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 = bparams.language->fontenc(bparams);
 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                 fontenc = running_font.language()->fontenc(bparams);
1399                 // "Script chars" need to embraced in \textcyrillic and \textgreek notwithstanding
1400                 // whether they are encodable or not (it only depends on the font encoding)
1401                 if (!runparams.isFullUnicode() && Encodings::isKnownScriptChar(c, script)) {
1402                         docstring const wrapper = from_ascii("\\" + script + "{");
1403                         docstring ltx = latex.first;
1404                         if (!prefixIs(ltx, wrapper))
1405                                 ltx = wrapper + latex.first + from_ascii("}");
1406                         column += writeScriptChars(bparams, runparams, os, ltx, running_change,
1407                                                    encoding, fontenc, i) - 1;
1408                 } else if (latex.second
1409                          && ((!prefixIs(nextlatex, '\\')
1410                                && !prefixIs(nextlatex, '{')
1411                                && !prefixIs(nextlatex, '}'))
1412                              || (nexttipas
1413                                  && !prefixIs(from_ascii(nexttipashortcut), '\\')))
1414                          && !tipas) {
1415                         // Prevent eating of a following
1416                         // space or command corruption by
1417                         // following characters
1418                         if (next == ' ' || next == '\0') {
1419                                 column += latex.first.length() + 1;
1420                                 os << latex.first << "{}";
1421                         } else {
1422                                 column += latex.first.length();
1423                                 os << latex.first << " ";
1424                         }
1425                 } else {
1426                         column += latex.first.length() - 1;
1427                         os << latex.first;
1428                 }
1429                 break;
1430         }
1431 }
1432
1433
1434 bool Paragraph::Private::latexSpecialT1(char_type const c, otexstream & os,
1435         pos_type i, unsigned int & column)
1436 {
1437         switch (c) {
1438         case '>':
1439         case '<':
1440                 os.put(c);
1441                 // In T1 encoding, these characters exist
1442                 // but we should avoid ligatures
1443                 if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
1444                         return true;
1445                 os << "\\textcompwordmark" << termcmd;
1446                 column += 19;
1447                 return true;
1448         case '|':
1449                 os.put(c);
1450                 return true;
1451         case '\"':
1452                 // soul.sty breaks with \char`\"
1453                 os << "\\textquotedbl" << termcmd;
1454                 column += 14;
1455                 return true;
1456         default:
1457                 return false;
1458         }
1459 }
1460
1461
1462 bool Paragraph::Private::latexSpecialTU(char_type const c, otexstream & os,
1463         pos_type i, unsigned int & column)
1464 {
1465         // TU encoding is currently on par with T1.
1466         return latexSpecialT1(c, os, i, column);
1467 }
1468
1469
1470 bool Paragraph::Private::latexSpecialT3(char_type const c, otexstream & os,
1471         pos_type /*i*/, unsigned int & column)
1472 {
1473         switch (c) {
1474         case '*':
1475         case '[':
1476         case ']':
1477         case '\"':
1478                 os.put(c);
1479                 return true;
1480         case '|':
1481                 os << "\\textvertline" << termcmd;
1482                 column += 14;
1483                 return true;
1484         default:
1485                 return false;
1486         }
1487 }
1488
1489
1490 void Paragraph::Private::validate(LaTeXFeatures & features) const
1491 {
1492         if (layout_->inpreamble && inset_owner_) {
1493                 // FIXME: Using a string stream here circumvents the encoding
1494                 // switching machinery of odocstream. Therefore the
1495                 // output is wrong if this paragraph contains content
1496                 // that needs to switch encoding.
1497                 Buffer const & buf = inset_owner_->buffer();
1498                 otexstringstream os;
1499                 os << layout_->preamble();
1500                 size_t const length = os.length();
1501                 TeXOnePar(buf, buf.text(), buf.getParFromID(owner_->id()).pit(), os,
1502                           features.runparams(), string(), 0, -1, true);
1503                 if (os.length() > length)
1504                         features.addPreambleSnippet(os.release(), true);
1505         }
1506
1507         if (features.runparams().flavor == OutputParams::HTML
1508             && layout_->htmltitle()) {
1509                 features.setHTMLTitle(owner_->asString(AS_STR_INSETS | AS_STR_SKIPDELETE));
1510         }
1511
1512         // check the params.
1513         if (!params_.spacing().isDefault())
1514                 features.require("setspace");
1515
1516         // then the layouts
1517         features.useLayout(layout_->name());
1518
1519         // then the fonts
1520         fontlist_.validate(features);
1521
1522         // then the indentation
1523         if (!params_.leftIndent().zero())
1524                 features.require("ParagraphLeftIndent");
1525
1526         // then the insets
1527         InsetList::const_iterator icit = insetlist_.begin();
1528         InsetList::const_iterator iend = insetlist_.end();
1529         for (; icit != iend; ++icit) {
1530                 if (icit->inset) {
1531                         features.inDeletedInset(owner_->isDeleted(icit->pos));
1532                         if (icit->inset->lyxCode() == FOOT_CODE) {
1533                                 // FIXME: an item inset would make things much easier.
1534                                 if ((layout_->latextype == LATEX_LIST_ENVIRONMENT
1535                                      || (layout_->latextype == LATEX_ITEM_ENVIRONMENT
1536                                          && layout_->margintype == MARGIN_FIRST_DYNAMIC))
1537                                     && (icit->pos < begin_of_body_
1538                                         || (icit->pos == begin_of_body_
1539                                             && (icit->pos == 0 || text_[icit->pos - 1] != ' '))))
1540                                         features.saveNoteEnv("description");
1541                         }
1542                         icit->inset->validate(features);
1543                         features.inDeletedInset(false);
1544                         if (layout_->needprotect &&
1545                             icit->inset->lyxCode() == FOOT_CODE)
1546                                 features.require("NeedLyXFootnoteCode");
1547                 }
1548         }
1549
1550         // then the contents
1551         BufferParams const bp = features.runparams().is_child
1552                 ? features.buffer().masterParams() : features.buffer().params();
1553         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1554                 char_type c = text_[i];
1555                 CharInfo const & ci = Encodings::unicodeCharInfo(c);
1556                 if (c == 0x0022) {
1557                         if (features.runparams().isFullUnicode() && bp.useNonTeXFonts)
1558                                 features.require("textquotedblp");
1559                         else if (bp.main_font_encoding() != "T1"
1560                                  || ((&owner_->getFontSettings(bp, i))->language()->internalFontEncoding()))
1561                                 features.require("textquotedbl");
1562                 } else if (ci.textfeature() && contains(ci.textpreamble(), '=')) {
1563                         // features that depend on the font or input encoding
1564                         string feats = ci.textpreamble();
1565                         string fontenc = (&owner_->getFontSettings(bp, i))->language()->fontenc(bp);
1566                         if (fontenc.empty())
1567                                 fontenc = features.runparams().main_fontenc;
1568                         while (!feats.empty()) {
1569                                 string feat;
1570                                 feats = split(feats, feat, ',');
1571                                 if (contains(feat, "!=")) {
1572                                         // a feature that is required except for the spcified
1573                                         // font or input encodings
1574                                         string realfeature;
1575                                         string const contexts = ltrim(split(feat, realfeature, '!'), "=");
1576                                         // multiple encodings are separated by semicolon
1577                                         vector<string> context = getVectorFromString(contexts, ";");
1578                                         // require feature if the context matches neither current font
1579                                         // nor input encoding
1580                                         if (std::find(context.begin(), context.end(), fontenc) == context.end()
1581                                             && std::find(context.begin(), context.end(),
1582                                                          features.runparams().encoding->name()) == context.end())
1583                                                 features.require(realfeature);
1584                                 } else if (contains(feat, '=')) {
1585                                         // a feature that is required only for the spcified
1586                                         // font or input encodings
1587                                         string realfeature;
1588                                         string const contexts = split(feat, realfeature, '=');
1589                                         // multiple encodings are separated by semicolon
1590                                         vector<string> context = getVectorFromString(contexts, ";");
1591                                         // require feature if the context matches either current font
1592                                         // or input encoding
1593                                         if (std::find(context.begin(), context.end(), fontenc) != context.end()
1594                                             || std::find(context.begin(), context.end(),
1595                                                          features.runparams().encoding->name()) != context.end())
1596                                                 features.require(realfeature);
1597                                 }
1598                         }
1599                 } else if (!bp.use_dash_ligatures
1600                            && (c == 0x2013 || c == 0x2014)
1601                            && bp.useNonTeXFonts
1602                            && features.runparams().flavor == OutputParams::XETEX)
1603                         // XeTeX's dash behaviour is determined via a global setting
1604                         features.require("xetexdashbreakstate");
1605                 BufferEncodings::validate(c, features);
1606         }
1607 }
1608
1609 /////////////////////////////////////////////////////////////////////
1610 //
1611 // Paragraph
1612 //
1613 /////////////////////////////////////////////////////////////////////
1614
1615 namespace {
1616         Layout const emptyParagraphLayout;
1617 }
1618
1619 Paragraph::Paragraph()
1620         : d(new Paragraph::Private(this, emptyParagraphLayout))
1621 {
1622         itemdepth = 0;
1623         d->params_.clear();
1624 }
1625
1626
1627 Paragraph::Paragraph(Paragraph const & par)
1628         : itemdepth(par.itemdepth),
1629         d(new Paragraph::Private(*par.d, this))
1630 {
1631         registerWords();
1632 }
1633
1634
1635 Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
1636         : itemdepth(par.itemdepth),
1637         d(new Paragraph::Private(*par.d, this, beg, end))
1638 {
1639         registerWords();
1640 }
1641
1642
1643 Paragraph & Paragraph::operator=(Paragraph const & par)
1644 {
1645         // needed as we will destroy the private part before copying it
1646         if (&par != this) {
1647                 itemdepth = par.itemdepth;
1648
1649                 deregisterWords();
1650                 delete d;
1651                 d = new Private(*par.d, this);
1652                 registerWords();
1653         }
1654         return *this;
1655 }
1656
1657
1658 Paragraph::~Paragraph()
1659 {
1660         deregisterWords();
1661         delete d;
1662 }
1663
1664
1665 namespace {
1666
1667 // this shall be called just before every "os << ..." action.
1668 void flushString(ostream & os, docstring & s)
1669 {
1670         os << to_utf8(s);
1671         s.erase();
1672 }
1673
1674 } // namespace
1675
1676
1677 void Paragraph::write(ostream & os, BufferParams const & bparams,
1678         depth_type & dth) const
1679 {
1680         // The beginning or end of a deeper (i.e. nested) area?
1681         if (dth != d->params_.depth()) {
1682                 if (d->params_.depth() > dth) {
1683                         while (d->params_.depth() > dth) {
1684                                 os << "\n\\begin_deeper";
1685                                 ++dth;
1686                         }
1687                 } else {
1688                         while (d->params_.depth() < dth) {
1689                                 os << "\n\\end_deeper";
1690                                 --dth;
1691                         }
1692                 }
1693         }
1694
1695         // First write the layout
1696         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1697
1698         d->params_.write(os);
1699
1700         Font font1(inherit_font, bparams.language);
1701
1702         Change running_change = Change(Change::UNCHANGED);
1703
1704         // this string is used as a buffer to avoid repetitive calls
1705         // to to_utf8(), which turn out to be expensive (JMarc)
1706         docstring write_buffer;
1707
1708         int column = 0;
1709         for (pos_type i = 0; i <= size(); ++i) {
1710
1711                 Change const & change = lookupChange(i);
1712                 if (change != running_change)
1713                         flushString(os, write_buffer);
1714                 Changes::lyxMarkChange(os, bparams, column, running_change, change);
1715                 running_change = change;
1716
1717                 if (i == size())
1718                         break;
1719
1720                 // Write font changes
1721                 Font font2 = getFontSettings(bparams, i);
1722                 if (font2 != font1) {
1723                         flushString(os, write_buffer);
1724                         font2.lyxWriteChanges(font1, os);
1725                         column = 0;
1726                         font1 = font2;
1727                 }
1728
1729                 char_type const c = d->text_[i];
1730                 switch (c) {
1731                 case META_INSET:
1732                         if (Inset const * inset = getInset(i)) {
1733                                 flushString(os, write_buffer);
1734                                 if (inset->directWrite()) {
1735                                         // international char, let it write
1736                                         // code directly so it's shorter in
1737                                         // the file
1738                                         inset->write(os);
1739                                 } else {
1740                                         if (i)
1741                                                 os << '\n';
1742                                         os << "\\begin_inset ";
1743                                         inset->write(os);
1744                                         os << "\n\\end_inset\n\n";
1745                                         column = 0;
1746                                 }
1747                                 // FIXME This can be removed again once the mystery
1748                                 // crash has been resolved.
