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);
 449                 if (endpos < last) {
 450                         endpos = last;
 451                         owner_->locateWord(last, endpos, WHOLE_WORD);
 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_(0), 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 void Paragraph::addChangesToBuffer(Buffer const & buf) const
 585 {
 586         d->changes_.updateBuffer(buf);
 587 }
 588
 589
 590 bool Paragraph::isChangeUpdateRequired() const
 591 {
 592         return d->changes_.isUpdateRequired();
 593 }
 594
 595
 596 bool Paragraph::isDeleted(pos_type start, pos_type end) const
 597 {
 598         LASSERT(start >= 0 && start <= size(), return false);
 599         LASSERT(end > start && end <= size() + 1, return false);
 600
 601         return d->changes_.isDeleted(start, end);
 602 }
 603
 604
 605 bool Paragraph::isChanged(pos_type start, pos_type end) const
 606 {
 607         LASSERT(start >= 0 && start <= size(), return false);
 608         LASSERT(end > start && end <= size() + 1, return false);
 609
 610         return d->changes_.isChanged(start, end);
 611 }
 612
 613
 614 bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const
 615 {
 616         // keep the logic here in sync with the logic of eraseChars()
 617         if (!trackChanges)
 618                 return true;
 619
 620         Change const & change = d->changes_.lookup(size());
 621         return change.inserted() && change.currentAuthor();
 622 }
 623
 624
 625 void Paragraph::setChange(Change const & change)
 626 {
 627         // beware of the imaginary end-of-par character!
 628         d->changes_.set(change, 0, size() + 1);
 629
 630         /*
 631          * Propagate the change recursively - but not in case of DELETED!
 632          *
 633          * Imagine that your co-author makes changes in an existing inset. He
 634          * sends your document to you and you come to the conclusion that the
 635          * inset should go completely. If you erase it, LyX must not delete all
 636          * text within the inset. Otherwise, the change tracked insertions of
 637          * your co-author get lost and there is no way to restore them later.
 638          *
 639          * Conclusion: An inset's content should remain untouched if you delete it
 640          */
 641
 642         if (!change.deleted()) {
 643                 for (pos_type pos = 0; pos < size(); ++pos) {
 644                         if (Inset * inset = getInset(pos))
 645                                 inset->setChange(change);
 646                 }
 647         }
 648 }
 649
 650
 651 void Paragraph::setChange(pos_type pos, Change const & change)
 652 {
 653         LASSERT(pos >= 0 && pos <= size(), return);
 654         d->changes_.set(change, pos);
 655
 656         // see comment in setChange(Change const &) above
 657         if (!change.deleted() && pos < size())
 658                         if (Inset * inset = getInset(pos))
 659                                 inset->setChange(change);
 660 }
 661
 662
 663 Change const & Paragraph::lookupChange(pos_type pos) const
 664 {
 665         LBUFERR(pos >= 0 && pos <= size());
 666         return d->changes_.lookup(pos);
 667 }
 668
 669
 670 void Paragraph::acceptChanges(pos_type start, pos_type end)
 671 {
 672         LASSERT(start >= 0 && start <= size(), return);
 673         LASSERT(end > start && end <= size() + 1, return);
 674
 675         for (pos_type pos = start; pos < end; ++pos) {
 676                 switch (lookupChange(pos).type) {
 677                         case Change::UNCHANGED:
 678                                 // accept changes in nested inset
 679                                 if (Inset * inset = getInset(pos))
 680                                         inset->acceptChanges();
 681                                 break;
 682
 683                         case Change::INSERTED:
 684                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 685                                 // also accept changes in nested inset
 686                                 if (Inset * inset = getInset(pos))
 687                                         inset->acceptChanges();
 688                                 break;
 689
 690                         case Change::DELETED:
 691                                 // Suppress access to non-existent
 692                                 // "end-of-paragraph char"
 693                                 if (pos < size()) {
 694                                         eraseChar(pos, false);
 695                                         --end;
 696                                         --pos;
 697                                 }
 698                                 break;
 699                 }
 700
 701         }
 702 }
 703
 704
 705 void Paragraph::rejectChanges(pos_type start, pos_type end)
 706 {
 707         LASSERT(start >= 0 && start <= size(), return);
 708         LASSERT(end > start && end <= size() + 1, return);
 709
 710         for (pos_type pos = start; pos < end; ++pos) {
 711                 switch (lookupChange(pos).type) {
 712                         case Change::UNCHANGED:
 713                                 // reject changes in nested inset
 714                                 if (Inset * inset = getInset(pos))
 715                                                 inset->rejectChanges();
 716                                 break;
 717
 718                         case Change::INSERTED:
 719                                 // Suppress access to non-existent
 720                                 // "end-of-paragraph char"
 721                                 if (pos < size()) {
 722                                         eraseChar(pos, false);
 723                                         --end;
 724                                         --pos;
 725                                 }
 726                                 break;
 727
 728                         case Change::DELETED:
 729                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 730
 731                                 // Do NOT reject changes within a deleted inset!
 732                                 // There may be insertions of a co-author inside of it!
 733
 734                                 break;
 735                 }
 736         }
 737 }
 738
 739
 740 void Paragraph::Private::insertChar(pos_type pos, char_type c,
 741                 Change const & change)
 742 {
 743         LASSERT(pos >= 0 && pos <= int(text_.size()), return);
 744
 745         // track change
 746         changes_.insert(change, pos);
 747
 748         // This is actually very common when parsing buffers (and
 749         // maybe inserting ascii text)
 750         if (pos == pos_type(text_.size())) {
 751                 // when appending characters, no need to update tables
 752                 text_.push_back(c);
 753                 // but we want spell checking
 754                 requestSpellCheck(pos);
 755                 return;
 756         }
 757
 758         text_.insert(text_.begin() + pos, c);
 759
 760         // Update the font table.
 761         fontlist_.increasePosAfterPos(pos);
 762
 763         // Update the insets
 764         insetlist_.increasePosAfterPos(pos);
 765
 766         // Update list of misspelled positions
 767         speller_state_.increasePosAfterPos(pos);
 768 }
 769
 770
 771 bool Paragraph::insertInset(pos_type pos, Inset * inset,
 772                                    Font const & font, Change const & change)
 773 {
 774         LASSERT(inset, return false);
 775         LASSERT(pos >= 0 && pos <= size(), return false);
 776
 777         // Paragraph::insertInset() can be used in cut/copy/paste operation where
 778         // d->inset_owner_ is not set yet.
 779         if (d->inset_owner_ && !d->inset_owner_->insetAllowed(inset->lyxCode()))
 780                 return false;
 781
 782         d->insertChar(pos, META_INSET, change);
 783         LASSERT(d->text_[pos] == META_INSET, return false);
 784
 785         // Add a new entry in the insetlist_.
 786         d->insetlist_.insert(inset, pos);
 787
 788         // Some insets require run of spell checker
 789         requestSpellCheck(pos);
 790         setFont(pos, font);
 791         return true;
 792 }
 793
 794
 795 bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
 796 {
 797         LASSERT(pos >= 0 && pos <= size(), return false);
 798
 799         // keep the logic here in sync with the logic of isMergedOnEndOfParDeletion()
 800
 801         if (trackChanges) {
 802                 Change change = d->changes_.lookup(pos);
 803
 804                 // set the character to DELETED if
 805                 //  a) it was previously unchanged or
 806                 //  b) it was inserted by a co-author
 807
 808                 if (!change.changed() ||
 809                       (change.inserted() && !change.currentAuthor())) {
 810                         setChange(pos, Change(Change::DELETED));
 811                         // request run of spell checker
 812                         requestSpellCheck(pos);
 813                         return false;
 814                 }
 815
 816                 if (change.deleted())
 817                         return false;
 818         }
 819
 820         // Don't physically access the imaginary end-of-paragraph character.
 821         // eraseChar() can only mark it as DELETED. A physical deletion of
 822         // end-of-par must be handled externally.
