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         /// This could go to ParagraphParameters if we want to.
 333         int startTeXParParams(BufferParams const &, otexstream &,
 334                               OutputParams const &) const;
 335
 336         /// This could go to ParagraphParameters if we want to.
 337         bool endTeXParParams(BufferParams const &, otexstream &,
 338                              OutputParams const &) const;
 339
 340         ///
 341         void latexInset(BufferParams const &,
 342                                    otexstream &,
 343                                    OutputParams &,
 344                                    Font & running_font,
 345                                    Font & basefont,
 346                                    Font const & outerfont,
 347                                    bool & open_font,
 348                                    Change & running_change,
 349                                    Layout const & style,
 350                                    pos_type & i,
 351                                    unsigned int & column);
 352
 353         ///
 354         void latexSpecialChar(
 355                                    otexstream & os,
 356                                    BufferParams const & bparams,
 357                                    OutputParams const & runparams,
 358                                    Font const & running_font,
 359                                    string & alien_script,
 360                                    Layout const & style,
 361                                    pos_type & i,
 362                                    pos_type end_pos,
 363                                    unsigned int & column);
 364
 365         ///
 366         bool latexSpecialT1(
 367                 char_type const c,
 368                 otexstream & os,
 369                 pos_type i,
 370                 unsigned int & column);
 371         ///
 372         bool latexSpecialTU(
 373                 char_type const c,
 374                 otexstream & os,
 375                 pos_type i,
 376                 unsigned int & column);
 377         ///
 378         bool latexSpecialT3(
 379                 char_type const c,
 380                 otexstream & os,
 381                 pos_type i,
 382                 unsigned int & column);
 383
 384         ///
 385         void validate(LaTeXFeatures & features) const;
 386
 387         /// Checks if the paragraph contains only text and no inset or font change.
 388         bool onlyText(Buffer const & buf, Font const & outerfont,
 389                       pos_type initial) const;
 390
 391         /// a vector of speller skip positions
 392         typedef vector<FontSpan> SkipPositions;
 393         typedef SkipPositions::const_iterator SkipPositionsIterator;
 394
 395         void appendSkipPosition(SkipPositions & skips, pos_type const pos) const;
 396
 397         Language * getSpellLanguage(pos_type const from) const;
 398
 399         Language * locateSpellRange(pos_type & from, pos_type & to,
 400                                     SkipPositions & skips) const;
 401
 402         bool hasSpellerChange() const
 403         {
 404                 SpellChecker::ChangeNumber speller_change_number = 0;
 405                 if (theSpellChecker())
 406                         speller_change_number = theSpellChecker()->changeNumber();
 407                 return speller_change_number > speller_state_.currentChangeNumber();
 408         }
 409
 410         bool ignoreWord(docstring const & word) const ;
 411
 412         void setMisspelled(pos_type from, pos_type to, SpellChecker::Result state)
 413         {
 414                 pos_type textsize = owner_->size();
 415                 // check for sane arguments
 416                 if (to <= from || from >= textsize)
 417                         return;
 418                 FontSpan fp = FontSpan(from, to - 1);
 419                 speller_state_.setRange(fp, state);
 420         }
 421
 422         void requestSpellCheck(pos_type pos)
 423         {
 424                 if (pos == -1)
 425                         speller_state_.needsCompleteRefresh(speller_state_.currentChangeNumber());
 426                 else
 427                         speller_state_.needsRefresh(pos);
 428         }
 429
 430         void readySpellCheck()
 431         {
 432                 speller_state_.needsRefresh(-1);
 433         }
 434
 435         bool needsSpellCheck() const
 436         {
 437                 return speller_state_.needsRefresh();
 438         }
 439
 440         void rangeOfSpellCheck(pos_type & first, pos_type & last) const
 441         {
 442                 speller_state_.refreshRange(first, last);
 443                 if (last == -1) {
 444                         last = owner_->size();
 445                         return;
 446                 }
 447                 pos_type endpos = last;
 448                 owner_->locateWord(first, endpos, WHOLE_WORD, true);
 449                 if (endpos < last) {
 450                         endpos = last;
 451                         owner_->locateWord(last, endpos, WHOLE_WORD, true);
 452                 }
 453                 last = endpos;
 454         }
 455
 456         int countSkips(SkipPositionsIterator & it, SkipPositionsIterator const et,
 457                             int & start) const
 458         {
 459                 int numskips = 0;
 460                 while (it != et && it->first < start) {
 461                         int skip = it->last - it->first + 1;
 462                         start += skip;
 463                         numskips += skip;
 464                         ++it;
 465                 }
 466                 return numskips;
 467         }
 468
 469         void markMisspelledWords(pos_type const & first, pos_type const & last,
 470                                                          SpellChecker::Result result,
 471                                                          docstring const & word,
 472                                                          SkipPositions const & skips);
 473
 474         InsetCode ownerCode() const
 475         {
 476                 return inset_owner_ ? inset_owner_->lyxCode() : NO_CODE;
 477         }
 478
 479         /// Which Paragraph owns us?
 480         Paragraph * owner_;
 481
 482         /// In which Inset?
 483         Inset const * inset_owner_;
 484
 485         ///
 486         FontList fontlist_;
 487
 488         ///
 489         int id_;
 490
 491         ///
 492         ParagraphParameters params_;
 493
 494         /// for recording and looking up changes
 495         Changes changes_;
 496
 497         ///
 498         InsetList insetlist_;
 499
 500         /// end of label
 501         pos_type begin_of_body_;
 502
 503         typedef docstring TextContainer;
 504         ///
 505         TextContainer text_;
 506
 507         typedef set<docstring> Words;
 508         typedef map<string, Words> LangWordsMap;
 509         ///
 510         LangWordsMap words_;
 511         ///
 512         Layout const * layout_;
 513         ///
 514         SpellCheckerState speller_state_;
 515 };
 516
 517
 518 Paragraph::Private::Private(Paragraph * owner, Layout const & layout)
 519         : owner_(owner), inset_owner_(nullptr), id_(-1), begin_of_body_(0), layout_(&layout)
 520 {
 521         text_.reserve(100);
 522 }
 523
 524
 525 //static
 526 int Paragraph::Private::make_id()
 527 {
 528         // The id is unique per session across buffers because it is used in
 529         // LFUN_PARAGRAPH_GOTO to switch to a different buffer, for instance in the
 530         // outliner.
 531         // (thread-safe)
 532         static atomic_uint next_id(0);
 533         return next_id++;
 534 }
 535
 536
 537 Paragraph::Private::Private(Private const & p, Paragraph * owner)
 538         : owner_(owner), inset_owner_(p.inset_owner_), fontlist_(p.fontlist_),
 539           id_(make_id()),
 540           params_(p.params_), changes_(p.changes_), insetlist_(p.insetlist_),
 541           begin_of_body_(p.begin_of_body_), text_(p.text_), words_(p.words_),
 542           layout_(p.layout_)
 543 {
 544         requestSpellCheck(p.text_.size());
 545 }
 546
 547
 548 Paragraph::Private::Private(Private const & p, Paragraph * owner,
 549         pos_type beg, pos_type end)
 550         : owner_(owner), inset_owner_(p.inset_owner_), id_(make_id()),
 551           params_(p.params_), changes_(p.changes_),
 552           insetlist_(p.insetlist_, beg, end),
 553           begin_of_body_(p.begin_of_body_), words_(p.words_),
 554           layout_(p.layout_)
 555 {
 556         if (beg >= pos_type(p.text_.size()))
 557                 return;
 558         text_ = p.text_.substr(beg, end - beg);
 559
 560         FontList::const_iterator fcit = fontlist_.begin();
 561         FontList::const_iterator fend = fontlist_.end();
 562         for (; fcit != fend; ++fcit) {
 563                 if (fcit->pos() < beg)
 564                         continue;
 565                 if (fcit->pos() >= end) {
 566                         // Add last entry in the fontlist_.
 567                         fontlist_.set(text_.size() - 1, fcit->font());
 568                         break;
 569                 }
 570                 // Add a new entry in the fontlist_.
 571                 fontlist_.set(fcit->pos() - beg, fcit->font());
 572         }
 573         requestSpellCheck(p.text_.size());
 574 }
 575
 576
 577 void Paragraph::addChangesToToc(DocIterator const & cdit, Buffer const & buf,
 578                                 bool output_active, TocBackend & backend) const
 579 {
 580         d->changes_.addToToc(cdit, buf, output_active, backend);
 581 }
 582
 583
 584 bool Paragraph::isDeleted(pos_type start, pos_type end) const
 585 {
 586         LASSERT(start >= 0 && start <= size(), return false);
 587         LASSERT(end > start && end <= size() + 1, return false);
 588
 589         return d->changes_.isDeleted(start, end);
 590 }
 591
 592
 593 bool Paragraph::isChanged(pos_type start, pos_type end) const
 594 {
 595         LASSERT(start >= 0 && start <= size(), return false);
 596         LASSERT(end > start && end <= size() + 1, return false);
 597
 598         return d->changes_.isChanged(start, end);
 599 }
 600
 601 // FIXME: Ideally the diverse isChanged() methods should account for that!
 602 bool Paragraph::hasChangedInsets(pos_type start, pos_type end) const
 603 {
 604         LASSERT(start >= 0 && start <= size(), return false);
 605         LASSERT(end > start && end <= size() + 1, return false);
 606
 607         for (auto const & icit : d->insetlist_) {
 608                 if (icit.pos < start)
 609                         continue;
 610                 if (icit.pos >= end)
 611                         break;
 612                 if (icit.inset && icit.inset->isChanged())
 613                         return true;
 614         }
 615         return false;
 616 }
 617
 618 bool Paragraph::isChanged() const
 619 {
 620         return d->changes_.isChanged();
 621 }
 622
 623
 624 bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const
 625 {
 626         // keep the logic here in sync with the logic of eraseChars()
 627         if (!trackChanges)
 628                 return true;
 629
 630         Change const & change = d->changes_.lookup(size());
 631         return change.inserted() && change.currentAuthor();
 632 }
 633
 634 Change Paragraph::parEndChange() const
 635 {
 636         return d->changes_.lookup(size());
 637 }
 638
 639
 640 void Paragraph::setChange(Change const & change)
 641 {
 642         // beware of the imaginary end-of-par character!
 643         d->changes_.set(change, 0, size() + 1);
 644
 645         /*
 646          * Propagate the change recursively - but not in case of DELETED!
 647          *
 648          * Imagine that your co-author makes changes in an existing inset. He
 649          * sends your document to you and you come to the conclusion that the
 650          * inset should go completely. If you erase it, LyX must not delete all
 651          * text within the inset. Otherwise, the change tracked insertions of
 652          * your co-author get lost and there is no way to restore them later.
 653          *
 654          * Conclusion: An inset's content should remain untouched if you delete it
 655          */
 656
 657         if (!change.deleted()) {
 658                 for (pos_type pos = 0; pos < size(); ++pos) {
 659                         if (Inset * inset = getInset(pos))
 660                                 inset->setChange(change);
 661                 }
 662         }
 663 }
 664
 665
 666 void Paragraph::setChange(pos_type pos, Change const & change)
 667 {
 668         LASSERT(pos >= 0 && pos <= size(), return);
 669         d->changes_.set(change, pos);
 670
 671         // see comment in setChange(Change const &) above
 672         if (!change.deleted() && pos < size())
 673                 if (Inset * inset = getInset(pos))
 674                         inset->setChange(change);
 675 }
 676
 677
 678 Change const & Paragraph::lookupChange(pos_type pos) const
 679 {
 680         LBUFERR(pos >= 0 && pos <= size());
 681         return d->changes_.lookup(pos);
 682 }
 683
 684
 685 void Paragraph::acceptChanges(pos_type start, pos_type end)
 686 {
 687         LASSERT(start >= 0 && start <= size(), return);
 688         LASSERT(end > start && end <= size() + 1, return);
 689
 690         for (pos_type pos = start; pos < end; ++pos) {
 691                 switch (lookupChange(pos).type) {
 692                         case Change::UNCHANGED:
 693                                 // accept changes in nested inset
 694                                 if (Inset * inset = getInset(pos))
 695                                         inset->acceptChanges();
 696                                 break;
 697
 698                         case Change::INSERTED:
 699                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 700                                 // also accept changes in nested inset
 701                                 if (Inset * inset = getInset(pos))
 702                                         inset->acceptChanges();
 703                                 break;
 704
 705                         case Change::DELETED:
 706                                 // Suppress access to non-existent
 707                                 // "end-of-paragraph char"
 708                                 if (pos < size()) {
 709                                         eraseChar(pos, false);
 710                                         --end;
 711                                         --pos;
 712                                 }
 713                                 break;
 714                 }
 715
 716         }
 717 }
 718
 719
 720 void Paragraph::rejectChanges(pos_type start, pos_type end)
 721 {
 722         LASSERT(start >= 0 && start <= size(), return);
 723         LASSERT(end > start && end <= size() + 1, return);
 724
 725         for (pos_type pos = start; pos < end; ++pos) {
 726                 switch (lookupChange(pos).type) {
 727                         case Change::UNCHANGED:
 728                                 // reject changes in nested inset
 729                                 if (Inset * inset = getInset(pos))
 730                                                 inset->rejectChanges();
 731                                 break;
 732
 733                         case Change::INSERTED:
 734                                 // Suppress access to non-existent
 735                                 // "end-of-paragraph char"
 736                                 if (pos < size()) {
 737                                         eraseChar(pos, false);
 738                                         --end;
 739                                         --pos;
 740                                 }
 741                                 break;
 742
 743                         case Change::DELETED:
 744                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 745
 746                                 // Do NOT reject changes within a deleted inset!
 747                                 // There may be insertions of a co-author inside of it!
 748
 749                                 break;
 750                 }
 751         }
 752 }
 753
 754
 755 void Paragraph::Private::insertChar(pos_type pos, char_type c,
 756                 Change const & change)
 757 {
 758         LASSERT(pos >= 0 && pos <= int(text_.size()), return);
 759
 760         // track change
 761         changes_.insert(change, pos);
 762
 763         // This is actually very common when parsing buffers (and
 764         // maybe inserting ascii text)
 765         if (pos == pos_type(text_.size())) {
 766                 // when appending characters, no need to update tables
 767                 text_.push_back(c);
 768                 // but we want spell checking
 769                 requestSpellCheck(pos);
 770                 return;
 771         }
 772
 773         text_.insert(text_.begin() + pos, c);
 774
 775         // Update the font table.
 776         fontlist_.increasePosAfterPos(pos);
 777
 778         // Update the insets
 779         insetlist_.increasePosAfterPos(pos);
 780
 781         // Update list of misspelled positions
 782         speller_state_.increasePosAfterPos(pos);
 783 }
 784
 785
 786 bool Paragraph::insertInset(pos_type pos, Inset * inset,
 787                                    Font const & font, Change const & change)
 788 {
 789         LASSERT(inset, return false);
 790         LASSERT(pos >= 0 && pos <= size(), return false);
 791
 792         // Paragraph::insertInset() can be used in cut/copy/paste operation where
 793         // d->inset_owner_ is not set yet.
 794         if (d->inset_owner_ && !d->inset_owner_->insetAllowed(inset->lyxCode()))
 795                 return false;
 796
 797         d->insertChar(pos, META_INSET, change);
 798         LASSERT(d->text_[pos] == META_INSET, return false);
 799
 800         // Add a new entry in the insetlist_.
 801         d->insetlist_.insert(inset, pos);
 802
 803         // Some insets require run of spell checker
 804         requestSpellCheck(pos);
 805         setFont(pos, font);
 806         return true;
 807 }
 808
 809
 810 bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
 811 {
 812         LASSERT(pos >= 0 && pos <= size(), return false);
 813
 814         // keep the logic here in sync with the logic of isMergedOnEndOfParDeletion()
 815
 816         if (trackChanges) {
 817                 Change change = d->changes_.lookup(pos);
 818
 819                 // set the character to DELETED if
 820                 //  a) it was previously unchanged or
 821                 //  b) it was inserted by a co-author
 822
 823                 if (!change.changed() ||
 824                       (change.inserted() && !change.currentAuthor())) {
 825                         setChange(pos, Change(Change::DELETED));
 826                         // request run of spell checker
 827                         requestSpellCheck(pos);
 828                         return false;
 829                 }
 830
 831                 if (change.deleted())
 832                         return false;
 833         }
 834
 835         // Don't physically access the imaginary end-of-paragraph character.
 836         // eraseChar() can only mark it as DELETED. A physical deletion of
 837         // end-of-par must be handled externally.
 838         if (pos == size()) {
 839                 return false;
 840         }
 841
 842         // track change
 843         d->changes_.erase(pos);
 844
 845         // if it is an inset, delete the inset entry
 846         if (d->text_[pos] == META_INSET)
 847                 d->insetlist_.erase(pos);
 848
 849         d->text_.erase(d->text_.begin() + pos);
 850
 851         // Update the fontlist_
 852         d->fontlist_.erase(pos);
 853
 854         // Update the insetlist_
 855         d->insetlist_.decreasePosAfterPos(pos);
 856
 857         // Update list of misspelled positions
 858         d->speller_state_.decreasePosAfterPos(pos);
 859         d->speller_state_.refreshLast(size());
 860
 861         return true;
 862 }
 863
 864
 865 int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
 866 {
 867         LASSERT(start >= 0 && start <= size(), return 0);
 868         LASSERT(end >= start && end <= size() + 1, return 0);
 869
 870         pos_type i = start;
 871         for (pos_type count = end - start; count; --count) {
 872                 if (!eraseChar(i, trackChanges))
 873                         ++i;
 874         }
 875         return end - i;
 876 }
 877
 878 // Handle combining characters
 879 int Paragraph::Private::latexSurrogatePair(BufferParams const & bparams,
 880                 otexstream & os, char_type c, char_type next,
 881                 OutputParams const & runparams)
 882 {
 883         // Writing next here may circumvent a possible font change between
 884         // c and next. Since next is only output if it forms a surrogate pair
 885         // with c we can ignore this:
 886         // A font change inside a surrogate pair does not make sense and is
 887         // hopefully impossible to input.
