii("{");
 939                 rb = from_ascii("}");
 940         }
 941
 942         os << scriptmacro << latex1 << lb << latex2 << rb << cb;
 943         return latex1.length() + latex2.length() + lb.length() + rb.length() + cb.length();
 944 }
 945
 946
 947 bool Paragraph::Private::simpleTeXBlanks(BufferParams const & bparams,
 948                                        OutputParams const & runparams,
 949                                        otexstream & os,
 950                                        pos_type i,
 951                                        unsigned int & column,
 952                                        Font const & font,
 953                                        Layout const & style)
 954 {
 955         if (style.pass_thru || runparams.pass_thru)
 956                 return false;
 957
 958         if (i + 1 < int(text_.size())) {
 959                 char_type next = text_[i + 1];
 960                 if (Encodings::isCombiningChar(next)) {
 961                         // This space has an accent, so we must always output it.
 962                         column += latexSurrogatePair(bparams, os, ' ', next, runparams) - 1;
 963                         return true;
 964                 }
 965         }
 966
 967         if (runparams.linelen > 0
 968             && column > runparams.linelen
 969             && i
 970             && text_[i - 1] != ' '
 971             && (i + 1 < int(text_.size()))
 972             // same in FreeSpacing mode
 973             && !owner_->isFreeSpacing()
 974             // In typewriter mode, we want to avoid
 975             // ! . ? : at the end of a line
 976             && !(font.fontInfo().family() == TYPEWRITER_FAMILY
 977                  && (text_[i - 1] == '.'
 978                      || text_[i - 1] == '?'
 979                      || text_[i - 1] == ':'
 980                      || text_[i - 1] == '!'))) {
 981                 os << '\n';
 982                 os.texrow().start(owner_->id(), i + 1);
 983                 column = 0;
 984         } else if (style.free_spacing) {
 985                 os << '~';
 986         } else {
 987                 os << ' ';
 988         }
 989         return false;
 990 }
 991
 992
 993 int Paragraph::Private::writeScriptChars(BufferParams const & bparams,
 994                                          OutputParams const & runparams,
 995                                          otexstream & os,
 996                                          docstring const & ltx,
 997                                          Change const & runningChange,
 998                                          Encoding const & encoding,
 999                                          string const fontenc,
1000                                          pos_type & i)
1001 {
1002         // FIXME: modifying i here is not very nice...
1003
1004         // We only arrive here when character text_[i] could not be translated
1005         // into the current latex encoding (or its latex translation has been forced,)
1006         // and it belongs to a known script.
1007         // Parameter ltx contains the latex translation of text_[i] as specified
1008         // in the unicodesymbols file and is something like "\textXXX{<spec>}".
1009         // The latex macro name "textXXX" specifies the script to which text_[i]
1010         // belongs and we use it in order to check whether characters from the
1011         // same script immediately follow, such that we can collect them in a
1012         // single "\textXXX" macro. So, we have to retain "\textXXX{<spec>"
1013         // for the first char but only "<spec>" for all subsequent chars.
1014         docstring::size_type const brace1 = ltx.find_first_of(from_ascii("{"));
1015         docstring::size_type const brace2 = ltx.find_last_of(from_ascii("}"));
1016         string script = to_ascii(ltx.substr(1, brace1 - 1));
1017         int pos = 0;
1018         int length = brace2;
1019         bool closing_brace = true;
1020         if (!Encodings::needsScriptWrapper(script, fontenc)) {
1021                 // Correct font encoding is being used, so we can avoid \text[greek|cyr].
1022                 pos = brace1 + 1;
1023                 length -= pos;
1024                 closing_brace = false;
1025         }
1026         os << ltx.substr(pos, length);
1027         int size = text_.size();
1028         while (i + 1 < size) {
1029                 char_type const next = text_[i + 1];
1030                 // Stop here if next character belongs to another script
1031                 // or there is a change in change tracking status.
1032                 if (!Encodings::isKnownScriptChar(next, script) ||
1033                     runningChange != owner_->lookupChange(i + 1))
1034                         break;
1035                 Font prev_font;
1036                 bool found = false;
1037                 FontList::const_iterator cit = fontlist_.begin();
1038                 FontList::const_iterator end = fontlist_.end();
1039                 for (; cit != end; ++cit) {
1040                         if (cit->pos() >= i && !found) {
1041                                 prev_font = cit->font();
1042                                 found = true;
1043                         }
1044                         if (cit->pos() >= i + 1)
1045                                 break;
1046                 }
1047                 // Stop here if there is a font attribute or encoding change.
1048                 if (found && cit != end && prev_font != cit->font())
1049                         break;
1050
1051                 // Check whether we have a combining pair
1052                 char_type next_next = '\0';
1053                 if (i + 2 < size) {
1054                         next_next = text_[i + 2];
1055                         if (Encodings::isCombiningChar(next_next)) {
1056                                 length += latexSurrogatePair(bparams, os, next, next_next, runparams) - 1;
1057                                 i += 2;
1058                                 continue;
1059                         }
1060                 }
1061
1062                 docstring const latex = encoding.latexChar(next).first;
1063                 docstring::size_type const b1 =
1064                                         latex.find_first_of(from_ascii("{"));
1065                 docstring::size_type const b2 =
1066                                         latex.find_last_of(from_ascii("}"));
1067                 int const len = b2 - b1 - 1;
1068                 os << latex.substr(b1 + 1, len);
1069                 length += len;
1070                 ++i;
1071         }
1072         if (closing_brace) {
1073                 os << '}';
1074                 ++length;
1075         }
1076         return length;
1077 }
1078
1079
1080 void Paragraph::Private::latexInset(BufferParams const & bparams,
1081                                     otexstream & os,
1082                                     OutputParams & runparams,
1083                                     Font & running_font,
1084                                     Font & basefont,
1085                                     Font const & outerfont,
1086                                     bool & open_font,
1087                                     Change & running_change,
1088                                     Layout const & style,
1089                                     pos_type & i,
1090                                     unsigned int & column)
1091 {
1092         Inset * inset = owner_->getInset(i);
1093         LBUFERR(inset);
1094
1095         if (style.pass_thru) {
1096                 odocstringstream ods;
1097                 inset->plaintext(ods, runparams);
1098                 os << ods.str();
1099                 return;
1100         }
1101
1102         // FIXME: move this to InsetNewline::latex
1103         if (inset->lyxCode() == NEWLINE_CODE || inset->lyxCode() == SEPARATOR_CODE) {
1104                 // newlines are handled differently here than
1105                 // the default in simpleTeXSpecialChars().
1106                 if (!style.newline_allowed) {
1107                         os << '\n';
1108                 } else {
1109                         if (open_font) {
1110                                 bool needPar = false;
1111                                 column += running_font.latexWriteEndChanges(
1112                                         os, bparams, runparams,
1113                                         basefont, basefont, needPar);
1114                                 open_font = false;
1115                         }
1116
1117                         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY)
1118                                 os << '~';
1119
1120                         basefont = owner_->getLayoutFont(bparams, outerfont);
1121                         running_font = basefont;
1122
1123                         if (runparams.moving_arg)
1124                                 os << "\\protect ";
1125
1126                 }
1127                 os.texrow().start(owner_->id(), i + 1);
1128                 column = 0;
1129         }
1130
1131         if (owner_->isDeleted(i)) {
1132                 if( ++runparams.inDeletedInset == 1)
1133                         runparams.changeOfDeletedInset = owner_->lookupChange(i);
1134         }
1135
1136         if (inset->canTrackChanges()) {
1137                 column += Changes::latexMarkChange(os, bparams, running_change,
1138                         Change(Change::UNCHANGED), runparams);
1139                 running_change = Change(Change::UNCHANGED);
1140         }
1141
1142         bool close = false;
1143         odocstream::pos_type const len = os.os().tellp();
1144
1145         if (inset->forceLTR()
1146             && !runparams.use_polyglossia
1147             && running_font.isRightToLeft()
1148             // ERT is an exception, it should be output with no
1149             // decorations at all
1150             && inset->lyxCode() != ERT_CODE) {
1151                 if (running_font.language()->lang() == "farsi")
1152                         os << "\\beginL" << termcmd;
1153                 else
1154                         os << "\\L{";
1155                 close = true;
1156         }
1157
1158         // FIXME: Bug: we can have an empty font change here!
1159         // if there has just been a font change, we are going to close it
1160         // right now, which means stupid latex code like \textsf{}. AFAIK,
1161         // this does not harm dvi output. A minor bug, thus (JMarc)
1162
1163         // Some insets cannot be inside a font change command.
1164         // However, even such insets *can* be placed in \L or \R
1165         // or their equivalents (for RTL language switches), so we don't
1166         // close the language in those cases.
1167         // ArabTeX, though, cannot handle this special behavior, it seems.
1168         bool arabtex = basefont.language()->lang() == "arabic_arabtex"
1169                 || running_font.language()->lang() == "arabic_arabtex";
1170         if (open_font && !inset->inheritFont()) {
1171                 bool needPar = false;
1172                 bool closeLanguage = arabtex
1173                         || basefont.isRightToLeft() == running_font.isRightToLeft();
1174                 unsigned int count = running_font.latexWriteEndChanges(os,
1175                                         bparams, runparams, basefont, basefont,
1176                                         needPar, closeLanguage);
1177                 column += count;
1178                 // if any font properties were closed, update the running_font,
1179                 // making sure, however, to leave the language as it was
1180                 if (count > 0) {
1181                         // FIXME: probably a better way to keep track of the old
1182                         // language, than copying the entire font?
1183                         Font const copy_font(running_font);
1184                         basefont = owner_->getLayoutFont(bparams, outerfont);
1185                         running_font = basefont;
1186                         if (!closeLanguage)
1187                                 running_font.setLanguage(copy_font.language());
1188                         // leave font open if language is still open
1189                         open_font = (running_font.language() == basefont.language());
1190                         if (closeLanguage)
1191                                 runparams.local_font = &basefont;
1192                 }
1193         }
1194
1195         size_t const previous_row_count = os.texrow().rows();
1196
1197         try {
1198                 runparams.lastid = id_;
1199                 runparams.lastpos = i;
1200                 inset->latex(os, runparams);
1201         } catch (EncodingException & e) {
1202                 // add location information and throw again.
1203                 e.par_id = id_;
1204                 e.pos = i;
1205                 throw(e);
1206         }
1207
1208         if (close) {
1209                 if (running_font.language()->lang() == "farsi")
1210                                 os << "\\endL" << termcmd;
1211                         else
1212                                 os << '}';
1213         }
1214
1215         if (os.texrow().rows() > previous_row_count) {
1216                 os.texrow().start(owner_->id(), i + 1);
1217                 column = 0;
1218         } else {
1219                 column += (unsigned int)(os.os().tellp() - len);
1220         }
1221
1222         if (owner_->isDeleted(i))
1223                 --runparams.inDeletedInset;
1224 }
1225
1226
1227 void Paragraph::Private::latexSpecialChar(otexstream & os,
1228                                           BufferParams const & bparams,
1229                                           OutputParams const & runparams,
1230                                           Font const & running_font,
1231                                           Change const & running_change,
1232                                           Layout const & style,
1233                                           pos_type & i,
1234                                           pos_type end_pos,
1235                                           unsigned int & column)
1236 {
1237         // With polyglossia, brackets and stuff need not be reversed
1238         // in RTL scripts (see bug #8251)
1239         char_type const c = (runparams.use_polyglossia) ?
1240                 owner_->getUChar(bparams, i) : text_[i];
1241
1242         if (style.pass_thru || runparams.pass_thru
1243             || contains(style.pass_thru_chars, c)
1244             || contains(runparams.pass_thru_chars, c)) {
1245                 if (c != '\0') {
1246                         Encoding const * const enc = runparams.encoding;
1247                         if (enc && !enc->encodable(c))
1248                                 throw EncodingException(c);
1249                         os.put(c);
1250                 }
1251                 return;
1252         }
1253
1254         // TIPA uses its own T3 encoding
1255         if (runparams.inIPA && latexSpecialT3(c, os, i, column))
1256                 return;
1257         // If T1 font encoding is used, use the special
1258         // characters it provides.
1259         // NOTE: Some languages reset the font encoding internally to a
1260         //       non-standard font encoding. If we are using such a language,
1261         //       we do not output special T1 chars.
1262         if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1263             && !runparams.isFullUnicode() && bparams.main_font_encoding() == "T1"
1264             && latexSpecialT1(c, os, i, column))
1265                 return;
1266         // NOTE: XeTeX and LuaTeX use EU1/2 (pre 2017) or TU (as of 2017) encoding
1267         else if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1268                  && runparams.isFullUnicode() && latexSpecialTU(c, os, i, column))
1269                      return;
1270
1271         // Otherwise, we use what LaTeX provides us.
1272         switch (c) {
1273         case '\\':
1274                 os << "\\textbackslash" << termcmd;
1275                 column += 15;
1276                 break;
1277         case '<':
1278                 os << "\\textless" << termcmd;
1279                 column += 10;
1280                 break;
1281         case '>':
1282                 os << "\\textgreater" << termcmd;
1283                 column += 13;
1284                 break;
1285         case '|':
1286                 os << "\\textbar" << termcmd;
1287                 column += 9;
1288                 break;
1289         case '-':
1290                 os << '-';
1291                 if (i + 1 < static_cast<pos_type>(text_.size()) &&
1292                     (end_pos == -1 || i + 1 < end_pos) &&
1293                     text_[i+1] == '-') {
1294                         // Prevent "--" becoming an en dash and "---" an em dash.
1295                         // (Within \ttfamily, "---" is merged to en dash + hyphen.)
1296                         os << "{}";
1297                         column += 2;
1298                 }
1299                 break;
1300         case '\"':
1301                 os << "\\textquotedbl" << termcmd;
1302                 column += 14;
1303                 break;
1304
1305         case '$': case '&':
1306         case '%': case '#': case '{':
1307         case '}': case '_':
1308                 os << '\\';
1309                 os.put(c);
1310                 column += 1;
1311                 break;
1312
1313         case '~':
1314                 os << "\\textasciitilde" << termcmd;
1315                 column += 16;
1316                 break;
1317
1318         case '^':
1319                 os << "\\textasciicircum" << termcmd;
1320                 column += 17;
1321                 break;
1322
1323         case '*':
1324         case '[':
1325         case ']':
1326                 // avoid being mistaken for optional arguments
1327                 os << '{';
1328                 os.put(c);
1329                 os << '}';
1330                 column += 2;
1331                 break;
1332
1333         case ' ':
1334                 // Blanks are printed before font switching.