 823         if (pos == size()) {
 824                 return false;
 825         }
 826
 827         // track change
 828         d->changes_.erase(pos);
 829
 830         // if it is an inset, delete the inset entry
 831         if (d->text_[pos] == META_INSET)
 832                 d->insetlist_.erase(pos);
 833
 834         d->text_.erase(d->text_.begin() + pos);
 835
 836         // Update the fontlist_
 837         d->fontlist_.erase(pos);
 838
 839         // Update the insetlist_
 840         d->insetlist_.decreasePosAfterPos(pos);
 841
 842         // Update list of misspelled positions
 843         d->speller_state_.decreasePosAfterPos(pos);
 844         d->speller_state_.refreshLast(size());
 845
 846         return true;
 847 }
 848
 849
 850 int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
 851 {
 852         LASSERT(start >= 0 && start <= size(), return 0);
 853         LASSERT(end >= start && end <= size() + 1, return 0);
 854
 855         pos_type i = start;
 856         for (pos_type count = end - start; count; --count) {
 857                 if (!eraseChar(i, trackChanges))
 858                         ++i;
 859         }
 860         return end - i;
 861 }
 862
 863 // Handle combining characters
 864 int Paragraph::Private::latexSurrogatePair(BufferParams const & bparams,
 865                 otexstream & os, char_type c, char_type next,
 866                 OutputParams const & runparams)
 867 {
 868         // Writing next here may circumvent a possible font change between
 869         // c and next. Since next is only output if it forms a surrogate pair
 870         // with c we can ignore this:
 871         // A font change inside a surrogate pair does not make sense and is
 872         // hopefully impossible to input.
 873         // FIXME: change tracking
 874         // Is this correct WRT change tracking?
 875         Encoding const & encoding = *(runparams.encoding);
 876         docstring latex1 = encoding.latexChar(next).first;
 877         if (runparams.inIPA) {
 878                 string const tipashortcut = Encodings::TIPAShortcut(next);
 879                 if (!tipashortcut.empty()) {
 880                         latex1 = from_ascii(tipashortcut);
 881                 }
 882         }
 883         docstring latex2 = encoding.latexChar(c).first;
 884
 885         if (bparams.useNonTeXFonts || docstring(1, next) == latex1) {
 886                 // Encoding supports the combination:
 887                 // output as is (combining char after base char).
 888                 os << latex2 << latex1;
 889                 return latex1.length() + latex2.length();
 890         }
 891
 892         os << latex1 << "{" << latex2 << "}";
 893         return latex1.length() + latex2.length() + 2;
 894 }
 895
 896
 897 bool Paragraph::Private::simpleTeXBlanks(BufferParams const & bparams,
 898                                        OutputParams const & runparams,
 899                                        otexstream & os,
 900                                        pos_type i,
 901                                        unsigned int & column,
 902                                        Font const & font,
 903                                        Layout const & style)
 904 {
 905         if (style.pass_thru || runparams.pass_thru)
 906                 return false;
 907
 908         if (i + 1 < int(text_.size())) {
 909                 char_type next = text_[i + 1];
 910                 if (Encodings::isCombiningChar(next)) {
 911                         // This space has an accent, so we must always output it.
 912                         column += latexSurrogatePair(bparams, os, ' ', next, runparams) - 1;
 913                         return true;
 914                 }
 915         }
 916
 917         if (runparams.linelen > 0
 918             && column > runparams.linelen
 919             && i
 920             && text_[i - 1] != ' '
 921             && (i + 1 < int(text_.size()))
 922             // same in FreeSpacing mode
 923             && !owner_->isFreeSpacing()
 924             // In typewriter mode, we want to avoid
 925             // ! . ? : at the end of a line
 926             && !(font.fontInfo().family() == TYPEWRITER_FAMILY
 927                  && (text_[i - 1] == '.'
 928                      || text_[i - 1] == '?'
 929                      || text_[i - 1] == ':'
 930                      || text_[i - 1] == '!'))) {
 931                 os << '\n';
 932                 os.texrow().start(owner_->id(), i + 1);
 933                 column = 0;
 934         } else if (style.free_spacing) {
 935                 os << '~';
 936         } else {
 937                 os << ' ';
 938         }
 939         return false;
 940 }
 941
 942
 943 void Paragraph::Private::latexInset(BufferParams const & bparams,
 944                                     otexstream & os,
 945                                     OutputParams & runparams,
 946                                     Font & running_font,
 947                                     Font & basefont,
 948                                     Font const & outerfont,
 949                                     bool & open_font,
 950                                     Change & running_change,
 951                                     Layout const & style,
 952                                     pos_type & i,
 953                                     unsigned int & column)
 954 {
 955         Inset * inset = owner_->getInset(i);
 956         LBUFERR(inset);
 957
 958         if (style.pass_thru) {
 959                 odocstringstream ods;
 960                 inset->plaintext(ods, runparams);
 961                 os << ods.str();
 962                 return;
 963         }
 964
 965         // FIXME: move this to InsetNewline::latex
 966         if (inset->lyxCode() == NEWLINE_CODE || inset->lyxCode() == SEPARATOR_CODE) {
 967                 // newlines are handled differently here than
 968                 // the default in simpleTeXSpecialChars().
 969                 if (!style.newline_allowed) {
 970                         os << '\n';
 971                 } else {
 972                         if (open_font) {
 973                                 bool needPar = false;
 974                                 column += running_font.latexWriteEndChanges(
 975                                         os, bparams, runparams,
 976                                         basefont, basefont, needPar);
 977                                 open_font = false;
 978                         }
 979
 980                         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY)
 981                                 os << '~';
 982
 983                         basefont = owner_->getLayoutFont(bparams, outerfont);
 984                         running_font = basefont;
 985
 986                         if (runparams.moving_arg)
 987                                 os << "\\protect ";
 988
 989                 }
 990                 os.texrow().start(owner_->id(), i + 1);
 991                 column = 0;
 992         }
 993
 994         if (owner_->isDeleted(i)) {
 995                 if( ++runparams.inDeletedInset == 1)
 996                         runparams.changeOfDeletedInset = owner_->lookupChange(i);
 997         }
 998
 999         if (inset->canTrackChanges()) {
1000                 column += Changes::latexMarkChange(os, bparams, running_change,
1001                         Change(Change::UNCHANGED), runparams);
1002                 running_change = Change(Change::UNCHANGED);
1003         }
1004
1005         bool close = false;
1006         odocstream::pos_type const len = os.os().tellp();
1007
1008         if (inset->forceLTR()
1009             && running_font.isRightToLeft()
1010             // ERT is an exception, it should be output with no
1011             // decorations at all
1012             && inset->lyxCode() != ERT_CODE) {
1013                 if (runparams.use_polyglossia) {
1014                         os << "\\LRE{";
1015                 } else if (running_font.language()->lang() == "farsi"
1016                            || running_font.language()->lang() == "arabic_arabi")
1017                         os << "\\textLR{" << termcmd;
1018                 else
1019                         os << "\\L{";
1020                 close = true;
1021         }
1022
1023         // FIXME: Bug: we can have an empty font change here!
1024         // if there has just been a font change, we are going to close it
1025         // right now, which means stupid latex code like \textsf{}. AFAIK,
1026         // this does not harm dvi output. A minor bug, thus (JMarc)
1027
1028         // Some insets cannot be inside a font change command.
1029         // However, even such insets *can* be placed in \L or \R
1030         // or their equivalents (for RTL language switches), so we don't
1031         // close the language in those cases.
1032         // ArabTeX, though, cannot handle this special behavior, it seems.
1033         bool arabtex = basefont.language()->lang() == "arabic_arabtex"
1034                 || running_font.language()->lang() == "arabic_arabtex";
1035         if (open_font && !inset->inheritFont()) {
1036                 bool needPar = false;
1037                 bool closeLanguage = arabtex
1038                         || basefont.isRightToLeft() == running_font.isRightToLeft();
1039                 unsigned int count = running_font.latexWriteEndChanges(os,
1040                                         bparams, runparams, basefont, basefont,
1041                                         needPar, closeLanguage);
1042                 column += count;
1043                 // if any font properties were closed, update the running_font,
1044                 // making sure, however, to leave the language as it was
1045                 if (count > 0) {
1046                         // FIXME: probably a better way to keep track of the old
1047                         // language, than copying the entire font?