1749                                 os << flush;
1750                         }
1751                         break;
1752                 case '\\':
1753                         flushString(os, write_buffer);
1754                         os << "\n\\backslash\n";
1755                         column = 0;
1756                         break;
1757                 case '.':
1758                         flushString(os, write_buffer);
1759                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1760                                 os << ".\n";
1761                                 column = 0;
1762                         } else
1763                                 os << '.';
1764                         break;
1765                 default:
1766                         if ((column > 70 && c == ' ')
1767                             || column > 79) {
1768                                 flushString(os, write_buffer);
1769                                 os << '\n';
1770                                 column = 0;
1771                         }
1772                         // this check is to amend a bug. LyX sometimes
1773                         // inserts '\0' this could cause problems.
1774                         if (c != '\0')
1775                                 write_buffer.push_back(c);
1776                         else
1777                                 LYXERR0("NUL char in structure.");
1778                         ++column;
1779                         break;
1780                 }
1781         }
1782
1783         flushString(os, write_buffer);
1784         os << "\n\\end_layout\n";
1785         // FIXME This can be removed again once the mystery
1786         // crash has been resolved.
1787         os << flush;
1788 }
1789
1790
1791 void Paragraph::validate(LaTeXFeatures & features) const
1792 {
1793         d->validate(features);
1794         bool fragile = features.runparams().moving_arg;
1795         fragile |= layout().needprotect;
1796         if (needsCProtection(fragile))
1797                 features.require("cprotect");
1798 }
1799
1800
1801 void Paragraph::insert(pos_type start, docstring const & str,
1802                        Font const & font, Change const & change)
1803 {
1804         for (size_t i = 0, n = str.size(); i != n ; ++i)
1805                 insertChar(start + i, str[i], font, change);
1806 }
1807
1808
1809 void Paragraph::appendChar(char_type c, Font const & font,
1810                 Change const & change)
1811 {
1812         // track change
1813         d->changes_.insert(change, d->text_.size());
1814         // when appending characters, no need to update tables
1815         d->text_.push_back(c);
1816         setFont(d->text_.size() - 1, font);
1817         d->requestSpellCheck(d->text_.size() - 1);
1818 }
1819
1820
1821 void Paragraph::appendString(docstring const & s, Font const & font,
1822                 Change const & change)
1823 {
1824         pos_type end = s.size();
1825         size_t oldsize = d->text_.size();
1826         size_t newsize = oldsize + end;
1827         size_t capacity = d->text_.capacity();
1828         if (newsize >= capacity)
1829                 d->text_.reserve(max(capacity + 100, newsize));
1830
1831         // when appending characters, no need to update tables
1832         d->text_.append(s);
1833
1834         // FIXME: Optimize this!
1835         for (size_t i = oldsize; i != newsize; ++i) {
1836                 // track change
1837                 d->changes_.insert(change, i);
1838                 d->requestSpellCheck(i);
1839         }
1840         d->fontlist_.set(oldsize, font);
1841         d->fontlist_.set(newsize - 1, font);
1842 }
1843
1844
1845 void Paragraph::insertChar(pos_type pos, char_type c,
1846                            bool trackChanges)
1847 {
1848         d->insertChar(pos, c, Change(trackChanges ?
1849                            Change::INSERTED : Change::UNCHANGED));
1850 }
1851
1852
1853 void Paragraph::insertChar(pos_type pos, char_type c,
1854                            Font const & font, bool trackChanges)
1855 {
1856         d->insertChar(pos, c, Change(trackChanges ?
1857                            Change::INSERTED : Change::UNCHANGED));
1858         setFont(pos, font);
1859 }
1860
1861
1862 void Paragraph::insertChar(pos_type pos, char_type c,
1863                            Font const & font, Change const & change)
1864 {
1865         d->insertChar(pos, c, change);
1866         setFont(pos, font);
1867 }
1868
1869
1870 void Paragraph::resetFonts(Font const & font)
1871 {
1872         d->fontlist_.clear();
1873         d->fontlist_.set(0, font);
1874         d->fontlist_.set(d->text_.size() - 1, font);
1875 }
1876
1877 // Gets uninstantiated font setting at position.
1878 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1879                                          pos_type pos) const
1880 {
1881         if (pos > size()) {
1882                 LYXERR0("pos: " << pos << " size: " << size());
1883                 LBUFERR(false);
1884         }
1885
1886         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1887         if (cit != d->fontlist_.end())
1888                 return cit->font();
1889
1890         if (pos == size() && !empty())
1891                 return getFontSettings(bparams, pos - 1);
1892
1893         // Optimisation: avoid a full font instantiation if there is no
1894         // language change from previous call.
1895         static Font previous_font;
1896         static Language const * previous_lang = 0;
1897         Language const * lang = getParLanguage(bparams);
1898         if (lang != previous_lang) {
1899                 previous_lang = lang;
1900                 previous_font = Font(inherit_font, lang);
1901         }
1902         return previous_font;
1903 }
1904
1905
1906 FontSpan Paragraph::fontSpan(pos_type pos) const
1907 {
1908         LBUFERR(pos <= size());
1909
1910         if (pos == size())
1911                 return FontSpan(pos, pos);
1912
1913         pos_type start = 0;
1914         FontList::const_iterator cit = d->fontlist_.begin();
1915         FontList::const_iterator end = d->fontlist_.end();
1916         for (; cit != end; ++cit) {
1917                 if (cit->pos() >= pos) {
1918                         if (pos >= beginOfBody())
1919                                 return FontSpan(max(start, beginOfBody()),
1920                                                 cit->pos());
1921                         else
1922                                 return FontSpan(start,
1923                                                 min(beginOfBody() - 1,
1924                                                          cit->pos()));
1925                 }
1926                 start = cit->pos() + 1;
1927         }
1928
1929         // This should not happen, but if so, we take no chances.
1930         LYXERR0("Paragraph::fontSpan: position not found in fontinfo table!");
1931         LASSERT(false, return FontSpan(pos, pos));
1932 }
1933
1934
1935 // Gets uninstantiated font setting at position 0
1936 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1937 {
1938         if (!empty() && !d->fontlist_.empty())
1939                 return d->fontlist_.begin()->font();
1940
1941         // Optimisation: avoid a full font instantiation if there is no
1942         // language change from previous call.
1943         static Font previous_font;
1944         static Language const * previous_lang = 0;
1945         if (bparams.language != previous_lang) {
1946                 previous_lang = bparams.language;
1947                 previous_font = Font(inherit_font, bparams.language);
1948         }
1949
1950         return previous_font;
1951 }
1952
1953
1954 // Gets the fully instantiated font at a given position in a paragraph
1955 // This is basically the same function as Text::GetFont() in text2.cpp.
1956 // The difference is that this one is used for generating the LaTeX file,
1957 // and thus cosmetic "improvements" are disallowed: This has to deliver
1958 // the true picture of the buffer. (Asger)
1959 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1960                                  Font const & outerfont) const
1961 {
1962         LBUFERR(pos >= 0);
1963
1964         Font font = getFontSettings(bparams, pos);
1965
1966         pos_type const body_pos = beginOfBody();
1967         FontInfo & fi = font.fontInfo();
1968         if (pos < body_pos)
1969                 fi.realize(d->layout_->labelfont);
1970         else
1971                 fi.realize(d->layout_->font);
1972
1973         fi.realize(outerfont.fontInfo());
1974         fi.realize(bparams.getFont().fontInfo());
1975
1976         return font;
1977 }
1978
1979
1980 Font const Paragraph::getLabelFont
1981         (BufferParams const & bparams, Font const & outerfont) const
1982 {
1983         FontInfo tmpfont = d->layout_->labelfont;
1984         tmpfont.realize(outerfont.fontInfo());
1985         tmpfont.realize(bparams.getFont().fontInfo());
1986         return Font(tmpfont, getParLanguage(bparams));
1987 }
1988
1989
1990 Font const Paragraph::getLayoutFont
1991         (BufferParams const & bparams, Font const & outerfont) const
1992 {
1993         FontInfo tmpfont = d->layout_->font;
1994         tmpfont.realize(outerfont.fontInfo());
1995         tmpfont.realize(bparams.getFont().fontInfo());
1996         return Font(tmpfont, getParLanguage(bparams));
1997 }
1998
1999
2000 char_type Paragraph::getUChar(BufferParams const & bparams, pos_type pos) const
2001 {
2002         char_type c = d->text_[pos];
2003         if (!getFontSettings(bparams, pos).isRightToLeft())
2004                 return c;
2005
2006         // FIXME: The arabic special casing is due to the difference of arabic
2007         // round brackets input introduced in r18599. Check if this should be
2008         // unified with Hebrew or at least if all bracket types should be
2009         // handled the same (file format change in either case).
2010         string const & lang = getFontSettings(bparams, pos).language()->lang();
2011         bool const arabic = lang == "arabic_arabtex" || lang == "arabic_arabi"
2012                 || lang == "farsi";
2013         char_type uc = c;
2014         switch (c) {
2015         case '(':
2016                 uc = arabic ? c : ')';
2017                 break;
2018         case ')':
2019                 uc = arabic ? c : '(';
2020                 break;
2021         case '[':
2022                 uc = ']';
2023                 break;
2024         case ']':
2025                 uc = '[';
2026                 break;
2027         case '{':
2028                 uc = '}';
2029                 break;
2030         case '}':
2031                 uc = '{';
2032                 break;
2033         case '<':
2034                 uc = '>';
2035                 break;
2036         case '>':
2037                 uc = '<';
2038                 break;
2039         }
2040
2041         return uc;
2042 }
2043
2044
2045 void Paragraph::setFont(pos_type pos, Font const & font)
2046 {
2047         LASSERT(pos <= size(), return);
2048
2049         // First, reduce font against layout/label font
2050         // Update: The setCharFont() routine in text2.cpp already
2051         // reduces font, so we don't need to do that here. (Asger)
2052
2053         d->fontlist_.set(pos, font);
2054 }
2055
2056
2057 void Paragraph::makeSameLayout(Paragraph const & par)
2058 {
2059         d->layout_ = par.d->layout_;
2060         d->params_ = par.d->params_;
2061 }
2062
2063
2064 bool Paragraph::stripLeadingSpaces(bool trackChanges)
2065 {
2066         if (isFreeSpacing())
2067                 return false;
2068
2069         int pos = 0;
2070         int count = 0;
2071
2072         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
2073                 if (eraseChar(pos, trackChanges))
2074                         ++count;
2075                 else
2076                         ++pos;
2077         }
2078
2079         return count > 0 || pos > 0;
2080 }
2081
2082
2083 bool Paragraph::hasSameLayout(Paragraph const & par) const
2084 {
2085         return par.d->layout_ == d->layout_
2086                 && d->params_.sameLayout(par.d->params_);
2087 }
2088
2089
2090 depth_type Paragraph::getDepth() const
2091 {
2092         return d->params_.depth();
2093 }
2094
2095
2096 depth_type Paragraph::getMaxDepthAfter() const
2097 {
2098         if (d->layout_->isEnvironment())
2099                 return d->params_.depth() + 1;
2100         else
2101                 return d->params_.depth();
2102 }
2103
2104
2105 LyXAlignment Paragraph::getAlign() const
2106 {
2107         if (d->params_.align() == LYX_ALIGN_LAYOUT)
2108                 return d->layout_->align;
2109         else
2110                 return d->params_.align();
2111 }
2112
2113
2114 docstring const & Paragraph::labelString() const
2115 {
2116         return d->params_.labelString();
2117 }
2118
2119
2120 // the next two functions are for the manual labels
2121 docstring const Paragraph::getLabelWidthString() const
2122 {
2123         if (d->layout_->margintype == MARGIN_MANUAL
2124             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
2125                 return d->params_.labelWidthString();
2126         else
2127                 return _("Senseless with this layout!");
2128 }
2129
2130
2131 void Paragraph::setLabelWidthString(docstring const & s)
2132 {
2133         d->params_.labelWidthString(s);
2134 }
2135
2136
2137 docstring Paragraph::expandLabel(Layout const & layout,
2138                 BufferParams const & bparams) const
2139 {
2140         return expandParagraphLabel(layout, bparams, true);
2141 }
2142
2143
2144 docstring Paragraph::expandDocBookLabel(Layout const & layout,
2145                 BufferParams const & bparams) const
2146 {
2147         return expandParagraphLabel(layout, bparams, false);
2148 }
2149
2150
2151 docstring Paragraph::expandParagraphLabel(Layout const & layout,
2152                 BufferParams const & bparams, bool process_appendix) const
2153 {
2154         DocumentClass const & tclass = bparams.documentClass();
2155         string const & lang = getParLanguage(bparams)->code();
2156         bool const in_appendix = process_appendix && d->params_.appendix();
2157         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
2158
2159         if (fmt.empty() && !layout.counter.empty())
2160                 return tclass.counters().theCounter(layout.counter, lang);
2161
2162         // handle 'inherited level parts' in 'fmt',
2163         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