1335                 // Sure? I am not! (try nice-latex)
1336                 // I am sure it's correct. LyX might be smarter
1337                 // in the future, but for now, nothing wrong is
1338                 // written. (Asger)
1339                 break;
1340
1341         case 0x2013:
1342         case 0x2014:
1343                 // XeTeX's dash behaviour is determined via a global setting
1344                 if (bparams.use_dash_ligatures
1345                     && owner_->getFontSettings(bparams, i).fontInfo().family() != TYPEWRITER_FAMILY
1346                     && !runparams.inIPA
1347                         // TODO #10961: && not in inset Flex Code
1348                         // TODO #10961: && not in layout LyXCode
1349                     && (!bparams.useNonTeXFonts || runparams.flavor != OutputParams::XETEX)) {
1350                         if (c == 0x2013) {
1351                                 // en-dash
1352                                 os << "--";
1353                                 column +=2;
1354                         } else {
1355                                 // em-dash
1356                                 os << "---";
1357                                 column +=3;
1358                         }
1359                         break;
1360                 }
1361                 // fall through
1362         default:
1363                 if (c == '\0')
1364                         return;
1365
1366                 Encoding const & encoding = *(runparams.encoding);
1367                 char_type next = '\0';
1368                 if (i + 1 < int(text_.size())) {
1369                         next = text_[i + 1];
1370                         if (Encodings::isCombiningChar(next)) {
1371                                 column += latexSurrogatePair(bparams, os, c, next, runparams) - 1;
1372                                 ++i;
1373                                 break;
1374                         }
1375                 }
1376                 string script;
1377                 pair<docstring, bool> latex = encoding.latexChar(c);
1378                 docstring nextlatex;
1379                 bool nexttipas = false;
1380                 string nexttipashortcut;
1381                 if (next != '\0' && next != META_INSET && encoding.encodable(next)) {
1382                         nextlatex = encoding.latexChar(next).first;
1383                         if (runparams.inIPA) {
1384                                 nexttipashortcut = Encodings::TIPAShortcut(next);
1385                                 nexttipas = !nexttipashortcut.empty();
1386                         }
1387                 }
1388                 bool tipas = false;
1389                 if (runparams.inIPA) {
1390                         string const tipashortcut = Encodings::TIPAShortcut(c);
1391                         if (!tipashortcut.empty()) {
1392                                 latex.first = from_ascii(tipashortcut);
1393                                 latex.second = false;
1394                                 tipas = true;
1395                         }
1396                 }
1397                 string fontenc;
1398                 if (running_font.language()->lang() == bparams.language->lang())
1399                         fontenc = runparams.main_fontenc;
1400                 else
1401                         fontenc = running_font.language()->fontenc(bparams);
1402                 // "Script chars" need to embraced in \textcyrillic and \textgreek notwithstanding
1403                 // whether they are encodable or not (it only depends on the font encoding)
1404                 if (!runparams.isFullUnicode() && Encodings::isKnownScriptChar(c, script)) {
1405                         docstring const wrapper = from_ascii("\\" + script + "{");
1406                         docstring ltx = latex.first;
1407                         if (!prefixIs(ltx, wrapper))
1408                                 ltx = wrapper + latex.first + from_ascii("}");
1409                         column += writeScriptChars(bparams, runparams, os, ltx, running_change,
1410                                                    encoding, fontenc, i) - 1;
1411                 } else if (latex.second
1412                          && ((!prefixIs(nextlatex, '\\')
1413                                && !prefixIs(nextlatex, '{')
1414                                && !prefixIs(nextlatex, '}'))
1415                              || (nexttipas
1416                                  && !prefixIs(from_ascii(nexttipashortcut), '\\')))
1417                          && !tipas) {
1418                         // Prevent eating of a following
1419                         // space or command corruption by
1420                         // following characters
1421                         if (next == ' ' || next == '\0') {
1422                                 column += latex.first.length() + 1;
1423                                 os << latex.first << "{}";
1424                         } else {
1425                                 column += latex.first.length();
1426                                 os << latex.first << " ";
1427                         }
1428                 } else {
1429                         column += latex.first.length() - 1;
1430                         os << latex.first;
1431                 }
1432                 break;
1433         }
1434 }
1435
1436
1437 bool Paragraph::Private::latexSpecialT1(char_type const c, otexstream & os,
1438         pos_type i, unsigned int & column)
1439 {
1440         switch (c) {
1441         case '>':
1442         case '<':
1443                 os.put(c);
1444                 // In T1 encoding, these characters exist
1445                 // but we should avoid ligatures
1446                 if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
1447                         return true;
1448                 os << "\\textcompwordmark" << termcmd;
1449                 column += 19;
1450                 return true;
1451         case '|':
1452                 os.put(c);
1453                 return true;
1454         case '\"':
1455                 // soul.sty breaks with \char`\"
1456                 os << "\\textquotedbl" << termcmd;
1457                 column += 14;
1458                 return true;
1459         default:
1460                 return false;
1461         }
1462 }
1463
1464
1465 bool Paragraph::Private::latexSpecialTU(char_type const c, otexstream & os,
1466         pos_type i, unsigned int & column)
1467 {
1468         // TU encoding is currently on par with T1.
1469         return latexSpecialT1(c, os, i, column);
1470 }
1471
1472
1473 bool Paragraph::Private::latexSpecialT3(char_type const c, otexstream & os,
1474         pos_type /*i*/, unsigned int & column)
1475 {
1476         switch (c) {
1477         case '*':
1478         case '[':
1479         case ']':
1480         case '\"':
1481                 os.put(c);
1482                 return true;
1483         case '|':
1484                 os << "\\textvertline" << termcmd;
1485                 column += 14;
1486                 return true;
1487         default:
1488                 return false;
1489         }
1490 }
1491
1492
1493 void Paragraph::Private::validate(LaTeXFeatures & features) const
1494 {
1495         if (layout_->inpreamble && inset_owner_) {
1496                 // FIXME: Using a string stream here circumvents the encoding
1497                 // switching machinery of odocstream. Therefore the
1498                 // output is wrong if this paragraph contains content
1499                 // that needs to switch encoding.
1500                 Buffer const & buf = inset_owner_->buffer();
1501                 otexstringstream os;
1502                 os << layout_->preamble();
1503                 size_t const length = os.length();
1504                 TeXOnePar(buf, buf.text(), buf.getParFromID(owner_->id()).pit(), os,
1505                           features.runparams(), string(), 0, -1, true);
1506                 if (os.length() > length)
1507                         features.addPreambleSnippet(os.release(), true);
1508         }
1509
1510         if (features.runparams().flavor == OutputParams::HTML
1511             && layout_->htmltitle()) {
1512                 features.setHTMLTitle(owner_->asString(AS_STR_INSETS | AS_STR_SKIPDELETE));
1513         }
1514
1515         // check the params.
1516         if (!params_.spacing().isDefault())
1517                 features.require("setspace");
1518
1519         // then the layouts
1520         features.useLayout(layout_->name());
1521
1522         // then the fonts
1523         fontlist_.validate(features);
1524
1525         // then the indentation
1526         if (!params_.leftIndent().zero())
1527                 features.require("ParagraphLeftIndent");
1528
1529         // then the insets
1530         InsetList::const_iterator icit = insetlist_.begin();
1531         InsetList::const_iterator iend = insetlist_.end();
1532         for (; icit != iend; ++icit) {
1533                 if (icit->inset) {
1534                         features.inDeletedInset(owner_->isDeleted(icit->pos));
1535                         if (icit->inset->lyxCode() == FOOT_CODE) {
1536                                 // FIXME: an item inset would make things much easier.
1537                                 if ((layout_->latextype == LATEX_LIST_ENVIRONMENT
1538                                      || (layout_->latextype == LATEX_ITEM_ENVIRONMENT
1539                                          && layout_->margintype == MARGIN_FIRST_DYNAMIC))
1540                                     && (icit->pos < begin_of_body_
1541                                         || (icit->pos == begin_of_body_
1542                                             && (icit->pos == 0 || text_[icit->pos - 1] != ' '))))
1543                                         features.saveNoteEnv("description");
1544                         }
1545                         icit->inset->validate(features);
1546                         features.inDeletedInset(false);
1547                         if (layout_->needprotect &&
1548                             icit->inset->lyxCode() == FOOT_CODE)
1549                                 features.require("NeedLyXFootnoteCode");
1550                 }
1551         }
1552
1553         // then the contents
1554         BufferParams const bp = features.runparams().is_child
1555                 ? features.buffer().masterParams() : features.buffer().params();
1556         string bscript = "textbaltic";
1557         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1558                 char_type c = text_[i];
1559                 if (c == 0x0022) {
1560                         if (features.runparams().isFullUnicode() && bp.useNonTeXFonts)
1561                                 features.require("textquotedblp");
1562                         else if (bp.main_font_encoding() != "T1"
1563                                  || ((&owner_->getFontSettings(bp, i))->language()->internalFontEncoding()))
1564                                 features.require("textquotedbl");
1565                 } else if (Encodings::isKnownScriptChar(c, bscript)){
1566                         string fontenc = (&owner_->getFontSettings(bp, i))->language()->fontenc(bp);
1567                         if (fontenc.empty())
1568                                 fontenc = features.runparams().main_fontenc;
1569                         if (Encodings::needsScriptWrapper("textbaltic", fontenc))
1570                                 features.require("textbalticdefs");
1571                 } else if (!bp.use_dash_ligatures
1572                            && (c == 0x2013 || c == 0x2014)
1573                            && bp.useNonTeXFonts
1574                            && features.runparams().flavor == OutputParams::XETEX)
1575                         // XeTeX's dash behaviour is determined via a global setting
1576                         features.require("xetexdashbreakstate");
1577                 BufferEncodings::validate(c, features);
1578         }
1579 }
1580
1581 /////////////////////////////////////////////////////////////////////
1582 //
1583 // Paragraph
1584 //
1585 /////////////////////////////////////////////////////////////////////
1586
1587 namespace {
1588         Layout const emptyParagraphLayout;
1589 }
1590
1591 Paragraph::Paragraph()
1592         : d(new Paragraph::Private(this, emptyParagraphLayout))
1593 {
1594         itemdepth = 0;
1595         d->params_.clear();
1596 }
1597
1598
1599 Paragraph::Paragraph(Paragraph const & par)
1600         : itemdepth(par.itemdepth),
1601         d(new Paragraph::Private(*par.d, this))
1602 {
1603         registerWords();
1604 }
1605
1606
1607 Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
1608         : itemdepth(par.itemdepth),
1609         d(new Paragraph::Private(*par.d, this, beg, end))
1610 {
1611         registerWords();
1612 }
1613
1614
1615 Paragraph & Paragraph::operator=(Paragraph const & par)
1616 {
1617         // needed as we will destroy the private part before copying it
1618         if (&par != this) {
1619                 itemdepth = par.itemdepth;
1620
1621                 deregisterWords();
1622                 delete d;
1623                 d = new Private(*par.d, this);
1624                 registerWords();
1625         }
1626         return *this;
1627 }
1628
1629
1630 Paragraph::~Paragraph()
1631 {
1632         deregisterWords();
1633         delete d;
1634 }
1635
1636
1637 namespace {
1638
1639 // this shall be called just before every "os << ..." action.
1640 void flushString(ostream & os, docstring & s)
1641 {
1642         os << to_utf8(s);
1643         s.erase();
1644 }
1645
1646 } // namespace
1647
1648
1649 void Paragraph::write(ostream & os, BufferParams const & bparams,
1650         depth_type & dth) const
1651 {
1652         // The beginning or end of a deeper (i.e. nested) area?
1653         if (dth != d->params_.depth()) {
1654                 if (d->params_.depth() > dth) {
1655                         while (d->params_.depth() > dth) {
1656                                 os << "\n\\begin_deeper";
1657                                 ++dth;
1658                         }
1659                 } else {
1660                         while (d->params_.depth() < dth) {
1661                                 os << "\n\\end_deeper";
1662                                 --dth;
1663                         }
1664                 }
1665         }
1666
1667         // First write the layout
1668         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1669
1670         d->params_.write(os);
1671
1672         Font font1(inherit_font, bparams.language);
1673
1674         Change running_change = Change(Change::UNCHANGED);
1675
1676         // this string is used as a buffer to avoid repetitive calls
1677         // to to_utf8(), which turn out to be expensive (JMarc)
1678         docstring write_buffer;
1679
1680         int column = 0;
1681         for (pos_type i = 0; i <= size(); ++i) {
1682
1683                 Change const & change = lookupChange(i);
1684                 if (change != running_change)
1685                         flushString(os, write_buffer);
1686                 Changes::lyxMarkChange(os, bparams, column, running_change, change);
1687                 running_change = change;
1688
1689                 if (i == size())
1690                         break;
1691
1692                 // Write font changes
1693                 Font font2 = getFontSettings(bparams, i);
1694                 if (font2 != font1) {
1695                         flushString(os, write_buffer);
1696                         font2.lyxWriteChanges(font1, os);
1697                         column = 0;
1698                         font1 = font2;
1699                 }
1700
1701                 char_type const c = d->text_[i];
1702                 switch (c) {
1703                 case META_INSET:
1704                         if (Inset const * inset = getInset(i)) {
1705                                 flushString(os, write_buffer);
1706                                 if (inset->directWrite()) {
1707                                         // international char, let it write
1708                                         // code directly so it's shorter in
1709                                         // the file
1710                                         inset->write(os);
1711                                 } else {
1712                                         if (i)
1713                                                 os << '\n';
1714                                         os << "\\begin_inset ";
1715                                         inset->write(os);
1716                                         os << "\n\\end_inset\n\n";
1717                                         column = 0;
1718                                 }
1719                                 // FIXME This can be removed again once the mystery
1720                                 // crash has been resolved.
1721                                 os << flush;
1722                         }
1723                         break;
1724                 case '\\':
1725                         flushString(os, write_buffer);
1726                         os << "\n\\backslash\n";
1727                         column = 0;
1728                         break;
1729                 case '.':
1730                         flushString(os, write_buffer);
1731                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1732                                 os << ".\n";
1733                                 column = 0;
1734                         } else
1735                                 os << '.';
1736                         break;
1737                 default:
1738                         if ((column > 70 && c == ' ')
1739                             || column > 79) {
1740                                 flushString(os, write_buffer);
1741                                 os << '\n';
1742                                 column = 0;
1743                         }
1744                         // this check is to amend a bug. LyX sometimes
1745                         // inserts '\0' this could cause problems.
1746                         if (c != '\0')
1747                                 write_buffer.push_back(c);
1748                         else
1749                                 LYXERR0("NUL char in structure.");
1750                         ++column;
1751                         break;
1752                 }
1753         }
1754
1755         flushString(os, write_buffer);
1756         os << "\n\\end_layout\n";
1757         // FIXME This can be removed again once the mystery
1758         // crash has been resolved.