1048                         Font const copy_font(running_font);
1049                         basefont = owner_->getLayoutFont(bparams, outerfont);
1050                         running_font = basefont;
1051                         if (!closeLanguage)
1052                                 running_font.setLanguage(copy_font.language());
1053                         // leave font open if language is still open
1054                         open_font = (running_font.language() == basefont.language());
1055                         if (closeLanguage)
1056                                 runparams.local_font = &basefont;
1057                 }
1058         }
1059
1060         size_t const previous_row_count = os.texrow().rows();
1061
1062         try {
1063                 runparams.lastid = id_;
1064                 runparams.lastpos = i;
1065                 inset->latex(os, runparams);
1066         } catch (EncodingException & e) {
1067                 // add location information and throw again.
1068                 e.par_id = id_;
1069                 e.pos = i;
1070                 throw(e);
1071         }
1072
1073         if (close)
1074                 os << '}';
1075
1076         if (os.texrow().rows() > previous_row_count) {
1077                 os.texrow().start(owner_->id(), i + 1);
1078                 column = 0;
1079         } else {
1080                 column += (unsigned int)(os.os().tellp() - len);
1081         }
1082
1083         if (owner_->isDeleted(i))
1084                 --runparams.inDeletedInset;
1085 }
1086
1087
1088 void Paragraph::Private::latexSpecialChar(otexstream & os,
1089                                           BufferParams const & bparams,
1090                                           OutputParams const & runparams,
1091                                           Font const & running_font,
1092                                           string & alien_script,
1093                                           Layout const & style,
1094                                           pos_type & i,
1095                                           pos_type end_pos,
1096                                           unsigned int & column)
1097 {
1098         char_type const c = owner_->getUChar(bparams, runparams, i);
1099
1100         if (style.pass_thru || runparams.pass_thru
1101             || contains(style.pass_thru_chars, c)
1102             || contains(runparams.pass_thru_chars, c)) {
1103                 if (c != '\0') {
1104                         Encoding const * const enc = runparams.encoding;
1105                         if (enc && !enc->encodable(c))
1106                                 throw EncodingException(c);
1107                         os.put(c);
1108                 }
1109                 return;
1110         }
1111
1112         // TIPA uses its own T3 encoding
1113         if (runparams.inIPA && latexSpecialT3(c, os, i, column))
1114                 return;
1115         // If T1 font encoding is used, use the special
1116         // characters it provides.
1117         // NOTE: Some languages reset the font encoding internally to a
1118         //       non-standard font encoding. If we are using such a language,
1119         //       we do not output special T1 chars.
1120         if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1121             && !runparams.isFullUnicode() && bparams.main_font_encoding() == "T1"
1122             && latexSpecialT1(c, os, i, column))
1123                 return;
1124         // NOTE: "fontspec" (non-TeX fonts) sets the font encoding to "TU" (untill 2017 "EU1" or "EU2")
1125         else if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1126                  && runparams.isFullUnicode() && latexSpecialTU(c, os, i, column))
1127                      return;
1128
1129         // Otherwise, we use what LaTeX provides us.
1130         switch (c) {
1131         case '\\':
1132                 os << "\\textbackslash" << termcmd;
1133                 column += 15;
1134                 break;
1135         case '<':
1136                 os << "\\textless" << termcmd;
1137                 column += 10;
1138                 break;
1139         case '>':
1140                 os << "\\textgreater" << termcmd;
1141                 column += 13;
1142                 break;
1143         case '|':
1144                 os << "\\textbar" << termcmd;
1145                 column += 9;
1146                 break;
1147         case '-':
1148                 os << '-';
1149                 if (i + 1 < static_cast<pos_type>(text_.size()) &&
1150                     (end_pos == -1 || i + 1 < end_pos) &&
1151                     text_[i+1] == '-') {
1152                         // Prevent "--" becoming an en dash and "---" an em dash.
1153                         // (Within \ttfamily, "---" is merged to en dash + hyphen.)
1154                         os << "{}";
1155                         column += 2;
1156                 }
1157                 break;
1158         case '\"':
1159                 os << "\\textquotedbl" << termcmd;
1160                 column += 14;
1161                 break;
1162
1163         case '$': case '&':
1164         case '%': case '#': case '{':
1165         case '}': case '_':
1166                 os << '\\';
1167                 os.put(c);
1168                 column += 1;
1169                 break;
1170
1171         case '~':
1172                 os << "\\textasciitilde" << termcmd;
1173                 column += 16;
1174                 break;
1175
1176         case '^':
1177                 os << "\\textasciicircum" << termcmd;
1178                 column += 17;
1179                 break;
1180
1181         case '*':
1182         case '[':
1183         case ']':
1184                 // avoid being mistaken for optional arguments
1185                 os << '{';
1186                 os.put(c);
1187                 os << '}';
1188                 column += 2;
1189                 break;
1190
1191         case ' ':
1192                 // Blanks are printed before font switching.
1193                 // Sure? I am not! (try nice-latex)
1194                 // I am sure it's correct. LyX might be smarter
1195                 // in the future, but for now, nothing wrong is
1196                 // written. (Asger)
1197                 break;
1198
1199         case 0x2013:
1200         case 0x2014:
1201                 // XeTeX's dash behaviour is determined via a global setting
1202                 if (bparams.use_dash_ligatures
1203                     && owner_->getFontSettings(bparams, i).fontInfo().family() != TYPEWRITER_FAMILY
1204                     && !runparams.inIPA
1205                         // TODO #10961: && not in inset Flex Code
1206                         // TODO #10961: && not in layout LyXCode
1207                     && (!bparams.useNonTeXFonts || runparams.flavor != OutputParams::XETEX)) {
1208                         if (c == 0x2013) {
1209                                 // en-dash
1210                                 os << "--";
1211                                 column +=2;
1212                         } else {
1213                                 // em-dash
1214                                 os << "---";
1215                                 column +=3;
1216                         }
1217                         break;
1218                 }
1219                 // fall through
1220         default:
1221                 if (c == '\0')
1222                         return;
1223
1224                 Encoding const & encoding = *(runparams.encoding);
1225                 char_type next = '\0';
1226                 if (i + 1 < int(text_.size())) {
1227                         next = text_[i + 1];
1228                         if (Encodings::isCombiningChar(next)) {
1229                                 column += latexSurrogatePair(bparams, os, c, next, runparams) - 1;
1230                                 ++i;
1231                                 break;
1232                         }
1233                 }
1234                 pair<docstring, bool> latex = encoding.latexChar(c);
1235                 docstring nextlatex;
1236                 bool nexttipas = false;
1237                 string nexttipashortcut;
1238                 if (next != '\0' && next != META_INSET && !encoding.encodable(next)) {
1239                         nextlatex = encoding.latexChar(next).first;
1240                         if (runparams.inIPA) {
1241                                 nexttipashortcut = Encodings::TIPAShortcut(next);
1242                                 nexttipas = !nexttipashortcut.empty();
1243                         }
1244                 }
1245                 bool tipas = false;
1246                 if (runparams.inIPA) {
1247                         string const tipashortcut = Encodings::TIPAShortcut(c);
1248                         if (!tipashortcut.empty()) {
1249                                 latex.first = from_ascii(tipashortcut);
1250                                 latex.second = false;
1251                                 tipas = true;
1252                         }
1253                 }
1254                 // eventually close "script wrapper" command (see `Paragraph::latex`)
1255                 if (!alien_script.empty()
1256                         && alien_script != Encodings::isKnownScriptChar(next)) {
1257                         column += latex.first.length();
1258                         alien_script.clear();
1259                         os << latex.first << "}";
1260                         break;
1261                 }
1262                 if (latex.second
1263                          && ((!prefixIs(nextlatex, '\\')
1264                                && !prefixIs(nextlatex, '{')
1265                                && !prefixIs(nextlatex, '}'))
1266                              || (nexttipas
1267                                  && !prefixIs(from_ascii(nexttipashortcut), '\\')))
1268                          && !tipas) {
1269                         // Prevent eating of a following space or command corruption by
1270                         // following characters
1271                         if (next == ' ' || next == '\0') {
1272                                 column += latex.first.length() + 1;
1273                                 os << latex.first << "{}";
1274                         } else {
1275                                 column += latex.first.length();
1276                                 os << latex.first << " ";
1277                         }
1278                 } else {
1279                         column += latex.first.length() - 1;
1280                         os << latex.first;
1281                 }
1282                 break;
1283         }
1284 }
1285
1286
1287 bool Paragraph::Private::latexSpecialT1(char_type const c, otexstream & os,
1288         pos_type i, unsigned int & column)
1289 {
1290         switch (c) {
1291         case '>':
1292         case '<':
1293                 os.put(c);
1294                 // In T1 encoding, these characters exist
1295                 // but we should avoid ligatures
1296                 if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
1297                         return true;
1298                 os << "\\textcompwordmark" << termcmd;
1299                 column += 19;
1300                 return true;
1301         case '|':
1302                 os.put(c);
1303                 return true;
1304         case '\"':
1305                 // soul.sty breaks with \char`\"
1306                 os << "\\textquotedbl" << termcmd;
1307                 column += 14;
1308                 return true;
1309         default:
1310                 return false;
1311         }
1312 }
1313
1314
1315 bool Paragraph::Private::latexSpecialTU(char_type const c, otexstream & os,
1316         pos_type i, unsigned int & column)
1317 {
1318         // TU encoding is currently on par with T1.