2164         size_t const i = fmt.find('@', 0);
2165         if (i != docstring::npos) {
2166                 size_t const j = fmt.find('@', i + 1);
2167                 if (j != docstring::npos) {
2168                         docstring parent(fmt, i + 1, j - i - 1);
2169                         docstring label = from_ascii("??");
2170                         if (tclass.hasLayout(parent))
2171                                 label = expandParagraphLabel(tclass[parent], bparams,
2172                                                       process_appendix);
2173                         fmt = docstring(fmt, 0, i) + label
2174                                 + docstring(fmt, j + 1, docstring::npos);
2175                 }
2176         }
2177
2178         return tclass.counters().counterLabel(fmt, lang);
2179 }
2180
2181
2182 void Paragraph::applyLayout(Layout const & new_layout)
2183 {
2184         d->layout_ = &new_layout;
2185         LyXAlignment const oldAlign = d->params_.align();
2186
2187         if (!(oldAlign & d->layout_->alignpossible)) {
2188                 frontend::Alert::warning(_("Alignment not permitted"),
2189                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
2190                 d->params_.align(LYX_ALIGN_LAYOUT);
2191         }
2192 }
2193
2194
2195 pos_type Paragraph::beginOfBody() const
2196 {
2197         return d->begin_of_body_;
2198 }
2199
2200
2201 void Paragraph::setBeginOfBody()
2202 {
2203         if (d->layout_->labeltype != LABEL_MANUAL) {
2204                 d->begin_of_body_ = 0;
2205                 return;
2206         }
2207
2208         // Unroll the first two cycles of the loop
2209         // and remember the previous character to
2210         // remove unnecessary getChar() calls
2211         pos_type i = 0;
2212         pos_type end = size();
2213         if (i < end && !(isNewline(i) || isEnvSeparator(i))) {
2214                 ++i;
2215                 if (i < end) {
2216                         char_type previous_char = d->text_[i];
2217                         if (!(isNewline(i) || isEnvSeparator(i))) {
2218                                 ++i;
2219                                 while (i < end && previous_char != ' ') {
2220                                         char_type temp = d->text_[i];
2221                                         if (isNewline(i) || isEnvSeparator(i))
2222                                                 break;
2223                                         ++i;
2224                                         previous_char = temp;
2225                                 }
2226                         }
2227                 }
2228         }
2229
2230         d->begin_of_body_ = i;
2231 }
2232
2233
2234 bool Paragraph::allowParagraphCustomization() const
2235 {
2236         return inInset().allowParagraphCustomization();
2237 }
2238
2239
2240 bool Paragraph::usePlainLayout() const
2241 {
2242         return inInset().usePlainLayout();
2243 }
2244
2245
2246 bool Paragraph::isPassThru() const
2247 {
2248         return inInset().isPassThru() || d->layout_->pass_thru;
2249 }
2250
2251 namespace {
2252
2253 // paragraphs inside floats need different alignment tags to avoid
2254 // unwanted space
2255
2256 bool noTrivlistCentering(InsetCode code)
2257 {
2258         return code == FLOAT_CODE
2259                || code == WRAP_CODE
2260                || code == CELL_CODE;
2261 }
2262
2263
2264 string correction(string const & orig)
2265 {
2266         if (orig == "flushleft")
2267                 return "raggedright";
2268         if (orig == "flushright")
2269                 return "raggedleft";
2270         if (orig == "center")
2271                 return "centering";
2272         return orig;
2273 }
2274
2275
2276 bool corrected_env(otexstream & os, string const & suffix, string const & env,
2277         InsetCode code, bool const lastpar, int & col)
2278 {
2279         string macro = suffix + "{";
2280         if (noTrivlistCentering(code)) {
2281                 if (lastpar) {
2282                         // the last paragraph in non-trivlist-aligned
2283                         // context is special (to avoid unwanted whitespace)
2284                         if (suffix == "\\begin") {
2285                                 macro = "\\" + correction(env) + "{}";
2286                                 os << from_ascii(macro);
2287                                 col += macro.size();
2288                                 return true;
2289                         }
2290                         return false;
2291                 }
2292                 macro += correction(env);
2293         } else
2294                 macro += env;
2295         macro += "}";
2296         if (suffix == "\\par\\end") {
2297                 os << breakln;
2298                 col = 0;
2299         }
2300         os << from_ascii(macro);
2301         col += macro.size();
2302         if (suffix == "\\begin") {
2303                 os << breakln;
2304                 col = 0;
2305         }
2306         return true;
2307 }
2308
2309 } // namespace
2310
2311
2312 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
2313                         otexstream & os, OutputParams const & runparams) const
2314 {
2315         int column = 0;
2316
2317         bool canindent =
2318                 (bparams.paragraph_separation == BufferParams::ParagraphIndentSeparation) ?
2319                         (layout_->toggle_indent != ITOGGLE_NEVER) :
2320                         (layout_->toggle_indent == ITOGGLE_ALWAYS);
2321
2322         if (canindent && params_.noindent() && !layout_->pass_thru) {
2323                 os << "\\noindent ";
2324                 column += 10;
2325         }
2326
2327         LyXAlignment const curAlign = params_.align();
2328
2329         if (curAlign == layout_->align)
2330                 return column;
2331
2332         switch (curAlign) {
2333         case LYX_ALIGN_NONE:
2334         case LYX_ALIGN_BLOCK:
2335         case LYX_ALIGN_LAYOUT:
2336         case LYX_ALIGN_SPECIAL:
2337         case LYX_ALIGN_DECIMAL:
2338                 break;
2339         case LYX_ALIGN_LEFT:
2340         case LYX_ALIGN_RIGHT:
2341         case LYX_ALIGN_CENTER:
2342                 if (runparams.moving_arg) {
2343                         os << "\\protect";
2344                         column += 8;
2345                 }
2346                 break;
2347         }
2348
2349         string const begin_tag = "\\begin";
2350         InsetCode code = ownerCode();
2351         bool const lastpar = runparams.isLastPar;
2352
2353         switch (curAlign) {
2354         case LYX_ALIGN_NONE:
2355         case LYX_ALIGN_BLOCK:
2356         case LYX_ALIGN_LAYOUT:
2357         case LYX_ALIGN_SPECIAL:
2358         case LYX_ALIGN_DECIMAL:
2359                 break;
2360         case LYX_ALIGN_LEFT: {
2361                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2362                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2363                 else
2364                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2365                 break;
2366         } case LYX_ALIGN_RIGHT: {
2367                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2368                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2369                 else
2370                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2371                 break;
2372         } case LYX_ALIGN_CENTER: {
2373                 corrected_env(os, begin_tag, "center", code, lastpar, column);
2374                 break;
2375         }
2376         }
2377
2378         return column;
2379 }
2380
2381
2382 bool Paragraph::Private::endTeXParParams(BufferParams const & bparams,
2383                         otexstream & os, OutputParams const & runparams) const
2384 {
2385         LyXAlignment const curAlign = params_.align();
2386
2387         if (curAlign == layout_->align)
2388                 return false;
2389
2390         switch (curAlign) {
2391         case LYX_ALIGN_NONE:
2392         case LYX_ALIGN_BLOCK:
2393         case LYX_ALIGN_LAYOUT:
2394         case LYX_ALIGN_SPECIAL:
2395         case LYX_ALIGN_DECIMAL:
2396                 break;
2397         case LYX_ALIGN_LEFT:
2398         case LYX_ALIGN_RIGHT:
2399         case LYX_ALIGN_CENTER:
2400                 if (runparams.moving_arg)
2401                         os << "\\protect";
2402                 break;
2403         }
2404
2405         bool output = false;
2406         int col = 0;
2407         string const end_tag = "\\par\\end";
2408         InsetCode code = ownerCode();
2409         bool const lastpar = runparams.isLastPar;
2410
2411         switch (curAlign) {
2412         case LYX_ALIGN_NONE:
2413         case LYX_ALIGN_BLOCK:
2414         case LYX_ALIGN_LAYOUT:
2415         case LYX_ALIGN_SPECIAL:
2416         case LYX_ALIGN_DECIMAL:
2417                 break;
2418         case LYX_ALIGN_LEFT: {
2419                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2420                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2421                 else
2422                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2423                 break;
2424         } case LYX_ALIGN_RIGHT: {
2425                 if (owner_->getParLanguage(bparams)->babel() != "hebrew")
2426                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2427                 else
2428                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2429                 break;
2430         } case LYX_ALIGN_CENTER: {
2431                 corrected_env(os, end_tag, "center", code, lastpar, col);
2432                 break;
2433         }
2434         }
2435
2436         return output || lastpar;
2437 }
2438
2439
2440 // This one spits out the text of the paragraph
2441 void Paragraph::latex(BufferParams const & bparams,
2442         Font const & outerfont,
2443         otexstream & os,
2444         OutputParams const & runparams,
2445         int start_pos, int end_pos, bool force) const
2446 {
2447         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
2448
2449         // FIXME This check should not be needed. Perhaps issue an
2450         // error if it triggers.
2451         Layout const & style = inInset().forcePlainLayout() ?
2452                 bparams.documentClass().plainLayout() : *d->layout_;
2453
2454         if (!force && style.inpreamble)
2455                 return;
2456
2457         bool const allowcust = allowParagraphCustomization();
2458
2459         // Current base font for all inherited font changes, without any
2460         // change caused by an individual character, except for the language:
2461         // It is set to the language of the first character.
2462         // As long as we are in the label, this font is the base font of the
2463         // label. Before the first body character it is set to the base font
2464         // of the body.
2465         Font basefont;
2466
2467         // Maybe we have to create a optional argument.
2468         pos_type body_pos = beginOfBody();
2469         unsigned int column = 0;
2470
2471         if (body_pos > 0) {
2472                 // the optional argument is kept in curly brackets in
2473                 // case it contains a ']'
2474                 // This is not strictly needed, but if this is changed it
2475                 // would be a file format change, and tex2lyx would need
2476                 // to be adjusted, since it unconditionally removes the
2477                 // braces when it parses \item.
2478                 os << "[{";
2479                 column += 2;
2480                 basefont = getLabelFont(bparams, outerfont);
2481         } else {
2482                 basefont = getLayoutFont(bparams, outerfont);
2483         }
2484
2485         // Which font is currently active?
2486         Font running_font(basefont);
2487         // Do we have an open font change?
2488         bool open_font = false;
2489
2490         Change runningChange = Change(Change::UNCHANGED);
2491
2492         Encoding const * const prev_encoding = runparams.encoding;
2493
2494         os.texrow().start(id(), 0);
2495
2496         // if the paragraph is empty, the loop will not be entered at all
2497         if (empty()) {
2498                 // For InTitle commands, we have already opened a group
2499                 // in output_latex::TeXOnePar.
2500                 if (style.isCommand() && !style.intitle) {
2501                         os << '{';
2502                         ++column;
2503                 }
2504                 if (!style.leftdelim().empty()) {
2505                         os << style.leftdelim();
2506                         column += style.leftdelim().size();
2507                 }
2508                 if (allowcust)
2509                         column += d->startTeXParParams(bparams, os, runparams);
2510         }
2511
2512         // Whether a \par can be issued for insets typeset inline with text.
2513         // Yes if greater than 0. This has to be static.
2514         THREAD_LOCAL_STATIC int parInline = 0;
2515
2516         for (pos_type i = 0; i < size(); ++i) {
2517                 // First char in paragraph or after label?
2518                 if (i == body_pos) {
2519                         if (body_pos > 0) {
2520                                 if (open_font) {
2521                                         bool needPar = false;
2522                                         column += running_font.latexWriteEndChanges(
2523                                                 os, bparams, runparams,
2524                                                 basefont, basefont, needPar);
2525                                         open_font = false;
2526                                 }
2527                                 basefont = getLayoutFont(bparams, outerfont);
2528                                 running_font = basefont;
2529
2530                                 column += Changes::latexMarkChange(os, bparams,
2531                                                 runningChange, Change(Change::UNCHANGED),
2532                                                 runparams);
2533                                 runningChange = Change(Change::UNCHANGED);
2534
2535                                 os << "}] ";
2536                                 column +=3;
2537                         }
2538                         // For InTitle commands, we have already opened a group
2539                         // in output_latex::TeXOnePar.