 888         // FIXME: change tracking
 889         // Is this correct WRT change tracking?
 890         Encoding const & encoding = *(runparams.encoding);
 891         docstring latex1 = encoding.latexChar(next).first;
 892         if (runparams.inIPA) {
 893                 string const tipashortcut = Encodings::TIPAShortcut(next);
 894                 if (!tipashortcut.empty()) {
 895                         latex1 = from_ascii(tipashortcut);
 896                 }
 897         }
 898         docstring latex2 = encoding.latexChar(c).first;
 899
 900         if (bparams.useNonTeXFonts || docstring(1, next) == latex1) {
 901                 // Encoding supports the combination:
 902                 // output as is (combining char after base char).
 903                 os << latex2 << latex1;
 904                 return latex1.length() + latex2.length();
 905         }
 906
 907         os << latex1 << "{" << latex2 << "}";
 908         return latex1.length() + latex2.length() + 2;
 909 }
 910
 911
 912 bool Paragraph::Private::simpleTeXBlanks(BufferParams const & bparams,
 913                                        OutputParams const & runparams,
 914                                        otexstream & os,
 915                                        pos_type i,
 916                                        unsigned int & column,
 917                                        Font const & font,
 918                                        Layout const & style)
 919 {
 920         if (style.pass_thru || runparams.pass_thru)
 921                 return false;
 922
 923         if (i + 1 < int(text_.size())) {
 924                 char_type next = text_[i + 1];
 925                 if (Encodings::isCombiningChar(next)) {
 926                         // This space has an accent, so we must always output it.
 927                         column += latexSurrogatePair(bparams, os, ' ', next, runparams) - 1;
 928                         return true;
 929                 }
 930         }
 931
 932         if (runparams.linelen > 0
 933             && column > runparams.linelen
 934             && i
 935             && text_[i - 1] != ' '
 936             && (i + 1 < int(text_.size()))
 937             // same in FreeSpacing mode
 938             && !owner_->isFreeSpacing()
 939             // In typewriter mode, we want to avoid
 940             // ! . ? : at the end of a line
 941             && !(font.fontInfo().family() == TYPEWRITER_FAMILY
 942                  && (text_[i - 1] == '.'
 943                      || text_[i - 1] == '?'
 944                      || text_[i - 1] == ':'
 945                      || text_[i - 1] == '!'))) {
 946                 os << '\n';
 947                 os.texrow().start(owner_->id(), i + 1);
 948                 column = 0;
 949         } else if (style.free_spacing) {
 950                 os << '~';
 951         } else {
 952                 os << ' ';
 953         }
 954         return false;
 955 }
 956
 957
 958 void Paragraph::Private::latexInset(BufferParams const & bparams,
 959                                     otexstream & os,
 960                                     OutputParams & runparams,
 961                                     Font & running_font,
 962                                     Font & basefont,
 963                                     Font const & outerfont,
 964                                     bool & open_font,
 965                                     Change & running_change,
 966                                     Layout const & style,
 967                                     pos_type & i,
 968                                     unsigned int & column)
 969 {
 970         Inset * inset = owner_->getInset(i);
 971         LBUFERR(inset);
 972
 973         if (style.pass_thru) {
 974                 odocstringstream ods;
 975                 inset->plaintext(ods, runparams);
 976                 os << ods.str();
 977                 return;
 978         }
 979
 980         // FIXME: move this to InsetNewline::latex
 981         if (inset->lyxCode() == NEWLINE_CODE || inset->lyxCode() == SEPARATOR_CODE) {
 982                 // newlines are handled differently here than
 983                 // the default in simpleTeXSpecialChars().
 984                 if (!style.newline_allowed) {
 985                         os << '\n';
 986                 } else {
 987                         if (open_font) {
 988                                 bool needPar = false;
 989                                 column += running_font.latexWriteEndChanges(
 990                                         os, bparams, runparams,
 991                                         basefont, basefont, needPar);
 992                                 open_font = false;
 993                         }
 994
 995                         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY)
 996                                 os << '~';
 997
 998                         basefont = owner_->getLayoutFont(bparams, outerfont);
 999                         running_font = basefont;
1000
1001                         if (runparams.moving_arg)
1002                                 os << "\\protect ";
1003
1004                 }
1005                 os.texrow().start(owner_->id(), i + 1);
1006                 column = 0;
1007         }
1008
1009         if (owner_->isDeleted(i)) {
1010                 if( ++runparams.inDeletedInset == 1)
1011                         runparams.changeOfDeletedInset = owner_->lookupChange(i);
1012         }
1013
1014         if (inset->canTrackChanges()) {
1015                 column += Changes::latexMarkChange(os, bparams, running_change,
1016                         Change(Change::UNCHANGED), runparams);
1017                 running_change = Change(Change::UNCHANGED);
1018         }
1019
1020         bool close = false;
1021         odocstream::pos_type const len = os.os().tellp();
1022
1023         if (inset->forceLTR(runparams)
1024             && running_font.isRightToLeft()
1025             // ERT is an exception, it should be output with no
1026             // decorations at all
1027             && inset->lyxCode() != ERT_CODE) {
1028                 if (runparams.use_polyglossia) {
1029                         os << "\\LRE{";
1030                 } else if (running_font.language()->lang() == "farsi"
1031                            || running_font.language()->lang() == "arabic_arabi")
1032                         os << "\\textLR{" << termcmd;
1033                 else
1034                         os << "\\L{";
1035                 close = true;
1036         }
1037
1038         // FIXME: Bug: we can have an empty font change here!
1039         // if there has just been a font change, we are going to close it
1040         // right now, which means stupid latex code like \textsf{}. AFAIK,
1041         // this does not harm dvi output. A minor bug, thus (JMarc)
1042
1043         // Some insets cannot be inside a font change command.
1044         // However, even such insets *can* be placed in \L or \R
1045         // or their equivalents (for RTL language switches), so we don't
1046         // close the language in those cases.
1047         // ArabTeX, though, cannot handle this special behavior, it seems.
1048         bool arabtex = basefont.language()->lang() == "arabic_arabtex"
1049                 || running_font.language()->lang() == "arabic_arabtex";
1050         if (open_font && !inset->inheritFont()) {
1051                 bool needPar = false;
1052                 bool closeLanguage = arabtex
1053                         || basefont.isRightToLeft() == running_font.isRightToLeft();
1054                 unsigned int count = running_font.latexWriteEndChanges(os,
1055                                         bparams, runparams, basefont, basefont,
1056                                         needPar, closeLanguage);
1057                 column += count;
1058                 // if any font properties were closed, update the running_font,
1059                 // making sure, however, to leave the language as it was
1060                 if (count > 0) {
1061                         // FIXME: probably a better way to keep track of the old
1062                         // language, than copying the entire font?
1063                         Font const copy_font(running_font);
1064                         basefont = owner_->getLayoutFont(bparams, outerfont);
1065                         running_font = basefont;
1066                         if (!closeLanguage)
1067                                 running_font.setLanguage(copy_font.language());
1068                         // leave font open if language is still open
1069                         open_font = (running_font.language() == basefont.language());
1070                         if (closeLanguage)
1071                                 runparams.local_font = &basefont;
1072                 }
1073         }
1074
1075         size_t const previous_row_count = os.texrow().rows();
1076
1077         try {
1078                 runparams.lastid = id_;
1079                 runparams.lastpos = i;
1080                 inset->latex(os, runparams);
1081         } catch (EncodingException & e) {
1082                 // add location information and throw again.
1083                 e.par_id = id_;
1084                 e.pos = i;
1085                 throw(e);
1086         }
1087
1088         if (close)
1089                 os << '}';
1090
1091         if (os.texrow().rows() > previous_row_count) {
1092                 os.texrow().start(owner_->id(), i + 1);
1093                 column = 0;
1094         } else {
1095                 column += (unsigned int)(os.os().tellp() - len);
1096         }
1097
1098         if (owner_->isDeleted(i))
1099                 --runparams.inDeletedInset;
1100 }
1101
1102
1103 void Paragraph::Private::latexSpecialChar(otexstream & os,
1104                                           BufferParams const & bparams,
1105                                           OutputParams const & runparams,
1106                                           Font const & running_font,
1107                                           string & alien_script,
1108                                           Layout const & style,
1109                                           pos_type & i,
1110                                           pos_type end_pos,
1111                                           unsigned int & column)
1112 {
1113         char_type const c = owner_->getUChar(bparams, runparams, i);
1114
1115         if (style.pass_thru || runparams.pass_thru
1116             || contains(style.pass_thru_chars, c)
1117             || contains(runparams.pass_thru_chars, c)) {
1118                 if (c != '\0') {
1119                         Encoding const * const enc = runparams.encoding;
1120                         if (enc && !enc->encodable(c))
1121                                 throw EncodingException(c);
1122                         os.put(c);
1123                 }
1124                 return;
1125         }
1126
1127         // TIPA uses its own T3 encoding
1128         if (runparams.inIPA && latexSpecialT3(c, os, i, column))
1129                 return;
1130         // If T1 font encoding is used, use the special
1131         // characters it provides.
1132         // NOTE: Some languages reset the font encoding internally to a
1133         //       non-standard font encoding. If we are using such a language,
1134         //       we do not output special T1 chars.
1135         if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1136             && !runparams.isFullUnicode() && bparams.main_font_encoding() == "T1"
1137             && latexSpecialT1(c, os, i, column))
1138                 return;
1139         // NOTE: "fontspec" (non-TeX fonts) sets the font encoding to "TU" (untill 2017 "EU1" or "EU2")
1140         else if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1141                  && runparams.isFullUnicode() && latexSpecialTU(c, os, i, column))
1142                      return;
1143
1144         // Otherwise, we use what LaTeX provides us.
1145         switch (c) {
1146         case '\\':
1147                 os << "\\textbackslash" << termcmd;
1148                 column += 15;
1149                 break;
1150         case '<':
1151                 os << "\\textless" << termcmd;
1152                 column += 10;
1153                 break;
1154         case '>':
1155                 os << "\\textgreater" << termcmd;
1156                 column += 13;
1157                 break;
1158         case '|':
1159                 os << "\\textbar" << termcmd;
1160                 column += 9;
1161                 break;
1162         case '-':
1163                 os << '-';
1164                 if (i + 1 < static_cast<pos_type>(text_.size()) &&
1165                     (end_pos == -1 || i + 1 < end_pos) &&
1166                     text_[i+1] == '-') {
1167                         // Prevent "--" becoming an en dash and "---" an em dash.
1168                         // (Within \ttfamily, "---" is merged to en dash + hyphen.)
1169                         os << "{}";
1170                         column += 2;
1171                 }
1172                 break;
1173         case '\"':
1174                 os << "\\textquotedbl" << termcmd;
1175                 column += 14;
1176                 break;
1177
1178         case '$': case '&':
1179         case '%': case '#': case '{':
1180         case '}': case '_':
1181                 os << '\\';
1182                 os.put(c);
1183                 column += 1;
1184                 break;
1185
1186         case '~':
1187                 os << "\\textasciitilde" << termcmd;
1188                 column += 16;
1189                 break;
1190
1191         case '^':
1192                 os << "\\textasciicircum" << termcmd;
1193                 column += 17;
1194                 break;
1195
1196         case '*':
1197         case '[':
1198         case ']':
1199                 // avoid being mistaken for optional arguments
1200                 os << '{';
1201                 os.put(c);
1202                 os << '}';
1203                 column += 2;
1204                 break;
1205
1206         case ' ':
1207                 // Blanks are printed before font switching.
1208                 // Sure? I am not! (try nice-latex)
1209                 // I am sure it's correct. LyX might be smarter
1210                 // in the future, but for now, nothing wrong is
1211                 // written. (Asger)
1212                 break;
1213
1214         case 0x2013:
1215         case 0x2014:
1216                 // XeTeX's dash behaviour is determined via a global setting
1217                 if (bparams.use_dash_ligatures
1218                     && owner_->getFontSettings(bparams, i).fontInfo().family() != TYPEWRITER_FAMILY
1219                     && !runparams.inIPA
1220                         // TODO #10961: && not in inset Flex Code
1221                         // TODO #10961: && not in layout LyXCode
1222                     && (!bparams.useNonTeXFonts || runparams.flavor != OutputParams::XETEX)) {
1223                         if (c == 0x2013) {
1224                                 // en-dash
1225                                 os << "--";
1226                                 column +=2;
1227                         } else {
1228                                 // em-dash
1229                                 os << "---";
1230                                 column +=3;
1231                         }
1232                         break;
1233                 }
1234                 // fall through
1235         default:
1236                 if (c == '\0')
1237                         return;
1238
1239                 Encoding const & encoding = *(runparams.encoding);
1240                 char_type next = '\0';
1241                 if (i + 1 < int(text_.size())) {
1242                         next = text_[i + 1];
1243                         if (Encodings::isCombiningChar(next)) {
1244                                 column += latexSurrogatePair(bparams, os, c, next, runparams) - 1;
1245                                 ++i;
1246                                 break;
1247                         }
1248                 }
1249                 pair<docstring, bool> latex = encoding.latexChar(c);
1250                 docstring nextlatex;
1251                 bool nexttipas = false;
1252                 string nexttipashortcut;
1253                 if (next != '\0' && next != META_INSET && !encoding.encodable(next)) {
1254                         nextlatex = encoding.latexChar(next).first;
1255                         if (runparams.inIPA) {
1256                                 nexttipashortcut = Encodings::TIPAShortcut(next);
1257                                 nexttipas = !nexttipashortcut.empty();
1258                         }
1259                 }
1260                 bool tipas = false;
1261                 if (runparams.inIPA) {
1262                         string const tipashortcut = Encodings::TIPAShortcut(c);
1263                         if (!tipashortcut.empty()) {
1264                                 latex.first = from_ascii(tipashortcut);
1265                                 latex.second = false;
1266                                 tipas = true;
1267                         }
1268                 }
1269                 // eventually close "script wrapper" command (see `Paragraph::latex`)
1270                 if (!alien_script.empty()
1271                         && alien_script != Encodings::isKnownScriptChar(next)) {
1272                         column += latex.first.length();
1273                         alien_script.clear();
1274                         os << latex.first << "}";
1275                         break;
1276                 }
1277                 if (latex.second
1278                          && ((!prefixIs(nextlatex, '\\')
1279                                && !prefixIs(nextlatex, '{')
1280                                && !prefixIs(nextlatex, '}'))
1281                              || (nexttipas
1282                                  && !prefixIs(from_ascii(nexttipashortcut), '\\')))
1283                          && !tipas) {
1284                         // Prevent eating of a following space or command corruption by
1285                         // following characters
1286                         if (next == ' ' || next == '\0') {
1287                                 column += latex.first.length() + 1;
1288                                 os << latex.first << "{}";
1289                         } else {
1290                                 column += latex.first.length();
1291                                 os << latex.first << " ";
1292                         }
1293                 } else {
1294                         column += latex.first.length() - 1;
1295                         os << latex.first;
1296                 }
1297                 break;
1298         }
1299 }
1300
1301
1302 bool Paragraph::Private::latexSpecialT1(char_type const c, otexstream & os,
1303         pos_type i, unsigned int & column)
1304 {
1305         switch (c) {
1306         case '>':
1307         case '<':
1308                 os.put(c);
1309                 // In T1 encoding, these characters exist
1310                 // but we should avoid ligatures
1311                 if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
1312                         return true;
1313                 os << "\\textcompwordmark" << termcmd;
1314                 column += 19;
1315                 return true;
1316         case '|':
1317                 os.put(c);
1318                 return true;
1319         case '\"':
1320                 // soul.sty breaks with \char`\"
1321                 os << "\\textquotedbl" << termcmd;
1322                 column += 14;
1323                 return true;
1324         default:
1325                 return false;
1326         }
1327 }
1328
1329
1330 bool Paragraph::Private::latexSpecialTU(char_type const c, otexstream & os,
1331         pos_type i, unsigned int & column)
1332 {
1333         // TU encoding is currently on par with T1.
1334         return latexSpecialT1(c, os, i, column);
1335 }
1336
1337
1338 bool Paragraph::Private::latexSpecialT3(char_type const c, otexstream & os,
1339         pos_type /*i*/, unsigned int & column)
1340 {
1341         switch (c) {
1342         case '*':
1343         case '[':
1344         case ']':
1345         case '\"':
1346                 os.put(c);
1347                 return true;
1348         case '|':
1349                 os << "\\textvertline" << termcmd;
1350                 column += 14;
1351                 return true;
1352         default:
1353                 return false;
1354         }
1355 }
1356
1357
1358 void Paragraph::Private::validate(LaTeXFeatures & features) const
1359 {
1360         if (layout_->inpreamble && inset_owner_) {
1361                 // FIXME: Using a string stream here circumvents the encoding
1362                 // switching machinery of odocstream. Therefore the
1363                 // output is wrong if this paragraph contains content
1364                 // that needs to switch encoding.