1759         os << flush;
1760 }
1761
1762
1763 void Paragraph::validate(LaTeXFeatures & features) const
1764 {
1765         d->validate(features);
1766         if (needsCProtection())
1767                 features.require("cprotect");
1768 }
1769
1770
1771 void Paragraph::insert(pos_type start, docstring const & str,
1772                        Font const & font, Change const & change)
1773 {
1774         for (size_t i = 0, n = str.size(); i != n ; ++i)
1775                 insertChar(start + i, str[i], font, change);
1776 }
1777
1778
1779 void Paragraph::appendChar(char_type c, Font const & font,
1780                 Change const & change)
1781 {
1782         // track change
1783         d->changes_.insert(change, d->text_.size());
1784         // when appending characters, no need to update tables
1785         d->text_.push_back(c);
1786         setFont(d->text_.size() - 1, font);
1787         d->requestSpellCheck(d->text_.size() - 1);
1788 }
1789
1790
1791 void Paragraph::appendString(docstring const & s, Font const & font,
1792                 Change const & change)
1793 {
1794         pos_type end = s.size();
1795         size_t oldsize = d->text_.size();
1796         size_t newsize = oldsize + end;
1797         size_t capacity = d->text_.capacity();
1798         if (newsize >= capacity)
1799                 d->text_.reserve(max(capacity + 100, newsize));
1800
1801         // when appending characters, no need to update tables
1802         d->text_.append(s);
1803
1804         // FIXME: Optimize this!
1805         for (size_t i = oldsize; i != newsize; ++i) {
1806                 // track change
1807                 d->changes_.insert(change, i);
1808                 d->requestSpellCheck(i);
1809         }
1810         d->fontlist_.set(oldsize, font);
1811         d->fontlist_.set(newsize - 1, font);
1812 }
1813
1814
1815 void Paragraph::insertChar(pos_type pos, char_type c,
1816                            bool trackChanges)
1817 {
1818         d->insertChar(pos, c, Change(trackChanges ?
1819                            Change::INSERTED : Change::UNCHANGED));
1820 }
1821
1822
1823 void Paragraph::insertChar(pos_type pos, char_type c,
1824                            Font const & font, bool trackChanges)
1825 {
1826         d->insertChar(pos, c, Change(trackChanges ?
1827                            Change::INSERTED : Change::UNCHANGED));
1828         setFont(pos, font);
1829 }
1830
1831
1832 void Paragraph::insertChar(pos_type pos, char_type c,
1833                            Font const & font, Change const & change)
1834 {
1835         d->insertChar(pos, c, change);
1836         setFont(pos, font);
1837 }
1838
1839
1840 void Paragraph::resetFonts(Font const & font)
1841 {
1842         d->fontlist_.clear();
1843         d->fontlist_.set(0, font);
1844         d->fontlist_.set(d->text_.size() - 1, font);
1845 }
1846
1847 // Gets uninstantiated font setting at position.
1848 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1849                                          pos_type pos) const
1850 {
1851         if (pos > size()) {
1852                 LYXERR0("pos: " << pos << " size: " << size());
1853                 LBUFERR(false);
1854         }
1855
1856         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1857         if (cit != d->fontlist_.end())
1858                 return cit->font();
1859
1860         if (pos == size() && !empty())
1861                 return getFontSettings(bparams, pos - 1);
1862
1863         // Optimisation: avoid a full font instantiation if there is no
1864         // language change from previous call.
1865         static Font previous_font;
1866         static Language const * previous_lang = 0;
1867         Language const * lang = getParLanguage(bparams);
1868         if (lang != previous_lang) {
1869                 previous_lang = lang;
1870                 previous_font = Font(inherit_font, lang);
1871         }
1872         return previous_font;
1873 }
1874
1875
1876 FontSpan Paragraph::fontSpan(pos_type pos) const
1877 {
1878         LBUFERR(pos <= size());
1879
1880         if (pos == size())
1881                 return FontSpan(pos, pos);
1882
1883         pos_type start = 0;
1884         FontList::const_iterator cit = d->fontlist_.begin();
1885         FontList::const_iterator end = d->fontlist_.end();
1886         for (; cit != end; ++cit) {
1887                 if (cit->pos() >= pos) {
1888                         if (pos >= beginOfBody())
1889                                 return FontSpan(max(start, beginOfBody()),
1890                                                 cit->pos());
1891                         else
1892                                 return FontSpan(start,
1893                                                 min(beginOfBody() - 1,
1894                                                          cit->pos()));
1895                 }
1896                 start = cit->pos() + 1;
1897         }
1898
1899         // This should not happen, but if so, we take no chances.
1900         LYXERR0("Paragraph::fontSpan: position not found in fontinfo table!");
1901         LASSERT(false, return FontSpan(pos, pos));
1902 }
1903
1904
1905 // Gets uninstantiated font setting at position 0
1906 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1907 {
1908         if (!empty() && !d->fontlist_.empty())
1909                 return d->fontlist_.begin()->font();
1910
1911         // Optimisation: avoid a full font instantiation if there is no
1912         // language change from previous call.
1913         static Font previous_font;
1914         static Language const * previous_lang = 0;
1915         if (bparams.language != previous_lang) {
1916                 previous_lang = bparams.language;
1917                 previous_font = Font(inherit_font, bparams.language);
1918         }
1919
1920         return previous_font;
1921 }
1922
1923
1924 // Gets the fully instantiated font at a given position in a paragraph
1925 // This is basically the same function as Text::GetFont() in text2.cpp.
1926 // The difference is that this one is used for generating the LaTeX file,
1927 // and thus cosmetic "improvements" are disallowed: This has to deliver
1928 // the true picture of the buffer. (Asger)
1929 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1930                                  Font const & outerfont) const
1931 {
1932         LBUFERR(pos >= 0);
1933
1934         Font font = getFontSettings(bparams, pos);
1935
1936         pos_type const body_pos = beginOfBody();
1937         FontInfo & fi = font.fontInfo();
1938         if (pos < body_pos)
1939                 fi.realize(d->layout_->labelfont);
1940         else
1941                 fi.realize(d->layout_->font);
1942
1943         fi.realize(outerfont.fontInfo());
1944         fi.realize(bparams.getFont().fontInfo());
1945
1946         return font;
1947 }
1948
1949
1950 Font const Paragraph::getLabelFont
1951         (BufferParams const & bparams, Font const & outerfont) const
1952 {
1953         FontInfo tmpfont = d->layout_->labelfont;
1954         tmpfont.realize(outerfont.fontInfo());
1955         tmpfont.realize(bparams.getFont().fontInfo());
1956         return Font(tmpfont, getParLanguage(bparams));
1957 }
1958
1959
1960 Font const Paragraph::getLayoutFont
1961         (BufferParams const & bparams, Font const & outerfont) const
1962 {
1963         FontInfo tmpfont = d->layout_->font;
1964         tmpfont.realize(outerfont.fontInfo());
1965         tmpfont.realize(bparams.getFont().fontInfo());
1966         return Font(tmpfont, getParLanguage(bparams));
1967 }
1968
1969
1970 char_type Paragraph::getUChar(BufferParams const & bparams, pos_type pos) const
1971 {
1972         char_type c = d->text_[pos];
1973         if (!getFontSettings(bparams, pos).isRightToLeft())
1974                 return c;
1975
1976         // FIXME: The arabic special casing is due to the difference of arabic
1977         // round brackets input introduced in r18599. Check if this should be
1978         // unified with Hebrew or at least if all bracket types should be
1979         // handled the same (file format change in either case).
1980         string const & lang = getFontSettings(bparams, pos).language()->lang();
1981         bool const arabic = lang == "arabic_arabtex" || lang == "arabic_arabi"
1982                 || lang == "farsi";
1983         char_type uc = c;
1984         switch (c) {
1985         case '(':
1986                 uc = arabic ? c : ')';
1987                 break;
1988         case ')':
1989                 uc = arabic ? c : '(';
1990                 break;
1991         case '[':
1992                 uc = ']';
1993                 break;
1994         case ']':
1995                 uc = '[';
1996                 break;
1997         case '{':
1998                 uc = '}';
1999                 break;
2000         case '}':
2001                 uc = '{';
2002                 break;
2003         case '<':
2004                 uc = '>';
2005                 break;
2006         case '>':
2007                 uc = '<';
2008                 break;
2009         }
2010
2011         return uc;
2012 }
2013
2014
2015 void Paragraph::setFont(pos_type pos, Font const & font)
2016 {
2017         LASSERT(pos <= size(), return);
2018
2019         // First, reduce font against layout/label font
2020         // Update: The setCharFont() routine in text2.cpp already
2021         // reduces font, so we don't need to do that here. (Asger)
2022
2023         d->fontlist_.set(pos, font);
2024 }
2025
2026
2027 void Paragraph::makeSameLayout(Paragraph const & par)
2028 {
2029         d->layout_ = par.d->layout_;
2030         d->params_ = par.d->params_;
2031 }
2032
2033
2034 bool Paragraph::stripLeadingSpaces(bool trackChanges)
2035 {
2036         if (isFreeSpacing())
2037                 return false;
2038
2039         int pos = 0;
2040         int count = 0;
2041
2042         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
2043                 if (eraseChar(pos, trackChanges))
2044                         ++count;
2045                 else
2046                         ++pos;
2047         }
2048
2049         return count > 0 || pos > 0;
2050 }
2051
2052
2053 bool Paragraph::hasSameLayout(Paragraph const & par) const
2054 {
2055         return par.d->layout_ == d->layout_
2056                 && d->params_.sameLayout(par.d->params_);
2057 }
2058
2059
2060 depth_type Paragraph::getDepth() const
2061 {
2062         return d->params_.depth();
2063 }
2064
2065
2066 depth_type Paragraph::getMaxDepthAfter() const
2067 {
2068         if (d->layout_->isEnvironment())
2069                 return d->params_.depth() + 1;
2070         else
2071                 return d->params_.depth();
2072 }
2073
2074
2075 LyXAlignment Paragraph::getAlign() const
2076 {
2077         if (d->params_.align() == LYX_ALIGN_LAYOUT)
2078                 return d->layout_->align;
2079         else
2080                 return d->params_.align();
2081 }
2082
2083
2084 docstring const & Paragraph::labelString() const
2085 {
2086         return d->params_.labelString();
2087 }
2088
2089
2090 // the next two functions are for the manual labels
2091 docstring const Paragraph::getLabelWidthString() const
2092 {
2093         if (d->layout_->margintype == MARGIN_MANUAL
2094             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
2095                 return d->params_.labelWidthString();
2096         else
2097                 return _("Senseless with this layout!");
2098 }
2099
2100
2101 void Paragraph::setLabelWidthString(docstring const & s)
2102 {
2103         d->params_.labelWidthString(s);
2104 }
2105
2106
2107 docstring Paragraph::expandLabel(Layout const & layout,
2108                 BufferParams const & bparams) const
2109 {
2110         return expandParagraphLabel(layout, bparams, true);
2111 }
2112
2113
2114 docstring Paragraph::expandDocBookLabel(Layout const & layout,
2115                 BufferParams const & bparams) const
2116 {
2117         return expandParagraphLabel(layout, bparams, false);
2118 }
2119
2120
2121 docstring Paragraph::expandParagraphLabel(Layout const & layout,
2122                 BufferParams const & bparams, bool process_appendix) const
2123 {
2124         DocumentClass const & tclass = bparams.documentClass();
2125         string const & lang = getParLanguage(bparams)->code();
2126         bool const in_appendix = process_appendix && d->params_.appendix();
2127         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
2128
2129         if (fmt.empty() && !layout.counter.empty())
2130                 return tclass.counters().theCounter(layout.counter, lang);
2131
2132         // handle 'inherited level parts' in 'fmt',
2133         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
2134         size_t const i = fmt.find('@', 0);
2135         if (i != docstring::npos) {
2136                 size_t const j = fmt.find('@', i + 1);
2137                 if (j != docstring::npos) {
2138                         docstring parent(fmt, i + 1, j - i - 1);
2139                         docstring label = from_ascii("??");
2140                         if (tclass.hasLayout(parent))
2141                                 label = expandParagraphLabel(tclass[parent], bparams,
2142                                                       process_appendix);
2143                         fmt = docstring(fmt, 0, i) + label
2144                                 + docstring(fmt, j + 1, docstring::npos);
2145                 }
2146         }
2147
2148         return tclass.counters().counterLabel(fmt, lang);
2149 }
2150
2151
2152 void Paragraph::applyLayout(Layout const & new_layout)
2153 {
2154         d->layout_ = &new_layout;
2155         LyXAlignment const oldAlign = d->params_.align();
2156
2157         if (!(oldAlign & d->layout_->alignpossible)) {
2158                 frontend::Alert::warning(_("Alignment not permitted"),
2159                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
2160                 d->params_.align(LYX_ALIGN_LAYOUT);
2161         }
2162 }
2163
2164
2165 pos_type Paragraph::beginOfBody() const
2166 {
2167         return d->begin_of_body_;
2168 }
2169
2170
2171 void Paragraph::setBeginOfBody()
2172 {
2173         if (d->layout_->labeltype != LABEL_MANUAL) {
2174                 d->begin_of_body_ = 0;
2175                 return;
2176         }
2177
2178         // Unroll the first two cycles of the loop
2179         // and remember the previous character to
2180         // remove unnecessary getChar() calls
2181         pos_type i = 0;
2182         pos_type end = size();
2183         if (i < end && !(isNewline(i) || isEnvSeparator(i))) {
2184                 ++i;
2185                 if (i < end) {
2186                         char_type previous_char = d->text_[i];
2187                         if (!(isNewline(i) || isEnvSeparator(i))) {
2188                                 ++i;
2189                                 while (i < end && previous_char != ' ') {
2190                                         char_type temp = d->text_[i];
2191                                         if (isNewline(i) || isEnvSeparator(i))
2192                                                 break;
2193                                         ++i;
2194                                         previous_char = temp;
2195                                 }
2196                         }
2197                 }
2198         }
2199
2200         d->begin_of_body_ = i;
2201 }
2202
2203
2204 bool Paragraph::allowParagraphCustomization() const
2205 {
2206         return inInset().allowParagraphCustomization();
2207 }
2208
2209
2210 bool Paragraph::usePlainLayout() const
2211 {
2212         return inInset().usePlainLayout();
2213 }
2214
2215
2216 bool Paragraph::isPassThru() const
2217 {
2218         return inInset().isPassThru() || d->layout_->pass_thru;
2219 }
2220
2221 namespace {
2222
2223 // paragraphs inside floats need different alignment tags to avoid
2224 // unwanted space
2225
2226 bool noTrivlistCentering(InsetCode code)
2227 {
2228         return code == FLOAT_CODE
2229                || code == WRAP_CODE
2230                || code == CELL_CODE;
2231 }
2232
2233
2234 string correction(string const & orig)
2235 {
2236         if (orig == "flushleft")
2237                 return "raggedright";
2238         if (orig == "flushright")
2239                 return "raggedleft";
2240         if (orig == "center")
2241                 return "centering";
2242         return orig;
2243 }
2244
2245
2246 bool corrected_env(otexstream & os, string const & suffix, string const & env,
2247         InsetCode code, bool const lastpar, int & col)
2248 {
2249         string macro = suffix + "{";
2250         if (noTrivlistCentering(code)) {
2251                 if (lastpar) {
2252                         // the last paragraph in non-trivlist-aligned
2253                         // context is special (to avoid unwanted whitespace)
2254                         if (suffix == "\\begin") {
2255                                 macro = "\\" + correction(env) + "{}";
2256                                 os << from_ascii(macro);
2257                                 col += macro.size();
2258                                 return true;
2259                         }
2260                         return false;
2261                 }
2262                 macro += correction(env);
2263         } else
2264                 macro += env;
2265         macro += "}";
2266         if (suffix == "\\par\\end") {
2267                 os << breakln;
2268                 col = 0;
2269         }
2270         os << from_ascii(macro);
2271         col += macro.size();
2272         if (suffix == "\\begin") {
2273                 os << breakln;
2274                 col = 0;
2275         }
2276         return true;
2277 }
2278
2279 } // namespace
2280
2281
2282 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
2283                         otexstream & os, OutputParams const & runparams) const
2284 {
2285         int column = 0;
2286
2287         bool canindent =
2288                 (bparams.paragraph_separation == BufferParams::ParagraphIndentSeparation) ?