2540                         if (style.isCommand() && !style.intitle) {
2541                                 os << '{';
2542                                 ++column;
2543                         }
2544
2545                         if (!style.leftdelim().empty()) {
2546                                 os << style.leftdelim();
2547                                 column += style.leftdelim().size();
2548                         }
2549
2550                         if (allowcust)
2551                                 column += d->startTeXParParams(bparams, os,
2552                                                             runparams);
2553                 }
2554
2555                 runparams.wasDisplayMath = runparams.inDisplayMath;
2556                 runparams.inDisplayMath = false;
2557                 bool deleted_display_math = false;
2558                 Change const & change = runparams.inDeletedInset
2559                         ? runparams.changeOfDeletedInset : lookupChange(i);
2560
2561                 // Check whether a display math inset follows
2562                 if (d->text_[i] == META_INSET
2563                     && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2564                         InsetMath const * im = getInset(i)->asInsetMath();
2565                         if (im && im->asHullInset()
2566                             && im->asHullInset()->outerDisplay()) {
2567                                 runparams.inDisplayMath = true;
2568                                 // runparams.inDeletedInset will be set by
2569                                 // latexInset later, but we need this info
2570                                 // before it is called. On the other hand, we
2571                                 // cannot set it here because it is a counter.
2572                                 deleted_display_math = isDeleted(i);
2573                         }
2574                         if (bparams.output_changes && deleted_display_math
2575                             && runningChange == change
2576                             && change.type == Change::DELETED
2577                             && !os.afterParbreak()) {
2578                                 // A display math in the same paragraph follows.
2579                                 // We have to close and then reopen \lyxdeleted,
2580                                 // otherwise the math will be shifted up.
2581                                 OutputParams rp = runparams;
2582                                 if (open_font) {
2583                                         bool needPar = false;
2584                                         column += running_font.latexWriteEndChanges(
2585                                                 os, bparams, rp, basefont,
2586                                                 basefont, needPar);
2587                                         open_font = false;
2588                                 }
2589                                 basefont = getLayoutFont(bparams, outerfont);
2590                                 running_font = basefont;
2591                                 column += Changes::latexMarkChange(os, bparams,
2592                                         Change(Change::INSERTED), change, rp);
2593                         }
2594                 }
2595
2596                 if (bparams.output_changes && runningChange != change) {
2597                         if (open_font) {
2598                                 bool needPar = false;
2599                                 column += running_font.latexWriteEndChanges(
2600                                                 os, bparams, runparams,
2601                                                 basefont, basefont, needPar);
2602                                 open_font = false;
2603                         }
2604                         basefont = getLayoutFont(bparams, outerfont);
2605                         running_font = basefont;
2606                         column += Changes::latexMarkChange(os, bparams, runningChange,
2607                                                            change, runparams);
2608                         runningChange = change;
2609                 }
2610
2611                 // do not output text which is marked deleted
2612                 // if change tracking output is disabled
2613                 if (!bparams.output_changes && change.deleted()) {
2614                         continue;
2615                 }
2616
2617                 ++column;
2618
2619                 // Fully instantiated font
2620                 Font const current_font = getFont(bparams, i, outerfont);
2621
2622                 Font const last_font = running_font;
2623
2624                 // Do we need to close the previous font?
2625                 if (open_font &&
2626                     (current_font != running_font ||
2627                      current_font.language() != running_font.language()))
2628                 {
2629                         bool needPar = false;
2630                         column += running_font.latexWriteEndChanges(
2631                                     os, bparams, runparams, basefont,
2632                                     (i == body_pos-1) ? basefont : current_font,
2633                                     needPar);
2634                         running_font = basefont;
2635                         open_font = false;
2636                 }
2637
2638                 string const running_lang = runparams.use_polyglossia ?
2639                         running_font.language()->polyglossia() : running_font.language()->babel();
2640                 // close babel's font environment before opening CJK.
2641                 string const lang_end_command = runparams.use_polyglossia ?
2642                         "\\end{$$lang}" : lyxrc.language_command_end;
2643                 bool const using_begin_end = runparams.use_polyglossia ||
2644                                                 !lang_end_command.empty();
2645                 if (!running_lang.empty() &&
2646                     (!using_begin_end || running_lang == openLanguageName()) &&
2647                     current_font.language()->encoding()->package() == Encoding::CJK) {
2648                                 string end_tag = subst(lang_end_command,
2649                                                         "$$lang",
2650                                                         running_lang);
2651                                 os << from_ascii(end_tag);
2652                                 column += end_tag.length();
2653                                 if (using_begin_end)
2654                                         popLanguageName();
2655                 }
2656
2657                 // Switch file encoding if necessary (and allowed)
2658                 if (!runparams.pass_thru && !style.pass_thru &&
2659                     runparams.encoding->package() != Encoding::none &&
2660                     current_font.language()->encoding()->package() != Encoding::none) {
2661                         pair<bool, int> const enc_switch =
2662                                 switchEncoding(os.os(), bparams, runparams,
2663                                         *(current_font.language()->encoding()));
2664                         if (enc_switch.first) {
2665                                 column += enc_switch.second;
2666                                 runparams.encoding = current_font.language()->encoding();
2667                         }
2668                 }
2669
2670                 char_type const c = d->text_[i];
2671
2672                 // A display math inset inside an ulem command will be output
2673                 // as a box of width \linewidth, so we have to either disable
2674                 // indentation if the inset starts a paragraph, or start a new
2675                 // line to accommodate such box. This has to be done before
2676                 // writing any font changing commands.
2677                 if (runparams.inDisplayMath && !deleted_display_math
2678                     && runparams.inulemcmd) {
2679                         if (os.afterParbreak())
2680                                 os << "\\noindent";
2681                         else
2682                                 os << "\\\\\n";
2683                 }
2684
2685                 // Do we need to change font?
2686                 if ((current_font != running_font ||
2687                      current_font.language() != running_font.language()) &&
2688                         i != body_pos - 1)
2689                 {
2690                         odocstringstream ods;
2691                         column += current_font.latexWriteStartChanges(ods, bparams,
2692                                                               runparams, basefont,
2693                                                               last_font);
2694                         // Check again for display math in ulem commands as a
2695                         // font change may also occur just before a math inset.
2696                         if (runparams.inDisplayMath && !deleted_display_math
2697                             && runparams.inulemcmd) {
2698                                 if (os.afterParbreak())
2699                                         os << "\\noindent";
2700                                 else
2701                                         os << "\\\\\n";
2702                         }
2703                         running_font = current_font;
2704                         open_font = true;
2705                         docstring fontchange = ods.str();
2706                         // check whether the fontchange ends with a \\textcolor
2707                         // modifier and the text starts with a space (bug 4473)
2708                         docstring const last_modifier = rsplit(fontchange, '\\');
2709                         if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2710                                 os << fontchange << from_ascii("{}");
2711                         // check if the fontchange ends with a trailing blank
2712                         // (like "\small " (see bug 3382)
2713                         else if (suffixIs(fontchange, ' ') && c == ' ')
2714                                 os << fontchange.substr(0, fontchange.size() - 1)
2715                                    << from_ascii("{}");
2716                         else
2717                                 os << fontchange;
2718                 }
2719
2720                 // FIXME: think about end_pos implementation...
2721                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2722                         // FIXME: integrate this case in latexSpecialChar
2723                         // Do not print the separation of the optional argument
2724                         // if style.pass_thru is false. This works because
2725                         // latexSpecialChar ignores spaces if
2726                         // style.pass_thru is false.
2727                         if (i != body_pos - 1) {
2728                                 if (d->simpleTeXBlanks(bparams, runparams, os,
2729                                                 i, column, current_font, style)) {
2730                                         // A surrogate pair was output. We
2731                                         // must not call latexSpecialChar
2732                                         // in this iteration, since it would output
2733                                         // the combining character again.
2734                                         ++i;
2735                                         continue;
2736                                 }
2737                         }
2738                 }
2739
2740                 OutputParams rp = runparams;
2741                 rp.free_spacing = style.free_spacing;
2742                 rp.local_font = &current_font;
2743                 rp.intitle = style.intitle;
2744
2745                 // Two major modes:  LaTeX or plain
2746                 // Handle here those cases common to both modes
2747                 // and then split to handle the two modes separately.
2748                 if (c == META_INSET) {
2749                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2750                                 // Greyedout notes and, in general, all insets
2751                                 // with InsetLayout::isDisplay() == false,
2752                                 // are typeset inline with the text. So, we
2753                                 // can add a \par to the last paragraph of
2754                                 // such insets only if nothing else follows.
2755                                 bool incremented = false;
2756                                 Inset const * inset = getInset(i);
2757                                 InsetText const * textinset = inset
2758                                                         ? inset->asInsetText()
2759                                                         : 0;
2760                                 if (i + 1 == size() && textinset
2761                                     && !inset->getLayout().isDisplay()) {
2762                                         ParagraphList const & pars =
2763                                                 textinset->text().paragraphs();
2764                                         pit_type const pit = pars.size() - 1;
2765                                         Font const lastfont =
2766                                                 pit < 0 || pars[pit].empty()
2767                                                 ? pars[pit].getLayoutFont(
2768                                                                 bparams,
2769                                                                 outerfont)
2770                                                 : pars[pit].getFont(bparams,
2771                                                         pars[pit].size() - 1,
2772                                                         outerfont);
2773                                         if (lastfont.fontInfo().size() !=
2774                                             basefont.fontInfo().size()) {
2775                                                 ++parInline;
2776                                                 incremented = true;
2777                                         }
2778                                 }
2779                                 d->latexInset(bparams, os, rp, running_font,
2780                                                 basefont, outerfont, open_font,
2781                                                 runningChange, style, i, column);
2782                                 if (incremented)
2783                                         --parInline;
2784                         }
2785                 } else if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2786                         try {
2787                                 d->latexSpecialChar(os, bparams, rp,
2788                                                     running_font, runningChange,
2789                                                     style, i, end_pos, column);
2790                         } catch (EncodingException & e) {
2791                                 if (runparams.dryrun) {
2792                                         os << "<" << _("LyX Warning: ")
2793                                            << _("uncodable character") << " '";
2794                                         os.put(c);
2795                                         os << "'>";
2796                                 } else {
2797                                         // add location information and throw again.
2798                                         e.par_id = id();
2799                                         e.pos = i;
2800                                         throw(e);
2801                                 }
2802                         }
2803                 }
2804
2805                 // Set the encoding to that returned from latexSpecialChar (see
2806                 // comment for encoding member in OutputParams.h)
2807                 runparams.encoding = rp.encoding;
2808
2809                 // Also carry on the info on a closed ulem command for insets
2810                 // such as Note that do not produce any output, so that no
2811                 // command is ever executed but its opening was recorded.
2812                 runparams.inulemcmd = rp.inulemcmd;
2813         }
2814
2815         // If we have an open font definition, we have to close it
2816         if (open_font) {
2817                 // Make sure that \\par is done with the font of the last
2818                 // character if this has another size as the default.
2819                 // This is necessary because LaTeX (and LyX on the screen)
2820                 // calculates the space between the baselines according
2821                 // to this font. (Matthias)
2822                 //
2823                 // We must not change the font for the last paragraph
2824                 // of non-multipar insets, tabular cells or commands,
2825                 // since this produces unwanted whitespace.
2826
2827                 Font const font = empty()
2828                         ? getLayoutFont(bparams, outerfont)
2829                         : getFont(bparams, size() - 1, outerfont);
2830
2831                 InsetText const * textinset = inInset().asInsetText();
2832
2833                 bool const maintext = textinset
2834                         ? textinset->text().isMainText()
2835                         : false;
2836
2837                 size_t const numpars = textinset
2838                         ? textinset->text().paragraphs().size()
2839                         : 0;
2840
2841                 bool needPar = false;
2842
2843                 if (style.resfont.size() != font.fontInfo().size()
2844                     && (!runparams.isLastPar || maintext
2845                         || (numpars > 1 && d->ownerCode() != CELL_CODE
2846                             && (inInset().getLayout().isDisplay()
2847                                 || parInline)))
2848                     && !style.isCommand()) {
2849                         needPar = true;
2850                 }
2851 #ifdef FIXED_LANGUAGE_END_DETECTION
2852                 if (next_) {
2853                         running_font.latexWriteEndChanges(os, bparams,
2854                                         runparams, basefont,
2855                                         next_->getFont(bparams, 0, outerfont),
2856                                                        needPar);
2857                 } else {
2858                         running_font.latexWriteEndChanges(os, bparams,
2859                                         runparams, basefont, basefont, needPar);
2860                 }
2861 #else
2862 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2863 //FIXME: there as we start another \selectlanguage with the next paragraph if
2864 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2865                 running_font.latexWriteEndChanges(os, bparams, runparams,
2866                                 basefont, basefont, needPar);
2867 #endif
2868                 if (needPar) {
2869                         // The \par could not be inserted at the same nesting
2870                         // level of the font size change, so do it now.