1365                 Buffer const & buf = inset_owner_->buffer();
1366                 otexstringstream os;
1367                 os << layout_->preamble();
1368                 size_t const length = os.length();
1369                 TeXOnePar(buf, *inset_owner_->getText(int(buf.getParFromID(owner_->id()).idx())),
1370                           buf.getParFromID(owner_->id()).pit(), os,
1371                           features.runparams(), string(), 0, -1, true);
1372                 if (os.length() > length)
1373                         features.addPreambleSnippet(os.release(), true);
1374         }
1375
1376         if (features.runparams().flavor == OutputParams::HTML
1377             && layout_->htmltitle()) {
1378                 features.setHTMLTitle(owner_->asString(AS_STR_INSETS | AS_STR_SKIPDELETE));
1379         }
1380
1381         // check the params.
1382         if (!params_.spacing().isDefault())
1383                 features.require("setspace");
1384
1385         // then the layouts
1386         features.useLayout(layout_->name());
1387
1388         // then the fonts
1389         fontlist_.validate(features);
1390
1391         // then the indentation
1392         if (!params_.leftIndent().zero())
1393                 features.require("ParagraphLeftIndent");
1394
1395         // then the insets
1396         InsetList::const_iterator icit = insetlist_.begin();
1397         InsetList::const_iterator iend = insetlist_.end();
1398         for (; icit != iend; ++icit) {
1399                 if (icit->inset) {
1400                         features.inDeletedInset(owner_->isDeleted(icit->pos));
1401                         if (icit->inset->lyxCode() == FOOT_CODE) {
1402                                 // FIXME: an item inset would make things much easier.
1403                                 if ((layout_->latextype == LATEX_LIST_ENVIRONMENT
1404                                      || (layout_->latextype == LATEX_ITEM_ENVIRONMENT
1405                                          && layout_->margintype == MARGIN_FIRST_DYNAMIC))
1406                                     && (icit->pos < begin_of_body_
1407                                         || (icit->pos == begin_of_body_
1408                                             && (icit->pos == 0 || text_[icit->pos - 1] != ' '))))
1409                                         features.saveNoteEnv("description");
1410                         }
1411                         icit->inset->validate(features);
1412                         features.inDeletedInset(false);
1413                         if (layout_->needprotect &&
1414                             icit->inset->lyxCode() == FOOT_CODE)
1415                                 features.require("NeedLyXFootnoteCode");
1416                 }
1417         }
1418
1419         // then the contents
1420         BufferParams const bp = features.runparams().is_child
1421                 ? features.buffer().masterParams() : features.buffer().params();
1422         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1423                 char_type c = text_[i];
1424                 CharInfo const & ci = Encodings::unicodeCharInfo(c);
1425                 if (c == 0x0022) {
1426                         if (features.runparams().isFullUnicode() && bp.useNonTeXFonts)
1427                                 features.require("textquotedblp");
1428                         else if (bp.main_font_encoding() != "T1"
1429                                  || ((&owner_->getFontSettings(bp, i))->language()->internalFontEncoding()))
1430                                 features.require("textquotedbl");
1431                 } else if (ci.textfeature() && contains(ci.textpreamble(), '=')) {
1432                         // features that depend on the font or input encoding
1433                         string feats = ci.textpreamble();
1434                         string fontenc = (&owner_->getFontSettings(bp, i))->language()->fontenc(bp);
1435                         if (fontenc.empty())
1436                                 fontenc = features.runparams().main_fontenc;
1437                         while (!feats.empty()) {
1438                                 string feat;
1439                                 feats = split(feats, feat, ',');
1440                                 if (contains(feat, "!=")) {
1441                                         // a feature that is required except for the spcified
1442                                         // font or input encodings
1443                                         string realfeature;
1444                                         string const contexts = ltrim(split(feat, realfeature, '!'), "=");
1445                                         // multiple encodings are separated by semicolon
1446                                         vector<string> context = getVectorFromString(contexts, ";");
1447                                         // require feature if the context matches neither current font
1448                                         // nor input encoding
1449                                         if (std::find(context.begin(), context.end(), fontenc) == context.end()
1450                                             && std::find(context.begin(), context.end(),
1451                                                          features.runparams().encoding->name()) == context.end())
1452                                                 features.require(realfeature);
1453                                 } else if (contains(feat, '=')) {
1454                                         // a feature that is required only for the spcified
1455                                         // font or input encodings
1456                                         string realfeature;
1457                                         string const contexts = split(feat, realfeature, '=');
1458                                         // multiple encodings are separated by semicolon
1459                                         vector<string> context = getVectorFromString(contexts, ";");
1460                                         // require feature if the context matches either current font
1461                                         // or input encoding
1462                                         if (std::find(context.begin(), context.end(), fontenc) != context.end()
1463                                             || std::find(context.begin(), context.end(),
1464                                                          features.runparams().encoding->name()) != context.end())
1465                                                 features.require(realfeature);
1466                                 }
1467                         }
1468                 } else if (!bp.use_dash_ligatures
1469                            && (c == 0x2013 || c == 0x2014)
1470                            && bp.useNonTeXFonts
1471                            && features.runparams().flavor == OutputParams::XETEX)
1472                         // XeTeX's dash behaviour is determined via a global setting
1473                         features.require("xetexdashbreakstate");
1474                 BufferEncodings::validate(c, features);
1475         }
1476 }
1477
1478 /////////////////////////////////////////////////////////////////////
1479 //
1480 // Paragraph
1481 //
1482 /////////////////////////////////////////////////////////////////////
1483
1484 namespace {
1485         Layout const emptyParagraphLayout;
1486 }
1487
1488 Paragraph::Paragraph()
1489         : d(new Paragraph::Private(this, emptyParagraphLayout))
1490 {
1491         itemdepth = 0;
1492         d->params_.clear();
1493 }
1494
1495
1496 Paragraph::Paragraph(Paragraph const & par)
1497         : itemdepth(par.itemdepth),
1498         d(new Paragraph::Private(*par.d, this))
1499 {
1500         registerWords();
1501 }
1502
1503
1504 Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
1505         : itemdepth(par.itemdepth),
1506         d(new Paragraph::Private(*par.d, this, beg, end))
1507 {
1508         registerWords();
1509 }
1510
1511
1512 Paragraph & Paragraph::operator=(Paragraph const & par)
1513 {
1514         // needed as we will destroy the private part before copying it
1515         if (&par != this) {
1516                 itemdepth = par.itemdepth;
1517
1518                 deregisterWords();
1519                 delete d;
1520                 d = new Private(*par.d, this);
1521                 registerWords();
1522         }
1523         return *this;
1524 }
1525
1526
1527 Paragraph::~Paragraph()
1528 {
1529         deregisterWords();
1530         delete d;
1531 }
1532
1533
1534 namespace {
1535
1536 // this shall be called just before every "os << ..." action.
1537 void flushString(ostream & os, docstring & s)
1538 {
1539         os << to_utf8(s);
1540         s.erase();
1541 }
1542
1543 } // namespace
1544
1545
1546 void Paragraph::write(ostream & os, BufferParams const & bparams,
1547         depth_type & dth) const
1548 {
1549         // The beginning or end of a deeper (i.e. nested) area?
1550         if (dth != d->params_.depth()) {
1551                 if (d->params_.depth() > dth) {
1552                         while (d->params_.depth() > dth) {
1553                                 os << "\n\\begin_deeper";
1554                                 ++dth;
1555                         }
1556                 } else {
1557                         while (d->params_.depth() < dth) {
1558                                 os << "\n\\end_deeper";
1559                                 --dth;
1560                         }
1561                 }
1562         }
1563
1564         // First write the layout
1565         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1566
1567         d->params_.write(os);
1568
1569         Font font1(inherit_font, bparams.language);
1570
1571         Change running_change = Change(Change::UNCHANGED);
1572
1573         // this string is used as a buffer to avoid repetitive calls
1574         // to to_utf8(), which turn out to be expensive (JMarc)
1575         docstring write_buffer;
1576
1577         int column = 0;
1578         for (pos_type i = 0; i <= size(); ++i) {
1579
1580                 Change const & change = lookupChange(i);
1581                 if (change != running_change)
1582                         flushString(os, write_buffer);
1583                 Changes::lyxMarkChange(os, bparams, column, running_change, change);
1584                 running_change = change;
1585
1586                 if (i == size())
1587                         break;
1588
1589                 // Write font changes
1590                 Font font2 = getFontSettings(bparams, i);
1591                 if (font2 != font1) {
1592                         flushString(os, write_buffer);
1593                         font2.lyxWriteChanges(font1, os);
1594                         column = 0;
1595                         font1 = font2;
1596                 }
1597
1598                 char_type const c = d->text_[i];
1599                 switch (c) {
1600                 case META_INSET:
1601                         if (Inset const * inset = getInset(i)) {
1602                                 flushString(os, write_buffer);
1603                                 if (inset->directWrite()) {
1604                                         // international char, let it write
1605                                         // code directly so it's shorter in
1606                                         // the file
1607                                         inset->write(os);
1608                                 } else {
1609                                         if (i)
1610                                                 os << '\n';
1611                                         os << "\\begin_inset ";
1612                                         inset->write(os);
1613                                         os << "\n\\end_inset\n\n";
1614                                         column = 0;
1615                                 }
1616                                 // FIXME This can be removed again once the mystery
1617                                 // crash has been resolved.
1618                                 os << flush;
1619                         }
1620                         break;
1621                 case '\\':
1622                         flushString(os, write_buffer);
1623                         os << "\n\\backslash\n";
1624                         column = 0;
1625                         break;
1626                 case '.':
1627                         flushString(os, write_buffer);
1628                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1629                                 os << ".\n";
1630                                 column = 0;
1631                         } else
1632                                 os << '.';
1633                         break;
1634                 default:
1635                         if ((column > 70 && c == ' ')
1636                             || column > 79) {
1637                                 flushString(os, write_buffer);
1638                                 os << '\n';
1639                                 column = 0;
1640                         }
1641                         // this check is to amend a bug. LyX sometimes
1642                         // inserts '\0' this could cause problems.
1643                         if (c != '\0')
1644                                 write_buffer.push_back(c);
1645                         else
1646                                 LYXERR0("NUL char in structure.");
1647                         ++column;
1648                         break;
1649                 }
1650         }
1651
1652         flushString(os, write_buffer);
1653         os << "\n\\end_layout\n";
1654         // FIXME This can be removed again once the mystery
1655         // crash has been resolved.
1656         os << flush;
1657 }
1658
1659
1660 void Paragraph::validate(LaTeXFeatures & features) const
1661 {
1662         d->validate(features);
1663         bool fragile = features.runparams().moving_arg;
1664         fragile |= layout().needprotect;
1665         if (inInset().getLayout().isNeedProtect())
1666                 fragile = true;
1667         if (needsCProtection(fragile))
1668                 features.require("cprotect");
1669 }
1670
1671
1672 void Paragraph::insert(pos_type start, docstring const & str,
1673                        Font const & font, Change const & change)
1674 {
1675         for (size_t i = 0, n = str.size(); i != n ; ++i)
1676                 insertChar(start + i, str[i], font, change);
1677 }
1678
1679
1680 void Paragraph::appendChar(char_type c, Font const & font,
1681                 Change const & change)
1682 {
1683         // track change
1684         d->changes_.insert(change, d->text_.size());
1685         // when appending characters, no need to update tables
1686         d->text_.push_back(c);
1687         setFont(d->text_.size() - 1, font);
1688         d->requestSpellCheck(d->text_.size() - 1);
1689 }
1690
1691
1692 void Paragraph::appendString(docstring const & s, Font const & font,
1693                 Change const & change)
1694 {
1695         pos_type end = s.size();
1696         size_t oldsize = d->text_.size();
1697         size_t newsize = oldsize + end;
1698         size_t capacity = d->text_.capacity();
1699         if (newsize >= capacity)
1700                 d->text_.reserve(max(capacity + 100, newsize));
1701
1702         // when appending characters, no need to update tables
1703         d->text_.append(s);
1704
1705         // FIXME: Optimize this!
1706         for (size_t i = oldsize; i != newsize; ++i) {
1707                 // track change
1708                 d->changes_.insert(change, i);
1709                 d->requestSpellCheck(i);
1710         }
1711         d->fontlist_.set(oldsize, font);
1712         d->fontlist_.set(newsize - 1, font);
1713 }
1714
1715
1716 void Paragraph::insertChar(pos_type pos, char_type c,
1717                            bool trackChanges)
1718 {
1719         d->insertChar(pos, c, Change(trackChanges ?
1720                            Change::INSERTED : Change::UNCHANGED));
1721 }
1722
1723
1724 void Paragraph::insertChar(pos_type pos, char_type c,
1725                            Font const & font, bool trackChanges)
1726 {
1727         d->insertChar(pos, c, Change(trackChanges ?
1728                            Change::INSERTED : Change::UNCHANGED));
1729         setFont(pos, font);
1730 }
1731
1732
1733 void Paragraph::insertChar(pos_type pos, char_type c,
1734                            Font const & font, Change const & change)
1735 {
1736         d->insertChar(pos, c, change);
1737         setFont(pos, font);
1738 }
1739
1740
1741 void Paragraph::resetFonts(Font const & font)
1742 {
1743         d->fontlist_.clear();
1744         d->fontlist_.set(0, font);
1745         d->fontlist_.set(d->text_.size() - 1, font);
1746 }
1747
1748 // Gets uninstantiated font setting at position.
1749 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1750                                          pos_type pos) const
1751 {
1752         if (pos > size()) {
1753                 LYXERR0("pos: " << pos << " size: " << size());
1754                 LBUFERR(false);
1755         }
1756
1757         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1758         if (cit != d->fontlist_.end())
1759                 return cit->font();
1760
1761         if (pos == size() && !empty())
1762                 return getFontSettings(bparams, pos - 1);
1763
1764         // Optimisation: avoid a full font instantiation if there is no
1765         // language change from previous call.
1766         static Font previous_font;
1767         static Language const * previous_lang = nullptr;
1768         Language const * lang = getParLanguage(bparams);
1769         if (lang != previous_lang) {
1770                 previous_lang = lang;
1771                 previous_font = Font(inherit_font, lang);
1772         }
1773         return previous_font;
1774 }
1775
1776
1777 FontSpan Paragraph::fontSpan(pos_type pos) const
1778 {
1779         LBUFERR(pos <= size());
1780
1781         if (pos == size())
1782                 return FontSpan(pos, pos);
1783
1784         pos_type start = 0;
1785         FontList::const_iterator cit = d->fontlist_.begin();
1786         FontList::const_iterator end = d->fontlist_.end();
1787         for (; cit != end; ++cit) {
1788                 if (cit->pos() >= pos) {
1789                         if (pos >= beginOfBody())
1790                                 return FontSpan(max(start, beginOfBody()),
1791                                                 cit->pos());
1792                         else
1793                                 return FontSpan(start,
1794                                                 min(beginOfBody() - 1,
1795                                                          cit->pos()));
1796                 }
1797                 start = cit->pos() + 1;
1798         }
1799
1800         // This should not happen, but if so, we take no chances.
1801         LYXERR0("Paragraph::fontSpan: position not found in fontinfo table!");
1802         LASSERT(false, return FontSpan(pos, pos));
1803 }
1804
1805
1806 // Gets uninstantiated font setting at position 0
1807 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1808 {
1809         if (!empty() && !d->fontlist_.empty())
1810                 return d->fontlist_.begin()->font();
1811
1812         // Optimisation: avoid a full font instantiation if there is no
1813         // language change from previous call.
1814         static Font previous_font;
1815         static Language const * previous_lang = nullptr;
1816         if (bparams.language != previous_lang) {
1817                 previous_lang = bparams.language;
1818                 previous_font = Font(inherit_font, bparams.language);
1819         }
1820
1821         return previous_font;
1822 }
1823
1824
1825 // Gets the fully instantiated font at a given position in a paragraph
1826 // This is basically the same function as Text::GetFont() in text2.cpp.
1827 // The difference is that this one is used for generating the LaTeX file,
1828 // and thus cosmetic "improvements" are disallowed: This has to deliver
1829 // the true picture of the buffer. (Asger)
1830 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1831                                  Font const & outerfont) const
1832 {
1833         LBUFERR(pos >= 0);
1834
1835         Font font = getFontSettings(bparams, pos);
1836
1837         pos_type const body_pos = beginOfBody();
1838         FontInfo & fi = font.fontInfo();
1839         if (pos < body_pos)
1840                 fi.realize(d->layout_->labelfont);
1841         else
1842                 fi.realize(d->layout_->font);
1843
1844         fi.realize(outerfont.fontInfo());
1845         fi.realize(bparams.getFont().fontInfo());
1846
1847         return font;
1848 }
1849
1850
1851 Font const Paragraph::getLabelFont
1852         (BufferParams const & bparams, Font const & outerfont) const
1853 {
1854         FontInfo tmpfont = d->layout_->labelfont;
1855         tmpfont.realize(outerfont.fontInfo());
1856         tmpfont.realize(bparams.getFont().fontInfo());
1857         return Font(tmpfont, getParLanguage(bparams));
1858 }
1859
1860
1861 Font const Paragraph::getLayoutFont
1862         (BufferParams const & bparams, Font const & outerfont) const
1863 {
1864         FontInfo tmpfont = d->layout_->font;
1865         tmpfont.realize(outerfont.fontInfo());
1866         tmpfont.realize(bparams.getFont().fontInfo());
1867         return Font(tmpfont, getParLanguage(bparams));
1868 }
1869
1870
1871 char_type Paragraph::getUChar(BufferParams const & bparams,
1872                               OutputParams const & rp,
1873                               pos_type pos) const
1874 {
1875         char_type c = d->text_[pos];
1876
1877         // Return unchanged character in LTR languages
1878         // or if we use poylglossia/bidi (XeTeX).