1319         return latexSpecialT1(c, os, i, column);
1320 }
1321
1322
1323 bool Paragraph::Private::latexSpecialT3(char_type const c, otexstream & os,
1324         pos_type /*i*/, unsigned int & column)
1325 {
1326         switch (c) {
1327         case '*':
1328         case '[':
1329         case ']':
1330         case '\"':
1331                 os.put(c);
1332                 return true;
1333         case '|':
1334                 os << "\\textvertline" << termcmd;
1335                 column += 14;
1336                 return true;
1337         default:
1338                 return false;
1339         }
1340 }
1341
1342
1343 void Paragraph::Private::validate(LaTeXFeatures & features) const
1344 {
1345         if (layout_->inpreamble && inset_owner_) {
1346                 // FIXME: Using a string stream here circumvents the encoding
1347                 // switching machinery of odocstream. Therefore the
1348                 // output is wrong if this paragraph contains content
1349                 // that needs to switch encoding.
1350                 Buffer const & buf = inset_owner_->buffer();
1351                 otexstringstream os;
1352                 os << layout_->preamble();
1353                 size_t const length = os.length();
1354                 TeXOnePar(buf, buf.text(), buf.getParFromID(owner_->id()).pit(), os,
1355                           features.runparams(), string(), 0, -1, true);
1356                 if (os.length() > length)
1357                         features.addPreambleSnippet(os.release(), true);
1358         }
1359
1360         if (features.runparams().flavor == OutputParams::HTML
1361             && layout_->htmltitle()) {
1362                 features.setHTMLTitle(owner_->asString(AS_STR_INSETS | AS_STR_SKIPDELETE));
1363         }
1364
1365         // check the params.
1366         if (!params_.spacing().isDefault())
1367                 features.require("setspace");
1368
1369         // then the layouts
1370         features.useLayout(layout_->name());
1371
1372         // then the fonts
1373         fontlist_.validate(features);
1374
1375         // then the indentation
1376         if (!params_.leftIndent().zero())
1377                 features.require("ParagraphLeftIndent");
1378
1379         // then the insets
1380         InsetList::const_iterator icit = insetlist_.begin();
1381         InsetList::const_iterator iend = insetlist_.end();
1382         for (; icit != iend; ++icit) {
1383                 if (icit->inset) {
1384                         features.inDeletedInset(owner_->isDeleted(icit->pos));
1385                         if (icit->inset->lyxCode() == FOOT_CODE) {
1386                                 // FIXME: an item inset would make things much easier.
1387                                 if ((layout_->latextype == LATEX_LIST_ENVIRONMENT
1388                                      || (layout_->latextype == LATEX_ITEM_ENVIRONMENT
1389                                          && layout_->margintype == MARGIN_FIRST_DYNAMIC))
1390                                     && (icit->pos < begin_of_body_
1391                                         || (icit->pos == begin_of_body_
1392                                             && (icit->pos == 0 || text_[icit->pos - 1] != ' '))))
1393                                         features.saveNoteEnv("description");
1394                         }
1395                         icit->inset->validate(features);
1396                         features.inDeletedInset(false);
1397                         if (layout_->needprotect &&
1398                             icit->inset->lyxCode() == FOOT_CODE)
1399                                 features.require("NeedLyXFootnoteCode");
1400                 }
1401         }
1402
1403         // then the contents
1404         BufferParams const bp = features.runparams().is_child
1405                 ? features.buffer().masterParams() : features.buffer().params();
1406         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1407                 char_type c = text_[i];
1408                 CharInfo const & ci = Encodings::unicodeCharInfo(c);
1409                 if (c == 0x0022) {
1410                         if (features.runparams().isFullUnicode() && bp.useNonTeXFonts)
1411                                 features.require("textquotedblp");
1412                         else if (bp.main_font_encoding() != "T1"
1413                                  || ((&owner_->getFontSettings(bp, i))->language()->internalFontEncoding()))
1414                                 features.require("textquotedbl");
1415                 } else if (ci.textfeature() && contains(ci.textpreamble(), '=')) {
1416                         // features that depend on the font or input encoding
1417                         string feats = ci.textpreamble();
1418                         string fontenc = (&owner_->getFontSettings(bp, i))->language()->fontenc(bp);
1419                         if (fontenc.empty())
1420                                 fontenc = features.runparams().main_fontenc;
1421                         while (!feats.empty()) {
1422                                 string feat;
1423                                 feats = split(feats, feat, ',');
1424                                 if (contains(feat, "!=")) {
1425                                         // a feature that is required except for the spcified
1426                                         // font or input encodings
1427                                         string realfeature;
1428                                         string const contexts = ltrim(split(feat, realfeature, '!'), "=");
1429                                         // multiple encodings are separated by semicolon
1430                                         vector<string> context = getVectorFromString(contexts, ";");
1431                                         // require feature if the context matches neither current font
1432                                         // nor input encoding
1433                                         if (std::find(context.begin(), context.end(), fontenc) == context.end()
1434                                             && std::find(context.begin(), context.end(),
1435                                                          features.runparams().encoding->name()) == context.end())
1436                                                 features.require(realfeature);
1437                                 } else if (contains(feat, '=')) {
1438                                         // a feature that is required only for the spcified
1439                                         // font or input encodings
1440                                         string realfeature;
1441                                         string const contexts = split(feat, realfeature, '=');
1442                                         // multiple encodings are separated by semicolon
1443                                         vector<string> context = getVectorFromString(contexts, ";");
1444                                         // require feature if the context matches either current font
1445                                         // or input encoding
1446                                         if (std::find(context.begin(), context.end(), fontenc) != context.end()
1447                                             || std::find(context.begin(), context.end(),
1448                                                          features.runparams().encoding->name()) != context.end())
1449                                                 features.require(realfeature);
1450                                 }
1451                         }
1452                 } else if (!bp.use_dash_ligatures
1453                            && (c == 0x2013 || c == 0x2014)
1454                            && bp.useNonTeXFonts
1455                            && features.runparams().flavor == OutputParams::XETEX)
1456                         // XeTeX's dash behaviour is determined via a global setting
1457                         features.require("xetexdashbreakstate");
1458                 BufferEncodings::validate(c, features);
1459         }
1460 }
1461
1462 /////////////////////////////////////////////////////////////////////
1463 //
1464 // Paragraph
1465 //
1466 /////////////////////////////////////////////////////////////////////
1467
1468 namespace {
1469         Layout const emptyParagraphLayout;
1470 }
1471
1472 Paragraph::Paragraph()
1473         : d(new Paragraph::Private(this, emptyParagraphLayout))
1474 {
1475         itemdepth = 0;
1476         d->params_.clear();
1477 }
1478
1479
1480 Paragraph::Paragraph(Paragraph const & par)
1481         : itemdepth(par.itemdepth),
1482         d(new Paragraph::Private(*par.d, this))
1483 {
1484         registerWords();
1485 }
1486
1487
1488 Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
1489         : itemdepth(par.itemdepth),
1490         d(new Paragraph::Private(*par.d, this, beg, end))
1491 {
1492         registerWords();
1493 }
1494
1495
1496 Paragraph & Paragraph::operator=(Paragraph const & par)
1497 {
1498         // needed as we will destroy the private part before copying it
1499         if (&par != this) {
1500                 itemdepth = par.itemdepth;
1501
1502                 deregisterWords();
1503                 delete d;
1504                 d = new Private(*par.d, this);
1505                 registerWords();
1506         }
1507         return *this;
1508 }
1509
1510
1511 Paragraph::~Paragraph()
1512 {
1513         deregisterWords();
1514         delete d;
1515 }
1516
1517
1518 namespace {
1519
1520 // this shall be called just before every "os << ..." action.