2871                         os << "{\\" << font.latexSize() << "\\par}";
2872                 }
2873         }
2874
2875         column += Changes::latexMarkChange(os, bparams, runningChange,
2876                                            Change(Change::UNCHANGED), runparams);
2877
2878         // Needed if there is an optional argument but no contents.
2879         if (body_pos > 0 && body_pos == size()) {
2880                 os << "}]~";
2881         }
2882
2883         if (!style.rightdelim().empty()) {
2884                 os << style.rightdelim();
2885                 column += style.rightdelim().size();
2886         }
2887
2888         if (allowcust && d->endTeXParParams(bparams, os, runparams)
2889             && runparams.encoding != prev_encoding) {
2890                 runparams.encoding = prev_encoding;
2891                 os << setEncoding(prev_encoding->iconvName());
2892         }
2893
2894         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2895 }
2896
2897
2898 bool Paragraph::emptyTag() const
2899 {
2900         for (pos_type i = 0; i < size(); ++i) {
2901                 if (Inset const * inset = getInset(i)) {
2902                         InsetCode lyx_code = inset->lyxCode();
2903                         // FIXME testing like that is wrong. What is
2904                         // the intent?
2905                         if (lyx_code != TOC_CODE &&
2906                             lyx_code != INCLUDE_CODE &&
2907                             lyx_code != GRAPHICS_CODE &&
2908                             lyx_code != ERT_CODE &&
2909                             lyx_code != LISTINGS_CODE &&
2910                             lyx_code != FLOAT_CODE &&
2911                             lyx_code != TABULAR_CODE) {
2912                                 return false;
2913                         }
2914                 } else {
2915                         char_type c = d->text_[i];
2916                         if (c != ' ' && c != '\t')
2917                                 return false;
2918                 }
2919         }
2920         return true;
2921 }
2922
2923
2924 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
2925         const
2926 {
2927         for (pos_type i = 0; i < size(); ++i) {
2928                 if (Inset const * inset = getInset(i)) {
2929                         InsetCode lyx_code = inset->lyxCode();
2930                         if (lyx_code == LABEL_CODE) {
2931                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2932                                 docstring const & id = il->getParam("name");
2933                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2934                         }
2935                 }
2936         }
2937         return string();
2938 }
2939
2940
2941 pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runparams)
2942         const
2943 {
2944         pos_type i;
2945         for (i = 0; i < size(); ++i) {
2946                 if (Inset const * inset = getInset(i)) {
2947                         inset->docbook(os, runparams);
2948                 } else {
2949                         char_type c = d->text_[i];
2950                         if (c == ' ')
2951                                 break;
2952                         os << sgml::escapeChar(c);
2953                 }
2954         }
2955         return i;
2956 }
2957
2958
2959 pos_type Paragraph::firstWordLyXHTML(XHTMLStream & xs, OutputParams const & runparams)
2960         const
2961 {
2962         pos_type i;
2963         for (i = 0; i < size(); ++i) {
2964                 if (Inset const * inset = getInset(i)) {
2965                         inset->xhtml(xs, runparams);
2966                 } else {
2967                         char_type c = d->text_[i];
2968                         if (c == ' ')
2969                                 break;
2970                         xs << c;
2971                 }
2972         }
2973         return i;
2974 }
2975
2976
2977 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2978 {
2979         Font font_old;
2980         pos_type size = text_.size();
2981         for (pos_type i = initial; i < size; ++i) {
2982                 Font font = owner_->getFont(buf.params(), i, outerfont);
2983                 if (text_[i] == META_INSET)
2984                         return false;
2985                 if (i != initial && font != font_old)
2986                         return false;
2987                 font_old = font;
2988         }
2989
2990         return true;
2991 }
2992
2993
2994 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
2995                                     odocstream & os,
2996                                     OutputParams const & runparams,
2997                                     Font const & outerfont,
2998                                     pos_type initial) const
2999 {
3000         bool emph_flag = false;
3001
3002         Layout const & style = *d->layout_;
3003         FontInfo font_old =
3004                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
3005
3006         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
3007                 os << "]]>";
3008
3009         // parsing main loop
3010         for (pos_type i = initial; i < size(); ++i) {
3011                 Font font = getFont(buf.params(), i, outerfont);
3012
3013                 // handle <emphasis> tag
3014                 if (font_old.emph() != font.fontInfo().emph()) {
3015                         if (font.fontInfo().emph() == FONT_ON) {
3016                                 os << "<emphasis>";
3017                                 emph_flag = true;
3018                         } else if (i != initial) {
3019                                 os << "</emphasis>";
3020                                 emph_flag = false;
3021                         }
3022                 }
3023
3024                 if (Inset const * inset = getInset(i)) {
3025                         inset->docbook(os, runparams);
3026                 } else {
3027                         char_type c = d->text_[i];
3028
3029                         if (style.pass_thru)
3030                                 os.put(c);
3031                         else
3032                                 os << sgml::escapeChar(c);
3033                 }
3034                 font_old = font.fontInfo();
3035         }
3036
3037         if (emph_flag) {
3038                 os << "</emphasis>";
3039         }
3040
3041         if (style.free_spacing)
3042                 os << '\n';
3043         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
3044                 os << "<![CDATA[";
3045 }
3046
3047
3048 namespace {
3049 void doFontSwitch(vector<html::FontTag> & tagsToOpen,
3050                   vector<html::EndFontTag> & tagsToClose,
3051                   bool & flag, FontState curstate, html::FontTypes type)
3052 {
3053         if (curstate == FONT_ON) {
3054                 tagsToOpen.push_back(html::FontTag(type));
3055                 flag = true;
3056         } else if (flag) {
3057                 tagsToClose.push_back(html::EndFontTag(type));
3058                 flag = false;
3059         }
3060 }
3061 } // namespace
3062
3063
3064 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
3065                                     XHTMLStream & xs,
3066                                     OutputParams const & runparams,
3067                                     Font const & outerfont,
3068                                     bool start_paragraph, bool close_paragraph,
3069                                     pos_type initial) const
3070 {
3071         docstring retval;
3072
3073         // track whether we have opened these tags
3074         bool emph_flag = false;
3075         bool bold_flag = false;
3076         bool noun_flag = false;
3077         bool ubar_flag = false;
3078         bool dbar_flag = false;
3079         bool sout_flag = false;
3080         bool xout_flag = false;
3081         bool wave_flag = false;
3082         // shape tags
3083         bool shap_flag = false;
3084         // family tags
3085         bool faml_flag = false;
3086         // size tags
3087         bool size_flag = false;
3088
3089         Layout const & style = *d->layout_;
3090
3091         if (start_paragraph)
3092                 xs.startDivision(allowEmpty());
3093
3094         FontInfo font_old =
3095                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
3096
3097         FontShape  curr_fs   = INHERIT_SHAPE;
3098         FontFamily curr_fam  = INHERIT_FAMILY;
3099         FontSize   curr_size = FONT_SIZE_INHERIT;
3100
3101         string const default_family =
3102                 buf.masterBuffer()->params().fonts_default_family;
3103
3104         vector<html::FontTag> tagsToOpen;
3105         vector<html::EndFontTag> tagsToClose;
3106
3107         // parsing main loop
3108         for (pos_type i = initial; i < size(); ++i) {
3109                 // let's not show deleted material in the output
3110                 if (isDeleted(i))
3111                         continue;
3112
3113                 Font const font = getFont(buf.masterBuffer()->params(), i, outerfont);
3114
3115                 // emphasis
3116                 FontState curstate = font.fontInfo().emph();
3117                 if (font_old.emph() != curstate)
3118                         doFontSwitch(tagsToOpen, tagsToClose, emph_flag, curstate, html::FT_EMPH);
3119
3120                 // noun
3121                 curstate = font.fontInfo().noun();
3122                 if (font_old.noun() != curstate)
3123                         doFontSwitch(tagsToOpen, tagsToClose, noun_flag, curstate, html::FT_NOUN);
3124
3125                 // underbar
3126                 curstate = font.fontInfo().underbar();
3127                 if (font_old.underbar() != curstate)
3128                         doFontSwitch(tagsToOpen, tagsToClose, ubar_flag, curstate, html::FT_UBAR);
3129
3130                 // strikeout
3131                 curstate = font.fontInfo().strikeout();
3132                 if (font_old.strikeout() != curstate)
3133                         doFontSwitch(tagsToOpen, tagsToClose, sout_flag, curstate, html::FT_SOUT);
3134
3135                 // xout
3136                 curstate = font.fontInfo().xout();
3137                 if (font_old.xout() != curstate)
3138                         doFontSwitch(tagsToOpen, tagsToClose, xout_flag, curstate, html::FT_XOUT);
3139
3140                 // double underbar
3141                 curstate = font.fontInfo().uuline();
3142                 if (font_old.uuline() != curstate)
3143                         doFontSwitch(tagsToOpen, tagsToClose, dbar_flag, curstate, html::FT_DBAR);
3144
3145                 // wavy line
3146                 curstate = font.fontInfo().uwave();
3147                 if (font_old.uwave() != curstate)
3148                         doFontSwitch(tagsToOpen, tagsToClose, wave_flag, curstate, html::FT_WAVE);
3149
3150                 // bold
3151                 // a little hackish, but allows us to reuse what we have.
3152                 curstate = (font.fontInfo().series() == BOLD_SERIES ? FONT_ON : FONT_OFF);
3153                 if (font_old.series() != font.fontInfo().series())
3154                         doFontSwitch(tagsToOpen, tagsToClose, bold_flag, curstate, html::FT_BOLD);
3155
3156                 // Font shape
3157                 curr_fs = font.fontInfo().shape();
3158                 FontShape old_fs = font_old.shape();
3159                 if (old_fs != curr_fs) {
3160                         if (shap_flag) {
3161                                 switch (old_fs) {
3162                                 case ITALIC_SHAPE:
3163                                         tagsToClose.push_back(html::EndFontTag(html::FT_ITALIC));
3164                                         break;
3165                                 case SLANTED_SHAPE:
3166                                         tagsToClose.push_back(html::EndFontTag(html::FT_SLANTED));
3167                                         break;
3168                                 case SMALLCAPS_SHAPE:
3169                                         tagsToClose.push_back(html::EndFontTag(html::FT_SMALLCAPS));
3170                                         break;
3171                                 case UP_SHAPE:
3172                                 case INHERIT_SHAPE:
3173                                         break;
3174                                 default:
3175                                         // the other tags are for internal use
3176                                         LATTEST(false);
3177                                         break;
3178                                 }
3179                                 shap_flag = false;
3180                         }
3181                         switch (curr_fs) {
3182                         case ITALIC_SHAPE:
3183                                 tagsToOpen.push_back(html::FontTag(html::FT_ITALIC));
3184                                 shap_flag = true;
3185                                 break;
3186                         case SLANTED_SHAPE:
3187                                 tagsToOpen.push_back(html::FontTag(html::FT_SLANTED));
3188                                 shap_flag = true;
3189                                 break;
3190                         case SMALLCAPS_SHAPE:
3191                                 tagsToOpen.push_back(html::FontTag(html::FT_SMALLCAPS));
3192                                 shap_flag = true;
3193                                 break;
3194                         case UP_SHAPE:
3195                         case INHERIT_SHAPE:
3196                                 break;
3197                         default:
3198                                 // the other tags are for internal use
3199                                 LATTEST(false);
3200                                 break;
3201                         }
3202                 }
3203
3204                 // Font family
3205                 curr_fam = font.fontInfo().family();
3206                 FontFamily old_fam = font_old.family();
3207                 if (old_fam != curr_fam) {
3208                         if (faml_flag) {
3209                                 switch (old_fam) {
3210                                 case ROMAN_FAMILY:
3211                                         tagsToClose.push_back(html::EndFontTag(html::FT_ROMAN));
3212                                         break;
3213                                 case SANS_FAMILY:
3214                                         tagsToClose.push_back(html::EndFontTag(html::FT_SANS));
3215                                         break;
3216                                 case TYPEWRITER_FAMILY:
3217                                         tagsToClose.push_back(html::EndFontTag(html::FT_TYPE));
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                                 faml_flag = false;
3227                         }
3228                         switch (curr_fam) {
3229                         case ROMAN_FAMILY:
3230                                 // we will treat a "default" font family as roman, since we have
3231                                 // no other idea what to do.