1879         if (rp.useBidiPackage()
1880             || !getFontSettings(bparams, pos).isRightToLeft())
1881                 return c;
1882
1883         // Without polyglossia/bidi, we need to account for some special cases.
1884         // FIXME This needs to be audited!
1885         // Check if:
1886         // * The input is as expected for all delimiters
1887         //   => checked for Hebrew!
1888         // * The output matches the display in the LyX workarea
1889         //   => checked for Hebrew!
1890         // * The special cases below are really necessary
1891         //   => checked for Hebrew!
1892         // * In arabic_arabi, brackets are transformed to Arabic
1893         //   Ornate Parentheses. Is this is really wanted?
1894
1895         string const & lang = getFontSettings(bparams, pos).language()->lang();
1896         char_type uc = c;
1897
1898         // 1. In the following languages, parentheses need to be reversed.
1899         //    Also with polyglodia/luabidi
1900         bool const reverseparens = (lang == "hebrew" || rp.use_polyglossia);
1901
1902         // 2. In the following languages, brackets don't need to be reversed.
1903         bool const reversebrackets = lang != "arabic_arabtex"
1904                         && lang != "arabic_arabi"
1905                         && lang != "farsi";
1906
1907         // Now swap delimiters if needed.
1908         switch (c) {
1909         case '(':
1910                 if (reverseparens)
1911                         uc = ')';
1912                 break;
1913         case ')':
1914                 if (reverseparens)
1915                         uc = '(';
1916                 break;
1917         case '[':
1918                 if (reversebrackets)
1919                         uc = ']';
1920                 break;
1921         case ']':
1922                 if (reversebrackets)
1923                         uc = '[';
1924                 break;
1925         case '{':
1926                 uc = '}';
1927                 break;
1928         case '}':
1929                 uc = '{';
1930                 break;
1931         case '<':
1932                 uc = '>';
1933                 break;
1934         case '>':
1935                 uc = '<';
1936                 break;
1937         }
1938
1939         return uc;
1940 }
1941
1942
1943 void Paragraph::setFont(pos_type pos, Font const & font)
1944 {
1945         LASSERT(pos <= size(), return);
1946
1947         // First, reduce font against layout/label font
1948         // Update: The setCharFont() routine in text2.cpp already
1949         // reduces font, so we don't need to do that here. (Asger)
1950
1951         d->fontlist_.set(pos, font);
1952 }
1953
1954
1955 void Paragraph::makeSameLayout(Paragraph const & par)
1956 {
1957         d->layout_ = par.d->layout_;
1958         d->params_ = par.d->params_;
1959 }
1960
1961
1962 bool Paragraph::stripLeadingSpaces(bool trackChanges)
1963 {
1964         if (isFreeSpacing())
1965                 return false;
1966
1967         int pos = 0;
1968         int count = 0;
1969
1970         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
1971                 if (eraseChar(pos, trackChanges))
1972                         ++count;
1973                 else
1974                         ++pos;
1975         }
1976
1977         return count > 0 || pos > 0;
1978 }
1979
1980
1981 bool Paragraph::hasSameLayout(Paragraph const & par) const
1982 {
1983         return par.d->layout_ == d->layout_
1984                 && d->params_.sameLayout(par.d->params_);
1985 }
1986
1987
1988 depth_type Paragraph::getDepth() const
1989 {
1990         return d->params_.depth();
1991 }
1992
1993
1994 depth_type Paragraph::getMaxDepthAfter() const
1995 {
1996         if (d->layout_->isEnvironment())
1997                 return d->params_.depth() + 1;
1998         else
1999                 return d->params_.depth();
2000 }
2001
2002
2003 LyXAlignment Paragraph::getAlign(BufferParams const & bparams) const
2004 {
2005         if (d->params_.align() == LYX_ALIGN_LAYOUT)
2006                 return getDefaultAlign(bparams);
2007         else
2008                 return d->params_.align();
2009 }
2010
2011
2012 LyXAlignment Paragraph::getDefaultAlign(BufferParams const & bparams) const
2013 {
2014         LyXAlignment res = layout().align;
2015         if (isRTL(bparams)) {
2016                 // Swap sides
2017                 if (res == LYX_ALIGN_LEFT)
2018                         res = LYX_ALIGN_RIGHT;
2019                 else if  (res == LYX_ALIGN_RIGHT)
2020                         res = LYX_ALIGN_LEFT;
2021         }
2022         return res;
2023 }
2024
2025
2026 docstring const & Paragraph::labelString() const
2027 {
2028         return d->params_.labelString();
2029 }
2030
2031
2032 // the next two functions are for the manual labels
2033 docstring const Paragraph::getLabelWidthString() const
2034 {
2035         if (d->layout_->margintype == MARGIN_MANUAL
2036             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
2037                 return d->params_.labelWidthString();
2038         else
2039                 return _("Senseless with this layout!");
2040 }
2041
2042
2043 void Paragraph::setLabelWidthString(docstring const & s)
2044 {
2045         d->params_.labelWidthString(s);
2046 }
2047
2048
2049 docstring Paragraph::expandLabel(Layout const & layout,
2050                 BufferParams const & bparams) const
2051 {
2052         return expandParagraphLabel(layout, bparams, true);
2053 }
2054
2055
2056 docstring Paragraph::expandDocBookLabel(Layout const & layout,
2057                 BufferParams const & bparams) const
2058 {
2059         return expandParagraphLabel(layout, bparams, false);
2060 }
2061
2062
2063 docstring Paragraph::expandParagraphLabel(Layout const & layout,
2064                 BufferParams const & bparams, bool process_appendix) const
2065 {
2066         DocumentClass const & tclass = bparams.documentClass();
2067         string const & lang = getParLanguage(bparams)->code();
2068         bool const in_appendix = process_appendix && d->params_.appendix();
2069         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
2070
2071         if (fmt.empty() && !layout.counter.empty())
2072                 return tclass.counters().theCounter(layout.counter, lang);
2073
2074         // handle 'inherited level parts' in 'fmt',
2075         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
2076         size_t const i = fmt.find('@', 0);
2077         if (i != docstring::npos) {
2078                 size_t const j = fmt.find('@', i + 1);
2079                 if (j != docstring::npos) {
2080                         docstring parent(fmt, i + 1, j - i - 1);
2081                         docstring label = from_ascii("??");
2082                         if (tclass.hasLayout(parent))
2083                                 label = expandParagraphLabel(tclass[parent], bparams,
2084                                                       process_appendix);
2085                         fmt = docstring(fmt, 0, i) + label
2086                                 + docstring(fmt, j + 1, docstring::npos);
2087                 }
2088         }
2089
2090         return tclass.counters().counterLabel(fmt, lang);
2091 }
2092
2093
2094 void Paragraph::applyLayout(Layout const & new_layout)
2095 {
2096         d->layout_ = &new_layout;
2097         LyXAlignment const oldAlign = d->params_.align();
2098
2099         if (!(oldAlign & d->layout_->alignpossible)) {
2100                 frontend::Alert::warning(_("Alignment not permitted"),
2101                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
2102                 d->params_.align(LYX_ALIGN_LAYOUT);
2103         }
2104 }
2105
2106
2107 pos_type Paragraph::beginOfBody() const
2108 {
2109         return d->begin_of_body_;
2110 }
2111
2112
2113 void Paragraph::setBeginOfBody()
2114 {
2115         if (d->layout_->labeltype != LABEL_MANUAL) {
2116                 d->begin_of_body_ = 0;
2117                 return;
2118         }
2119
2120         // Unroll the first two cycles of the loop
2121         // and remember the previous character to
2122         // remove unnecessary getChar() calls
2123         pos_type i = 0;
2124         pos_type end = size();
2125         bool prev_char_deleted = false;
2126         if (i < end && (!(isNewline(i) || isEnvSeparator(i)) || isDeleted(i))) {
2127                 ++i;
2128                 if (i < end) {
2129                         char_type previous_char = d->text_[i];
2130                         if (!(isNewline(i) || isEnvSeparator(i))) {
2131                                 ++i;
2132                                 while (i < end && (previous_char != ' ' || prev_char_deleted)) {
2133                                         char_type temp = d->text_[i];
2134                                         prev_char_deleted = isDeleted(i);
2135                                         if (!isDeleted(i) && (isNewline(i) || isEnvSeparator(i)))
2136                                                 break;
2137                                         ++i;
2138                                         previous_char = temp;
2139                                 }
2140                         }
2141                 }
2142         }
2143
2144         d->begin_of_body_ = i;
2145 }
2146
2147
2148 bool Paragraph::allowParagraphCustomization() const
2149 {
2150         return inInset().allowParagraphCustomization();
2151 }
2152
2153
2154 bool Paragraph::usePlainLayout() const
2155 {
2156         return inInset().usePlainLayout();
2157 }
2158
2159
2160 bool Paragraph::isPassThru() const
2161 {
2162         return inInset().isPassThru() || d->layout_->pass_thru;
2163 }
2164
2165 namespace {
2166
2167 // paragraphs inside floats need different alignment tags to avoid
2168 // unwanted space
2169
2170 bool noTrivlistCentering(InsetCode code)
2171 {
2172         return code == FLOAT_CODE
2173                || code == WRAP_CODE
2174                || code == CELL_CODE;
2175 }
2176
2177
2178 string correction(string const & orig)
2179 {
2180         if (orig == "flushleft")
2181                 return "raggedright";
2182         if (orig == "flushright")
2183                 return "raggedleft";
2184         if (orig == "center")
2185                 return "centering";
2186         return orig;
2187 }
2188
2189
2190 bool corrected_env(otexstream & os, string const & suffix, string const & env,
2191         InsetCode code, bool const lastpar, int & col)
2192 {
2193         string macro = suffix + "{";
2194         if (noTrivlistCentering(code)) {
2195                 if (lastpar) {
2196                         // the last paragraph in non-trivlist-aligned
2197                         // context is special (to avoid unwanted whitespace)
2198                         if (suffix == "\\begin") {
2199                                 macro = "\\" + correction(env) + "{}";
2200                                 os << from_ascii(macro);
2201                                 col += macro.size();
2202                                 return true;
2203                         }
2204                         return false;
2205                 }
2206                 macro += correction(env);
2207         } else
2208                 macro += env;
2209         macro += "}";
2210         if (suffix == "\\par\\end") {
2211                 os << breakln;
2212                 col = 0;
2213         }
2214         os << from_ascii(macro);
2215         col += macro.size();
2216         if (suffix == "\\begin") {
2217                 os << breakln;
2218                 col = 0;
2219         }
2220         return true;
2221 }
2222
2223 } // namespace
2224
2225
2226 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
2227                         otexstream & os, OutputParams const & runparams) const
2228 {
2229         int column = 0;
2230
2231         bool canindent =
2232                 (bparams.paragraph_separation == BufferParams::ParagraphIndentSeparation) ?
2233                         (layout_->toggle_indent != ITOGGLE_NEVER) :
2234                         (layout_->toggle_indent == ITOGGLE_ALWAYS);
2235
2236         if (canindent && params_.noindent() && !layout_->pass_thru) {
2237                 os << "\\noindent ";
2238                 column += 10;
2239         }
2240
2241         LyXAlignment const curAlign = params_.align();
2242
2243         if (curAlign == layout_->align)
2244                 return column;
2245
2246         switch (curAlign) {
2247         case LYX_ALIGN_NONE:
2248         case LYX_ALIGN_BLOCK:
2249         case LYX_ALIGN_LAYOUT:
2250         case LYX_ALIGN_SPECIAL:
2251         case LYX_ALIGN_DECIMAL:
2252                 break;
2253         case LYX_ALIGN_LEFT:
2254         case LYX_ALIGN_RIGHT:
2255         case LYX_ALIGN_CENTER:
2256                 if (runparams.moving_arg) {
2257                         os << "\\protect";
2258                         column += 8;
2259                 }
2260                 break;
2261         }
2262
2263         string const begin_tag = "\\begin";
2264         InsetCode code = ownerCode();
2265         bool const lastpar = runparams.isLastPar;
2266         // RTL in classic (PDF)LaTeX (without the Bidi package)
2267         // Luabibdi (used by LuaTeX) behaves like classic
2268         bool const rtl_classic = owner_->getParLanguage(bparams)->rightToLeft()
2269                 && !runparams.useBidiPackage();
2270
2271         switch (curAlign) {
2272         case LYX_ALIGN_NONE:
2273         case LYX_ALIGN_BLOCK:
2274         case LYX_ALIGN_LAYOUT:
2275         case LYX_ALIGN_SPECIAL:
2276         case LYX_ALIGN_DECIMAL:
2277                 break;
2278         case LYX_ALIGN_LEFT: {
2279                 if (rtl_classic)
2280                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2281                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2282                 else
2283                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2284                 break;
2285         } case LYX_ALIGN_RIGHT: {
2286                 if (rtl_classic)
2287                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2288                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2289                 else
2290                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2291                 break;
2292         } case LYX_ALIGN_CENTER: {
2293                 corrected_env(os, begin_tag, "center", code, lastpar, column);
2294                 break;
2295         }
2296         }
2297
2298         return column;
2299 }
2300
2301
2302 bool Paragraph::Private::endTeXParParams(BufferParams const & bparams,
2303                         otexstream & os, OutputParams const & runparams) const
2304 {
2305         LyXAlignment const curAlign = params_.align();
2306
2307         if (curAlign == layout_->align)
2308                 return false;
2309
2310         switch (curAlign) {
2311         case LYX_ALIGN_NONE:
2312         case LYX_ALIGN_BLOCK:
2313         case LYX_ALIGN_LAYOUT:
2314         case LYX_ALIGN_SPECIAL:
2315         case LYX_ALIGN_DECIMAL:
2316                 break;
2317         case LYX_ALIGN_LEFT:
2318         case LYX_ALIGN_RIGHT:
2319         case LYX_ALIGN_CENTER:
2320                 if (runparams.moving_arg)
2321                         os << "\\protect";
2322                 break;
2323         }
2324
2325         bool output = false;
2326         int col = 0;
2327         string const end_tag = "\\par\\end";
2328         InsetCode code = ownerCode();
2329         bool const lastpar = runparams.isLastPar;
2330         // RTL in classic (PDF)LaTeX (without the Bidi package)
2331         // Luabibdi (used by LuaTeX) behaves like classic
2332         bool const rtl_classic = owner_->getParLanguage(bparams)->rightToLeft()
2333                 && !runparams.useBidiPackage();
2334
2335         switch (curAlign) {
2336         case LYX_ALIGN_NONE:
2337         case LYX_ALIGN_BLOCK:
2338         case LYX_ALIGN_LAYOUT:
2339         case LYX_ALIGN_SPECIAL:
2340         case LYX_ALIGN_DECIMAL:
2341                 break;
2342         case LYX_ALIGN_LEFT: {
2343                 if (rtl_classic)
2344                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2345                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2346                 else
2347                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2348                 break;
2349         } case LYX_ALIGN_RIGHT: {
2350                 if (rtl_classic)
2351                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2352                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2353                 else
2354                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2355                 break;
2356         } case LYX_ALIGN_CENTER: {
2357                 corrected_env(os, end_tag, "center", code, lastpar, col);
2358                 break;
2359         }
2360         }
2361
2362         return output || lastpar;
2363 }
2364
2365
2366 // This one spits out the text of the paragraph
2367 void Paragraph::latex(BufferParams const & bparams,
2368         Font const & outerfont,
2369         otexstream & os,
2370         OutputParams const & runparams,
2371         int start_pos, int end_pos, bool force) const
2372 {
2373         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
2374
2375         // FIXME This check should not be needed. Perhaps issue an
2376         // error if it triggers.
2377         Layout const & style = inInset().forcePlainLayout() ?
2378                 bparams.documentClass().plainLayout() : *d->layout_;
2379
2380         if (!force && style.inpreamble)
2381                 return;
2382
2383         bool const allowcust = allowParagraphCustomization();
2384
2385         // Current base font for all inherited font changes, without any
2386         // change caused by an individual character, except for the language:
2387         // It is set to the language of the first character.
2388         // As long as we are in the label, this font is the base font of the
2389         // label. Before the first body character it is set to the base font
2390         // of the body.
2391         Font basefont;
2392
2393         // If there is an open font-encoding changing command (script wrapper),
2394         // alien_script is set to its name
2395         string alien_script;
2396         string script;
2397
2398         // Maybe we have to create a optional argument.
2399         pos_type body_pos = beginOfBody();
2400         unsigned int column = 0;
2401
2402         if (body_pos > 0) {
2403                 // the optional argument is kept in curly brackets in
2404                 // case it contains a ']'
2405                 // This is not strictly needed, but if this is changed it
2406                 // would be a file format change, and tex2lyx would need
2407                 // to be adjusted, since it unconditionally removes the
2408                 // braces when it parses \item.
2409                 os << "[{";
2410                 column += 2;
2411                 basefont = getLabelFont(bparams, outerfont);
2412         } else {
2413                 basefont = getLayoutFont(bparams, outerfont);
2414         }
2415
2416         // Which font is currently active?