1521 void flushString(ostream & os, docstring & s)
1522 {
1523         os << to_utf8(s);
1524         s.erase();
1525 }
1526
1527 } // namespace
1528
1529
1530 void Paragraph::write(ostream & os, BufferParams const & bparams,
1531         depth_type & dth) const
1532 {
1533         // The beginning or end of a deeper (i.e. nested) area?
1534         if (dth != d->params_.depth()) {
1535                 if (d->params_.depth() > dth) {
1536                         while (d->params_.depth() > dth) {
1537                                 os << "\n\\begin_deeper";
1538                                 ++dth;
1539                         }
1540                 } else {
1541                         while (d->params_.depth() < dth) {
1542                                 os << "\n\\end_deeper";
1543                                 --dth;
1544                         }
1545                 }
1546         }
1547
1548         // First write the layout
1549         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1550
1551         d->params_.write(os);
1552
1553         Font font1(inherit_font, bparams.language);
1554
1555         Change running_change = Change(Change::UNCHANGED);
1556
1557         // this string is used as a buffer to avoid repetitive calls
1558         // to to_utf8(), which turn out to be expensive (JMarc)
1559         docstring write_buffer;
1560
1561         int column = 0;
1562         for (pos_type i = 0; i <= size(); ++i) {
1563
1564                 Change const & change = lookupChange(i);
1565                 if (change != running_change)
1566                         flushString(os, write_buffer);
1567                 Changes::lyxMarkChange(os, bparams, column, running_change, change);
1568                 running_change = change;
1569
1570                 if (i == size())
1571                         break;
1572
1573                 // Write font changes
1574                 Font font2 = getFontSettings(bparams, i);
1575                 if (font2 != font1) {
1576                         flushString(os, write_buffer);
1577                         font2.lyxWriteChanges(font1, os);
1578                         column = 0;
1579                         font1 = font2;
1580                 }
1581
1582                 char_type const c = d->text_[i];
1583                 switch (c) {
1584                 case META_INSET:
1585                         if (Inset const * inset = getInset(i)) {
1586                                 flushString(os, write_buffer);
1587                                 if (inset->directWrite()) {
1588                                         // international char, let it write
1589                                         // code directly so it's shorter in
1590                                         // the file
1591                                         inset->write(os);
1592                                 } else {
1593                                         if (i)
1594                                                 os << '\n';
1595                                         os << "\\begin_inset ";
1596                                         inset->write(os);
1597                                         os << "\n\\end_inset\n\n";
1598                                         column = 0;
1599                                 }
1600                                 // FIXME This can be removed again once the mystery
1601                                 // crash has been resolved.
1602                                 os << flush;
1603                         }
1604                         break;
1605                 case '\\':
1606                         flushString(os, write_buffer);
1607                         os << "\n\\backslash\n";
1608                         column = 0;
1609                         break;
1610                 case '.':
1611                         flushString(os, write_buffer);
1612                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1613                                 os << ".\n";
1614                                 column = 0;
1615                         } else
1616                                 os << '.';
1617                         break;
1618                 default:
1619                         if ((column > 70 && c == ' ')
1620                             || column > 79) {
1621                                 flushString(os, write_buffer);
1622                                 os << '\n';
1623                                 column = 0;
1624                         }
1625                         // this check is to amend a bug. LyX sometimes
1626                         // inserts '\0' this could cause problems.
1627                         if (c != '\0')
1628                                 write_buffer.push_back(c);
1629                         else
1630                                 LYXERR0("NUL char in structure.");
1631                         ++column;
1632                         break;
1633                 }
1634         }
1635
1636         flushString(os, write_buffer);
1637         os << "\n\\end_layout\n";
1638         // FIXME This can be removed again once the mystery
1639         // crash has been resolved.
1640         os << flush;
1641 }
1642
1643
1644 void Paragraph::validate(LaTeXFeatures & features) const
1645 {
1646         d->validate(features);
1647         bool fragile = features.runparams().moving_arg;
1648         fragile |= layout().needprotect;
1649         if (needsCProtection(fragile))
1650                 features.require("cprotect");
1651 }
1652
1653
1654 void Paragraph::insert(pos_type start, docstring const & str,
1655                        Font const & font, Change const & change)
1656 {
1657         for (size_t i = 0, n = str.size(); i != n ; ++i)
1658                 insertChar(start + i, str[i], font, change);
1659 }
1660
1661
1662 void Paragraph::appendChar(char_type c, Font const & font,
1663                 Change const & change)
1664 {
1665         // track change
1666         d->changes_.insert(change, d->text_.size());
1667         // when appending characters, no need to update tables
1668         d->text_.push_back(c);
1669         setFont(d->text_.size() - 1, font);
1670         d->requestSpellCheck(d->text_.size() - 1);
1671 }
1672
1673
1674 void Paragraph::appendString(docstring const & s, Font const & font,
1675                 Change const & change)
1676 {
1677         pos_type end = s.size();
1678         size_t oldsize = d->text_.size();
1679         size_t newsize = oldsize + end;
1680         size_t capacity = d->text_.capacity();
1681         if (newsize >= capacity)
1682                 d->text_.reserve(max(capacity + 100, newsize));
1683
1684         // when appending characters, no need to update tables
1685         d->text_.append(s);
1686
1687         // FIXME: Optimize this!
1688         for (size_t i = oldsize; i != newsize; ++i) {
1689                 // track change
1690                 d->changes_.insert(change, i);
1691                 d->requestSpellCheck(i);
1692         }
1693         d->fontlist_.set(oldsize, font);
1694         d->fontlist_.set(newsize - 1, font);
1695 }
1696
1697
1698 void Paragraph::insertChar(pos_type pos, char_type c,
1699                            bool trackChanges)
1700 {
1701         d->insertChar(pos, c, Change(trackChanges ?
1702                            Change::INSERTED : Change::UNCHANGED));
1703 }
1704
1705
1706 void Paragraph::insertChar(pos_type pos, char_type c,
1707                            Font const & font, bool trackChanges)
1708 {
1709         d->insertChar(pos, c, Change(trackChanges ?
1710                            Change::INSERTED : Change::UNCHANGED));
1711         setFont(pos, font);
1712 }
1713
1714
1715 void Paragraph::insertChar(pos_type pos, char_type c,
1716                            Font const & font, Change const & change)
1717 {
1718         d->insertChar(pos, c, change);
1719         setFont(pos, font);
1720 }
1721
1722
1723 void Paragraph::resetFonts(Font const & font)
1724 {
1725         d->fontlist_.clear();
1726         d->fontlist_.set(0, font);
1727         d->fontlist_.set(d->text_.size() - 1, font);
1728 }
1729
1730 // Gets uninstantiated font setting at position.
1731 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1732                                          pos_type pos) const
1733 {
1734         if (pos > size()) {
1735                 LYXERR0("pos: " << pos << " size: " << size());
1736                 LBUFERR(false);
1737         }
1738
1739         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1740         if (cit != d->fontlist_.end())
1741                 return cit->font();
1742
1743         if (pos == size() && !empty())
1744                 return getFontSettings(bparams, pos - 1);
1745
1746         // Optimisation: avoid a full font instantiation if there is no
1747         // language change from previous call.