2289                         (layout_->toggle_indent != ITOGGLE_NEVER) :
2290                         (layout_->toggle_indent == ITOGGLE_ALWAYS);
2291
2292         if (canindent && params_.noindent() && !layout_->pass_thru) {
2293                 os << "\\noindent ";
2294                 column += 10;
2295         }
2296
2297         LyXAlignment const curAlign = params_.align();
2298
2299         if (curAlign == layout_->align)
2300                 return column;
2301
2302         switch (curAlign) {
2303         case LYX_ALIGN_NONE:
2304         case LYX_ALIGN_BLOCK:
2305         case LYX_ALIGN_LAYOUT:
2306         case LYX_ALIGN_SPECIAL:
2307         case LYX_ALIGN_DECIMAL:
2308                 break;
2309         case LYX_ALIGN_LEFT:
2310         case LYX_ALIGN_RIGHT:
2311         case LYX_ALIGN_CENTER:
2312                 if (runparams.moving_arg) {
2313                         os << "\\protect";
2314                         column += 8;
2315                 }
2316                 break;
2317         }
2318
2319         string const begin_tag = "\\begin";
2320         InsetCode code = ownerCode();
2321         bool const lastpar = runparams.isLastPar;
2322
2323         switch (curAlign) {
2324         case LYX_ALIGN_NONE:
2325         case LYX_ALIGN_BLOCK:
2326         case LYX_ALIGN_LAYOUT:
2327         case LYX_ALIGN_SPECIAL:
2328         case LYX_ALIGN_DECIMAL:
2329                 break;
2330         case LYX_ALIGN_LEFT: {
2331                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2332                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2333                 else
2334                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2335                 break;
2336         } case LYX_ALIGN_RIGHT: {
2337                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2338                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2339                 else
2340                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2341                 break;
2342         } case LYX_ALIGN_CENTER: {
2343                 corrected_env(os, begin_tag, "center", code, lastpar, column);
2344                 break;
2345         }
2346         }
2347
2348         return column;
2349 }
2350
2351
2352 bool Paragraph::Private::endTeXParParams(BufferParams const & bparams,
2353                         otexstream & os, OutputParams const & runparams) const
2354 {
2355         LyXAlignment const curAlign = params_.align();
2356
2357         if (curAlign == layout_->align)
2358                 return false;
2359
2360         switch (curAlign) {
2361         case LYX_ALIGN_NONE:
2362         case LYX_ALIGN_BLOCK:
2363         case LYX_ALIGN_LAYOUT:
2364         case LYX_ALIGN_SPECIAL:
2365         case LYX_ALIGN_DECIMAL:
2366                 break;
2367         case LYX_ALIGN_LEFT:
2368         case LYX_ALIGN_RIGHT:
2369         case LYX_ALIGN_CENTER:
2370                 if (runparams.moving_arg)
2371                         os << "\\protect";
2372                 break;
2373         }
2374
2375         bool output = false;
2376         int col = 0;
2377         string const end_tag = "\\par\\end";
2378         InsetCode code = ownerCode();
2379         bool const lastpar = runparams.isLastPar;
2380
2381         switch (curAlign) {
2382         case LYX_ALIGN_NONE:
2383         case LYX_ALIGN_BLOCK:
2384         case LYX_ALIGN_LAYOUT:
2385         case LYX_ALIGN_SPECIAL:
2386         case LYX_ALIGN_DECIMAL:
2387                 break;
2388         case LYX_ALIGN_LEFT: {
2389                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2390                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2391                 else
2392                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2393                 break;
2394         } case LYX_ALIGN_RIGHT: {
2395                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2396                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2397                 else
2398                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2399                 break;
2400         } case LYX_ALIGN_CENTER: {
2401                 corrected_env(os, end_tag, "center", code, lastpar, col);
2402                 break;
2403         }
2404         }
2405
2406         return output || lastpar;
2407 }
2408
2409
2410 // This one spits out the text of the paragraph
2411 void Paragraph::latex(BufferParams const & bparams,
2412         Font const & outerfont,
2413         otexstream & os,
2414         OutputParams const & runparams,
2415         int start_pos, int end_pos, bool force) const
2416 {
2417         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
2418
2419         // FIXME This check should not be needed. Perhaps issue an
2420         // error if it triggers.
2421         Layout const & style = inInset().forcePlainLayout() ?
2422                 bparams.documentClass().plainLayout() : *d->layout_;
2423
2424         if (!force && style.inpreamble)
2425                 return;
2426
2427         bool const allowcust = allowParagraphCustomization();
2428
2429         // Current base font for all inherited font changes, without any
2430         // change caused by an individual character, except for the language:
2431         // It is set to the language of the first character.
2432         // As long as we are in the label, this font is the base font of the
2433         // label. Before the first body character it is set to the base font
2434         // of the body.
2435         Font basefont;
2436
2437         // Maybe we have to create a optional argument.
2438         pos_type body_pos = beginOfBody();
2439         unsigned int column = 0;
2440
2441         if (body_pos > 0) {
2442                 // the optional argument is kept in curly brackets in
2443                 // case it contains a ']'
2444                 // This is not strictly needed, but if this is changed it
2445                 // would be a file format change, and tex2lyx would need
2446                 // to be adjusted, since it unconditionally removes the
2447                 // braces when it parses \item.
2448                 os << "[{";
2449                 column += 2;
2450                 basefont = getLabelFont(bparams, outerfont);
2451         } else {
2452                 basefont = getLayoutFont(bparams, outerfont);
2453         }
2454
2455         // Which font is currently active?
2456         Font running_font(basefont);
2457         // Do we have an open font change?
2458         bool open_font = false;
2459
2460         Change runningChange = Change(Change::UNCHANGED);
2461
2462         Encoding const * const prev_encoding = runparams.encoding;
2463
2464         os.texrow().start(id(), 0);
2465
2466         // if the paragraph is empty, the loop will not be entered at all
2467         if (empty()) {
2468                 // For InTitle commands, we have already opened a group
2469                 // in output_latex::TeXOnePar.
2470                 if (style.isCommand() && !style.intitle) {
2471                         os << '{';
2472                         ++column;
2473                 }
2474                 if (!style.leftdelim().empty()) {
2475                         os << style.leftdelim();
2476                         column += style.leftdelim().size();
2477                 }
2478                 if (allowcust)
2479                         column += d->startTeXParParams(bparams, os, runparams);
2480         }
2481
2482         // Whether a \par can be issued for insets typeset inline with text.
2483         // Yes if greater than 0. This has to be static.
2484         THREAD_LOCAL_STATIC int parInline = 0;
2485
2486         for (pos_type i = 0; i < size(); ++i) {
2487                 // First char in paragraph or after label?
2488                 if (i == body_pos) {
2489                         if (body_pos > 0) {
2490                                 if (open_font) {
2491                                         bool needPar = false;
2492                                         column += running_font.latexWriteEndChanges(
2493                                                 os, bparams, runparams,
2494                                                 basefont, basefont, needPar);
2495                                         open_font = false;
2496                                 }
2497                                 basefont = getLayoutFont(bparams, outerfont);
2498                                 running_font = basefont;
2499
2500                                 column += Changes::latexMarkChange(os, bparams,
2501                                                 runningChange, Change(Change::UNCHANGED),
2502                                                 runparams);
2503                                 runningChange = Change(Change::UNCHANGED);
2504
2505                                 os << "}] ";
2506                                 column +=3;
2507                         }
2508                         // For InTitle commands, we have already opened a group
2509                         // in output_latex::TeXOnePar.
2510                         if (style.isCommand() && !style.intitle) {
2511                                 os << '{';
2512                                 ++column;
2513                         }
2514
2515                         if (!style.leftdelim().empty()) {
2516                                 os << style.leftdelim();
2517                                 column += style.leftdelim().size();
2518                         }
2519
2520                         if (allowcust)
2521                                 column += d->startTeXParParams(bparams, os,
2522                                                             runparams);
2523                 }
2524
2525                 runparams.wasDisplayMath = runparams.inDisplayMath;
2526                 runparams.inDisplayMath = false;
2527                 bool deleted_display_math = false;
2528                 Change const & change = runparams.inDeletedInset
2529                         ? runparams.changeOfDeletedInset : lookupChange(i);
2530
2531                 // Check whether a display math inset follows
2532                 if (d->text_[i] == META_INSET
2533                     && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2534                         InsetMath const * im = getInset(i)->asInsetMath();
2535                         if (im && im->asHullInset()
2536                             && im->asHullInset()->outerDisplay()) {
2537                                 runparams.inDisplayMath = true;
2538                                 // runparams.inDeletedInset will be set by
2539                                 // latexInset later, but we need this info
2540                                 // before it is called. On the other hand, we
2541                                 // cannot set it here because it is a counter.
2542                                 deleted_display_math = isDeleted(i);
2543                         }
2544                         if (bparams.output_changes && deleted_display_math
2545                             && runningChange == change
2546                             && change.type == Change::DELETED
2547                             && !os.afterParbreak()) {
2548                                 // A display math in the same paragraph follows.
2549                                 // We have to close and then reopen \lyxdeleted,
2550                                 // otherwise the math will be shifted up.
2551                                 OutputParams rp = runparams;
2552                                 if (open_font) {
2553                                         bool needPar = false;
2554                                         column += running_font.latexWriteEndChanges(
2555                                                 os, bparams, rp, basefont,
2556                                                 basefont, needPar);
2557                                         open_font = false;
2558                                 }
2559                                 basefont = getLayoutFont(bparams, outerfont);
2560                                 running_font = basefont;
2561                                 column += Changes::latexMarkChange(os, bparams,
2562                                         Change(Change::INSERTED), change, rp);
2563                         }
2564                 }
2565
2566                 if (bparams.output_changes && runningChange != change) {
2567                         if (open_font) {
2568                                 bool needPar = false;
2569                                 column += running_font.latexWriteEndChanges(
2570                                                 os, bparams, runparams,
2571                                                 basefont, basefont, needPar);
2572                                 open_font = false;
2573                         }
2574                         basefont = getLayoutFont(bparams, outerfont);
2575                         running_font = basefont;
2576                         column += Changes::latexMarkChange(os, bparams, runningChange,
2577                                                            change, runparams);
2578                         runningChange = change;
2579                 }
2580
2581                 // do not output text which is marked deleted
2582                 // if change tracking output is disabled
2583                 if (!bparams.output_changes && change.deleted()) {
2584                         continue;
2585                 }
2586
2587                 ++column;
2588
2589                 // Fully instantiated font
2590                 Font const current_font = getFont(bparams, i, outerfont);
2591
2592                 Font const last_font = running_font;
2593
2594                 // Do we need to close the previous font?
2595                 if (open_font &&
2596                     (current_font != running_font ||
2597                      current_font.language() != running_font.language()))
2598                 {
2599                         bool needPar = false;
2600                         column += running_font.latexWriteEndChanges(
2601                                     os, bparams, runparams, basefont,
2602                                     (i == body_pos-1) ? basefont : current_font,
2603                                     needPar);
2604                         running_font = basefont;
2605                         open_font = false;
2606                 }
2607
2608                 string const running_lang = runparams.use_polyglossia ?
2609                         running_font.language()->polyglossia() : running_font.language()->babel();
2610                 // close babel's font environment before opening CJK.
2611                 string const lang_end_command = runparams.use_polyglossia ?
2612                         "\\end{$$lang}" : lyxrc.language_command_end;
2613                 bool const using_begin_end = runparams.use_polyglossia ||
2614                                                 !lang_end_command.empty();
2615                 if (!running_lang.empty() &&
2616                     (!using_begin_end || running_lang == openLanguageName()) &&
2617                     current_font.language()->encoding()->package() == Encoding::CJK) {
2618                                 string end_tag = subst(lang_end_command,
2619                                                         "$$lang",
2620                                                         running_lang);
2621                                 os << from_ascii(end_tag);
2622                                 column += end_tag.length();
2623                                 if (using_begin_end)
2624                                         popLanguageName();
2625                 }
2626
2627                 // Switch file encoding if necessary (and allowed)
2628                 if (!runparams.pass_thru && !style.pass_thru &&
2629                     runparams.encoding->package() != Encoding::none &&
2630                     current_font.language()->encoding()->package() != Encoding::none) {
2631                         pair<bool, int> const enc_switch =
2632                                 switchEncoding(os.os(), bparams, runparams,
2633                                         *(current_font.language()->encoding()));
2634                         if (enc_switch.first) {
2635                                 column += enc_switch.second;
2636                                 runparams.encoding = current_font.language()->encoding();
2637                         }
2638                 }
2639
2640                 char_type const c = d->text_[i];
2641
2642                 // A display math inset inside an ulem command will be output
2643                 // as a box of width \linewidth, so we have to either disable
2644                 // indentation if the inset starts a paragraph, or start a new
2645                 // line to accommodate such box. This has to be done before
2646                 // writing any font changing commands.