2417         Font running_font(basefont);
2418         // Do we have an open font change?
2419         bool open_font = false;
2420
2421         Change runningChange = Change(Change::UNCHANGED);
2422
2423         Encoding const * const prev_encoding = runparams.encoding;
2424
2425         os.texrow().start(id(), 0);
2426
2427         // if the paragraph is empty, the loop will not be entered at all
2428         if (empty()) {
2429                 // For InTitle commands, we have already opened a group
2430                 // in output_latex::TeXOnePar.
2431                 if (style.isCommand() && !style.intitle) {
2432                         os << '{';
2433                         ++column;
2434                 }
2435                 if (!style.leftdelim().empty()) {
2436                         os << style.leftdelim();
2437                         column += style.leftdelim().size();
2438                 }
2439                 if (allowcust)
2440                         column += d->startTeXParParams(bparams, os, runparams);
2441         }
2442
2443         // Whether a \par can be issued for insets typeset inline with text.
2444         // Yes if greater than 0. This has to be static.
2445         THREAD_LOCAL_STATIC int parInline = 0;
2446
2447         for (pos_type i = 0; i < size(); ++i) {
2448                 // First char in paragraph or after label?
2449                 if (i == body_pos) {
2450                         if (body_pos > 0) {
2451                                 if (open_font) {
2452                                         bool needPar = false;
2453                                         column += running_font.latexWriteEndChanges(
2454                                                 os, bparams, runparams,
2455                                                 basefont, basefont, needPar);
2456                                         open_font = false;
2457                                 }
2458                                 basefont = getLayoutFont(bparams, outerfont);
2459                                 running_font = basefont;
2460
2461                                 column += Changes::latexMarkChange(os, bparams,
2462                                                 runningChange, Change(Change::UNCHANGED),
2463                                                 runparams);
2464                                 runningChange = Change(Change::UNCHANGED);
2465
2466                                 os << "}] ";
2467                                 column +=3;
2468                         }
2469                         // For InTitle commands, we have already opened a group
2470                         // in output_latex::TeXOnePar.
2471                         if (style.isCommand() && !style.intitle) {
2472                                 os << '{';
2473                                 ++column;
2474                         }
2475
2476                         if (!style.leftdelim().empty()) {
2477                                 os << style.leftdelim();
2478                                 column += style.leftdelim().size();
2479                         }
2480
2481                         if (allowcust)
2482                                 column += d->startTeXParParams(bparams, os,
2483                                                             runparams);
2484                 }
2485
2486                 runparams.wasDisplayMath = runparams.inDisplayMath;
2487                 runparams.inDisplayMath = false;
2488                 bool deleted_display_math = false;
2489                 Change const & change = runparams.inDeletedInset
2490                         ? runparams.changeOfDeletedInset : lookupChange(i);
2491
2492                 char_type const c = d->text_[i];
2493
2494                 // Check whether a display math inset follows
2495                 if (c == META_INSET
2496                     && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2497                         if (isDeleted(i))
2498                                 runparams.ctObject = getInset(i)->CtObject(runparams);
2499
2500                         InsetMath const * im = getInset(i)->asInsetMath();
2501                         if (im && im->asHullInset()
2502                             && im->asHullInset()->outerDisplay()) {
2503                                 runparams.inDisplayMath = true;
2504                                 // runparams.inDeletedInset will be set by
2505                                 // latexInset later, but we need this info
2506                                 // before it is called. On the other hand, we
2507                                 // cannot set it here because it is a counter.
2508                                 deleted_display_math = isDeleted(i);
2509                         }
2510                         if (bparams.output_changes && deleted_display_math
2511                             && runningChange == change
2512                             && change.type == Change::DELETED
2513                             && !os.afterParbreak()) {
2514                                 // A display math in the same paragraph follows.
2515                                 // We have to close and then reopen \lyxdeleted,
2516                                 // otherwise the math will be shifted up.
2517                                 OutputParams rp = runparams;
2518                                 if (open_font) {
2519                                         bool needPar = false;
2520                                         column += running_font.latexWriteEndChanges(
2521                                                 os, bparams, rp, basefont,
2522                                                 basefont, needPar);
2523                                         open_font = false;
2524                                 }
2525                                 basefont = (body_pos > i) ? getLabelFont(bparams, outerfont)
2526                                                           : getLayoutFont(bparams, outerfont);
2527                                 running_font = basefont;
2528                                 column += Changes::latexMarkChange(os, bparams,
2529                                         Change(Change::INSERTED), change, rp);
2530                         }
2531                 }
2532
2533                 if (bparams.output_changes && runningChange != change) {
2534                         if (!alien_script.empty()) {
2535                                 column += 1;
2536                                 os << "}";
2537                                 alien_script.clear();
2538                         }
2539                         if (open_font) {
2540                                 bool needPar = false;
2541                                 column += running_font.latexWriteEndChanges(
2542                                                 os, bparams, runparams,
2543                                                 basefont, basefont, needPar);
2544                                 open_font = false;
2545                         }
2546                         basefont = (body_pos > i) ? getLabelFont(bparams, outerfont)
2547                                                   : getLayoutFont(bparams, outerfont);
2548                         running_font = basefont;
2549                         column += Changes::latexMarkChange(os, bparams, runningChange,
2550                                                            change, runparams);
2551                         runningChange = change;
2552                 }
2553
2554                 // do not output text which is marked deleted
2555                 // if change tracking output is disabled
2556                 if (!bparams.output_changes && change.deleted()) {
2557                         continue;
2558                 }
2559
2560                 ++column;
2561
2562                 // Fully instantiated font
2563                 Font const current_font = getFont(bparams, i, outerfont);
2564
2565                 Font const last_font = running_font;
2566                 bool const in_ct_deletion = (bparams.output_changes
2567                                   && runningChange == change
2568                                   && change.type == Change::DELETED
2569                                   && !os.afterParbreak());
2570
2571                 // Do we need to close the previous font?
2572                 if (open_font &&
2573                     (current_font != running_font ||
2574                      current_font.language() != running_font.language()))
2575                 {
2576                         // ensure there is no open script-wrapper
2577                         if (!alien_script.empty()) {
2578                                 column += 1;
2579                                 os << "}";
2580                                 alien_script.clear();
2581                         }
2582                         bool needPar = false;
2583                         if (in_ct_deletion) {
2584                                 // We have to close and then reopen \lyxdeleted,
2585                                 // as strikeout needs to be on lowest level.
2586                                 os << '}';
2587                                 column += 1;
2588                         }
2589                         column += running_font.latexWriteEndChanges(
2590                                     os, bparams, runparams, basefont,
2591                                     (i == body_pos-1) ? basefont : current_font,
2592                                     needPar);
2593                         if (in_ct_deletion) {
2594                                 // We have to close and then reopen \lyxdeleted,
2595                                 // as strikeout needs to be on lowest level.
2596                                 OutputParams rp = runparams;
2597                                 column += Changes::latexMarkChange(os, bparams,
2598                                         Change(Change::UNCHANGED), Change(Change::DELETED), rp);
2599                         }
2600                         running_font = basefont;
2601                         open_font = false;
2602                 }
2603
2604                 // if necessary, close language environment before opening CJK
2605                 string const running_lang = running_font.language()->babel();
2606                 string const lang_end_command = lyxrc.language_command_end;
2607                 if (!lang_end_command.empty() && !bparams.useNonTeXFonts
2608                         && !running_lang.empty()
2609                         && running_lang == openLanguageName()
2610                         && current_font.language()->encoding()->package() == Encoding::CJK) {
2611                         string end_tag = subst(lang_end_command, "$$lang", running_lang);
2612                         os << from_ascii(end_tag);
2613                         column += end_tag.length();
2614                         popLanguageName();
2615                 }
2616
2617                 // Switch file encoding if necessary (and allowed)
2618                 if (!runparams.pass_thru && !style.pass_thru &&
2619                     runparams.encoding->package() != Encoding::none &&
2620                     current_font.language()->encoding()->package() != Encoding::none) {
2621                         pair<bool, int> const enc_switch =
2622                                 switchEncoding(os.os(), bparams, runparams,
2623                                         *(current_font.language()->encoding()));
2624                         if (enc_switch.first) {
2625                                 column += enc_switch.second;
2626                                 runparams.encoding = current_font.language()->encoding();
2627                         }
2628                 }
2629
2630                 // A display math inset inside an ulem command will be output
2631                 // as a box of width \linewidth, so we have to either disable
2632                 // indentation if the inset starts a paragraph, or start a new
2633                 // line to accommodate such box. This has to be done before
2634                 // writing any font changing commands.
2635                 if (runparams.inDisplayMath && !deleted_display_math
2636                     && runparams.inulemcmd) {
2637                         if (os.afterParbreak())
2638                                 os << "\\noindent";
2639                         else
2640                                 os << "\\\\\n";
2641                 }
2642
2643                 // Do we need to change font?
2644                 if ((current_font != running_font ||
2645                      current_font.language() != running_font.language())
2646                     && i != body_pos - 1)
2647                 {
2648                         if (in_ct_deletion) {
2649                                 // We have to close and then reopen \lyxdeleted,
2650                                 // as strikeout needs to be on lowest level.
2651                                 bool needPar = false;
2652                                 OutputParams rp = runparams;
2653                                 column += running_font.latexWriteEndChanges(
2654                                         os, bparams, rp, basefont,
2655                                         basefont, needPar);
2656                                 os << '}';
2657                                 column += 1;
2658                         }
2659                         odocstringstream ods;
2660                         column += current_font.latexWriteStartChanges(ods, bparams,
2661                                                               runparams, basefont,
2662                                                               last_font);
2663                         // Check again for display math in ulem commands as a
2664                         // font change may also occur just before a math inset.
2665                         if (runparams.inDisplayMath && !deleted_display_math
2666                             && runparams.inulemcmd) {
2667                                 if (os.afterParbreak())
2668                                         os << "\\noindent";
2669                                 else
2670                                         os << "\\\\\n";
2671                         }
2672                         running_font = current_font;
2673                         open_font = true;
2674                         docstring fontchange = ods.str();
2675                         // check whether the fontchange ends with a \\textcolor
2676                         // modifier and the text starts with a space (bug 4473)
2677                         docstring const last_modifier = rsplit(fontchange, '\\');
2678                         if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2679                                 os << fontchange << from_ascii("{}");
2680                         // check if the fontchange ends with a trailing blank
2681                         // (like "\small " (see bug 3382)
2682                         else if (suffixIs(fontchange, ' ') && c == ' ')
2683                                 os << fontchange.substr(0, fontchange.size() - 1)
2684                                    << from_ascii("{}");
2685                         else
2686                                 os << fontchange;
2687                         if (in_ct_deletion) {
2688                                 // We have to close and then reopen \lyxdeleted,
2689                                 // as strikeout needs to be on lowest level.
2690                                 OutputParams rp = runparams;
2691                                 column += Changes::latexMarkChange(os, bparams,
2692                                         Change(Change::UNCHANGED), change, rp);
2693                         }
2694                 }
2695
2696                 // FIXME: think about end_pos implementation...
2697                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2698                         // FIXME: integrate this case in latexSpecialChar
2699                         // Do not print the separation of the optional argument
2700                         // if style.pass_thru is false. This works because
2701                         // latexSpecialChar ignores spaces if
2702                         // style.pass_thru is false.
2703                         if (i != body_pos - 1) {
2704                                 if (d->simpleTeXBlanks(bparams, runparams, os,
2705                                                 i, column, current_font, style)) {
2706                                         // A surrogate pair was output. We
2707                                         // must not call latexSpecialChar
2708                                         // in this iteration, since it would output
2709                                         // the combining character again.
2710                                         ++i;
2711                                         continue;
2712                                 }
2713                         }
2714                 }
2715
2716                 OutputParams rp = runparams;
2717                 rp.free_spacing = style.free_spacing;
2718                 rp.local_font = &current_font;
2719                 rp.intitle = style.intitle;
2720
2721                 // Two major modes:  LaTeX or plain
2722                 // Handle here those cases common to both modes
2723                 // and then split to handle the two modes separately.
2724                 if (c == META_INSET) {
2725                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2726                                 // Greyedout notes and, in general, all insets
2727                                 // with InsetLayout::isDisplay() == false,
2728                                 // are typeset inline with the text. So, we
2729                                 // can add a \par to the last paragraph of
2730                                 // such insets only if nothing else follows.
2731                                 bool incremented = false;
2732                                 Inset const * inset = getInset(i);
2733                                 InsetText const * textinset = inset
2734                                                         ? inset->asInsetText()
2735                                                         : nullptr;
2736                                 if (i + 1 == size() && textinset
2737                                     && !inset->getLayout().isDisplay()) {
2738                                         ParagraphList const & pars =
2739                                                 textinset->text().paragraphs();
2740                                         pit_type const pit = pars.size() - 1;
2741                                         Font const lastfont =
2742                                                 pit < 0 || pars[pit].empty()
2743                                                 ? pars[pit].getLayoutFont(
2744                                                                 bparams,
2745                                                                 outerfont)
2746                                                 : pars[pit].getFont(bparams,
2747                                                         pars[pit].size() - 1,
2748                                                         outerfont);
2749                                         if (lastfont.fontInfo().size() !=
2750                                             basefont.fontInfo().size()) {
2751                                                 ++parInline;
2752                                                 incremented = true;
2753                                         }
2754                                 }
2755                                 d->latexInset(bparams, os, rp, running_font,
2756                                                 basefont, outerfont, open_font,
2757                                                 runningChange, style, i, column);
2758                                 if (incremented)
2759                                         --parInline;
2760
2761                                 if (runparams.ctObject == OutputParams::CT_DISPLAYOBJECT
2762                                     || runparams.ctObject == OutputParams::CT_UDISPLAYOBJECT) {
2763                                         // Close \lyx*deleted and force its
2764                                         // reopening (if needed)
2765                                         os << '}';
2766                                         column++;
2767                                         runningChange = Change(Change::UNCHANGED);
2768                                         runparams.ctObject = OutputParams::CT_NORMAL;
2769                                 }
2770                         }
2771                 } else if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2772                         if (!bparams.useNonTeXFonts)
2773                           script = Encodings::isKnownScriptChar(c);
2774                         if (script != alien_script) {
2775                                 if (!alien_script.empty()) {
2776                                         os << "}";
2777                                         alien_script.clear();
2778                                 }
2779                                 string fontenc = running_font.language()->fontenc(bparams);
2780                                 if (!script.empty()
2781                                         && !Encodings::fontencSupportsScript(fontenc, script)) {
2782                                         column += script.length() + 2;
2783                                         os << "\\" << script << "{";
2784                                         alien_script = script;
2785                                 }
2786                         }
2787                         try {
2788                                 d->latexSpecialChar(os, bparams, rp, running_font,
2789                                                                         alien_script, 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                 // And finally, pass the post_macros upstream
2815                 runparams.post_macro = rp.post_macro;
2816         }
2817
2818         // Close wrapper for alien script
2819         if (!alien_script.empty()) {
2820                 os << "}";
2821                 alien_script.clear();
2822         }
2823
2824         // If we have an open font definition, we have to close it
2825         if (open_font) {
2826                 // Make sure that \\par is done with the font of the last
2827                 // character if this has another size as the default.
2828                 // This is necessary because LaTeX (and LyX on the screen)
2829                 // calculates the space between the baselines according
2830                 // to this font. (Matthias)
2831                 //
2832                 // We must not change the font for the last paragraph
2833                 // of non-multipar insets, tabular cells or commands,
2834                 // since this produces unwanted whitespace.
2835
2836                 Font const font = empty()
2837                         ? getLayoutFont(bparams, outerfont)
2838                         : getFont(bparams, size() - 1, outerfont);
2839
2840                 InsetText const * textinset = inInset().asInsetText();
2841
2842                 bool const maintext = textinset
2843                         ? textinset->text().isMainText()
2844                         : false;
2845
2846                 size_t const numpars = textinset
2847                         ? textinset->text().paragraphs().size()
2848                         : 0;
2849
2850                 bool needPar = false;
2851
2852                 if (style.resfont.size() != font.fontInfo().size()
2853                     && (!runparams.isLastPar || maintext
2854                         || (numpars > 1 && d->ownerCode() != CELL_CODE
2855                             && (inInset().getLayout().isDisplay()
2856                                 || parInline)))
2857                     && !style.isCommand()) {
2858                         needPar = true;
2859                 }
2860 #ifdef FIXED_LANGUAGE_END_DETECTION
2861                 if (next_) {
2862                         running_font.latexWriteEndChanges(os, bparams,
2863                                         runparams, basefont,
2864                                         next_->getFont(bparams, 0, outerfont),
2865                                                        needPar);
2866                 } else {
2867                         running_font.latexWriteEndChanges(os, bparams,
2868                                         runparams, basefont, basefont, needPar);
2869                 }
2870 #else
2871 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2872 //FIXME: there as we start another \selectlanguage with the next paragraph if
2873 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2874                 running_font.latexWriteEndChanges(os, bparams, runparams,
2875                                 basefont, basefont, needPar);
2876 #endif
2877                 if (needPar) {
2878                         // The \par could not be inserted at the same nesting
2879                         // level of the font size change, so do it now.
2880                         os << "{\\" << font.latexSize() << "\\par}";
2881                 }
2882         }
2883
2884         column += Changes::latexMarkChange(os, bparams, runningChange,
2885                                            Change(Change::UNCHANGED), runparams);
2886
2887         // Needed if there is an optional argument but no contents.