1748         static Font previous_font;
1749         static Language const * previous_lang = 0;
1750         Language const * lang = getParLanguage(bparams);
1751         if (lang != previous_lang) {
1752                 previous_lang = lang;
1753                 previous_font = Font(inherit_font, lang);
1754         }
1755         return previous_font;
1756 }
1757
1758
1759 FontSpan Paragraph::fontSpan(pos_type pos) const
1760 {
1761         LBUFERR(pos <= size());
1762
1763         if (pos == size())
1764                 return FontSpan(pos, pos);
1765
1766         pos_type start = 0;
1767         FontList::const_iterator cit = d->fontlist_.begin();
1768         FontList::const_iterator end = d->fontlist_.end();
1769         for (; cit != end; ++cit) {
1770                 if (cit->pos() >= pos) {
1771                         if (pos >= beginOfBody())
1772                                 return FontSpan(max(start, beginOfBody()),
1773                                                 cit->pos());
1774                         else
1775                                 return FontSpan(start,
1776                                                 min(beginOfBody() - 1,
1777                                                          cit->pos()));
1778                 }
1779                 start = cit->pos() + 1;
1780         }
1781
1782         // This should not happen, but if so, we take no chances.
1783         LYXERR0("Paragraph::fontSpan: position not found in fontinfo table!");
1784         LASSERT(false, return FontSpan(pos, pos));
1785 }
1786
1787
1788 // Gets uninstantiated font setting at position 0
1789 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1790 {
1791         if (!empty() && !d->fontlist_.empty())
1792                 return d->fontlist_.begin()->font();
1793
1794         // Optimisation: avoid a full font instantiation if there is no
1795         // language change from previous call.
1796         static Font previous_font;
1797         static Language const * previous_lang = 0;
1798         if (bparams.language != previous_lang) {
1799                 previous_lang = bparams.language;
1800                 previous_font = Font(inherit_font, bparams.language);
1801         }
1802
1803         return previous_font;
1804 }
1805
1806
1807 // Gets the fully instantiated font at a given position in a paragraph
1808 // This is basically the same function as Text::GetFont() in text2.cpp.
1809 // The difference is that this one is used for generating the LaTeX file,
1810 // and thus cosmetic "improvements" are disallowed: This has to deliver
1811 // the true picture of the buffer. (Asger)
1812 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1813                                  Font const & outerfont) const
1814 {
1815         LBUFERR(pos >= 0);
1816
1817         Font font = getFontSettings(bparams, pos);
1818
1819         pos_type const body_pos = beginOfBody();
1820         FontInfo & fi = font.fontInfo();
1821         if (pos < body_pos)
1822                 fi.realize(d->layout_->labelfont);
1823         else
1824                 fi.realize(d->layout_->font);
1825
1826         fi.realize(outerfont.fontInfo());
1827         fi.realize(bparams.getFont().fontInfo());
1828
1829         return font;
1830 }
1831
1832
1833 Font const Paragraph::getLabelFont
1834         (BufferParams const & bparams, Font const & outerfont) const
1835 {
1836         FontInfo tmpfont = d->layout_->labelfont;
1837         tmpfont.realize(outerfont.fontInfo());
1838         tmpfont.realize(bparams.getFont().fontInfo());
1839         return Font(tmpfont, getParLanguage(bparams));
1840 }
1841
1842
1843 Font const Paragraph::getLayoutFont
1844         (BufferParams const & bparams, Font const & outerfont) const
1845 {
1846         FontInfo tmpfont = d->layout_->font;
1847         tmpfont.realize(outerfont.fontInfo());
1848         tmpfont.realize(bparams.getFont().fontInfo());
1849         return Font(tmpfont, getParLanguage(bparams));
1850 }
1851
1852
1853 char_type Paragraph::getUChar(BufferParams const & bparams,
1854                               OutputParams const & rp,
1855                               pos_type pos) const
1856 {
1857         char_type c = d->text_[pos];
1858
1859         // Return unchanged character in LTR languages
1860         // or if we use poylglossia/bidi.
1861         if (rp.use_polyglossia || !getFontSettings(bparams, pos).isRightToLeft())
1862                 return c;
1863
1864         // Without polyglossia/bidi, we need to account for some special cases.
1865         // FIXME This needs to be audited!
1866         // Check if:
1867         // * The input is as expected for all delimiters
1868         //   => checked for Hebrew!
1869         // * The output matches the display in the LyX workarea
1870         //   => checked for Hebrew!
1871         // * The special cases below are really necessary
1872         //   => checked for Hebrew!
1873         // * In arabic_arabi, brackets are transformed to Arabic
1874         //   Ornate Parentheses. Is this is really wanted?
1875
1876         string const & lang = getFontSettings(bparams, pos).language()->lang();
1877         char_type uc = c;
1878
1879         // 1. In the following languages, parentheses need to be reversed.
1880         bool const reverseparens = lang == "hebrew";
1881
1882         // 2. In the following languages, brackets don't need to be reversed.
1883         bool const reversebrackets = lang != "arabic_arabtex"
1884                         && lang != "arabic_arabi"
1885                         && lang != "farsi";
1886
1887         // Now swap delimiters if needed.
1888         switch (c) {
1889         case '(':
1890                 if (reverseparens)
1891                         uc = ')';
1892                 break;
1893         case ')':
1894                 if (reverseparens)
1895                         uc = '(';
1896                 break;
1897         case '[':
1898                 if (reversebrackets)
1899                         uc = ']';
1900                 break;
1901         case ']':
1902                 if (reversebrackets)
1903                         uc = '[';
1904                 break;
1905         case '{':
1906                 uc = '}';
1907                 break;
1908         case '}':
1909                 uc = '{';
1910                 break;
1911         case '<':
1912                 uc = '>';
1913                 break;
1914         case '>':
1915                 uc = '<';
1916                 break;
1917         }
1918
1919         return uc;
1920 }
1921
1922
1923 void Paragraph::setFont(pos_type pos, Font const & font)
1924 {
1925         LASSERT(pos <= size(), return);
1926
1927         // First, reduce font against layout/label font
1928         // Update: The setCharFont() routine in text2.cpp already
1929         // reduces font, so we don't need to do that here. (Asger)
1930
1931         d->fontlist_.set(pos, font);
1932 }
1933
1934
1935 void Paragraph::makeSameLayout(Paragraph const & par)
1936 {
1937         d->layout_ = par.d->layout_;
1938         d->params_ = par.d->params_;
1939 }
1940
1941
1942 bool Paragraph::stripLeadingSpaces(bool trackChanges)
1943 {
1944         if (isFreeSpacing())
1945                 return false;
1946
1947         int pos = 0;
1948         int count = 0;
1949
1950         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
1951                 if (eraseChar(pos, trackChanges))
1952                         ++count;
1953                 else
1954                         ++pos;
1955         }
1956
1957         return count > 0 || pos > 0;
1958 }
1959
1960
1961 bool Paragraph::hasSameLayout(Paragraph const & par) const
1962 {
1963         return par.d->layout_ == d->layout_
1964                 && d->params_.sameLayout(par.d->params_);