2647                 if (runparams.inDisplayMath && !deleted_display_math
2648                     && runparams.inulemcmd) {
2649                         if (os.afterParbreak())
2650                                 os << "\\noindent";
2651                         else
2652                                 os << "\\\\\n";
2653                 }
2654
2655                 // Do we need to change font?
2656                 if ((current_font != running_font ||
2657                      current_font.language() != running_font.language()) &&
2658                         i != body_pos - 1)
2659                 {
2660                         odocstringstream ods;
2661                         column += current_font.latexWriteStartChanges(ods, bparams,
2662                                                               runparams, basefont,
2663                                                               last_font);
2664                         // Check again for display math in ulem commands as a
2665                         // font change may also occur just before a math inset.
2666                         if (runparams.inDisplayMath && !deleted_display_math
2667                             && runparams.inulemcmd) {
2668                                 if (os.afterParbreak())
2669                                         os << "\\noindent";
2670                                 else
2671                                         os << "\\\\\n";
2672                         }
2673                         running_font = current_font;
2674                         open_font = true;
2675                         docstring fontchange = ods.str();
2676                         // check whether the fontchange ends with a \\textcolor
2677                         // modifier and the text starts with a space (bug 4473)
2678                         docstring const last_modifier = rsplit(fontchange, '\\');
2679                         if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2680                                 os << fontchange << from_ascii("{}");
2681                         // check if the fontchange ends with a trailing blank
2682                         // (like "\small " (see bug 3382)
2683                         else if (suffixIs(fontchange, ' ') && c == ' ')
2684                                 os << fontchange.substr(0, fontchange.size() - 1)
2685                                    << from_ascii("{}");
2686                         else
2687                                 os << fontchange;
2688                 }
2689
2690                 // FIXME: think about end_pos implementation...
2691                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2692                         // FIXME: integrate this case in latexSpecialChar
2693                         // Do not print the separation of the optional argument
2694                         // if style.pass_thru is false. This works because
2695                         // latexSpecialChar ignores spaces if
2696                         // style.pass_thru is false.
2697                         if (i != body_pos - 1) {
2698                                 if (d->simpleTeXBlanks(bparams, runparams, os,
2699                                                 i, column, current_font, style)) {
2700                                         // A surrogate pair was output. We
2701                                         // must not call latexSpecialChar
2702                                         // in this iteration, since it would output
2703                                         // the combining character again.
2704                                         ++i;
2705                                         continue;
2706                                 }
2707                         }
2708                 }
2709
2710                 OutputParams rp = runparams;
2711                 rp.free_spacing = style.free_spacing;
2712                 rp.local_font = &current_font;
2713                 rp.intitle = style.intitle;
2714
2715                 // Two major modes:  LaTeX or plain
2716                 // Handle here those cases common to both modes
2717                 // and then split to handle the two modes separately.
2718                 if (c == META_INSET) {
2719                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2720                                 // Greyedout notes and, in general, all insets
2721                                 // with InsetLayout::isDisplay() == false,
2722                                 // are typeset inline with the text. So, we
2723                                 // can add a \par to the last paragraph of
2724                                 // such insets only if nothing else follows.
2725                                 bool incremented = false;
2726                                 Inset const * inset = getInset(i);
2727                                 InsetText const * textinset = inset
2728                                                         ? inset->asInsetText()
2729                                                         : 0;
2730                                 if (i + 1 == size() && textinset
2731                                     && !inset->getLayout().isDisplay()) {
2732                                         ParagraphList const & pars =
2733                                                 textinset->text().paragraphs();
2734                                         pit_type const pit = pars.size() - 1;
2735                                         Font const lastfont =
2736                                                 pit < 0 || pars[pit].empty()
2737                                                 ? pars[pit].getLayoutFont(
2738                                                                 bparams,
2739                                                                 outerfont)
2740                                                 : pars[pit].getFont(bparams,
2741                                                         pars[pit].size() - 1,
2742                                                         outerfont);
2743                                         if (lastfont.fontInfo().size() !=
2744                                             basefont.fontInfo().size()) {
2745                                                 ++parInline;
2746                                                 incremented = true;
2747                                         }
2748                                 }
2749                                 d->latexInset(bparams, os, rp, running_font,
2750                                                 basefont, outerfont, open_font,
2751                                                 runningChange, style, i, column);
2752                                 if (incremented)
2753                                         --parInline;
2754                         }
2755                 } else if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2756                         try {
2757                                 d->latexSpecialChar(os, bparams, rp,
2758                                                     running_font, runningChange,
2759                                                     style, i, end_pos, column);
2760                         } catch (EncodingException & e) {
2761                                 if (runparams.dryrun) {
2762                                         os << "<" << _("LyX Warning: ")
2763                                            << _("uncodable character") << " '";
2764                                         os.put(c);
2765                                         os << "'>";
2766                                 } else {
2767                                         // add location information and throw again.
2768                                         e.par_id = id();
2769                                         e.pos = i;
2770                                         throw(e);
2771                                 }
2772                         }
2773                 }
2774
2775                 // Set the encoding to that returned from latexSpecialChar (see
2776                 // comment for encoding member in OutputParams.h)
2777                 runparams.encoding = rp.encoding;
2778
2779                 // Also carry on the info on a closed ulem command for insets
2780                 // such as Note that do not produce any output, so that no
2781                 // command is ever executed but its opening was recorded.
2782                 runparams.inulemcmd = rp.inulemcmd;
2783         }
2784
2785         // If we have an open font definition, we have to close it
2786         if (open_font) {
2787                 // Make sure that \\par is done with the font of the last
2788                 // character if this has another size as the default.
2789                 // This is necessary because LaTeX (and LyX on the screen)
2790                 // calculates the space between the baselines according
2791                 // to this font. (Matthias)
2792                 //
2793                 // We must not change the font for the last paragraph
2794                 // of non-multipar insets, tabular cells or commands,
2795                 // since this produces unwanted whitespace.
2796
2797                 Font const font = empty()
2798                         ? getLayoutFont(bparams, outerfont)
2799                         : getFont(bparams, size() - 1, outerfont);
2800
2801                 InsetText const * textinset = inInset().asInsetText();
2802
2803                 bool const maintext = textinset
2804                         ? textinset->text().isMainText()
2805                         : false;
2806
2807                 size_t const numpars = textinset
2808                         ? textinset->text().paragraphs().size()
2809                         : 0;
2810
2811                 bool needPar = false;
2812
2813                 if (style.resfont.size() != font.fontInfo().size()
2814                     && (!runparams.isLastPar || maintext
2815                         || (numpars > 1 && d->ownerCode() != CELL_CODE
2816                             && (inInset().getLayout().isDisplay()
2817                                 || parInline)))
2818                     && !style.isCommand()) {
2819                         needPar = true;
2820                 }
2821 #ifdef FIXED_LANGUAGE_END_DETECTION
2822                 if (next_) {
2823                         running_font.latexWriteEndChanges(os, bparams,
2824                                         runparams, basefont,
2825                                         next_->getFont(bparams, 0, outerfont),
2826                                                        needPar);
2827                 } else {
2828                         running_font.latexWriteEndChanges(os, bparams,
2829                                         runparams, basefont, basefont, needPar);
2830                 }
2831 #else
2832 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2833 //FIXME: there as we start another \selectlanguage with the next paragraph if
2834 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2835                 running_font.latexWriteEndChanges(os, bparams, runparams,
2836                                 basefont, basefont, needPar);
2837 #endif
2838                 if (needPar) {
2839                         // The \par could not be inserted at the same nesting
2840                         // level of the font size change, so do it now.
2841                         os << "{\\" << font.latexSize() << "\\par}";
2842                 }
2843         }
2844
2845         column += Changes::latexMarkChange(os, bparams, runningChange,
2846                                            Change(Change::UNCHANGED), runparams);
2847
2848         // Needed if there is an optional argument but no contents.
2849         if (body_pos > 0 && body_pos == size()) {
2850                 os << "}]~";
2851         }
2852
2853         if (!style.rightdelim().empty()) {
2854                 os << style.rightdelim();
2855                 column += style.rightdelim().size();
2856         }
2857
2858         if (allowcust && d->endTeXParParams(bparams, os, runparams)
2859             && runparams.encoding != prev_encoding) {
2860                 runparams.encoding = prev_encoding;
2861                 os << setEncoding(prev_encoding->iconvName());
2862         }
2863
2864         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2865 }
2866
2867
2868 bool Paragraph::emptyTag() const
2869 {
2870         for (pos_type i = 0; i < size(); ++i) {
2871                 if (Inset const * inset = getInset(i)) {
2872                         InsetCode lyx_code = inset->lyxCode();
2873                         // FIXME testing like that is wrong. What is
2874                         // the intent?
2875                         if (lyx_code != TOC_CODE &&
2876                             lyx_code != INCLUDE_CODE &&
2877                             lyx_code != GRAPHICS_CODE &&
2878                             lyx_code != ERT_CODE &&
2879                             lyx_code != LISTINGS_CODE &&
2880                             lyx_code != FLOAT_CODE &&
2881                             lyx_code != TABULAR_CODE) {
2882                                 return false;
2883                         }
2884                 } else {
2885                         char_type c = d->text_[i];
2886                         if (c != ' ' && c != '\t')
2887                                 return false;
2888                 }
2889         }
2890         return true;
2891 }
2892
2893
2894 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
2895         const
2896 {
2897         for (pos_type i = 0; i < size(); ++i) {
2898                 if (Inset const * inset = getInset(i)) {
2899                         InsetCode lyx_code = inset->lyxCode();
2900                         if (lyx_code == LABEL_CODE) {
2901                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2902                                 docstring const & id = il->getParam("name");
2903                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2904                         }
2905                 }
2906         }
2907         return string();
2908 }
2909
2910
2911 pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runparams)
2912         const
2913 {
2914         pos_type i;
2915         for (i = 0; i < size(); ++i) {
2916                 if (Inset const * inset = getInset(i)) {
2917                         inset->docbook(os, runparams);
2918                 } else {
2919                         char_type c = d->text_[i];
2920                         if (c == ' ')
2921                                 break;
2922                         os << sgml::escapeChar(c);
2923                 }
2924         }
2925         return i;
2926 }
2927
2928
2929 pos_type Paragraph::firstWordLyXHTML(XHTMLStream & xs, OutputParams const & runparams)
2930         const
2931 {
2932         pos_type i;
2933         for (i = 0; i < size(); ++i) {
2934                 if (Inset const * inset = getInset(i)) {
2935                         inset->xhtml(xs, runparams);
2936                 } else {
2937                         char_type c = d->text_[i];
2938                         if (c == ' ')
2939                                 break;
2940                         xs << c;
2941                 }
2942         }
2943         return i;
2944 }
2945
2946
2947 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2948 {
2949         Font font_old;
2950         pos_type size = text_.size();
2951         for (pos_type i = initial; i < size; ++i) {
2952                 Font font = owner_->getFont(buf.params(), i, outerfont);
2953                 if (text_[i] == META_INSET)
2954                         return false;
2955                 if (i != initial && font != font_old)
2956                         return false;
2957                 font_old = font;
2958         }
2959
2960         return true;
2961 }
2962
2963
2964 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
2965                                     odocstream & os,
2966                                     OutputParams const & runparams,
2967                                     Font const & outerfont,
2968                                     pos_type initial) const
2969 {
2970         bool emph_flag = false;
2971
2972         Layout const & style = *d->layout_;
2973         FontInfo font_old =
2974                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2975
2976         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2977                 os << "]]>";
2978
2979         // parsing main loop
2980         for (pos_type i = initial; i < size(); ++i) {
2981                 Font font = getFont(buf.params(), i, outerfont);
2982
2983                 // handle <emphasis> tag
2984                 if (font_old.emph() != font.fontInfo().emph()) {
2985                         if (font.fontInfo().emph() == FONT_ON) {
2986                                 os << "<emphasis>";
2987                                 emph_flag = true;
2988                         } else if (i != initial) {
2989                                 os << "</emphasis>";
2990                                 emph_flag = false;
2991                         }
2992                 }
2993
2994                 if (Inset const * inset = getInset(i)) {
2995                         inset->docbook(os, runparams);
2996                 } else {
2997                         char_type c = d->text_[i];
2998
2999                         if (style.pass_thru)
3000                                 os.put(c);
3001                         else
3002                                 os << sgml::escapeChar(c);
3003                 }
3004                 font_old = font.fontInfo();
3005         }
3006
3007         if (emph_flag) {
3008                 os << "</emphasis>";
3009         }
3010
3011         if (style.free_spacing)
3012                 os << '\n';
3013         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
3014                 os << "<![CDATA[";
3015 }
3016
3017
3018 namespace {
3019 void doFontSwitch(vector<html::FontTag> & tagsToOpen,
3020                   vector<html::EndFontTag> & tagsToClose,
3021                   bool & flag, FontState curstate, html::FontTypes type)
3022 {
3023         if (curstate == FONT_ON) {
3024                 tagsToOpen.push_back(html::FontTag(type));
3025                 flag = true;
3026         } else if (flag) {
3027                 tagsToClose.push_back(html::EndFontTag(type));
3028                 flag = false;
3029         }
3030 }
3031 } // namespace
3032
3033
3034 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
3035                                     XHTMLStream & xs,
3036                                     OutputParams const & runparams,
3037                                     Font const & outerfont,
3038                                     bool start_paragraph, bool close_paragraph,
3039                                     pos_type initial) const
3040 {
3041         docstring retval;
3042
3043         // track whether we have opened these tags
3044         bool emph_flag = false;
3045         bool bold_flag = false;
3046         bool noun_flag = false;
3047         bool ubar_flag = false;
3048         bool dbar_flag = false;
3049         bool sout_flag = false;
3050         bool xout_flag = false;
3051         bool wave_flag = false;
3052         // shape tags
3053         bool shap_flag = false;
3054         // family tags
3055         bool faml_flag = false;
3056         // size tags
3057         bool size_flag = false;
3058
3059         Layout const & style = *d->layout_;
3060
3061         if (start_paragraph)
3062                 xs.startDivision(allowEmpty());
3063
3064         FontInfo font_old =
3065                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
3066
3067         FontShape  curr_fs   = INHERIT_SHAPE;
3068         FontFamily curr_fam  = INHERIT_FAMILY;
3069         FontSize   curr_size = FONT_SIZE_INHERIT;
3070
3071         string const default_family =
3072                 buf.masterBuffer()->params().fonts_default_family;
3073
3074         vector<html::FontTag> tagsToOpen;
3075         vector<html::EndFontTag> tagsToClose;
3076
3077         // parsing main loop
3078         for (pos_type i = initial; i < size(); ++i) {
3079                 // let's not show deleted material in the output
3080                 if (isDeleted(i))
3081                         continue;
3082
3083                 Font const font = getFont(buf.masterBuffer()->params(), i, outerfont);
3084
3085                 // emphasis
3086                 FontState curstate = font.fontInfo().emph();
3087                 if (font_old.emph() != curstate)
3088                         doFontSwitch(tagsToOpen, tagsToClose, emph_flag, curstate, html::FT_EMPH);
3089
3090                 // noun
3091                 curstate = font.fontInfo().noun();
3092                 if (font_old.noun() != curstate)
3093                         doFontSwitch(tagsToOpen, tagsToClose, noun_flag, curstate, html::FT_NOUN);
3094
3095                 // underbar
3096                 curstate = font.fontInfo().underbar();
3097                 if (font_old.underbar() != curstate)
3098                         doFontSwitch(tagsToOpen, tagsToClose, ubar_flag, curstate, html::FT_UBAR);
3099
3100                 // strikeout
3101                 curstate = font.fontInfo().strikeout();
3102                 if (font_old.strikeout() != curstate)
3103                         doFontSwitch(tagsToOpen, tagsToClose, sout_flag, curstate, html::FT_SOUT);
3104
3105                 // xout
3106                 curstate = font.fontInfo().xout();
3107                 if (font_old.xout() != curstate)
3108                         doFontSwitch(tagsToOpen, tagsToClose, xout_flag, curstate, html::FT_XOUT);
3109
3110                 // double underbar
3111                 curstate = font.fontInfo().uuline();
3112                 if (font_old.uuline() != curstate)
3113                         doFontSwitch(tagsToOpen, tagsToClose, dbar_flag, curstate, html::FT_DBAR);
3114
3115                 // wavy line
3116                 curstate = font.fontInfo().uwave();
3117                 if (font_old.uwave() != curstate)
3118                         doFontSwitch(tagsToOpen, tagsToClose, wave_flag, curstate, html::FT_WAVE);
3119
3120                 // bold
3121                 // a little hackish, but allows us to reuse what we have.