2888         if (body_pos > 0 && body_pos == size()) {
2889                 os << "}]~";
2890         }
2891
2892         if (!style.rightdelim().empty()) {
2893                 os << style.rightdelim();
2894                 column += style.rightdelim().size();
2895         }
2896
2897         if (allowcust && d->endTeXParParams(bparams, os, runparams)
2898             && runparams.encoding != prev_encoding) {
2899                 runparams.encoding = prev_encoding;
2900                 os << setEncoding(prev_encoding->iconvName());
2901         }
2902
2903         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2904 }
2905
2906
2907 bool Paragraph::emptyTag() const
2908 {
2909         for (pos_type i = 0; i < size(); ++i) {
2910                 if (Inset const * inset = getInset(i)) {
2911                         InsetCode lyx_code = inset->lyxCode();
2912                         // FIXME testing like that is wrong. What is
2913                         // the intent?
2914                         if (lyx_code != TOC_CODE &&
2915                             lyx_code != INCLUDE_CODE &&
2916                             lyx_code != GRAPHICS_CODE &&
2917                             lyx_code != ERT_CODE &&
2918                             lyx_code != LISTINGS_CODE &&
2919                             lyx_code != FLOAT_CODE &&
2920                             lyx_code != TABULAR_CODE) {
2921                                 return false;
2922                         }
2923                 } else {
2924                         char_type c = d->text_[i];
2925                         if (c != ' ' && c != '\t')
2926                                 return false;
2927                 }
2928         }
2929         return true;
2930 }
2931
2932
2933 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
2934         const
2935 {
2936         for (pos_type i = 0; i < size(); ++i) {
2937                 if (Inset const * inset = getInset(i)) {
2938                         InsetCode lyx_code = inset->lyxCode();
2939                         if (lyx_code == LABEL_CODE) {
2940                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2941                                 docstring const & id = il->getParam("name");
2942                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2943                         }
2944                 }
2945         }
2946         return string();
2947 }
2948
2949
2950 pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runparams)
2951         const
2952 {
2953         pos_type i;
2954         for (i = 0; i < size(); ++i) {
2955                 if (Inset const * inset = getInset(i)) {
2956                         inset->docbook(os, runparams);
2957                 } else {
2958                         char_type c = d->text_[i];
2959                         if (c == ' ')
2960                                 break;
2961                         os << sgml::escapeChar(c);
2962                 }
2963         }
2964         return i;
2965 }
2966
2967
2968 pos_type Paragraph::firstWordLyXHTML(XHTMLStream & xs, OutputParams const & runparams)
2969         const
2970 {
2971         pos_type i;
2972         for (i = 0; i < size(); ++i) {
2973                 if (Inset const * inset = getInset(i)) {
2974                         inset->xhtml(xs, runparams);
2975                 } else {
2976                         char_type c = d->text_[i];
2977                         if (c == ' ')
2978                                 break;
2979                         xs << c;
2980                 }
2981         }
2982         return i;
2983 }
2984
2985
2986 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2987 {
2988         Font font_old;
2989         pos_type size = text_.size();
2990         for (pos_type i = initial; i < size; ++i) {
2991                 Font font = owner_->getFont(buf.params(), i, outerfont);
2992                 if (text_[i] == META_INSET)
2993                         return false;
2994                 if (i != initial && font != font_old)
2995                         return false;
2996                 font_old = font;
2997         }
2998
2999         return true;
3000 }
3001
3002
3003 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
3004                                     odocstream & os,
3005                                     OutputParams const & runparams,
3006                                     Font const & outerfont,
3007                                     pos_type initial) const
3008 {
3009         bool emph_flag = false;
3010
3011         Layout const & style = *d->layout_;
3012         FontInfo font_old =
3013                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
3014
3015         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
3016                 os << "]]>";
3017
3018         // parsing main loop
3019         for (pos_type i = initial; i < size(); ++i) {
3020                 Font font = getFont(buf.params(), i, outerfont);
3021
3022                 // handle <emphasis> tag
3023                 if (font_old.emph() != font.fontInfo().emph()) {
3024                         if (font.fontInfo().emph() == FONT_ON) {
3025                                 os << "<emphasis>";
3026                                 emph_flag = true;
3027                         } else if (i != initial) {
3028                                 os << "</emphasis>";
3029                                 emph_flag = false;
3030                         }
3031                 }
3032
3033                 if (Inset const * inset = getInset(i)) {
3034                         inset->docbook(os, runparams);
3035                 } else {
3036                         char_type c = d->text_[i];
3037
3038                         if (style.pass_thru)
3039                                 os.put(c);
3040                         else
3041                                 os << sgml::escapeChar(c);
3042                 }
3043                 font_old = font.fontInfo();
3044         }
3045
3046         if (emph_flag) {
3047                 os << "</emphasis>";
3048         }
3049
3050         if (style.free_spacing)
3051                 os << '\n';
3052         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
3053                 os << "<![CDATA[";
3054 }
3055
3056
3057 namespace {
3058 void doFontSwitch(vector<html::FontTag> & tagsToOpen,
3059                   vector<html::EndFontTag> & tagsToClose,
3060                   bool & flag, FontState curstate, html::FontTypes type)
3061 {
3062         if (curstate == FONT_ON) {
3063                 tagsToOpen.push_back(html::FontTag(type));
3064                 flag = true;
3065         } else if (flag) {
3066                 tagsToClose.push_back(html::EndFontTag(type));
3067                 flag = false;
3068         }
3069 }
3070 } // namespace
3071
3072
3073 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
3074                                     XHTMLStream & xs,
3075                                     OutputParams const & runparams,
3076                                     Font const & outerfont,
3077                                     bool start_paragraph, bool close_paragraph,
3078                                     pos_type initial) const
3079 {
3080         docstring retval;
3081
3082         // track whether we have opened these tags
3083         bool emph_flag = false;
3084         bool bold_flag = false;
3085         bool noun_flag = false;
3086         bool ubar_flag = false;
3087         bool dbar_flag = false;
3088         bool sout_flag = false;
3089         bool xout_flag = false;
3090         bool wave_flag = false;
3091         // shape tags
3092         bool shap_flag = false;
3093         // family tags
3094         bool faml_flag = false;
3095         // size tags
3096         bool size_flag = false;
3097
3098         Layout const & style = *d->layout_;
3099
3100         if (start_paragraph)
3101                 xs.startDivision(allowEmpty());
3102
3103         FontInfo font_old =
3104                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
3105
3106         FontShape  curr_fs   = INHERIT_SHAPE;
3107         FontFamily curr_fam  = INHERIT_FAMILY;
3108         FontSize   curr_size = INHERIT_SIZE;
3109
3110         string const default_family =
3111                 buf.masterBuffer()->params().fonts_default_family;
3112
3113         vector<html::FontTag> tagsToOpen;
3114         vector<html::EndFontTag> tagsToClose;
3115
3116         // parsing main loop
3117         for (pos_type i = initial; i < size(); ++i) {
3118                 // let's not show deleted material in the output
3119                 if (isDeleted(i))
3120                         continue;
3121
3122                 Font const font = getFont(buf.masterBuffer()->params(), i, outerfont);
3123
3124                 // emphasis
3125                 FontState curstate = font.fontInfo().emph();
3126                 if (font_old.emph() != curstate)
3127                         doFontSwitch(tagsToOpen, tagsToClose, emph_flag, curstate, html::FT_EMPH);
3128
3129                 // noun
3130                 curstate = font.fontInfo().noun();
3131                 if (font_old.noun() != curstate)
3132                         doFontSwitch(tagsToOpen, tagsToClose, noun_flag, curstate, html::FT_NOUN);
3133
3134                 // underbar
3135                 curstate = font.fontInfo().underbar();
3136                 if (font_old.underbar() != curstate)
3137                         doFontSwitch(tagsToOpen, tagsToClose, ubar_flag, curstate, html::FT_UBAR);
3138
3139                 // strikeout
3140                 curstate = font.fontInfo().strikeout();
3141                 if (font_old.strikeout() != curstate)
3142                         doFontSwitch(tagsToOpen, tagsToClose, sout_flag, curstate, html::FT_SOUT);
3143
3144                 // xout
3145                 curstate = font.fontInfo().xout();
3146                 if (font_old.xout() != curstate)
3147                         doFontSwitch(tagsToOpen, tagsToClose, xout_flag, curstate, html::FT_XOUT);
3148
3149                 // double underbar
3150                 curstate = font.fontInfo().uuline();
3151                 if (font_old.uuline() != curstate)
3152                         doFontSwitch(tagsToOpen, tagsToClose, dbar_flag, curstate, html::FT_DBAR);
3153
3154                 // wavy line
3155                 curstate = font.fontInfo().uwave();
3156                 if (font_old.uwave() != curstate)
3157                         doFontSwitch(tagsToOpen, tagsToClose, wave_flag, curstate, html::FT_WAVE);
3158
3159                 // bold
3160                 // a little hackish, but allows us to reuse what we have.
3161                 curstate = (font.fontInfo().series() == BOLD_SERIES ? FONT_ON : FONT_OFF);
3162                 if (font_old.series() != font.fontInfo().series())
3163                         doFontSwitch(tagsToOpen, tagsToClose, bold_flag, curstate, html::FT_BOLD);
3164
3165                 // Font shape
3166                 curr_fs = font.fontInfo().shape();
3167                 FontShape old_fs = font_old.shape();
3168                 if (old_fs != curr_fs) {
3169                         if (shap_flag) {
3170                                 switch (old_fs) {
3171                                 case ITALIC_SHAPE:
3172                                         tagsToClose.push_back(html::EndFontTag(html::FT_ITALIC));
3173                                         break;
3174                                 case SLANTED_SHAPE:
3175                                         tagsToClose.push_back(html::EndFontTag(html::FT_SLANTED));
3176                                         break;
3177                                 case SMALLCAPS_SHAPE:
3178                                         tagsToClose.push_back(html::EndFontTag(html::FT_SMALLCAPS));
3179                                         break;
3180                                 case UP_SHAPE:
3181                                 case INHERIT_SHAPE:
3182                                         break;
3183                                 default:
3184                                         // the other tags are for internal use
3185                                         LATTEST(false);
3186                                         break;
3187                                 }
3188                                 shap_flag = false;
3189                         }
3190                         switch (curr_fs) {
3191                         case ITALIC_SHAPE:
3192                                 tagsToOpen.push_back(html::FontTag(html::FT_ITALIC));
3193                                 shap_flag = true;
3194                                 break;
3195                         case SLANTED_SHAPE:
3196                                 tagsToOpen.push_back(html::FontTag(html::FT_SLANTED));
3197                                 shap_flag = true;
3198                                 break;
3199                         case SMALLCAPS_SHAPE:
3200                                 tagsToOpen.push_back(html::FontTag(html::FT_SMALLCAPS));
3201                                 shap_flag = true;
3202                                 break;
3203                         case UP_SHAPE:
3204                         case INHERIT_SHAPE:
3205                                 break;
3206                         default:
3207                                 // the other tags are for internal use
3208                                 LATTEST(false);
3209                                 break;
3210                         }
3211                 }
3212
3213                 // Font family
3214                 curr_fam = font.fontInfo().family();
3215                 FontFamily old_fam = font_old.family();
3216                 if (old_fam != curr_fam) {
3217                         if (faml_flag) {
3218                                 switch (old_fam) {
3219                                 case ROMAN_FAMILY:
3220                                         tagsToClose.push_back(html::EndFontTag(html::FT_ROMAN));
3221                                         break;
3222                                 case SANS_FAMILY:
3223                                         tagsToClose.push_back(html::EndFontTag(html::FT_SANS));
3224                                         break;
3225                                 case TYPEWRITER_FAMILY:
3226                                         tagsToClose.push_back(html::EndFontTag(html::FT_TYPE));
3227                                         break;
3228                                 case INHERIT_FAMILY:
3229                                         break;
3230                                 default:
3231                                         // the other tags are for internal use
3232                                         LATTEST(false);
3233                                         break;
3234                                 }
3235                                 faml_flag = false;
3236                         }
3237                         switch (curr_fam) {
3238                         case ROMAN_FAMILY:
3239                                 // we will treat a "default" font family as roman, since we have
3240                                 // no other idea what to do.
3241                                 if (default_family != "rmdefault" && default_family != "default") {
3242                                         tagsToOpen.push_back(html::FontTag(html::FT_ROMAN));
3243                                         faml_flag = true;
3244                                 }
3245                                 break;
3246                         case SANS_FAMILY:
3247                                 if (default_family != "sfdefault") {
3248                                         tagsToOpen.push_back(html::FontTag(html::FT_SANS));
3249                                         faml_flag = true;
3250                                 }
3251                                 break;
3252                         case TYPEWRITER_FAMILY:
3253                                 if (default_family != "ttdefault") {
3254                                         tagsToOpen.push_back(html::FontTag(html::FT_TYPE));
3255                                         faml_flag = true;
3256                                 }
3257                                 break;
3258                         case INHERIT_FAMILY:
3259                                 break;
3260                         default:
3261                                 // the other tags are for internal use
3262                                 LATTEST(false);
3263                                 break;
3264                         }
3265                 }
3266
3267                 // Font size
3268                 curr_size = font.fontInfo().size();
3269                 FontSize old_size = font_old.size();
3270                 if (old_size != curr_size) {
3271                         if (size_flag) {
3272                                 switch (old_size) {
3273                                 case TINY_SIZE:
3274                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_TINY));
3275                                         break;
3276                                 case SCRIPT_SIZE:
3277                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SCRIPT));
3278                                         break;
3279                                 case FOOTNOTE_SIZE:
3280                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_FOOTNOTE));
3281                                         break;
3282                                 case SMALL_SIZE:
3283                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SMALL));
3284                                         break;
3285                                 case LARGE_SIZE:
3286                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGE));
3287                                         break;
3288                                 case LARGER_SIZE:
3289                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGER));
3290                                         break;
3291                                 case LARGEST_SIZE:
3292                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGEST));
3293                                         break;
3294                                 case HUGE_SIZE:
3295                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGE));
3296                                         break;
3297                                 case HUGER_SIZE:
3298                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGER));
3299                                         break;
3300                                 case INCREASE_SIZE:
3301                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_INCREASE));
3302                                         break;
3303                                 case DECREASE_SIZE:
3304                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_DECREASE));
3305                                         break;
3306                                 case INHERIT_SIZE:
3307                                 case NORMAL_SIZE:
3308                                         break;
3309                                 default:
3310                                         // the other tags are for internal use
3311                                         LATTEST(false);
3312                                         break;
3313                                 }
3314                                 size_flag = false;
3315                         }
3316                         switch (curr_size) {
3317                         case TINY_SIZE:
3318                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_TINY));
3319                                 size_flag = true;
3320                                 break;
3321                         case SCRIPT_SIZE:
3322                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SCRIPT));
3323                                 size_flag = true;
3324                                 break;
3325                         case FOOTNOTE_SIZE:
3326                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_FOOTNOTE));
3327                                 size_flag = true;
3328                                 break;
3329                         case SMALL_SIZE:
3330                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SMALL));
3331                                 size_flag = true;
3332                                 break;
3333                         case LARGE_SIZE:
3334                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGE));
3335                                 size_flag = true;
3336                                 break;
3337                         case LARGER_SIZE:
3338                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGER));
3339                                 size_flag = true;
3340                                 break;
3341                         case LARGEST_SIZE:
3342                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGEST));
3343                                 size_flag = true;
3344                                 break;
3345                         case HUGE_SIZE:
3346                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGE));
3347                                 size_flag = true;
3348                                 break;
3349                         case HUGER_SIZE:
3350                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGER));
3351                                 size_flag = true;
3352                                 break;
3353                         case INCREASE_SIZE:
3354                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_INCREASE));
3355                                 size_flag = true;
3356                                 break;
3357                         case DECREASE_SIZE:
3358                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_DECREASE));
3359                                 size_flag = true;
3360                                 break;
3361                         case NORMAL_SIZE:
3362                         case INHERIT_SIZE:
3363                                 break;
3364                         default:
3365                                 // the other tags are for internal use
3366                                 LATTEST(false);
3367                                 break;
3368                         }
3369                 }
3370
3371                 // FIXME XHTML
3372                 // Other such tags? What about the other text ranges?
3373
3374                 vector<html::EndFontTag>::const_iterator cit = tagsToClose.begin();
3375                 vector<html::EndFontTag>::const_iterator cen = tagsToClose.end();
3376                 for (; cit != cen; ++cit)
3377                         xs << *cit;
3378
3379                 vector<html::FontTag>::const_iterator sit = tagsToOpen.begin();
3380                 vector<html::FontTag>::const_iterator sen = tagsToOpen.end();
3381                 for (; sit != sen; ++sit)
3382                         xs << *sit;
3383
3384                 tagsToClose.clear();
3385                 tagsToOpen.clear();
3386
3387                 Inset const * inset = getInset(i);
3388                 if (inset) {
3389                         if (!runparams.for_toc || inset->isInToc()) {
3390                                 OutputParams np = runparams;
3391                                 np.local_font = &font;
3392                                 // If the paragraph has size 1, then we are in the "special
3393                                 // case" where we do not output the containing paragraph info
3394                                 if (!inset->getLayout().htmlisblock() && size() != 1)
3395                                         np.html_in_par = true;
3396                                 retval += inset->xhtml(xs, np);
3397                         }
3398                 } else {
3399                         char_type c = getUChar(buf.masterBuffer()->params(),
3400                                                runparams, i);
3401                         if (c == ' ' && (style.free_spacing || runparams.free_spacing))
3402                                 xs << XHTMLStream::ESCAPE_NONE << "&nbsp;";
3403                         else
3404                                 xs << c;
3405                 }
3406                 font_old = font.fontInfo();
3407         }
3408
3409         // FIXME XHTML
3410         // I'm worried about what happens if a branch, say, is itself
3411         // wrapped in some font stuff. I think that will not work.