1965 }
1966
1967
1968 depth_type Paragraph::getDepth() const
1969 {
1970         return d->params_.depth();
1971 }
1972
1973
1974 depth_type Paragraph::getMaxDepthAfter() const
1975 {
1976         if (d->layout_->isEnvironment())
1977                 return d->params_.depth() + 1;
1978         else
1979                 return d->params_.depth();
1980 }
1981
1982
1983 LyXAlignment Paragraph::getAlign() const
1984 {
1985         if (d->params_.align() == LYX_ALIGN_LAYOUT)
1986                 return d->layout_->align;
1987         else
1988                 return d->params_.align();
1989 }
1990
1991
1992 docstring const & Paragraph::labelString() const
1993 {
1994         return d->params_.labelString();
1995 }
1996
1997
1998 // the next two functions are for the manual labels
1999 docstring const Paragraph::getLabelWidthString() const
2000 {
2001         if (d->layout_->margintype == MARGIN_MANUAL
2002             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
2003                 return d->params_.labelWidthString();
2004         else
2005                 return _("Senseless with this layout!");
2006 }
2007
2008
2009 void Paragraph::setLabelWidthString(docstring const & s)
2010 {
2011         d->params_.labelWidthString(s);
2012 }
2013
2014
2015 docstring Paragraph::expandLabel(Layout const & layout,
2016                 BufferParams const & bparams) const
2017 {
2018         return expandParagraphLabel(layout, bparams, true);
2019 }
2020
2021
2022 docstring Paragraph::expandDocBookLabel(Layout const & layout,
2023                 BufferParams const & bparams) const
2024 {
2025         return expandParagraphLabel(layout, bparams, false);
2026 }
2027
2028
2029 docstring Paragraph::expandParagraphLabel(Layout const & layout,
2030                 BufferParams const & bparams, bool process_appendix) const
2031 {
2032         DocumentClass const & tclass = bparams.documentClass();
2033         string const & lang = getParLanguage(bparams)->code();
2034         bool const in_appendix = process_appendix && d->params_.appendix();
2035         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
2036
2037         if (fmt.empty() && !layout.counter.empty())
2038                 return tclass.counters().theCounter(layout.counter, lang);
2039
2040         // handle 'inherited level parts' in 'fmt',
2041         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
2042         size_t const i = fmt.find('@', 0);
2043         if (i != docstring::npos) {
2044                 size_t const j = fmt.find('@', i + 1);
2045                 if (j != docstring::npos) {
2046                         docstring parent(fmt, i + 1, j - i - 1);
2047                         docstring label = from_ascii("??");
2048                         if (tclass.hasLayout(parent))
2049                                 label = expandParagraphLabel(tclass[parent], bparams,
2050                                                       process_appendix);
2051                         fmt = docstring(fmt, 0, i) + label
2052                                 + docstring(fmt, j + 1, docstring::npos);
2053                 }
2054         }
2055
2056         return tclass.counters().counterLabel(fmt, lang);
2057 }
2058
2059
2060 void Paragraph::applyLayout(Layout const & new_layout)
2061 {
2062         d->layout_ = &new_layout;
2063         LyXAlignment const oldAlign = d->params_.align();
2064
2065         if (!(oldAlign & d->layout_->alignpossible)) {
2066                 frontend::Alert::warning(_("Alignment not permitted"),
2067                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
2068                 d->params_.align(LYX_ALIGN_LAYOUT);
2069         }
2070 }
2071
2072
2073 pos_type Paragraph::beginOfBody() const
2074 {
2075         return d->begin_of_body_;
2076 }
2077
2078
2079 void Paragraph::setBeginOfBody()
2080 {
2081         if (d->layout_->labeltype != LABEL_MANUAL) {
2082                 d->begin_of_body_ = 0;
2083                 return;
2084         }
2085
2086         // Unroll the first two cycles of the loop
2087         // and remember the previous character to
2088         // remove unnecessary getChar() calls
2089         pos_type i = 0;
2090         pos_type end = size();
2091         if (i < end && !(isNewline(i) || isEnvSeparator(i))) {
2092                 ++i;
2093                 if (i < end) {
2094                         char_type previous_char = d->text_[i];
2095                         if (!(isNewline(i) || isEnvSeparator(i))) {
2096                                 ++i;
2097                                 while (i < end && previous_char != ' ') {
2098                                         char_type temp = d->text_[i];
2099                                         if (isNewline(i) || isEnvSeparator(i))
2100                                                 break;
2101                                         ++i;
2102                                         previous_char = temp;
2103                                 }
2104                         }
2105                 }
2106         }
2107
2108         d->begin_of_body_ = i;
2109 }
2110
2111
2112 bool Paragraph::allowParagraphCustomization() const
2113 {
2114         return inInset().allowParagraphCustomization();
2115 }
2116
2117
2118 bool Paragraph::usePlainLayout() const
2119 {
2120         return inInset().usePlainLayout();
2121 }
2122
2123
2124 bool Paragraph::isPassThru() const
2125 {
2126         return inInset().isPassThru() || d->layout_->pass_thru;
2127 }
2128
2129 namespace {
2130
2131 // paragraphs inside floats need different alignment tags to avoid
2132 // unwanted space
2133
2134 bool noTrivlistCentering(InsetCode code)
2135 {
2136         return code == FLOAT_CODE
2137                || code == WRAP_CODE
2138                || code == CELL_CODE;
2139 }
2140
2141
2142 string correction(string const & orig)
2143 {
2144         if (orig == "flushleft")
2145                 return "raggedright";
2146         if (orig == "flushright")
2147                 return "raggedleft";
2148         if (orig == "center")
2149                 return "centering";
2150         return orig;
2151 }
2152
2153
2154 bool corrected_env(otexstream & os, string const & suffix, string const & env,
2155         InsetCode code, bool const lastpar, int & col)
2156 {
2157         string macro = suffix + "{";
2158         if (noTrivlistCentering(code)) {
2159                 if (lastpar) {
2160                         // the last paragraph in non-trivlist-aligned
2161                         // context is special (to avoid unwanted whitespace)
2162                         if (suffix == "\\begin") {
2163                                 macro = "\\" + correction(env) + "{}";
2164                                 os << from_ascii(macro);
2165                                 col += macro.size();
2166                                 return true;
2167                         }
2168                         return false;
2169                 }
2170                 macro += correction(env);
2171         } else
2172                 macro += env;
2173         macro += "}";
2174         if (suffix == "\\par\\end") {
2175                 os << breakln;
2176                 col = 0;
2177         }
2178         os << from_ascii(macro);
2179         col += macro.size();
2180         if (suffix == "\\begin") {
2181                 os << breakln;
2182                 col = 0;
2183         }
2184         return true;
2185 }
2186
2187 } // namespace
2188
2189
2190 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
2191                         otexstream & os, OutputParams const & runparams) const
2192 {
2193         int column = 0;
2194
2195         bool canindent =
2196                 (bparams.paragraph_separation == BufferParams::ParagraphIndentSeparation) ?