3232                                 if (default_family != "rmdefault" && default_family != "default") {
3233                                         tagsToOpen.push_back(html::FontTag(html::FT_ROMAN));
3234                                         faml_flag = true;
3235                                 }
3236                                 break;
3237                         case SANS_FAMILY:
3238                                 if (default_family != "sfdefault") {
3239                                         tagsToOpen.push_back(html::FontTag(html::FT_SANS));
3240                                         faml_flag = true;
3241                                 }
3242                                 break;
3243                         case TYPEWRITER_FAMILY:
3244                                 if (default_family != "ttdefault") {
3245                                         tagsToOpen.push_back(html::FontTag(html::FT_TYPE));
3246                                         faml_flag = true;
3247                                 }
3248                                 break;
3249                         case INHERIT_FAMILY:
3250                                 break;
3251                         default:
3252                                 // the other tags are for internal use
3253                                 LATTEST(false);
3254                                 break;
3255                         }
3256                 }
3257
3258                 // Font size
3259                 curr_size = font.fontInfo().size();
3260                 FontSize old_size = font_old.size();
3261                 if (old_size != curr_size) {
3262                         if (size_flag) {
3263                                 switch (old_size) {
3264                                 case FONT_SIZE_TINY:
3265                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_TINY));
3266                                         break;
3267                                 case FONT_SIZE_SCRIPT:
3268                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SCRIPT));
3269                                         break;
3270                                 case FONT_SIZE_FOOTNOTE:
3271                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_FOOTNOTE));
3272                                         break;
3273                                 case FONT_SIZE_SMALL:
3274                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SMALL));
3275                                         break;
3276                                 case FONT_SIZE_LARGE:
3277                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGE));
3278                                         break;
3279                                 case FONT_SIZE_LARGER:
3280                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGER));
3281                                         break;
3282                                 case FONT_SIZE_LARGEST:
3283                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGEST));
3284                                         break;
3285                                 case FONT_SIZE_HUGE:
3286                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGE));
3287                                         break;
3288                                 case FONT_SIZE_HUGER:
3289                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGER));
3290                                         break;
3291                                 case FONT_SIZE_INCREASE:
3292                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_INCREASE));
3293                                         break;
3294                                 case FONT_SIZE_DECREASE:
3295                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_DECREASE));
3296                                         break;
3297                                 case FONT_SIZE_INHERIT:
3298                                 case FONT_SIZE_NORMAL:
3299                                         break;
3300                                 default:
3301                                         // the other tags are for internal use
3302                                         LATTEST(false);
3303                                         break;
3304                                 }
3305                                 size_flag = false;
3306                         }
3307                         switch (curr_size) {
3308                         case FONT_SIZE_TINY:
3309                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_TINY));
3310                                 size_flag = true;
3311                                 break;
3312                         case FONT_SIZE_SCRIPT:
3313                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SCRIPT));
3314                                 size_flag = true;
3315                                 break;
3316                         case FONT_SIZE_FOOTNOTE:
3317                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_FOOTNOTE));
3318                                 size_flag = true;
3319                                 break;
3320                         case FONT_SIZE_SMALL:
3321                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SMALL));
3322                                 size_flag = true;
3323                                 break;
3324                         case FONT_SIZE_LARGE:
3325                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGE));
3326                                 size_flag = true;
3327                                 break;
3328                         case FONT_SIZE_LARGER:
3329                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGER));
3330                                 size_flag = true;
3331                                 break;
3332                         case FONT_SIZE_LARGEST:
3333                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGEST));
3334                                 size_flag = true;
3335                                 break;
3336                         case FONT_SIZE_HUGE:
3337                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGE));
3338                                 size_flag = true;
3339                                 break;
3340                         case FONT_SIZE_HUGER:
3341                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGER));
3342                                 size_flag = true;
3343                                 break;
3344                         case FONT_SIZE_INCREASE:
3345                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_INCREASE));
3346                                 size_flag = true;
3347                                 break;
3348                         case FONT_SIZE_DECREASE:
3349                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_DECREASE));
3350                                 size_flag = true;
3351                                 break;
3352                         case FONT_SIZE_NORMAL:
3353                         case FONT_SIZE_INHERIT:
3354                                 break;
3355                         default:
3356                                 // the other tags are for internal use
3357                                 LATTEST(false);
3358                                 break;
3359                         }
3360                 }
3361
3362                 // FIXME XHTML
3363                 // Other such tags? What about the other text ranges?
3364
3365                 vector<html::EndFontTag>::const_iterator cit = tagsToClose.begin();
3366                 vector<html::EndFontTag>::const_iterator cen = tagsToClose.end();
3367                 for (; cit != cen; ++cit)
3368                         xs << *cit;
3369
3370                 vector<html::FontTag>::const_iterator sit = tagsToOpen.begin();
3371                 vector<html::FontTag>::const_iterator sen = tagsToOpen.end();
3372                 for (; sit != sen; ++sit)
3373                         xs << *sit;
3374
3375                 tagsToClose.clear();
3376                 tagsToOpen.clear();
3377
3378                 Inset const * inset = getInset(i);
3379                 if (inset) {
3380                         if (!runparams.for_toc || inset->isInToc()) {
3381                                 OutputParams np = runparams;
3382                                 np.local_font = &font;
3383                                 // If the paragraph has size 1, then we are in the "special
3384                                 // case" where we do not output the containing paragraph info
3385                                 if (!inset->getLayout().htmlisblock() && size() != 1)
3386                                         np.html_in_par = true;
3387                                 retval += inset->xhtml(xs, np);
3388                         }
3389                 } else {
3390                         char_type c = getUChar(buf.masterBuffer()->params(), i);
3391                         xs << c;
3392                 }
3393                 font_old = font.fontInfo();
3394         }
3395
3396         // FIXME XHTML
3397         // I'm worried about what happens if a branch, say, is itself
3398         // wrapped in some font stuff. I think that will not work.
3399         xs.closeFontTags();
3400         if (close_paragraph)
3401                 xs.endDivision();
3402
3403         return retval;
3404 }
3405
3406
3407 bool Paragraph::isHfill(pos_type pos) const
3408 {
3409         Inset const * inset = getInset(pos);
3410         return inset && inset->isHfill();
3411 }
3412
3413
3414 bool Paragraph::isNewline(pos_type pos) const
3415 {
3416         // U+2028 LINE SEPARATOR
3417         // U+2029 PARAGRAPH SEPARATOR
3418         char_type const c = d->text_[pos];
3419         if (c == 0x2028 || c == 0x2029)
3420                 return true;
3421         Inset const * inset = getInset(pos);
3422         return inset && inset->lyxCode() == NEWLINE_CODE;
3423 }
3424
3425
3426 bool Paragraph::isEnvSeparator(pos_type pos) const
3427 {
3428         Inset const * inset = getInset(pos);
3429         return inset && inset->lyxCode() == SEPARATOR_CODE;
3430 }
3431
3432
3433 bool Paragraph::isLineSeparator(pos_type pos) const
3434 {
3435         char_type const c = d->text_[pos];
3436         if (isLineSeparatorChar(c))
3437                 return true;
3438         Inset const * inset = getInset(pos);
3439         return inset && inset->isLineSeparator();
3440 }
3441
3442
3443 bool Paragraph::isWordSeparator(pos_type pos) const
3444 {
3445         if (pos == size())
3446                 return true;
3447         if (Inset const * inset = getInset(pos))
3448                 return !inset->isLetter();
3449         // if we have a hard hyphen (no en- or emdash) or apostrophe
3450         // we pass this to the spell checker
3451         // FIXME: this method is subject to change, visit
3452         // https://bugzilla.mozilla.org/show_bug.cgi?id=355178
3453         // to get an impression how complex this is.
3454         if (isHardHyphenOrApostrophe(pos))
3455                 return false;
3456         char_type const c = d->text_[pos];
3457         // We want to pass the escape chars to the spellchecker
3458         docstring const escape_chars = from_utf8(lyxrc.spellchecker_esc_chars);
3459         return !isLetterChar(c) && !isDigitASCII(c) && !contains(escape_chars, c);
3460 }
3461
3462
3463 bool Paragraph::isHardHyphenOrApostrophe(pos_type pos) const
3464 {
3465         pos_type const psize = size();
3466         if (pos >= psize)
3467                 return false;
3468         char_type const c = d->text_[pos];
3469         if (c != '-' && c != '\'')
3470                 return false;
3471         int nextpos = pos + 1;
3472         int prevpos = pos > 0 ? pos - 1 : 0;
3473         if ((nextpos == psize || isSpace(nextpos))
3474                 && (pos == 0 || isSpace(prevpos)))
3475                 return false;
3476         return true;
3477 }
3478
3479
3480 bool Paragraph::needsCProtection(bool const fragile) const
3481 {
3482         // first check the layout of the paragraph, but only in insets
3483         InsetText const * textinset = inInset().asInsetText();
3484         bool const maintext = textinset
3485                 ? textinset->text().isMainText()
3486                 : false;
3487
3488         if (!maintext && layout().needcprotect) {
3489                 // Environments need cprotection regardless the content
3490                 if (layout().latextype == LATEX_ENVIRONMENT)
3491                         return true;
3492
3493                 // Commands need cprotection if they contain specific chars
3494                 int const nchars_escape = 9;
3495                 static char_type const chars_escape[nchars_escape] = {
3496                         '&', '_', '$', '%', '#', '^', '{', '}', '\\'};
3497
3498                 docstring const pars = asString();
3499                 for (int k = 0; k < nchars_escape; k++) {
3500                         if (contains(pars, chars_escape[k]))
3501                                 return true;
3502                 }
3503         }
3504
3505         // now check whether we have insets that need cprotection
3506         pos_type size = d->text_.size();
3507         for (pos_type i = 0; i < size; ++i)
3508                 if (isInset(i) && getInset(i)->needsCProtection(maintext, fragile))
3509                         return true;
3510
3511         return false;
3512 }
3513
3514
3515 FontSpan const & Paragraph::getSpellRange(pos_type pos) const
3516 {
3517         return d->speller_state_.getRange(pos);
3518 }
3519
3520
3521 bool Paragraph::isChar(pos_type pos) const
3522 {
3523         if (Inset const * inset = getInset(pos))
3524                 return inset->isChar();
3525         char_type const c = d->text_[pos];
3526         return !isLetterChar(c) && !isDigitASCII(c) && !lyx::isSpace(c);
3527 }
3528
3529
3530 bool Paragraph::isSpace(pos_type pos) const
3531 {
3532         if (Inset const * inset = getInset(pos))
3533                 return inset->isSpace();
3534         char_type const c = d->text_[pos];
3535         return lyx::isSpace(c);
3536 }
3537
3538
3539 Language const *
3540 Paragraph::getParLanguage(BufferParams const & bparams) const
3541 {
3542         if (!empty())
3543                 return getFirstFontSettings(bparams).language();
3544         // FIXME: we should check the prev par as well (Lgb)
3545         return bparams.language;
3546 }
3547
3548
3549 bool Paragraph::isRTL(BufferParams const & bparams) const
3550 {
3551         return getParLanguage(bparams)->rightToLeft()
3552                 && !inInset().getLayout().forceLTR();
3553 }
3554
3555
3556 void Paragraph::changeLanguage(BufferParams const & bparams,
3557                                Language const * from, Language const * to)
3558 {
3559         // change language including dummy font change at the end
3560         for (pos_type i = 0; i <= size(); ++i) {
3561                 Font font = getFontSettings(bparams, i);
3562                 if (font.language() == from) {
3563                         font.setLanguage(to);
3564                         setFont(i, font);
3565                         d->requestSpellCheck(i);
3566                 }
3567         }
3568 }
3569
3570
3571 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
3572 {
3573         Language const * doc_language = bparams.language;
3574         for (auto const & f : d->fontlist_)
3575                 if (f.font().language() != ignore_language &&
3576                     f.font().language() != latex_language &&
3577                     f.font().language() != doc_language)
3578                         return true;
3579         return false;
3580 }
3581
3582
3583 void Paragraph::getLanguages(std::set<Language const *> & langs) const
3584 {
3585         for (auto const & f : d->fontlist_) {
3586                 Language const * lang = f.font().language();
3587                 if (lang != ignore_language &&
3588                     lang != latex_language)
3589                         langs.insert(lang);
3590         }
3591 }
3592
3593
3594 docstring Paragraph::asString(int options) const
3595 {
3596         return asString(0, size(), options);
3597 }
3598
3599
3600 docstring Paragraph::asString(pos_type beg, pos_type end, int options, const OutputParams *runparams) const
3601 {
3602         odocstringstream os;
3603
3604         if (beg == 0
3605             && options & AS_STR_LABEL
3606             && !d->params_.labelString().empty())
3607                 os << d->params_.labelString() << ' ';
3608
3609         for (pos_type i = beg; i < end; ++i) {
3610                 if ((options & AS_STR_SKIPDELETE) && isDeleted(i))
3611                         continue;
3612                 char_type const c = d->text_[i];
3613                 if (isPrintable(c) || c == '\t'
3614                     || (c == '\n' && (options & AS_STR_NEWLINES)))
3615                         os.put(c);
3616                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
3617                         if (c == META_INSET && (options & AS_STR_PLAINTEXT)) {
3618                                 LASSERT(runparams != 0, return docstring());
3619                                 getInset(i)->plaintext(os, *runparams);
3620                         } else {
3621                                 getInset(i)->toString(os);
3622                         }
3623                 }
3624         }
3625
3626         return os.str();
3627 }
3628
3629
3630 void Paragraph::forOutliner(docstring & os, size_t const maxlen,
3631                             bool const shorten, bool const label) const
3632 {
3633         size_t tmplen = shorten ? maxlen + 1 : maxlen;
3634         if (label && !labelString().empty())
3635                 os += labelString() + ' ';
3636         if (!layout().isTocCaption())
3637                 return;
3638         for (pos_type i = 0; i < size() && os.length() < tmplen; ++i) {
3639                 if (isDeleted(i))
3640                         continue;
3641                 char_type const c = d->text_[i];
3642                 if (isPrintable(c))
3643                         os += c;
3644                 else if (c == META_INSET)
3645                         getInset(i)->forOutliner(os, tmplen, false);
3646         }
3647         if (shorten)
3648                 Text::shortenForOutliner(os, maxlen);
3649 }
3650
3651
3652 void Paragraph::setInsetOwner(Inset const * inset)
3653 {
3654         d->inset_owner_ = inset;
3655 }
3656
3657
3658 int Paragraph::id() const
3659 {
3660         return d->id_;
3661 }
3662
3663
3664 void Paragraph::setId(int id)
3665 {
3666         d->id_ = id;
3667 }
3668
3669
3670 Layout const & Paragraph::layout() const
3671 {
3672         return *d->layout_;
3673 }
3674
3675
3676 void Paragraph::setLayout(Layout const & layout)
3677 {
3678         d->layout_ = &layout;
3679 }
3680
3681
3682 void Paragraph::setDefaultLayout(DocumentClass const & tc)
3683 {
3684         setLayout(tc.defaultLayout());
3685 }
3686
3687
3688 void Paragraph::setPlainLayout(DocumentClass const & tc)
3689 {
3690         setLayout(tc.plainLayout());
3691 }
3692
3693
3694 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
3695 {
3696         if (usePlainLayout())
3697                 setPlainLayout(tclass);
3698         else
3699                 setDefaultLayout(tclass);
3700 }
3701
3702
3703 Inset const & Paragraph::inInset() const
3704 {
3705         LBUFERR(d->inset_owner_);
3706         return *d->inset_owner_;
3707 }
3708
3709
3710 ParagraphParameters & Paragraph::params()
3711 {
3712         return d->params_;
3713 }
3714
3715
3716 ParagraphParameters const & Paragraph::params() const
3717 {
3718         return d->params_;
3719 }
3720
3721
3722 bool Paragraph::isFreeSpacing() const
3723 {
3724         if (d->layout_->free_spacing)
3725                 return true;
3726         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
3727 }
3728
3729
3730 bool Paragraph::allowEmpty() const
3731 {
3732         if (d->layout_->keepempty)
3733                 return true;
3734         return d->inset_owner_ && d->inset_owner_->allowEmpty();
3735 }
3736
3737
3738 bool Paragraph::brokenBiblio() const
3739 {
3740         // there is a problem if there is no bibitem at position 0 or
3741         // if there is another bibitem in the paragraph.