3412         xs.closeFontTags();
3413         if (close_paragraph)
3414                 xs.endDivision();
3415
3416         return retval;
3417 }
3418
3419
3420 bool Paragraph::isHfill(pos_type pos) const
3421 {
3422         Inset const * inset = getInset(pos);
3423         return inset && inset->isHfill();
3424 }
3425
3426
3427 bool Paragraph::isNewline(pos_type pos) const
3428 {
3429         // U+2028 LINE SEPARATOR
3430         // U+2029 PARAGRAPH SEPARATOR
3431         char_type const c = d->text_[pos];
3432         if (c == 0x2028 || c == 0x2029)
3433                 return true;
3434         Inset const * inset = getInset(pos);
3435         return inset && inset->lyxCode() == NEWLINE_CODE;
3436 }
3437
3438
3439 bool Paragraph::isEnvSeparator(pos_type pos) const
3440 {
3441         Inset const * inset = getInset(pos);
3442         return inset && inset->lyxCode() == SEPARATOR_CODE;
3443 }
3444
3445
3446 bool Paragraph::isLineSeparator(pos_type pos) const
3447 {
3448         char_type const c = d->text_[pos];
3449         if (isLineSeparatorChar(c))
3450                 return true;
3451         Inset const * inset = getInset(pos);
3452         return inset && inset->isLineSeparator();
3453 }
3454
3455
3456 bool Paragraph::isWordSeparator(pos_type pos, bool const ignore_deleted) const
3457 {
3458         if (pos == size())
3459                 return true;
3460         if (ignore_deleted && isDeleted(pos))
3461                 return false;
3462         if (Inset const * inset = getInset(pos))
3463                 return !inset->isLetter();
3464         // if we have a hard hyphen (no en- or emdash) or apostrophe
3465         // we pass this to the spell checker
3466         // FIXME: this method is subject to change, visit
3467         // https://bugzilla.mozilla.org/show_bug.cgi?id=355178
3468         // to get an impression how complex this is.
3469         if (isHardHyphenOrApostrophe(pos))
3470                 return false;
3471         char_type const c = d->text_[pos];
3472         // We want to pass the escape chars to the spellchecker
3473         docstring const escape_chars = from_utf8(lyxrc.spellchecker_esc_chars);
3474         return !isLetterChar(c) && !isDigitASCII(c) && !contains(escape_chars, c);
3475 }
3476
3477
3478 bool Paragraph::isHardHyphenOrApostrophe(pos_type pos) const
3479 {
3480         pos_type const psize = size();
3481         if (pos >= psize)
3482                 return false;
3483         char_type const c = d->text_[pos];
3484         if (c != '-' && c != '\'')
3485                 return false;
3486         int nextpos = pos + 1;
3487         int prevpos = pos > 0 ? pos - 1 : 0;
3488         if ((nextpos == psize || isSpace(nextpos))
3489                 && (pos == 0 || isSpace(prevpos)))
3490                 return false;
3491         return true;
3492 }
3493
3494
3495 bool Paragraph::needsCProtection(bool const fragile) const
3496 {
3497         // first check the layout of the paragraph, but only in insets
3498         InsetText const * textinset = inInset().asInsetText();
3499         bool const maintext = textinset
3500                 ? textinset->text().isMainText()
3501                 : false;
3502
3503         if (!maintext && layout().needcprotect) {
3504                 // Environments need cprotection regardless the content
3505                 if (layout().latextype == LATEX_ENVIRONMENT)
3506                         return true;
3507
3508                 // Commands need cprotection if they contain specific chars
3509                 int const nchars_escape = 9;
3510                 static char_type const chars_escape[nchars_escape] = {
3511                         '&', '_', '$', '%', '#', '^', '{', '}', '\\'};
3512
3513                 docstring const pars = asString();
3514                 for (int k = 0; k < nchars_escape; k++) {
3515                         if (contains(pars, chars_escape[k]))
3516                                 return true;
3517                 }
3518         }
3519
3520         // now check whether we have insets that need cprotection
3521         pos_type size = pos_type(d->text_.size());
3522         for (pos_type i = 0; i < size; ++i) {
3523                 if (!isInset(i))
3524                         continue;
3525                 Inset const * ins = getInset(i);
3526                 if (ins->needsCProtection(maintext, fragile))
3527                         return true;
3528                 if (ins->getLayout().latextype() == InsetLayout::ENVIRONMENT)
3529                         // Environments need cprotection regardless the content
3530                         return true;
3531                 // Now check math environments
3532                 InsetMath const * im = getInset(i)->asInsetMath();
3533                 if (!im || im->cell(0).empty())
3534                         continue;
3535                 switch(im->cell(0)[0]->lyxCode()) {
3536                 case MATH_AMSARRAY_CODE:
3537                 case MATH_SUBSTACK_CODE:
3538                 case MATH_ENV_CODE:
3539                 case MATH_XYMATRIX_CODE:
3540                         // these need cprotection
3541                         return true;
3542                 default:
3543                         break;
3544                 }
3545         }
3546
3547         return false;
3548 }
3549
3550
3551 FontSpan const & Paragraph::getSpellRange(pos_type pos) const
3552 {
3553         return d->speller_state_.getRange(pos);
3554 }
3555
3556
3557 bool Paragraph::isChar(pos_type pos) const
3558 {
3559         if (Inset const * inset = getInset(pos))
3560                 return inset->isChar();
3561         char_type const c = d->text_[pos];
3562         return !isLetterChar(c) && !isDigitASCII(c) && !lyx::isSpace(c);
3563 }
3564
3565
3566 bool Paragraph::isSpace(pos_type pos) const
3567 {
3568         if (Inset const * inset = getInset(pos))
3569                 return inset->isSpace();
3570         char_type const c = d->text_[pos];
3571         return lyx::isSpace(c);
3572 }
3573
3574
3575 Language const *
3576 Paragraph::getParLanguage(BufferParams const & bparams) const
3577 {
3578         if (!empty())
3579                 return getFirstFontSettings(bparams).language();
3580         // FIXME: we should check the prev par as well (Lgb)
3581         return bparams.language;
3582 }
3583
3584
3585 bool Paragraph::isRTL(BufferParams const & bparams) const
3586 {
3587         return getParLanguage(bparams)->rightToLeft()
3588                 && !inInset().getLayout().forceLTR();
3589 }
3590
3591
3592 void Paragraph::changeLanguage(BufferParams const & bparams,
3593                                Language const * from, Language const * to)
3594 {
3595         // change language including dummy font change at the end
3596         for (pos_type i = 0; i <= size(); ++i) {
3597                 Font font = getFontSettings(bparams, i);
3598                 if (font.language() == from) {
3599                         font.setLanguage(to);
3600                         setFont(i, font);
3601                         d->requestSpellCheck(i);
3602                 }
3603         }
3604 }
3605
3606
3607 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
3608 {
3609         Language const * doc_language = bparams.language;
3610         for (auto const & f : d->fontlist_)
3611                 if (f.font().language() != ignore_language &&
3612                     f.font().language() != latex_language &&
3613                     f.font().language() != doc_language)
3614                         return true;
3615         return false;
3616 }
3617
3618
3619 void Paragraph::getLanguages(std::set<Language const *> & langs) const
3620 {
3621         for (auto const & f : d->fontlist_) {
3622                 Language const * lang = f.font().language();
3623                 if (lang != ignore_language &&
3624                     lang != latex_language)
3625                         langs.insert(lang);
3626         }
3627 }
3628
3629
3630 docstring Paragraph::asString(int options) const
3631 {
3632         return asString(0, size(), options);
3633 }
3634
3635
3636 docstring Paragraph::asString(pos_type beg, pos_type end, int options, const OutputParams *runparams) const
3637 {
3638         odocstringstream os;
3639
3640         if (beg == 0
3641             && options & AS_STR_LABEL
3642             && !d->params_.labelString().empty())
3643                 os << d->params_.labelString() << ' ';
3644
3645         for (pos_type i = beg; i < end; ++i) {
3646                 if ((options & AS_STR_SKIPDELETE) && isDeleted(i))
3647                         continue;
3648                 char_type const c = d->text_[i];
3649                 if (isPrintable(c) || c == '\t'
3650                     || (c == '\n' && (options & AS_STR_NEWLINES)))
3651                         os.put(c);
3652                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
3653                         if (c == META_INSET && (options & AS_STR_PLAINTEXT)) {
3654                                 LASSERT(runparams != nullptr, return docstring());
3655                                 getInset(i)->plaintext(os, *runparams);
3656                         } else {
3657                                 getInset(i)->toString(os);
3658                         }
3659                 }
3660         }
3661
3662         return os.str();
3663 }
3664
3665
3666 void Paragraph::forOutliner(docstring & os, size_t const maxlen,
3667                             bool const shorten, bool const label) const
3668 {
3669         size_t tmplen = shorten ? maxlen + 1 : maxlen;
3670         if (label && !labelString().empty())
3671                 os += labelString() + ' ';
3672         if (!layout().isTocCaption())
3673                 return;
3674         for (pos_type i = 0; i < size() && os.length() < tmplen; ++i) {
3675                 if (isDeleted(i))
3676                         continue;
3677                 char_type const c = d->text_[i];
3678                 if (isPrintable(c))
3679                         os += c;
3680                 else if (c == META_INSET)
3681                         getInset(i)->forOutliner(os, tmplen, false);
3682         }
3683         if (shorten)
3684                 Text::shortenForOutliner(os, maxlen);
3685 }
3686
3687
3688 void Paragraph::setInsetOwner(Inset const * inset)
3689 {
3690         d->inset_owner_ = inset;
3691 }
3692
3693
3694 int Paragraph::id() const
3695 {
3696         return d->id_;
3697 }
3698
3699
3700 void Paragraph::setId(int id)
3701 {
3702         d->id_ = id;
3703 }
3704
3705
3706 Layout const & Paragraph::layout() const
3707 {
3708         return *d->layout_;
3709 }
3710
3711
3712 void Paragraph::setLayout(Layout const & layout)
3713 {
3714         d->layout_ = &layout;
3715 }
3716
3717
3718 void Paragraph::setDefaultLayout(DocumentClass const & tc)
3719 {
3720         setLayout(tc.defaultLayout());
3721 }
3722
3723
3724 void Paragraph::setPlainLayout(DocumentClass const & tc)
3725 {
3726         setLayout(tc.plainLayout());
3727 }
3728
3729
3730 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
3731 {
3732         if (usePlainLayout())
3733                 setPlainLayout(tclass);
3734         else
3735                 setDefaultLayout(tclass);
3736 }
3737
3738
3739 Inset const & Paragraph::inInset() const
3740 {
3741         LBUFERR(d->inset_owner_);
3742         return *d->inset_owner_;
3743 }
3744
3745
3746 ParagraphParameters & Paragraph::params()
3747 {
3748         return d->params_;
3749 }
3750
3751
3752 ParagraphParameters const & Paragraph::params() const
3753 {
3754         return d->params_;
3755 }
3756
3757
3758 bool Paragraph::isFreeSpacing() const
3759 {
3760         if (d->layout_->free_spacing)
3761                 return true;
3762         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
3763 }
3764
3765
3766 bool Paragraph::allowEmpty() const
3767 {
3768         if (d->layout_->keepempty)
3769                 return true;
3770         return d->inset_owner_ && d->inset_owner_->allowEmpty();
3771 }
3772
3773
3774 bool Paragraph::brokenBiblio() const
3775 {
3776         // There is a problem if there is no bibitem at position 0 in
3777         // paragraphs that need one, if there is another bibitem in the
3778         // paragraph or if this paragraph is not supposed to have
3779         // a bibitem inset at all.
3780         return ((d->layout_->labeltype == LABEL_BIBLIO
3781                 && (d->insetlist_.find(BIBITEM_CODE) != 0
3782                     || d->insetlist_.find(BIBITEM_CODE, 1) > 0))
3783                 || (d->layout_->labeltype != LABEL_BIBLIO
3784                     && d->insetlist_.find(BIBITEM_CODE) != -1));
3785 }
3786
3787
3788 int Paragraph::fixBiblio(Buffer const & buffer)
3789 {
3790         // FIXME: when there was already an inset at 0, the return value is 1,
3791         // which does not tell whether another inset has been remove; the
3792         // cursor cannot be correctly updated.
3793
3794         bool const track_changes = buffer.params().track_changes;
3795         int bibitem_pos = d->insetlist_.find(BIBITEM_CODE);
3796
3797         // The case where paragraph is not BIBLIO
3798         if (d->layout_->labeltype != LABEL_BIBLIO) {
3799                 if (bibitem_pos == -1)
3800                         // No InsetBibitem => OK
3801                         return 0;
3802                 // There is an InsetBibitem: remove it!
3803                 d->insetlist_.release(bibitem_pos);
3804                 eraseChar(bibitem_pos, track_changes);
3805                 return (bibitem_pos == 0) ? -1 : -bibitem_pos;
3806         }
3807
3808         bool const hasbibitem0 = bibitem_pos == 0;
3809         if (hasbibitem0) {
3810                 bibitem_pos = d->insetlist_.find(BIBITEM_CODE, 1);
3811                 // There was an InsetBibitem at pos 0,
3812                 // and no other one => OK
3813                 if (bibitem_pos == -1)
3814                         return 0;
3815                 // there is a bibitem at the 0 position, but since
3816                 // there is a second one, we copy the second on the
3817                 // first. We're assuming there are at most two of
3818                 // these, which there should be.
3819                 // FIXME: why does it make sense to do that rather
3820                 // than keep the first? (JMarc)
3821                 Inset * inset = releaseInset(bibitem_pos);
3822                 d->insetlist_.begin()->inset = inset;
3823                 return -bibitem_pos;
3824         }
3825
3826         // We need to create an inset at the beginning
3827         Inset * inset = nullptr;
3828         if (bibitem_pos > 0) {
3829                 // there was one somewhere in the paragraph, let's move it
3830                 inset = d->insetlist_.release(bibitem_pos);
3831                 eraseChar(bibitem_pos, track_changes);
3832         } else
3833                 // make a fresh one
3834                 inset = new InsetBibitem(const_cast<Buffer *>(&buffer),
3835                                          InsetCommandParams(BIBITEM_CODE));
3836
3837         Font font(inherit_font, buffer.params().language);
3838         insertInset(0, inset, font, Change(track_changes ? Change::INSERTED
3839                                                    : Change::UNCHANGED));
3840
3841         // This is needed to get the counters right
3842         buffer.updateBuffer();
3843         return 1;
3844 }
3845
3846
3847 void Paragraph::checkAuthors(AuthorList const & authorList)
3848 {
3849         d->changes_.checkAuthors(authorList);
3850 }
3851
3852
3853 bool Paragraph::isChanged(pos_type pos) const
3854 {
3855         return lookupChange(pos).changed();
3856 }
3857
3858
3859 bool Paragraph::isInserted(pos_type pos) const
3860 {
3861         return lookupChange(pos).inserted();
3862 }
3863
3864
3865 bool Paragraph::isDeleted(pos_type pos) const
3866 {
3867         return lookupChange(pos).deleted();
3868 }
3869
3870
3871 InsetList const & Paragraph::insetList() const
3872 {
3873         return d->insetlist_;
3874 }
3875
3876
3877 void Paragraph::setInsetBuffers(Buffer & b)
3878 {
3879         d->insetlist_.setBuffer(b);
3880 }
3881
3882
3883 void Paragraph::resetBuffer()
3884 {
3885         d->insetlist_.resetBuffer();
3886 }
3887
3888
3889 Inset * Paragraph::releaseInset(pos_type pos)
3890 {
3891         Inset * inset = d->insetlist_.release(pos);
3892         /// does not honour change tracking!
3893         eraseChar(pos, false);
3894         return inset;
3895 }
3896
3897
3898 Inset * Paragraph::getInset(pos_type pos)
3899 {
3900         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3901                  ? d->insetlist_.get(pos) : nullptr;
3902 }
3903
3904
3905 Inset const * Paragraph::getInset(pos_type pos) const
3906 {
3907         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3908                  ? d->insetlist_.get(pos) : nullptr;
3909 }
3910
3911
3912 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
3913                 pos_type & right, TextCase action)
3914 {
3915         // process sequences of modified characters; in change
3916         // tracking mode, this approach results in much better
3917         // usability than changing case on a char-by-char basis
3918         // We also need to track the current font, since font
3919         // changes within sequences can occur.
3920         vector<pair<char_type, Font> > changes;
3921
3922         bool const trackChanges = bparams.track_changes;
3923
3924         bool capitalize = true;
3925
3926         for (; pos < right; ++pos) {
3927                 char_type oldChar = d->text_[pos];
3928                 char_type newChar = oldChar;
3929
3930                 // ignore insets and don't play with deleted text!