3122                 curstate = (font.fontInfo().series() == BOLD_SERIES ? FONT_ON : FONT_OFF);
3123                 if (font_old.series() != font.fontInfo().series())
3124                         doFontSwitch(tagsToOpen, tagsToClose, bold_flag, curstate, html::FT_BOLD);
3125
3126                 // Font shape
3127                 curr_fs = font.fontInfo().shape();
3128                 FontShape old_fs = font_old.shape();
3129                 if (old_fs != curr_fs) {
3130                         if (shap_flag) {
3131                                 switch (old_fs) {
3132                                 case ITALIC_SHAPE:
3133                                         tagsToClose.push_back(html::EndFontTag(html::FT_ITALIC));
3134                                         break;
3135                                 case SLANTED_SHAPE:
3136                                         tagsToClose.push_back(html::EndFontTag(html::FT_SLANTED));
3137                                         break;
3138                                 case SMALLCAPS_SHAPE:
3139                                         tagsToClose.push_back(html::EndFontTag(html::FT_SMALLCAPS));
3140                                         break;
3141                                 case UP_SHAPE:
3142                                 case INHERIT_SHAPE:
3143                                         break;
3144                                 default:
3145                                         // the other tags are for internal use
3146                                         LATTEST(false);
3147                                         break;
3148                                 }
3149                                 shap_flag = false;
3150                         }
3151                         switch (curr_fs) {
3152                         case ITALIC_SHAPE:
3153                                 tagsToOpen.push_back(html::FontTag(html::FT_ITALIC));
3154                                 shap_flag = true;
3155                                 break;
3156                         case SLANTED_SHAPE:
3157                                 tagsToOpen.push_back(html::FontTag(html::FT_SLANTED));
3158                                 shap_flag = true;
3159                                 break;
3160                         case SMALLCAPS_SHAPE:
3161                                 tagsToOpen.push_back(html::FontTag(html::FT_SMALLCAPS));
3162                                 shap_flag = true;
3163                                 break;
3164                         case UP_SHAPE:
3165                         case INHERIT_SHAPE:
3166                                 break;
3167                         default:
3168                                 // the other tags are for internal use
3169                                 LATTEST(false);
3170                                 break;
3171                         }
3172                 }
3173
3174                 // Font family
3175                 curr_fam = font.fontInfo().family();
3176                 FontFamily old_fam = font_old.family();
3177                 if (old_fam != curr_fam) {
3178                         if (faml_flag) {
3179                                 switch (old_fam) {
3180                                 case ROMAN_FAMILY:
3181                                         tagsToClose.push_back(html::EndFontTag(html::FT_ROMAN));
3182                                         break;
3183                                 case SANS_FAMILY:
3184                                         tagsToClose.push_back(html::EndFontTag(html::FT_SANS));
3185                                         break;
3186                                 case TYPEWRITER_FAMILY:
3187                                         tagsToClose.push_back(html::EndFontTag(html::FT_TYPE));
3188                                         break;
3189                                 case INHERIT_FAMILY:
3190                                         break;
3191                                 default:
3192                                         // the other tags are for internal use
3193                                         LATTEST(false);
3194                                         break;
3195                                 }
3196                                 faml_flag = false;
3197                         }
3198                         switch (curr_fam) {
3199                         case ROMAN_FAMILY:
3200                                 // we will treat a "default" font family as roman, since we have
3201                                 // no other idea what to do.
3202                                 if (default_family != "rmdefault" && default_family != "default") {
3203                                         tagsToOpen.push_back(html::FontTag(html::FT_ROMAN));
3204                                         faml_flag = true;
3205                                 }
3206                                 break;
3207                         case SANS_FAMILY:
3208                                 if (default_family != "sfdefault") {
3209                                         tagsToOpen.push_back(html::FontTag(html::FT_SANS));
3210                                         faml_flag = true;
3211                                 }
3212                                 break;
3213                         case TYPEWRITER_FAMILY:
3214                                 if (default_family != "ttdefault") {
3215                                         tagsToOpen.push_back(html::FontTag(html::FT_TYPE));
3216                                         faml_flag = true;
3217                                 }
3218                                 break;
3219                         case INHERIT_FAMILY:
3220                                 break;
3221                         default:
3222                                 // the other tags are for internal use
3223                                 LATTEST(false);
3224                                 break;
3225                         }
3226                 }
3227
3228                 // Font size
3229                 curr_size = font.fontInfo().size();
3230                 FontSize old_size = font_old.size();
3231                 if (old_size != curr_size) {
3232                         if (size_flag) {
3233                                 switch (old_size) {
3234                                 case FONT_SIZE_TINY:
3235                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_TINY));
3236                                         break;
3237                                 case FONT_SIZE_SCRIPT:
3238                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SCRIPT));
3239                                         break;
3240                                 case FONT_SIZE_FOOTNOTE:
3241                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_FOOTNOTE));
3242                                         break;
3243                                 case FONT_SIZE_SMALL:
3244                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SMALL));
3245                                         break;
3246                                 case FONT_SIZE_LARGE:
3247                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGE));
3248                                         break;
3249                                 case FONT_SIZE_LARGER:
3250                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGER));
3251                                         break;
3252                                 case FONT_SIZE_LARGEST:
3253                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGEST));
3254                                         break;
3255                                 case FONT_SIZE_HUGE:
3256                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGE));
3257                                         break;
3258                                 case FONT_SIZE_HUGER:
3259                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGER));
3260                                         break;
3261                                 case FONT_SIZE_INCREASE:
3262                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_INCREASE));
3263                                         break;
3264                                 case FONT_SIZE_DECREASE:
3265                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_DECREASE));
3266                                         break;
3267                                 case FONT_SIZE_INHERIT:
3268                                 case FONT_SIZE_NORMAL:
3269                                         break;
3270                                 default:
3271                                         // the other tags are for internal use
3272                                         LATTEST(false);
3273                                         break;
3274                                 }
3275                                 size_flag = false;
3276                         }
3277                         switch (curr_size) {
3278                         case FONT_SIZE_TINY:
3279                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_TINY));
3280                                 size_flag = true;
3281                                 break;
3282                         case FONT_SIZE_SCRIPT:
3283                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SCRIPT));
3284                                 size_flag = true;
3285                                 break;
3286                         case FONT_SIZE_FOOTNOTE:
3287                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_FOOTNOTE));
3288                                 size_flag = true;
3289                                 break;
3290                         case FONT_SIZE_SMALL:
3291                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SMALL));
3292                                 size_flag = true;
3293                                 break;
3294                         case FONT_SIZE_LARGE:
3295                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGE));
3296                                 size_flag = true;
3297                                 break;
3298                         case FONT_SIZE_LARGER:
3299                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGER));
3300                                 size_flag = true;
3301                                 break;
3302                         case FONT_SIZE_LARGEST:
3303                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGEST));
3304                                 size_flag = true;
3305                                 break;
3306                         case FONT_SIZE_HUGE:
3307                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGE));
3308                                 size_flag = true;
3309                                 break;
3310                         case FONT_SIZE_HUGER:
3311                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGER));
3312                                 size_flag = true;
3313                                 break;
3314                         case FONT_SIZE_INCREASE:
3315                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_INCREASE));
3316                                 size_flag = true;
3317                                 break;
3318                         case FONT_SIZE_DECREASE:
3319                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_DECREASE));
3320                                 size_flag = true;
3321                                 break;
3322                         case FONT_SIZE_NORMAL:
3323                         case FONT_SIZE_INHERIT:
3324                                 break;
3325                         default:
3326                                 // the other tags are for internal use
3327                                 LATTEST(false);
3328                                 break;
3329                         }
3330                 }
3331
3332                 // FIXME XHTML
3333                 // Other such tags? What about the other text ranges?
3334
3335                 vector<html::EndFontTag>::const_iterator cit = tagsToClose.begin();
3336                 vector<html::EndFontTag>::const_iterator cen = tagsToClose.end();
3337                 for (; cit != cen; ++cit)
3338                         xs << *cit;
3339
3340                 vector<html::FontTag>::const_iterator sit = tagsToOpen.begin();
3341                 vector<html::FontTag>::const_iterator sen = tagsToOpen.end();
3342                 for (; sit != sen; ++sit)
3343                         xs << *sit;
3344
3345                 tagsToClose.clear();
3346                 tagsToOpen.clear();
3347
3348                 Inset const * inset = getInset(i);
3349                 if (inset) {
3350                         if (!runparams.for_toc || inset->isInToc()) {
3351                                 OutputParams np = runparams;
3352                                 np.local_font = &font;
3353                                 // If the paragraph has size 1, then we are in the "special
3354                                 // case" where we do not output the containing paragraph info
3355                                 if (!inset->getLayout().htmlisblock() && size() != 1)
3356                                         np.html_in_par = true;
3357                                 retval += inset->xhtml(xs, np);
3358                         }
3359                 } else {
3360                         char_type c = getUChar(buf.masterBuffer()->params(), i);
3361                         xs << c;
3362                 }
3363                 font_old = font.fontInfo();
3364         }
3365
3366         // FIXME XHTML
3367         // I'm worried about what happens if a branch, say, is itself
3368         // wrapped in some font stuff. I think that will not work.
3369         xs.closeFontTags();
3370         if (close_paragraph)
3371                 xs.endDivision();
3372
3373         return retval;
3374 }
3375
3376
3377 bool Paragraph::isHfill(pos_type pos) const
3378 {
3379         Inset const * inset = getInset(pos);
3380         return inset && inset->isHfill();
3381 }
3382
3383
3384 bool Paragraph::isNewline(pos_type pos) const
3385 {
3386         // U+2028 LINE SEPARATOR
3387         // U+2029 PARAGRAPH SEPARATOR
3388         char_type const c = d->text_[pos];
3389         if (c == 0x2028 || c == 0x2029)
3390                 return true;
3391         Inset const * inset = getInset(pos);
3392         return inset && inset->lyxCode() == NEWLINE_CODE;
3393 }
3394
3395
3396 bool Paragraph::isEnvSeparator(pos_type pos) const
3397 {
3398         Inset const * inset = getInset(pos);
3399         return inset && inset->lyxCode() == SEPARATOR_CODE;
3400 }
3401
3402
3403 bool Paragraph::isLineSeparator(pos_type pos) const
3404 {
3405         char_type const c = d->text_[pos];
3406         if (isLineSeparatorChar(c))
3407                 return true;
3408         Inset const * inset = getInset(pos);
3409         return inset && inset->isLineSeparator();
3410 }
3411
3412
3413 bool Paragraph::isWordSeparator(pos_type pos) const
3414 {
3415         if (pos == size())
3416                 return true;
3417         if (Inset const * inset = getInset(pos))
3418                 return !inset->isLetter();
3419         // if we have a hard hyphen (no en- or emdash) or apostrophe
3420         // we pass this to the spell checker
3421         // FIXME: this method is subject to change, visit
3422         // https://bugzilla.mozilla.org/show_bug.cgi?id=355178
3423         // to get an impression how complex this is.