3742         return d->layout_->labeltype == LABEL_BIBLIO
3743                 && (d->insetlist_.find(BIBITEM_CODE) != 0
3744                     || d->insetlist_.find(BIBITEM_CODE, 1) > 0);
3745 }
3746
3747
3748 int Paragraph::fixBiblio(Buffer const & buffer)
3749 {
3750         // FIXME: What about the case where paragraph is not BIBLIO
3751         // but there is an InsetBibitem?
3752         // FIXME: when there was already an inset at 0, the return value is 1,
3753         // which does not tell whether another inset has been remove; the
3754         // cursor cannot be correctly updated.
3755
3756         if (d->layout_->labeltype != LABEL_BIBLIO)
3757                 return 0;
3758
3759         bool const track_changes = buffer.params().track_changes;
3760         int bibitem_pos = d->insetlist_.find(BIBITEM_CODE);
3761         bool const hasbibitem0 = bibitem_pos == 0;
3762
3763         if (hasbibitem0) {
3764                 bibitem_pos = d->insetlist_.find(BIBITEM_CODE, 1);
3765                 // There was an InsetBibitem at pos 0, and no other one => OK
3766                 if (bibitem_pos == -1)
3767                         return 0;
3768                 // there is a bibitem at the 0 position, but since
3769                 // there is a second one, we copy the second on the
3770                 // first. We're assuming there are at most two of
3771                 // these, which there should be.
3772                 // FIXME: why does it make sense to do that rather
3773                 // than keep the first? (JMarc)
3774                 Inset * inset = releaseInset(bibitem_pos);
3775                 d->insetlist_.begin()->inset = inset;
3776                 return -bibitem_pos;
3777         }
3778
3779         // We need to create an inset at the beginning
3780         Inset * inset = 0;
3781         if (bibitem_pos > 0) {
3782                 // there was one somewhere in the paragraph, let's move it
3783                 inset = d->insetlist_.release(bibitem_pos);
3784                 eraseChar(bibitem_pos, track_changes);
3785         } else
3786                 // make a fresh one
3787                 inset = new InsetBibitem(const_cast<Buffer *>(&buffer),
3788                                          InsetCommandParams(BIBITEM_CODE));
3789
3790         Font font(inherit_font, buffer.params().language);
3791         insertInset(0, inset, font, Change(track_changes ? Change::INSERTED
3792                                                    : Change::UNCHANGED));
3793
3794         return 1;
3795 }
3796
3797
3798 void Paragraph::checkAuthors(AuthorList const & authorList)
3799 {
3800         d->changes_.checkAuthors(authorList);
3801 }
3802
3803
3804 bool Paragraph::isChanged(pos_type pos) const
3805 {
3806         return lookupChange(pos).changed();
3807 }
3808
3809
3810 bool Paragraph::isInserted(pos_type pos) const
3811 {
3812         return lookupChange(pos).inserted();
3813 }
3814
3815
3816 bool Paragraph::isDeleted(pos_type pos) const
3817 {
3818         return lookupChange(pos).deleted();
3819 }
3820
3821
3822 InsetList const & Paragraph::insetList() const
3823 {
3824         return d->insetlist_;
3825 }
3826
3827
3828 void Paragraph::setInsetBuffers(Buffer & b)
3829 {
3830         d->insetlist_.setBuffer(b);
3831 }
3832
3833
3834 void Paragraph::resetBuffer()
3835 {
3836         d->insetlist_.resetBuffer();
3837 }
3838
3839
3840 Inset * Paragraph::releaseInset(pos_type pos)
3841 {
3842         Inset * inset = d->insetlist_.release(pos);
3843         /// does not honour change tracking!
3844         eraseChar(pos, false);
3845         return inset;
3846 }
3847
3848
3849 Inset * Paragraph::getInset(pos_type pos)
3850 {
3851         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3852                  ? d->insetlist_.get(pos) : 0;
3853 }
3854
3855
3856 Inset const * Paragraph::getInset(pos_type pos) const
3857 {
3858         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3859                  ? d->insetlist_.get(pos) : 0;
3860 }
3861
3862
3863 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
3864                 pos_type & right, TextCase action)
3865 {
3866         // process sequences of modified characters; in change
3867         // tracking mode, this approach results in much better
3868         // usability than changing case on a char-by-char basis
3869         // We also need to track the current font, since font
3870         // changes within sequences can occur.
3871         vector<pair<char_type, Font> > changes;
3872
3873         bool const trackChanges = bparams.track_changes;
3874
3875         bool capitalize = true;
3876
3877         for (; pos < right; ++pos) {
3878                 char_type oldChar = d->text_[pos];
3879                 char_type newChar = oldChar;
3880
3881                 // ignore insets and don't play with deleted text!
3882                 if (oldChar != META_INSET && !isDeleted(pos)) {
3883                         switch (action) {
3884                                 case text_lowercase:
3885                                         newChar = lowercase(oldChar);
3886                                         break;
3887                                 case text_capitalization:
3888                                         if (capitalize) {
3889                                                 newChar = uppercase(oldChar);
3890                                                 capitalize = false;
3891                                         }
3892                                         break;
3893                                 case text_uppercase:
3894                                         newChar = uppercase(oldChar);
3895                                         break;
3896                         }
3897                 }
3898
3899                 if (isWordSeparator(pos) || isDeleted(pos)) {
3900                         // permit capitalization again
3901                         capitalize = true;
3902                 }
3903
3904                 if (oldChar != newChar) {
3905                         changes.push_back(make_pair(newChar, getFontSettings(bparams, pos)));
3906                         if (pos != right - 1)
3907                                 continue;
3908                         // step behind the changing area
3909                         pos++;
3910                 }
3911
3912                 int erasePos = pos - changes.size();
3913                 for (size_t i = 0; i < changes.size(); i++) {
3914                         insertChar(pos, changes[i].first,
3915                                    changes[i].second,
3916                                    trackChanges);
3917                         if (!eraseChar(erasePos, trackChanges)) {
3918                                 ++erasePos;
3919                                 ++pos; // advance
3920                                 ++right; // expand selection
3921                         }
3922                 }
3923                 changes.clear();
3924         }
3925 }
3926
3927
3928 int Paragraph::find(docstring const & str, bool cs, bool mw,
3929                 pos_type start_pos, bool del) const
3930 {
3931         pos_type pos = start_pos;
3932         int const strsize = str.length();
3933         int i = 0;
3934         pos_type const parsize = d->text_.size();
3935         for (i = 0; i < strsize && pos < parsize; ++i, ++pos) {
3936                 // Ignore "invisible" letters such as ligature breaks
3937                 // and hyphenation chars while searching
3938                 while (pos < parsize - 1 && isInset(pos)) {
3939                         odocstringstream os;
3940                         getInset(pos)->toString(os);
3941                         if (!getInset(pos)->isLetter() || !os.str().empty())
3942                                 break;
3943                         pos++;
3944                 }
3945                 if (cs && str[i] != d->text_[pos])
3946                         break;
3947                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos]))
3948                         break;
3949                 if (!del && isDeleted(pos))
3950                         break;
3951         }
3952
3953         if (i != strsize)
3954                 return 0;
3955
3956         // if necessary, check whether string matches word
3957         if (mw) {
3958                 if (start_pos > 0 && !isWordSeparator(start_pos - 1))
3959                         return 0;
3960                 if (pos < parsize
3961                         && !isWordSeparator(pos))
3962                         return 0;
3963         }
3964
3965         return pos - start_pos;
3966 }
3967
3968
3969 char_type Paragraph::getChar(pos_type pos) const
3970 {
3971         return d->text_[pos];
3972 }
3973
3974
3975 pos_type Paragraph::size() const
3976 {
3977         return d->text_.size();
3978 }
3979
3980
3981 bool Paragraph::empty() const
3982 {
3983         return d->text_.empty();
3984 }
3985
3986
3987 bool Paragraph::isInset(pos_type pos) const
3988 {
3989         return d->text_[pos] == META_INSET;
3990 }
3991
3992
3993 bool Paragraph::isSeparator(pos_type pos) const
3994 {
3995         //FIXME: Are we sure this can be the only separator?
3996         return d->text_[pos] == ' ';
3997 }
3998
3999
4000 void Paragraph::deregisterWords()
4001 {
4002         Private::LangWordsMap::const_iterator itl = d->words_.begin();
4003         Private::LangWordsMap::const_iterator ite = d->words_.end();
4004         for (; itl != ite; ++itl) {
4005                 WordList & wl = theWordList(itl->first);
4006                 Private::Words::const_iterator it = (itl->second).begin();
4007                 Private::Words::const_iterator et = (itl->second).end();
4008                 for (; it != et; ++it)
4009                         wl.remove(*it);
4010         }
4011         d->words_.clear();
4012 }
4013
4014
4015 void Paragraph::locateWord(pos_type & from, pos_type & to,
4016         word_location const loc) const
4017 {
4018         switch (loc) {
4019         case WHOLE_WORD_STRICT:
4020                 if (from == 0 || from == size()
4021                     || isWordSeparator(from)
4022                     || isWordSeparator(from - 1)) {
4023                         to = from;
4024                         return;
4025                 }
4026                 // fall through
4027
4028         case WHOLE_WORD:
4029                 // If we are already at the beginning of a word, do nothing
4030                 if (!from || isWordSeparator(from - 1))
4031                         break;
4032                 // fall through
4033
4034         case PREVIOUS_WORD:
4035                 // always move the cursor to the beginning of previous word
4036                 while (from && !isWordSeparator(from - 1))
4037                         --from;
4038                 break;
4039         case NEXT_WORD:
4040                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
4041                 break;
4042         case PARTIAL_WORD:
4043                 // no need to move the 'from' cursor
4044                 break;
4045         }
4046         to = from;
4047         while (to < size() && !isWordSeparator(to))
4048                 ++to;
4049 }
4050
4051
4052 void Paragraph::collectWords()
4053 {
4054         for (pos_type pos = 0; pos < size(); ++pos) {
4055                 if (isWordSeparator(pos))
4056                         continue;
4057                 pos_type from = pos;
4058                 locateWord(from, pos, WHOLE_WORD);
4059                 // Work around MSVC warning: The statement
4060                 // if (pos < from + lyxrc.completion_minlength)
4061                 // triggers a signed vs. unsigned warning.