3931                 if (oldChar != META_INSET && !isDeleted(pos)) {
3932                         switch (action) {
3933                                 case text_lowercase:
3934                                         newChar = lowercase(oldChar);
3935                                         break;
3936                                 case text_capitalization:
3937                                         if (capitalize) {
3938                                                 newChar = uppercase(oldChar);
3939                                                 capitalize = false;
3940                                         }
3941                                         break;
3942                                 case text_uppercase:
3943                                         newChar = uppercase(oldChar);
3944                                         break;
3945                         }
3946                 }
3947
3948                 if (isWordSeparator(pos) || isDeleted(pos)) {
3949                         // permit capitalization again
3950                         capitalize = true;
3951                 }
3952
3953                 if (oldChar != newChar) {
3954                         changes.push_back(make_pair(newChar, getFontSettings(bparams, pos)));
3955                         if (pos != right - 1)
3956                                 continue;
3957                         // step behind the changing area
3958                         pos++;
3959                 }
3960
3961                 int erasePos = pos - changes.size();
3962                 for (size_t i = 0; i < changes.size(); i++) {
3963                         insertChar(pos, changes[i].first,
3964                                    changes[i].second,
3965                                    trackChanges);
3966                         if (!eraseChar(erasePos, trackChanges)) {
3967                                 ++erasePos;
3968                                 ++pos; // advance
3969                                 ++right; // expand selection
3970                         }
3971                 }
3972                 changes.clear();
3973         }
3974 }
3975
3976
3977 int Paragraph::find(docstring const & str, bool cs, bool mw,
3978                 pos_type start_pos, bool del) const
3979 {
3980         pos_type pos = start_pos;
3981         int const strsize = str.length();
3982         int i = 0;
3983         pos_type const parsize = d->text_.size();
3984         for (i = 0; i < strsize && pos < parsize; ++i, ++pos) {
3985                 // Ignore "invisible" letters such as ligature breaks
3986                 // and hyphenation chars while searching
3987                 while (pos < parsize - 1 && isInset(pos)) {
3988                         odocstringstream os;
3989                         getInset(pos)->toString(os);
3990                         if (!getInset(pos)->isLetter() || !os.str().empty())
3991                                 break;
3992                         pos++;
3993                 }
3994                 if (cs && str[i] != d->text_[pos])
3995                         break;
3996                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos]))
3997                         break;
3998                 if (!del && isDeleted(pos))
3999                         break;
4000         }
4001
4002         if (i != strsize)
4003                 return 0;
4004
4005         // if necessary, check whether string matches word
4006         if (mw) {
4007                 if (start_pos > 0 && !isWordSeparator(start_pos - 1))
4008                         return 0;
4009                 if (pos < parsize
4010                         && !isWordSeparator(pos))
4011                         return 0;
4012         }
4013
4014         return pos - start_pos;
4015 }
4016
4017
4018 char_type Paragraph::getChar(pos_type pos) const
4019 {
4020         return d->text_[pos];
4021 }
4022
4023
4024 pos_type Paragraph::size() const
4025 {
4026         return d->text_.size();
4027 }
4028
4029
4030 bool Paragraph::empty() const
4031 {
4032         return d->text_.empty();
4033 }
4034
4035
4036 bool Paragraph::isInset(pos_type pos) const
4037 {
4038         return d->text_[pos] == META_INSET;
4039 }
4040
4041
4042 bool Paragraph::isSeparator(pos_type pos) const
4043 {
4044         //FIXME: Are we sure this can be the only separator?
4045         return d->text_[pos] == ' ';
4046 }
4047
4048
4049 void Paragraph::deregisterWords()
4050 {
4051         Private::LangWordsMap::const_iterator itl = d->words_.begin();
4052         Private::LangWordsMap::const_iterator ite = d->words_.end();
4053         for (; itl != ite; ++itl) {
4054                 WordList & wl = theWordList(itl->first);
4055                 Private::Words::const_iterator it = (itl->second).begin();
4056                 Private::Words::const_iterator et = (itl->second).end();
4057                 for (; it != et; ++it)
4058                         wl.remove(*it);
4059         }
4060         d->words_.clear();
4061 }
4062
4063
4064 void Paragraph::locateWord(pos_type & from, pos_type & to,
4065         word_location const loc, bool const ignore_deleted) const
4066 {
4067         switch (loc) {
4068         case WHOLE_WORD_STRICT:
4069                 if (from == 0 || from == size()
4070                     || isWordSeparator(from, ignore_deleted)
4071                     || isWordSeparator(from - 1, ignore_deleted)) {
4072                         to = from;
4073                         return;
4074                 }
4075                 // fall through
4076
4077         case WHOLE_WORD:
4078                 // If we are already at the beginning of a word, do nothing
4079                 if (!from || isWordSeparator(from - 1, ignore_deleted))
4080                         break;
4081                 // fall through
4082
4083         case PREVIOUS_WORD:
4084                 // always move the cursor to the beginning of previous word
4085                 while (from && !isWordSeparator(from - 1, ignore_deleted))
4086                         --from;
4087                 break;
4088         case NEXT_WORD:
4089                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
4090                 break;
4091         case PARTIAL_WORD:
4092                 // no need to move the 'from' cursor
4093                 break;
4094         }
4095         to = from;
4096         while (to < size() && !isWordSeparator(to, ignore_deleted))
4097                 ++to;
4098 }
4099
4100
4101 void Paragraph::collectWords()
4102 {
4103         for (pos_type pos = 0; pos < size(); ++pos) {
4104                 if (isWordSeparator(pos))
4105                         continue;
4106                 pos_type from = pos;
4107                 locateWord(from, pos, WHOLE_WORD);
4108                 // Work around MSVC warning: The statement
4109                 // if (pos < from + lyxrc.completion_minlength)
4110                 // triggers a signed vs. unsigned warning.
4111                 // I don't know why this happens, it could be a MSVC bug, or
4112                 // related to LLP64 (windows) vs. LP64 (unix) programming
4113                 // model, or the C++ standard might be ambigous in the section
4114                 // defining the "usual arithmetic conversions". However, using
4115                 // a temporary variable is safe and works on all compilers.
4116                 pos_type const endpos = from + lyxrc.completion_minlength;
4117                 if (pos < endpos)
4118                         continue;
4119                 FontList::const_iterator cit = d->fontlist_.fontIterator(from);
4120                 if (cit == d->fontlist_.end())
4121                         return;
4122                 Language const * lang = cit->font().language();
4123                 docstring const word = asString(from, pos, AS_STR_NONE);
4124                 d->words_[lang->lang()].insert(word);
4125         }
4126 }
4127
4128
4129 void Paragraph::registerWords()
4130 {
4131         Private::LangWordsMap::const_iterator itl = d->words_.begin();
4132         Private::LangWordsMap::const_iterator ite = d->words_.end();
4133         for (; itl != ite; ++itl) {
4134                 WordList & wl = theWordList(itl->first);
4135                 Private::Words::const_iterator it = (itl->second).begin();
4136                 Private::Words::const_iterator et = (itl->second).end();
4137                 for (; it != et; ++it)
4138                         wl.insert(*it);
4139         }
4140 }
4141
4142
4143 void Paragraph::updateWords()
4144 {
4145         deregisterWords();
4146         collectWords();
4147         registerWords();
4148 }
4149
4150
4151 void Paragraph::Private::appendSkipPosition(SkipPositions & skips, pos_type const pos) const
4152 {
4153         SkipPositionsIterator begin = skips.begin();
4154         SkipPositions::iterator end = skips.end();
4155         if (pos > 0 && begin < end) {
4156                 --end;
4157                 if (end->last == pos - 1) {
4158                         end->last = pos;
4159                         return;
4160                 }
4161         }
4162         skips.insert(end, FontSpan(pos, pos));
4163 }
4164
4165
4166 Language * Paragraph::Private::locateSpellRange(
4167         pos_type & from, pos_type & to,
4168         SkipPositions & skips) const
4169 {
4170         // skip leading white space
4171         while (from < to && owner_->isWordSeparator(from))
4172                 ++from;
4173         // don't check empty range
4174         if (from >= to)
4175                 return nullptr;
4176         // get current language
4177         Language * lang = getSpellLanguage(from);
4178         pos_type last = from;
4179         bool samelang = true;
4180         bool sameinset = true;
4181         while (last < to && samelang && sameinset) {
4182                 // hop to end of word
4183                 while (last < to && !owner_->isWordSeparator(last)) {
4184                         if (owner_->getInset(last)) {
4185                                 appendSkipPosition(skips, last);
4186                         } else if (owner_->isDeleted(last)) {
4187                                 appendSkipPosition(skips, last);
4188                         }
4189                         ++last;
4190                 }
4191                 // hop to next word while checking for insets
4192                 while (sameinset && last < to && owner_->isWordSeparator(last)) {
4193                         if (Inset const * inset = owner_->getInset(last))
4194                                 sameinset = inset->isChar() && inset->isLetter();
4195                         if (sameinset && owner_->isDeleted(last)) {
4196                                 appendSkipPosition(skips, last);
4197                         }
4198                         if (sameinset)
4199                                 last++;
4200                 }
4201                 if (sameinset && last < to) {
4202                         // now check for language change
4203                         samelang = lang == getSpellLanguage(last);
4204                 }
4205         }
4206         // if language change detected backstep is needed
4207         if (!samelang)
4208                 --last;
4209         to = last;
4210         return lang;
4211 }
4212
4213
4214 Language * Paragraph::Private::getSpellLanguage(pos_type const from) const
4215 {
4216         Language * lang =
4217                 const_cast<Language *>(owner_->getFontSettings(
4218                         inset_owner_->buffer().params(), from).language());
4219         if (lang == inset_owner_->buffer().params().language
4220                 && !lyxrc.spellchecker_alt_lang.empty()) {
4221                 string lang_code;
4222                 string const lang_variety =
4223                         split(lyxrc.spellchecker_alt_lang, lang_code, '-');
4224                 lang->setCode(lang_code);
4225                 lang->setVariety(lang_variety);
4226         }
4227         return lang;
4228 }
4229
4230
4231 void Paragraph::requestSpellCheck(pos_type pos)
4232 {
4233         d->requestSpellCheck(pos);
4234 }
4235
4236
4237 bool Paragraph::needsSpellCheck() const
4238 {
4239         SpellChecker::ChangeNumber speller_change_number = 0;
4240         if (theSpellChecker())
4241                 speller_change_number = theSpellChecker()->changeNumber();
4242         if (speller_change_number > d->speller_state_.currentChangeNumber()) {
4243                 d->speller_state_.needsCompleteRefresh(speller_change_number);
4244         }
4245         return d->needsSpellCheck();
4246 }
4247
4248
4249 bool Paragraph::Private::ignoreWord(docstring const & word) const
4250 {
4251         // Ignore words with digits
4252         // FIXME: make this customizable
4253         // (note that some checkers ignore words with digits by default)
4254         docstring::const_iterator cit = word.begin();
4255         docstring::const_iterator const end = word.end();
4256         for (; cit != end; ++cit) {
4257                 if (isNumber((*cit)))
4258                         return true;
4259         }
4260         return false;
4261 }
4262
4263
4264 SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
4265         WordLangTuple & wl, docstring_list & suggestions,
4266         bool do_suggestion, bool check_learned) const
4267 {
4268         SpellChecker::Result result = SpellChecker::WORD_OK;
4269         SpellChecker * speller = theSpellChecker();
4270         if (!speller)
4271                 return result;
4272
4273         if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
4274                 return result;
4275
4276         locateWord(from, to, WHOLE_WORD, true);
4277         if (from == to || from >= size())
4278                 return result;
4279
4280         docstring word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
4281         Language * lang = d->getSpellLanguage(from);
4282
4283         if (getFontSettings(d->inset_owner_->buffer().params(), from).fontInfo().nospellcheck() == FONT_ON)
4284                 return result;
4285
4286         wl = WordLangTuple(word, lang);
4287
4288         if (word.empty())
4289                 return result;
4290
4291         if (needsSpellCheck() || check_learned) {
4292                 pos_type end = to;
4293                 if (!d->ignoreWord(word)) {
4294                         bool const trailing_dot = to < size() && d->text_[to] == '.';
4295                         result = speller->check(wl);
4296                         if (SpellChecker::misspelled(result) && trailing_dot) {
4297                                 wl = WordLangTuple(word.append(from_ascii(".")), lang);
4298                                 result = speller->check(wl);
4299                                 if (!SpellChecker::misspelled(result)) {
4300                                         LYXERR(Debug::GUI, "misspelled word is correct with dot: \"" <<
4301                                            word << "\" [" <<
4302                                            from << ".." << to << "]");
4303                                 } else {
4304                                         // spell check with dot appended failed too
4305                                         // restore original word/lang value
4306                                         word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
4307                                         wl = WordLangTuple(word, lang);
4308                                 }
4309                         }
4310                 }
4311                 if (!SpellChecker::misspelled(result)) {
4312                         // area up to the begin of the next word is not misspelled
4313                         while (end < size() && isWordSeparator(end))
4314                                 ++end;
4315                 }
4316                 d->setMisspelled(from, end, result);
4317         } else {
4318                 result = d->speller_state_.getState(from);
4319         }
4320
4321         if (do_suggestion)
4322                 suggestions.clear();
4323
4324         if (SpellChecker::misspelled(result)) {
4325                 LYXERR(Debug::GUI, "misspelled word: \"" <<
4326                            word << "\" [" <<
4327                            from << ".." << to << "]");
4328                 if (do_suggestion)
4329                         speller->suggest(wl, suggestions);
4330         }
4331         return result;
4332 }
4333
4334
4335 void Paragraph::anonymize()
4336 {
4337         // This is a very crude anonymization for now
4338         for (char_type & c : d->text_)
4339                 if (isLetterChar(c) || isNumber(c))
4340                         c = 'a';
4341 }
4342
4343
4344 void Paragraph::Private::markMisspelledWords(
4345         pos_type const & first, pos_type const & last,
4346         SpellChecker::Result result,
4347         docstring const & word,
4348         SkipPositions const & skips)
4349 {
4350         if (!SpellChecker::misspelled(result)) {
4351                 setMisspelled(first, last, SpellChecker::WORD_OK);
4352                 return;
4353         }
4354         int snext = first;
4355         SpellChecker * speller = theSpellChecker();
4356         // locate and enumerate the error positions
4357         int nerrors = speller->numMisspelledWords();
4358         int numskipped = 0;
4359         SkipPositionsIterator it = skips.begin();
4360         SkipPositionsIterator et = skips.end();
4361         for (int index = 0; index < nerrors; ++index) {
4362                 int wstart;
4363                 int wlen = 0;
4364                 speller->misspelledWord(index, wstart, wlen);
4365                 /// should not happen if speller supports range checks
4366                 if (!wlen) continue;
4367                 docstring const misspelled = word.substr(wstart, wlen);
4368                 wstart += first + numskipped;
4369                 if (snext < wstart) {
4370                         /// mark the range of correct spelling
4371                         numskipped += countSkips(it, et, wstart);
4372                         setMisspelled(snext,
4373                                 wstart - 1, SpellChecker::WORD_OK);
4374                 }
4375                 snext = wstart + wlen;
4376                 numskipped += countSkips(it, et, snext);
4377                 /// mark the range of misspelling
4378                 setMisspelled(wstart, snext, result);
4379                 LYXERR(Debug::GUI, "misspelled word: \"" <<
4380                            misspelled << "\" [" <<
4381                            wstart << ".." << (snext-1) << "]");
4382                 ++snext;
4383         }
4384         if (snext <= last) {
4385                 /// mark the range of correct spelling at end
4386                 setMisspelled(snext, last, SpellChecker::WORD_OK);
4387         }
4388 }
4389
4390
4391 void Paragraph::spellCheck() const
4392 {
4393         SpellChecker * speller = theSpellChecker();
4394         if (!speller || empty() ||!needsSpellCheck())
4395                 return;
4396         pos_type start;
4397         pos_type endpos;
4398         d->rangeOfSpellCheck(start, endpos);
4399         if (speller->canCheckParagraph()) {
4400                 // loop until we leave the range
4401                 for (pos_type first = start; first < endpos; ) {
4402                         pos_type last = endpos;
4403                         Private::SkipPositions skips;
4404                         Language * lang = d->locateSpellRange(first, last, skips);
4405                         if (first >= endpos)
4406                                 break;
4407                         // start the spell checker on the unit of meaning
4408                         docstring word = asString(first, last, AS_STR_INSETS + AS_STR_SKIPDELETE);
4409                         WordLangTuple wl = WordLangTuple(word, lang);
4410                         SpellChecker::Result result = word.size() ?
4411                                 speller->check(wl) : SpellChecker::WORD_OK;
4412                         d->markMisspelledWords(first, last, result, word, skips);
4413                         first = ++last;
4414                 }
4415         } else {
4416                 static docstring_list suggestions;
4417                 pos_type to = endpos;
4418                 while (start < endpos) {
4419                         WordLangTuple wl;
4420                         spellCheck(start, to, wl, suggestions, false);
4421                         start = to + 1;
4422                 }
4423         }
4424         d->readySpellCheck();
4425 }
4426
4427
4428 bool Paragraph::isMisspelled(pos_type pos, bool check_boundary) const
4429 {
4430         bool result = SpellChecker::misspelled(d->speller_state_.getState(pos));
4431         if (result || pos <= 0 || pos > size())
4432                 return result;
4433         if (check_boundary && (pos == size() || isWordSeparator(pos)))
4434                 result = SpellChecker::misspelled(d->speller_state_.getState(pos - 1));
4435         return result;
4436 }
4437
4438
4439 string Paragraph::magicLabel() const
4440 {
4441         stringstream ss;
4442         ss << "magicparlabel-" << id();
4443         return ss.str();
4444 }
4445
4446
4447 } // namespace lyx