3424         if (isHardHyphenOrApostrophe(pos))
3425                 return false;
3426         char_type const c = d->text_[pos];
3427         // We want to pass the escape chars to the spellchecker
3428         docstring const escape_chars = from_utf8(lyxrc.spellchecker_esc_chars);
3429         return !isLetterChar(c) && !isDigitASCII(c) && !contains(escape_chars, c);
3430 }
3431
3432
3433 bool Paragraph::isHardHyphenOrApostrophe(pos_type pos) const
3434 {
3435         pos_type const psize = size();
3436         if (pos >= psize)
3437                 return false;
3438         char_type const c = d->text_[pos];
3439         if (c != '-' && c != '\'')
3440                 return false;
3441         int nextpos = pos + 1;
3442         int prevpos = pos > 0 ? pos - 1 : 0;
3443         if ((nextpos == psize || isSpace(nextpos))
3444                 && (pos == 0 || isSpace(prevpos)))
3445                 return false;
3446         return true;
3447 }
3448
3449
3450 bool Paragraph::needsCProtection() const
3451 {
3452         // first check the layout of the paragraph, but only in insets
3453         InsetText const * textinset = inInset().asInsetText();
3454         bool const maintext = textinset
3455                 ? textinset->text().isMainText()
3456                 : false;
3457
3458         if (!maintext && layout().needcprotect) {
3459                 // Environments need cprotection regardless the content
3460                 if (layout().latextype == LATEX_ENVIRONMENT)
3461                         return true;
3462
3463                 // Commands need cprotection if they contain specific chars
3464                 int const nchars_escape = 9;
3465                 static char_type const chars_escape[nchars_escape] = {
3466                         '&', '_', '$', '%', '#', '^', '{', '}', '\\'};
3467
3468                 docstring const pars = asString();
3469                 for (int k = 0; k < nchars_escape; k++) {
3470                         if (contains(pars, chars_escape[k]))
3471                                 return true;
3472                 }
3473         }
3474
3475         // now check whether we have insets that need cprotection
3476         pos_type size = d->text_.size();
3477         for (pos_type i = 0; i < size; ++i)
3478                 if (isInset(i) && getInset(i)->needsCProtection(maintext))
3479                         return true;
3480
3481         return false;
3482 }
3483
3484
3485 FontSpan const & Paragraph::getSpellRange(pos_type pos) const
3486 {
3487         return d->speller_state_.getRange(pos);
3488 }
3489
3490
3491 bool Paragraph::isChar(pos_type pos) const
3492 {
3493         if (Inset const * inset = getInset(pos))
3494                 return inset->isChar();
3495         char_type const c = d->text_[pos];
3496         return !isLetterChar(c) && !isDigitASCII(c) && !lyx::isSpace(c);
3497 }
3498
3499
3500 bool Paragraph::isSpace(pos_type pos) const
3501 {
3502         if (Inset const * inset = getInset(pos))
3503                 return inset->isSpace();
3504         char_type const c = d->text_[pos];
3505         return lyx::isSpace(c);
3506 }
3507
3508
3509 Language const *
3510 Paragraph::getParLanguage(BufferParams const & bparams) const
3511 {
3512         if (!empty())
3513                 return getFirstFontSettings(bparams).language();
3514         // FIXME: we should check the prev par as well (Lgb)
3515         return bparams.language;
3516 }
3517
3518
3519 bool Paragraph::isRTL(BufferParams const & bparams) const
3520 {
3521         return getParLanguage(bparams)->rightToLeft()
3522                 && !inInset().getLayout().forceLTR();
3523 }
3524
3525
3526 void Paragraph::changeLanguage(BufferParams const & bparams,
3527                                Language const * from, Language const * to)
3528 {
3529         // change language including dummy font change at the end
3530         for (pos_type i = 0; i <= size(); ++i) {
3531                 Font font = getFontSettings(bparams, i);
3532                 if (font.language() == from) {
3533                         font.setLanguage(to);
3534                         setFont(i, font);
3535                         d->requestSpellCheck(i);
3536                 }
3537         }
3538 }
3539
3540
3541 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
3542 {
3543         Language const * doc_language = bparams.language;
3544         for (auto const & f : d->fontlist_)
3545                 if (f.font().language() != ignore_language &&
3546                     f.font().language() != latex_language &&
3547                     f.font().language() != doc_language)
3548                         return true;
3549         return false;
3550 }
3551
3552
3553 void Paragraph::getLanguages(std::set<Language const *> & langs) const
3554 {
3555         for (auto const & f : d->fontlist_) {
3556                 Language const * lang = f.font().language();
3557                 if (lang != ignore_language &&
3558                     lang != latex_language)
3559                         langs.insert(lang);
3560         }
3561 }
3562
3563
3564 docstring Paragraph::asString(int options) const
3565 {
3566         return asString(0, size(), options);
3567 }
3568
3569
3570 docstring Paragraph::asString(pos_type beg, pos_type end, int options, const OutputParams *runparams) const
3571 {
3572         odocstringstream os;
3573
3574         if (beg == 0
3575             && options & AS_STR_LABEL
3576             && !d->params_.labelString().empty())
3577                 os << d->params_.labelString() << ' ';
3578
3579         for (pos_type i = beg; i < end; ++i) {
3580                 if ((options & AS_STR_SKIPDELETE) && isDeleted(i))
3581                         continue;
3582                 char_type const c = d->text_[i];
3583                 if (isPrintable(c) || c == '\t'
3584                     || (c == '\n' && (options & AS_STR_NEWLINES)))
3585                         os.put(c);
3586                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
3587                         if (c == META_INSET && (options & AS_STR_PLAINTEXT)) {
3588                                 LASSERT(runparams != 0, return docstring());
3589                                 getInset(i)->plaintext(os, *runparams);
3590                         } else {
3591                                 getInset(i)->toString(os);
3592                         }
3593                 }
3594         }
3595
3596         return os.str();
3597 }
3598
3599
3600 void Paragraph::forOutliner(docstring & os, size_t const maxlen,
3601                             bool const shorten, bool const label) const
3602 {
3603         size_t tmplen = shorten ? maxlen + 1 : maxlen;
3604         if (label && !labelString().empty())
3605                 os += labelString() + ' ';
3606         if (!layout().isTocCaption())
3607                 return;
3608         for (pos_type i = 0; i < size() && os.length() < tmplen; ++i) {
3609                 if (isDeleted(i))
3610                         continue;
3611                 char_type const c = d->text_[i];
3612                 if (isPrintable(c))
3613                         os += c;
3614                 else if (c == META_INSET)
3615                         getInset(i)->forOutliner(os, tmplen, false);
3616         }
3617         if (shorten)
3618                 Text::shortenForOutliner(os, maxlen);
3619 }
3620
3621
3622 void Paragraph::setInsetOwner(Inset const * inset)
3623 {
3624         d->inset_owner_ = inset;
3625 }
3626
3627
3628 int Paragraph::id() const
3629 {
3630         return d->id_;
3631 }
3632
3633
3634 void Paragraph::setId(int id)
3635 {
3636         d->id_ = id;
3637 }
3638
3639
3640 Layout const & Paragraph::layout() const
3641 {
3642         return *d->layout_;
3643 }
3644
3645
3646 void Paragraph::setLayout(Layout const & layout)
3647 {
3648         d->layout_ = &layout;
3649 }
3650
3651
3652 void Paragraph::setDefaultLayout(DocumentClass const & tc)
3653 {
3654         setLayout(tc.defaultLayout());
3655 }
3656
3657
3658 void Paragraph::setPlainLayout(DocumentClass const & tc)
3659 {
3660         setLayout(tc.plainLayout());
3661 }
3662
3663
3664 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
3665 {
3666         if (usePlainLayout())
3667                 setPlainLayout(tclass);
3668         else
3669                 setDefaultLayout(tclass);
3670 }
3671
3672
3673 Inset const & Paragraph::inInset() const
3674 {
3675         LBUFERR(d->inset_owner_);
3676         return *d->inset_owner_;
3677 }
3678
3679
3680 ParagraphParameters & Paragraph::params()
3681 {
3682         return d->params_;
3683 }
3684
3685
3686 ParagraphParameters const & Paragraph::params() const
3687 {
3688         return d->params_;
3689 }
3690
3691
3692 bool Paragraph::isFreeSpacing() const
3693 {
3694         if (d->layout_->free_spacing)
3695                 return true;
3696         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
3697 }
3698
3699
3700 bool Paragraph::allowEmpty() const
3701 {
3702         if (d->layout_->keepempty)
3703                 return true;
3704         return d->inset_owner_ && d->inset_owner_->allowEmpty();
3705 }
3706
3707
3708 bool Paragraph::brokenBiblio() const
3709 {
3710         // there is a problem if there is no bibitem at position 0 or
3711         // if there is another bibitem in the paragraph.
3712         return d->layout_->labeltype == LABEL_BIBLIO
3713                 && (d->insetlist_.find(BIBITEM_CODE) != 0
3714                     || d->insetlist_.find(BIBITEM_CODE, 1) > 0);
3715 }
3716
3717
3718 int Paragraph::fixBiblio(Buffer const & buffer)
3719 {
3720         // FIXME: What about the case where paragraph is not BIBLIO
3721         // but there is an InsetBibitem?
3722         // FIXME: when there was already an inset at 0, the return value is 1,
3723         // which does not tell whether another inset has been remove; the
3724         // cursor cannot be correctly updated.
3725
3726         if (d->layout_->labeltype != LABEL_BIBLIO)
3727                 return 0;
3728
3729         bool const track_changes = buffer.params().track_changes;
3730         int bibitem_pos = d->insetlist_.find(BIBITEM_CODE);
3731         bool const hasbibitem0 = bibitem_pos == 0;
3732
3733         if (hasbibitem0) {
3734                 bibitem_pos = d->insetlist_.find(BIBITEM_CODE, 1);
3735                 // There was an InsetBibitem at pos 0, and no other one => OK
3736                 if (bibitem_pos == -1)
3737                         return 0;
3738                 // there is a bibitem at the 0 position, but since
3739                 // there is a second one, we copy the second on the
3740                 // first. We're assuming there are at most two of
3741                 // these, which there should be.
3742                 // FIXME: why does it make sense to do that rather
3743                 // than keep the first? (JMarc)
3744                 Inset * inset = releaseInset(bibitem_pos);
3745                 d->insetlist_.begin()->inset = inset;
3746                 return -bibitem_pos;
3747         }
3748
3749         // We need to create an inset at the beginning
3750         Inset * inset = 0;
3751         if (bibitem_pos > 0) {
3752                 // there was one somewhere in the paragraph, let's move it
3753                 inset = d->insetlist_.release(bibitem_pos);
3754                 eraseChar(bibitem_pos, track_changes);
3755         } else
3756                 // make a fresh one
3757                 inset = new InsetBibitem(const_cast<Buffer *>(&buffer),
3758                                          InsetCommandParams(BIBITEM_CODE));
3759
3760         Font font(inherit_font, buffer.params().language);
3761         insertInset(0, inset, font, Change(track_changes ? Change::INSERTED
3762                                                    : Change::UNCHANGED));
3763
3764         return 1;
3765 }
3766
3767
3768 void Paragraph::checkAuthors(AuthorList const & authorList)
3769 {
3770         d->changes_.checkAuthors(authorList);
3771 }
3772
3773
3774 bool Paragraph::isChanged(pos_type pos) const
3775 {
3776         return lookupChange(pos).changed();
3777 }
3778
3779
3780 bool Paragraph::isInserted(pos_type pos) const
3781 {
3782         return lookupChange(pos).inserted();
3783 }
3784
3785
3786 bool Paragraph::isDeleted(pos_type pos) const
3787 {
3788         return lookupChange(pos).deleted();
3789 }
3790
3791
3792 InsetList const & Paragraph::insetList() const
3793 {
3794         return d->insetlist_;
3795 }
3796
3797
3798 void Paragraph::setInsetBuffers(Buffer & b)
3799 {
3800         d->insetlist_.setBuffer(b);
3801 }
3802
3803
3804 void Paragraph::resetBuffer()
3805 {
3806         d->insetlist_.resetBuffer();
3807 }
3808
3809
3810 Inset * Paragraph::releaseInset(pos_type pos)
3811 {
3812         Inset * inset = d->insetlist_.release(pos);
3813         /// does not honour change tracking!
3814         eraseChar(pos, false);
3815         return inset;
3816 }
3817
3818
3819 Inset * Paragraph::getInset(pos_type pos)
3820 {
3821         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3822                  ? d->insetlist_.get(pos) : 0;
3823 }
3824
3825
3826 Inset const * Paragraph::getInset(pos_type pos) const
3827 {
3828         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3829                  ? d->insetlist_.get(pos) : 0;
3830 }
3831
3832
3833 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
3834                 pos_type & right, TextCase action)
3835 {
3836         // process sequences of modified characters; in change
3837         // tracking mode, this approach results in much better
3838         // usability than changing case on a char-by-char basis
3839         // We also need to track the current font, since font
3840         // changes within sequences can occur.
3841         vector<pair<char_type, Font> > changes;
3842
3843         bool const trackChanges = bparams.track_changes;
3844
3845         bool capitalize = true;
3846
3847         for (; pos < right; ++pos) {
3848                 char_type oldChar = d->text_[pos];
3849                 char_type newChar = oldChar;
3850
3851                 // ignore insets and don't play with deleted text!
3852                 if (oldChar != META_INSET && !isDeleted(pos)) {
3853                         switch (action) {
3854                                 case text_lowercase:
3855                                         newChar = lowercase(oldChar);
3856                                         break;
3857                                 case text_capitalization:
3858                                         if (capitalize) {
3859                                                 newChar = uppercase(oldChar);
3860                                                 capitalize = false;
3861                                         }
3862                                         break;
3863                                 case text_uppercase:
3864                                         newChar = uppercase(oldChar);
3865                                         break;
3866                         }
3867                 }
3868
3869                 if (isWordSeparator(pos) || isDeleted(pos)) {
3870                         // permit capitalization again
3871                         capitalize = true;
3872                 }
3873
3874                 if (oldChar != newChar) {
3875                         changes.push_back(make_pair(newChar, getFontSettings(bparams, pos)));
3876                         if (pos != right - 1)
3877                                 continue;
3878                         // step behind the changing area
3879                         pos++;
3880                 }
3881
3882                 int erasePos = pos - changes.size();
3883                 for (size_t i = 0; i < changes.size(); i++) {
3884                         insertChar(pos, changes[i].first,
3885                                    changes[i].second,
3886                                    trackChanges);
3887                         if (!eraseChar(erasePos, trackChanges)) {
3888                                 ++erasePos;
3889                                 ++pos; // advance
3890                                 ++right; // expand selection
3891                         }
3892                 }
3893                 changes.clear();
3894         }
3895 }
3896
3897
3898 int Paragraph::find(docstring const & str, bool cs, bool mw,
3899                 pos_type start_pos, bool del) const
3900 {
3901         pos_type pos = start_pos;
3902         int const strsize = str.length();
3903         int i = 0;
3904         pos_type const parsize = d->text_.size();
3905         for (i = 0; i < strsize && pos < parsize; ++i, ++pos) {
3906                 // Ignore "invisible" letters such as ligature breaks
3907                 // and hyphenation chars while searching
3908                 while (pos < parsize - 1 && isInset(pos)) {
3909                         odocstringstream os;
3910                         getInset(pos)->toString(os);
3911                         if (!getInset(pos)->isLetter() || !os.str().empty())
3912                                 break;
3913                         pos++;
3914                 }
3915                 if (cs && str[i] != d->text_[pos])
3916                         break;
3917                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos]))
3918                         break;
3919                 if (!del && isDeleted(pos))
3920                         break;
3921         }
3922
3923         if (i != strsize)
3924                 return 0;
3925
3926         // if necessary, check whether string matches word
3927         if (mw) {
3928                 if (start_pos > 0 && !isWordSeparator(start_pos - 1))
3929                         return 0;
3930                 if (pos < parsize
3931                         && !isWordSeparator(pos))
3932                         return 0;
3933         }
3934
3935         return pos - start_pos;
3936 }
3937
3938
3939 char_type Paragraph::getChar(pos_type pos) const
3940 {
3941         return d->text_[pos];
3942 }
3943
3944
3945 pos_type Paragraph::size() const
3946 {
3947         return d->text_.size();
3948 }
3949
3950
3951 bool Paragraph::empty() const
3952 {
3953         return d->text_.empty();
3954 }
3955
3956
3957 bool Paragraph::isInset(pos_type pos) const
3958 {
3959         return d->text_[pos] == META_INSET;
3960 }
3961
3962
3963 bool Paragraph::isSeparator(pos_type pos) const
3964 {
3965         //FIXME: Are we sure this can be the only separator?