4062                 // I don't know why this happens, it could be a MSVC bug, or
4063                 // related to LLP64 (windows) vs. LP64 (unix) programming
4064                 // model, or the C++ standard might be ambigous in the section
4065                 // defining the "usual arithmetic conversions". However, using
4066                 // a temporary variable is safe and works on all compilers.
4067                 pos_type const endpos = from + lyxrc.completion_minlength;
4068                 if (pos < endpos)
4069                         continue;
4070                 FontList::const_iterator cit = d->fontlist_.fontIterator(from);
4071                 if (cit == d->fontlist_.end())
4072                         return;
4073                 Language const * lang = cit->font().language();
4074                 docstring const word = asString(from, pos, AS_STR_NONE);
4075                 d->words_[lang->lang()].insert(word);
4076         }
4077 }
4078
4079
4080 void Paragraph::registerWords()
4081 {
4082         Private::LangWordsMap::const_iterator itl = d->words_.begin();
4083         Private::LangWordsMap::const_iterator ite = d->words_.end();
4084         for (; itl != ite; ++itl) {
4085                 WordList & wl = theWordList(itl->first);
4086                 Private::Words::const_iterator it = (itl->second).begin();
4087                 Private::Words::const_iterator et = (itl->second).end();
4088                 for (; it != et; ++it)
4089                         wl.insert(*it);
4090         }
4091 }
4092
4093
4094 void Paragraph::updateWords()
4095 {
4096         deregisterWords();
4097         collectWords();
4098         registerWords();
4099 }
4100
4101
4102 void Paragraph::Private::appendSkipPosition(SkipPositions & skips, pos_type const pos) const
4103 {
4104         SkipPositionsIterator begin = skips.begin();
4105         SkipPositions::iterator end = skips.end();
4106         if (pos > 0 && begin < end) {
4107                 --end;
4108                 if (end->last == pos - 1) {
4109                         end->last = pos;
4110                         return;
4111                 }
4112         }
4113         skips.insert(end, FontSpan(pos, pos));
4114 }
4115
4116
4117 Language * Paragraph::Private::locateSpellRange(
4118         pos_type & from, pos_type & to,
4119         SkipPositions & skips) const
4120 {
4121         // skip leading white space
4122         while (from < to && owner_->isWordSeparator(from))
4123                 ++from;
4124         // don't check empty range
4125         if (from >= to)
4126                 return 0;
4127         // get current language
4128         Language * lang = getSpellLanguage(from);
4129         pos_type last = from;
4130         bool samelang = true;
4131         bool sameinset = true;
4132         while (last < to && samelang && sameinset) {
4133                 // hop to end of word
4134                 while (last < to && !owner_->isWordSeparator(last)) {
4135                         if (owner_->getInset(last)) {
4136                                 appendSkipPosition(skips, last);
4137                         } else if (owner_->isDeleted(last)) {
4138                                 appendSkipPosition(skips, last);
4139                         }
4140                         ++last;
4141                 }
4142                 // hop to next word while checking for insets
4143                 while (sameinset && last < to && owner_->isWordSeparator(last)) {
4144                         if (Inset const * inset = owner_->getInset(last))
4145                                 sameinset = inset->isChar() && inset->isLetter();
4146                         if (sameinset && owner_->isDeleted(last)) {
4147                                 appendSkipPosition(skips, last);
4148                         }
4149                         if (sameinset)
4150                                 last++;
4151                 }
4152                 if (sameinset && last < to) {
4153                         // now check for language change
4154                         samelang = lang == getSpellLanguage(last);
4155                 }
4156         }
4157         // if language change detected backstep is needed
4158         if (!samelang)
4159                 --last;
4160         to = last;
4161         return lang;
4162 }
4163
4164
4165 Language * Paragraph::Private::getSpellLanguage(pos_type const from) const
4166 {
4167         Language * lang =
4168                 const_cast<Language *>(owner_->getFontSettings(
4169                         inset_owner_->buffer().params(), from).language());
4170         if (lang == inset_owner_->buffer().params().language
4171                 && !lyxrc.spellchecker_alt_lang.empty()) {
4172                 string lang_code;
4173                 string const lang_variety =
4174                         split(lyxrc.spellchecker_alt_lang, lang_code, '-');
4175                 lang->setCode(lang_code);
4176                 lang->setVariety(lang_variety);
4177         }
4178         return lang;
4179 }
4180
4181
4182 void Paragraph::requestSpellCheck(pos_type pos)
4183 {
4184         d->requestSpellCheck(pos);
4185 }
4186
4187
4188 bool Paragraph::needsSpellCheck() const
4189 {
4190         SpellChecker::ChangeNumber speller_change_number = 0;
4191         if (theSpellChecker())
4192                 speller_change_number = theSpellChecker()->changeNumber();
4193         if (speller_change_number > d->speller_state_.currentChangeNumber()) {
4194                 d->speller_state_.needsCompleteRefresh(speller_change_number);
4195         }
4196         return d->needsSpellCheck();
4197 }
4198
4199
4200 bool Paragraph::Private::ignoreWord(docstring const & word) const
4201 {
4202         // Ignore words with digits
4203         // FIXME: make this customizable
4204         // (note that some checkers ignore words with digits by default)
4205         docstring::const_iterator cit = word.begin();
4206         docstring::const_iterator const end = word.end();
4207         for (; cit != end; ++cit) {
4208                 if (isNumber((*cit)))
4209                         return true;
4210         }
4211         return false;
4212 }
4213
4214
4215 SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
4216         WordLangTuple & wl, docstring_list & suggestions,
4217         bool do_suggestion, bool check_learned) const
4218 {
4219         SpellChecker::Result result = SpellChecker::WORD_OK;
4220         SpellChecker * speller = theSpellChecker();
4221         if (!speller)
4222                 return result;
4223
4224         if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
4225                 return result;
4226
4227         locateWord(from, to, WHOLE_WORD);
4228         if (from == to || from >= size())
4229                 return result;
4230
4231         docstring word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
4232         Language * lang = d->getSpellLanguage(from);
4233
4234         wl = WordLangTuple(word, lang);
4235
4236         if (word.empty())
4237                 return result;
4238
4239         if (needsSpellCheck() || check_learned) {
4240                 pos_type end = to;
4241                 if (!d->ignoreWord(word)) {
4242                         bool const trailing_dot = to < size() && d->text_[to] == '.';
4243                         result = speller->check(wl);
4244                         if (SpellChecker::misspelled(result) && trailing_dot) {
4245                                 wl = WordLangTuple(word.append(from_ascii(".")), lang);
4246                                 result = speller->check(wl);
4247                                 if (!SpellChecker::misspelled(result)) {
4248                                         LYXERR(Debug::GUI, "misspelled word is correct with dot: \"" <<
4249                                            word << "\" [" <<
4250                                            from << ".." << to << "]");
4251                                 } else {
4252                                         // spell check with dot appended failed too
4253                                         // restore original word/lang value
4254                                         word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
4255                                         wl = WordLangTuple(word, lang);
4256                                 }
4257                         }
4258                 }
4259                 if (!SpellChecker::misspelled(result)) {
4260                         // area up to the begin of the next word is not misspelled
4261                         while (end < size() && isWordSeparator(end))
4262                                 ++end;
4263                 }
4264                 d->setMisspelled(from, end, result);
4265         } else {
4266                 result = d->speller_state_.getState(from);
4267         }
4268
4269         if (do_suggestion)
4270                 suggestions.clear();
4271
4272         if (SpellChecker::misspelled(result)) {
4273                 LYXERR(Debug::GUI, "misspelled word: \"" <<
4274                            word << "\" [" <<
4275                            from << ".." << to << "]");
4276                 if (do_suggestion)
4277                         speller->suggest(wl, suggestions);
4278         }
4279         return result;
4280 }
4281
4282
4283 void Paragraph::anonymize()
4284 {
4285         // This is a very crude anonymization for now
4286         for (char_type & c : d->text_)
4287                 if (isLetterChar(c) || isNumber(c))
4288                         c = 'a';
4289 }
4290
4291
4292 void Paragraph::Private::markMisspelledWords(
4293         pos_type const & first, pos_type const & last,
4294         SpellChecker::Result result,
4295         docstring const & word,
4296         SkipPositions const & skips)
4297 {
4298         if (!SpellChecker::misspelled(result)) {
4299                 setMisspelled(first, last, SpellChecker::WORD_OK);
4300                 return;
4301         }
4302         int snext = first;
4303         SpellChecker * speller = theSpellChecker();
4304         // locate and enumerate the error positions
4305         int nerrors = speller->numMisspelledWords();
4306         int numskipped = 0;
4307         SkipPositionsIterator it = skips.begin();
4308         SkipPositionsIterator et = skips.end();
4309         for (int index = 0; index < nerrors; ++index) {
4310                 int wstart;
4311                 int wlen = 0;
4312                 speller->misspelledWord(index, wstart, wlen);
4313                 /// should not happen if speller supports range checks
4314                 if (!wlen) continue;
4315                 docstring const misspelled = word.substr(wstart, wlen);
4316                 wstart += first + numskipped;
4317                 if (snext < wstart) {
4318                         /// mark the range of correct spelling
4319                         numskipped += countSkips(it, et, wstart);
4320                         setMisspelled(snext,
4321                                 wstart - 1, SpellChecker::WORD_OK);
4322                 }
4323                 snext = wstart + wlen;
4324                 numskipped += countSkips(it, et, snext);
4325                 /// mark the range of misspelling
4326                 setMisspelled(wstart, snext, result);
4327                 LYXERR(Debug::GUI, "misspelled word: \"" <<
4328                            misspelled << "\" [" <<
4329                            wstart << ".." << (snext-1) << "]");
4330                 ++snext;
4331         }
4332         if (snext <= last) {
4333                 /// mark the range of correct spelling at end
4334                 setMisspelled(snext, last, SpellChecker::WORD_OK);
4335         }
4336 }
4337
4338
4339 void Paragraph::spellCheck() const
4340 {
4341         SpellChecker * speller = theSpellChecker();
4342         if (!speller || empty() ||!needsSpellCheck())
4343                 return;
4344         pos_type start;
4345         pos_type endpos;
4346         d->rangeOfSpellCheck(start, endpos);
4347         if (speller->canCheckParagraph()) {
4348                 // loop until we leave the range
4349                 for (pos_type first = start; first < endpos; ) {
4350                         pos_type last = endpos;
4351                         Private::SkipPositions skips;
4352                         Language * lang = d->locateSpellRange(first, last, skips);
4353                         if (first >= endpos)
4354                                 break;
4355                         // start the spell checker on the unit of meaning
4356                         docstring word = asString(first, last, AS_STR_INSETS + AS_STR_SKIPDELETE);
4357                         WordLangTuple wl = WordLangTuple(word, lang);
4358                         SpellChecker::Result result = word.size() ?
4359                                 speller->check(wl) : SpellChecker::WORD_OK;
4360                         d->markMisspelledWords(first, last, result, word, skips);
4361                         first = ++last;
4362                 }
4363         } else {
4364                 static docstring_list suggestions;
4365                 pos_type to = endpos;
4366                 while (start < endpos) {
4367                         WordLangTuple wl;
4368                         spellCheck(start, to, wl, suggestions, false);
4369                         start = to + 1;
4370                 }
4371         }
4372         d->readySpellCheck();
4373 }
4374
4375
4376 bool Paragraph::isMisspelled(pos_type pos, bool check_boundary) const
4377 {
4378         bool result = SpellChecker::misspelled(d->speller_state_.getState(pos));
4379         if (result || pos <= 0 || pos > size())
4380                 return result;
4381         if (check_boundary && (pos == size() || isWordSeparator(pos)))
4382                 result = SpellChecker::misspelled(d->speller_state_.getState(pos - 1));
4383         return result;
4384 }
4385
4386
4387 string Paragraph::magicLabel() const
4388 {
4389         stringstream ss;
4390         ss << "magicparlabel-" << id();
4391         return ss.str();
4392 }
4393
4394
4395 } // namespace lyx