3966         return d->text_[pos] == ' ';
3967 }
3968
3969
3970 void Paragraph::deregisterWords()
3971 {
3972         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3973         Private::LangWordsMap::const_iterator ite = d->words_.end();
3974         for (; itl != ite; ++itl) {
3975                 WordList & wl = theWordList(itl->first);
3976                 Private::Words::const_iterator it = (itl->second).begin();
3977                 Private::Words::const_iterator et = (itl->second).end();
3978                 for (; it != et; ++it)
3979                         wl.remove(*it);
3980         }
3981         d->words_.clear();
3982 }
3983
3984
3985 void Paragraph::locateWord(pos_type & from, pos_type & to,
3986         word_location const loc) const
3987 {
3988         switch (loc) {
3989         case WHOLE_WORD_STRICT:
3990                 if (from == 0 || from == size()
3991                     || isWordSeparator(from)
3992                     || isWordSeparator(from - 1)) {
3993                         to = from;
3994                         return;
3995                 }
3996                 // fall through
3997
3998         case WHOLE_WORD:
3999                 // If we are already at the beginning of a word, do nothing
4000                 if (!from || isWordSeparator(from - 1))
4001                         break;
4002                 // fall through
4003
4004         case PREVIOUS_WORD:
4005                 // always move the cursor to the beginning of previous word
4006                 while (from && !isWordSeparator(from - 1))
4007                         --from;
4008                 break;
4009         case NEXT_WORD:
4010                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
4011                 break;
4012         case PARTIAL_WORD:
4013                 // no need to move the 'from' cursor
4014                 break;
4015         }
4016         to = from;
4017         while (to < size() && !isWordSeparator(to))
4018                 ++to;
4019 }
4020
4021
4022 void Paragraph::collectWords()
4023 {
4024         for (pos_type pos = 0; pos < size(); ++pos) {
4025                 if (isWordSeparator(pos))
4026                         continue;
4027                 pos_type from = pos;
4028                 locateWord(from, pos, WHOLE_WORD);
4029                 // Work around MSVC warning: The statement
4030                 // if (pos < from + lyxrc.completion_minlength)
4031                 // triggers a signed vs. unsigned warning.
4032                 // I don't know why this happens, it could be a MSVC bug, or
4033                 // related to LLP64 (windows) vs. LP64 (unix) programming
4034                 // model, or the C++ standard might be ambigous in the section
4035                 // defining the "usual arithmetic conversions". However, using
4036                 // a temporary variable is safe and works on all compilers.
4037                 pos_type const endpos = from + lyxrc.completion_minlength;
4038                 if (pos < endpos)
4039                         continue;
4040                 FontList::const_iterator cit = d->fontlist_.fontIterator(from);
4041                 if (cit == d->fontlist_.end())
4042                         return;
4043                 Language const * lang = cit->font().language();
4044                 docstring const word = asString(from, pos, AS_STR_NONE);
4045                 d->words_[lang->lang()].insert(word);
4046         }
4047 }
4048
4049
4050 void Paragraph::registerWords()
4051 {
4052         Private::LangWordsMap::const_iterator itl = d->words_.begin();
4053         Private::LangWordsMap::const_iterator ite = d->words_.end();
4054         for (; itl != ite; ++itl) {
4055                 WordList & wl = theWordList(itl->first);
4056                 Private::Words::const_iterator it = (itl->second).begin();
4057                 Private::Words::const_iterator et = (itl->second).end();
4058                 for (; it != et; ++it)
4059                         wl.insert(*it);
4060         }
4061 }
4062
4063
4064 void Paragraph::updateWords()
4065 {
4066         deregisterWords();
4067         collectWords();
4068         registerWords();
4069 }
4070
4071
4072 void Paragraph::Private::appendSkipPosition(SkipPositions & skips, pos_type const pos) const
4073 {
4074         SkipPositionsIterator begin = skips.begin();
4075         SkipPositions::iterator end = skips.end();
4076         if (pos > 0 && begin < end) {
4077                 --end;
4078                 if (end->last == pos - 1) {
4079                         end->last = pos;
4080                         return;
4081                 }
4082         }
4083         skips.insert(end, FontSpan(pos, pos));
4084 }
4085
4086
4087 Language * Paragraph::Private::locateSpellRange(
4088         pos_type & from, pos_type & to,
4089         SkipPositions & skips) const
4090 {
4091         // skip leading white space
4092         while (from < to && owner_->isWordSeparator(from))
4093                 ++from;
4094         // don't check empty range
4095         if (from >= to)
4096                 return 0;
4097         // get current language
4098         Language * lang = getSpellLanguage(from);
4099         pos_type last = from;
4100         bool samelang = true;
4101         bool sameinset = true;
4102         while (last < to && samelang && sameinset) {
4103                 // hop to end of word
4104                 while (last < to && !owner_->isWordSeparator(last)) {
4105                         if (owner_->getInset(last)) {
4106                                 appendSkipPosition(skips, last);
4107                         } else if (owner_->isDeleted(last)) {
4108                                 appendSkipPosition(skips, last);
4109                         }
4110                         ++last;
4111                 }
4112                 // hop to next word while checking for insets
4113                 while (sameinset && last < to && owner_->isWordSeparator(last)) {
4114                         if (Inset const * inset = owner_->getInset(last))
4115                                 sameinset = inset->isChar() && inset->isLetter();
4116                         if (sameinset && owner_->isDeleted(last)) {
4117                                 appendSkipPosition(skips, last);
4118                         }
4119                         if (sameinset)
4120                                 last++;
4121                 }
4122                 if (sameinset && last < to) {
4123                         // now check for language change
4124                         samelang = lang == getSpellLanguage(last);
4125                 }
4126         }
4127         // if language change detected backstep is needed
4128         if (!samelang)
4129                 --last;
4130         to = last;
4131         return lang;
4132 }
4133
4134
4135 Language * Paragraph::Private::getSpellLanguage(pos_type const from) const
4136 {
4137         Language * lang =
4138                 const_cast<Language *>(owner_->getFontSettings(
4139                         inset_owner_->buffer().params(), from).language());
4140         if (lang == inset_owner_->buffer().params().language
4141                 && !lyxrc.spellchecker_alt_lang.empty()) {
4142                 string lang_code;
4143                 string const lang_variety =
4144                         split(lyxrc.spellchecker_alt_lang, lang_code, '-');
4145                 lang->setCode(lang_code);
4146                 lang->setVariety(lang_variety);
4147         }
4148         return lang;
4149 }
4150
4151
4152 void Paragraph::requestSpellCheck(pos_type pos)
4153 {
4154         d->requestSpellCheck(pos);
4155 }
4156
4157
4158 bool Paragraph::needsSpellCheck() const
4159 {
4160         SpellChecker::ChangeNumber speller_change_number = 0;
4161         if (theSpellChecker())
4162                 speller_change_number = theSpellChecker()->changeNumber();
4163         if (speller_change_number > d->speller_state_.currentChangeNumber()) {
4164                 d->speller_state_.needsCompleteRefresh(speller_change_number);
4165         }
4166         return d->needsSpellCheck();
4167 }
4168
4169
4170 bool Paragraph::Private::ignoreWord(docstring const & word) const
4171 {
4172         // Ignore words with digits
4173         // FIXME: make this customizable
4174         // (note that some checkers ignore words with digits by default)
4175         docstring::const_iterator cit = word.begin();
4176         docstring::const_iterator const end = word.end();
4177         for (; cit != end; ++cit) {
4178                 if (isNumber((*cit)))
4179                         return true;
4180         }
4181         return false;
4182 }
4183
4184
4185 SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
4186         WordLangTuple & wl, docstring_list & suggestions,
4187         bool do_suggestion, bool check_learned) const
4188 {
4189         SpellChecker::Result result = SpellChecker::WORD_OK;
4190         SpellChecker * speller = theSpellChecker();
4191         if (!speller)
4192                 return result;
4193
4194         if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
4195                 return result;
4196
4197         locateWord(from, to, WHOLE_WORD);
4198         if (from == to || from >= size())
4199                 return result;
4200
4201         docstring word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
4202         Language * lang = d->getSpellLanguage(from);
4203
4204         wl = WordLangTuple(word, lang);
4205
4206         if (word.empty())
4207                 return result;
4208
4209         if (needsSpellCheck() || check_learned) {
4210                 pos_type end = to;
4211                 if (!d->ignoreWord(word)) {
4212                         bool const trailing_dot = to < size() && d->text_[to] == '.';
4213                         result = speller->check(wl);
4214                         if (SpellChecker::misspelled(result) && trailing_dot) {
4215                                 wl = WordLangTuple(word.append(from_ascii(".")), lang);
4216                                 result = speller->check(wl);
4217                                 if (!SpellChecker::misspelled(result)) {
4218                                         LYXERR(Debug::GUI, "misspelled word is correct with dot: \"" <<
4219                                            word << "\" [" <<
4220                                            from << ".." << to << "]");
4221                                 } else {
4222                                         // spell check with dot appended failed too
4223                                         // restore original word/lang value
4224                                         word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
4225                                         wl = WordLangTuple(word, lang);
4226                                 }
4227                         }
4228                 }
4229                 if (!SpellChecker::misspelled(result)) {
4230                         // area up to the begin of the next word is not misspelled
4231                         while (end < size() && isWordSeparator(end))
4232                                 ++end;
4233                 }
4234                 d->setMisspelled(from, end, result);
4235         } else {
4236                 result = d->speller_state_.getState(from);
4237         }
4238
4239         if (do_suggestion)
4240                 suggestions.clear();
4241
4242         if (SpellChecker::misspelled(result)) {
4243                 LYXERR(Debug::GUI, "misspelled word: \"" <<
4244                            word << "\" [" <<
4245                            from << ".." << to << "]");
4246                 if (do_suggestion)
4247                         speller->suggest(wl, suggestions);
4248         }
4249         return result;
4250 }
4251
4252
4253 void Paragraph::anonymize()
4254 {
4255         // This is a very crude anonymization for now
4256         for (char_type & c : d->text_)
4257                 if (isLetterChar(c) || isNumber(c))
4258                         c = 'a';
4259 }
4260
4261
4262 void Paragraph::Private::markMisspelledWords(
4263         pos_type const & first, pos_type const & last,
4264         SpellChecker::Result result,
4265         docstring const & word,
4266         SkipPositions const & skips)
4267 {
4268         if (!SpellChecker::misspelled(result)) {
4269                 setMisspelled(first, last, SpellChecker::WORD_OK);
4270                 return;
4271         }
4272         int snext = first;
4273         SpellChecker * speller = theSpellChecker();
4274         // locate and enumerate the error positions
4275         int nerrors = speller->numMisspelledWords();
4276         int numskipped = 0;
4277         SkipPositionsIterator it = skips.begin();
4278         SkipPositionsIterator et = skips.end();
4279         for (int index = 0; index < nerrors; ++index) {
4280                 int wstart;
4281                 int wlen = 0;
4282                 speller->misspelledWord(index, wstart, wlen);
4283                 /// should not happen if speller supports range checks
4284                 if (!wlen) continue;
4285                 docstring const misspelled = word.substr(wstart, wlen);
4286                 wstart += first + numskipped;
4287                 if (snext < wstart) {
4288                         /// mark the range of correct spelling
4289                         numskipped += countSkips(it, et, wstart);
4290                         setMisspelled(snext,
4291                                 wstart - 1, SpellChecker::WORD_OK);
4292                 }
4293                 snext = wstart + wlen;
4294                 numskipped += countSkips(it, et, snext);
4295                 /// mark the range of misspelling
4296                 setMisspelled(wstart, snext, result);
4297                 LYXERR(Debug::GUI, "misspelled word: \"" <<
4298                            misspelled << "\" [" <<
4299                            wstart << ".." << (snext-1) << "]");
4300                 ++snext;
4301         }
4302         if (snext <= last) {
4303                 /// mark the range of correct spelling at end
4304                 setMisspelled(snext, last, SpellChecker::WORD_OK);
4305         }
4306 }
4307
4308
4309 void Paragraph::spellCheck() const
4310 {
4311         SpellChecker * speller = theSpellChecker();
4312         if (!speller || empty() ||!needsSpellCheck())
4313                 return;
4314         pos_type start;
4315         pos_type endpos;
4316         d->rangeOfSpellCheck(start, endpos);
4317         if (speller->canCheckParagraph()) {
4318                 // loop until we leave the range
4319                 for (pos_type first = start; first < endpos; ) {
4320                         pos_type last = endpos;
4321                         Private::SkipPositions skips;
4322                         Language * lang = d->locateSpellRange(first, last, skips);
4323                         if (first >= endpos)
4324                                 break;
4325                         // start the spell checker on the unit of meaning
4326                         docstring word = asString(first, last, AS_STR_INSETS + AS_STR_SKIPDELETE);
4327                         WordLangTuple wl = WordLangTuple(word, lang);
4328                         SpellChecker::Result result = word.size() ?
4329                                 speller->check(wl) : SpellChecker::WORD_OK;
4330                         d->markMisspelledWords(first, last, result, word, skips);
4331                         first = ++last;
4332                 }
4333         } else {
4334                 static docstring_list suggestions;
4335                 pos_type to = endpos;
4336                 while (start < endpos) {
4337                         WordLangTuple wl;
4338                         spellCheck(start, to, wl, suggestions, false);
4339                         start = to + 1;
4340                 }
4341         }
4342         d->readySpellCheck();
4343 }
4344
4345
4346 bool Paragraph::isMisspelled(pos_type pos, bool check_boundary) const
4347 {
4348         bool result = SpellChecker::misspelled(d->speller_state_.getState(pos));
4349         if (result || pos <= 0 || pos > size())
4350                 return result;
4351         if (check_boundary && (pos == size() || isWordSeparator(pos)))
4352                 result = SpellChecker::misspelled(d->speller_state_.getState(pos - 1));
4353         return result;
4354 }
4355
4356
4357 string Paragraph::magicLabel() const
4358 {
4359         stringstream ss;
4360         ss << "magicparlabel-" << id();
4361         return ss.str();
4362 }
4363
4364
4365 } // namespace lyx