2197                         (layout_->toggle_indent != ITOGGLE_NEVER) :
2198                         (layout_->toggle_indent == ITOGGLE_ALWAYS);
2199
2200         if (canindent && params_.noindent() && !layout_->pass_thru) {
2201                 os << "\\noindent ";
2202                 column += 10;
2203         }
2204
2205         LyXAlignment const curAlign = params_.align();
2206
2207         if (curAlign == layout_->align)
2208                 return column;
2209
2210         switch (curAlign) {
2211         case LYX_ALIGN_NONE:
2212         case LYX_ALIGN_BLOCK:
2213         case LYX_ALIGN_LAYOUT:
2214         case LYX_ALIGN_SPECIAL:
2215         case LYX_ALIGN_DECIMAL:
2216                 break;
2217         case LYX_ALIGN_LEFT:
2218         case LYX_ALIGN_RIGHT:
2219         case LYX_ALIGN_CENTER:
2220                 if (runparams.moving_arg) {
2221                         os << "\\protect";
2222                         column += 8;
2223                 }
2224                 break;
2225         }
2226
2227         string const begin_tag = "\\begin";
2228         InsetCode code = ownerCode();
2229         bool const lastpar = runparams.isLastPar;
2230         // RTL in classic (PDF)LaTeX (without the Bidi package)
2231         bool const rtl_classic = owner_->getParLanguage(bparams)->rightToLeft()
2232                 && !runparams.use_polyglossia;
2233
2234         switch (curAlign) {
2235         case LYX_ALIGN_NONE:
2236         case LYX_ALIGN_BLOCK:
2237         case LYX_ALIGN_LAYOUT:
2238         case LYX_ALIGN_SPECIAL:
2239         case LYX_ALIGN_DECIMAL:
2240                 break;
2241         case LYX_ALIGN_LEFT: {
2242                 if (rtl_classic)
2243                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2244                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2245                 else
2246                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2247                 break;
2248         } case LYX_ALIGN_RIGHT: {
2249                 if (rtl_classic)
2250                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2251                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2252                 else
2253                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2254                 break;
2255         } case LYX_ALIGN_CENTER: {
2256                 corrected_env(os, begin_tag, "center", code, lastpar, column);
2257                 break;
2258         }
2259         }
2260
2261         return column;
2262 }
2263
2264
2265 bool Paragraph::Private::endTeXParParams(BufferParams const & bparams,
2266                         otexstream & os, OutputParams const & runparams) const
2267 {
2268         LyXAlignment const curAlign = params_.align();
2269
2270         if (curAlign == layout_->align)
2271                 return false;
2272
2273         switch (curAlign) {
2274         case LYX_ALIGN_NONE:
2275         case LYX_ALIGN_BLOCK:
2276         case LYX_ALIGN_LAYOUT:
2277         case LYX_ALIGN_SPECIAL:
2278         case LYX_ALIGN_DECIMAL:
2279                 break;
2280         case LYX_ALIGN_LEFT:
2281         case LYX_ALIGN_RIGHT:
2282         case LYX_ALIGN_CENTER:
2283                 if (runparams.moving_arg)
2284                         os << "\\protect";
2285                 break;
2286         }
2287
2288         bool output = false;
2289         int col = 0;
2290         string const end_tag = "\\par\\end";
2291         InsetCode code = ownerCode();
2292         bool const lastpar = runparams.isLastPar;
2293         // RTL in classic (PDF)LaTeX (without the Bidi package)
2294         bool const rtl_classic = owner_->getParLanguage(bparams)->rightToLeft()
2295                 && !runparams.use_polyglossia;
2296
2297         switch (curAlign) {
2298         case LYX_ALIGN_NONE:
2299         case LYX_ALIGN_BLOCK:
2300         case LYX_ALIGN_LAYOUT:
2301         case LYX_ALIGN_SPECIAL:
2302         case LYX_ALIGN_DECIMAL:
2303                 break;
2304         case LYX_ALIGN_LEFT: {
2305                 if (rtl_classic)
2306                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2307                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2308                 else
2309                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2310                 break;
2311         } case LYX_ALIGN_RIGHT: {
2312                 if (rtl_classic)
2313                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2314                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2315                 else
2316                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2317                 break;
2318         } case LYX_ALIGN_CENTER: {
2319                 corrected_env(os, end_tag, "center", code, lastpar, col);
2320                 break;
2321         }
2322         }
2323
2324         return output || lastpar;
2325 }
2326
2327
2328 // This one spits out the text of the paragraph
2329 void Paragraph::latex(BufferParams const & bparams,
2330         Font const & outerfont,
2331         otexstream & os,
2332         OutputParams const & runparams,
2333         int start_pos, int end_pos, bool force) const
2334 {
2335         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
2336
2337         // FIXME This check should not be needed. Perhaps issue an
2338         // error if it triggers.
2339         Layout const & style = inInset().forcePlainLayout() ?
2340                 bparams.documentClass().plainLayout() : *d->layout_;
2341
2342         if (!force && style.inpreamble)
2343                 return;
2344
2345         bool const allowcust = allowParagraphCustomization();
2346
2347         // Current base font for all inherited font changes, without any
2348         // change caused by an individual character, except for the language:
2349         // It is set to the language of the first character.
2350         // As long as we are in the label, this font is the base font of the
2351         // label. Before the first body character it is set to the base font
2352         // of the body.
2353         Font basefont;
2354
2355         // If there is an open font-encoding changing command (script wrapper),
2356         // alien_script is set to its name
2357         string alien_script;
2358         string script;
2359
2360         // Maybe we have to create a optional argument.
2361         pos_type body_pos = beginOfBody();
2362         unsigned int column = 0;
2363
2364         if (body_pos > 0) {
2365                 // the optional argument is kept in curly brackets in
2366                 // case it contains a ']'
2367                 // This is not strictly needed, but if this is changed it
2368                 // would be a file format change, and tex2lyx would need
2369                 // to be adjusted, since it unconditionally removes the
2370                 // braces when it parses \item.
2371                 os << "[{";
2372                 column += 2;
2373                 basefont = getLabelFont(bparams, outerfont);
2374         } else {
2375                 basefont = getLayoutFont(bparams, outerfont);
2376         }
2377
2378         // Which font is currently active?
2379         Font running_font(basefont);
2380         // Do we have an open font change?
2381         bool open_font = false;
2382
2383         Change runningChange = Change(Change::UNCHANGED);
2384
2385         Encoding const * const prev_encoding = runparams.encoding;
2386
2387         os.texrow().start(id(), 0);
2388
2389         // if the paragraph is empty, the loop will not be entered at all
2390         if (empty()) {
2391                 // For InTitle commands, we have already opened a group
2392                 // in output_latex::TeXOnePar.
2393                 if (style.isCommand() && !style.intitle) {
2394                         os << '{';
2395                         ++column;
2396                 }
2397                 if (!style.leftdelim().empty()) {
2398                         os << style.leftdelim();
2399                         column += style.leftdelim().size();
2400                 }
2401                 if (allowcust)
2402                         column += d->startTeXParParams(bparams, os, runparams);
2403         }
2404
2405         // Whether a \par can be issued for insets typeset inline with text.
2406         // Yes if greater than 0. This has to be static.
2407         THREAD_LOCAL_STATIC int parInline = 0;
2408
2409         for (pos_type i = 0; i < size(); ++i) {
2410                 // First char in paragraph or after label?
2411                 if (i == body_pos) {
2412                         if (body_pos > 0) {
2413                                 if (open_font) {
2414                                         bool needPar = false;
2415                                         column += running_font.latexWriteEndChanges(
2416                                                 os, bparams, runparams,
2417                                                 basefont, basefont, needPar);
2418                                         open_font = false;
2419                                 }
2420                                 basefont = getLayoutFont(bparams, outerfont);
2421                                 running_font = basefont;
2422
2423                                 column += Changes::latexMarkChange(os, bparams,
2424                                                 runningChange, Change(Change::UNCHANGED),
2425                                                 runparams);
2426                                 runningChange = Change(Change::UNCHANGED);
2427
2428                                 os << "}] ";
2429                                 column +=3;
2430                         }
2431                         // For InTitle commands, we have already opened a group
2432                         // in output_latex::TeXOnePar.
2433                         if (style.isCommand() && !style.intitle) {
2434                                 os << '{';
2435                                 ++column;
2436                         }
2437
2438                         if (!style.leftdelim().empty()) {
2439                                 os << style.leftdelim();
2440                                 column += style.leftdelim().size();
2441                         }
2442
2443                         if (allowcust)
2444                                 column += d->startTeXParParams(bparams, os,
2445                                                             runparams);
2446                 }
2447
2448                 runparams.wasDisplayMath = runparams.inDisplayMath;
2449                 runparams.inDisplayMath = false;
2450                 bool deleted_display_math = false;
2451                 Change const & change = runparams.inDeletedInset
2452                         ? runparams.changeOfDeletedInset : lookupChange(i);
2453
2454                 // Check whether a display math inset follows
2455                 if (d->text_[i] == META_INSET
2456                     && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2457                         InsetMath const * im = getInset(i)->asInsetMath();
2458                         if (im && im->asHullInset()
2459                             && im->asHullInset()->outerDisplay()) {
2460                                 runparams.inDisplayMath = true;
2461                                 // runparams.inDeletedInset will be set by
2462                                 // latexInset later, but we need this info
2463                                 // before it is called. On the other hand, we
2464                                 // cannot set it here because it is a counter.
2465                                 deleted_display_math = isDeleted(i);
2466                         }
2467                         if (bparams.output_changes && deleted_display_math
2468                             && runningChange == change
2469                             && change.type == Change::DELETED
2470                             && !os.afterParbreak()) {
2471                                 // A display math in the same paragraph follows.
2472                                 // We have to close and then reopen \lyxdeleted,
2473                                 // otherwise the math will be shifted up.
2474                                 OutputParams rp = runparams;
2475                                 if (open_font) {
2476                                         bool needPar = false;
2477                                         column += running_font.latexWriteEndChanges(
2478                                                 os, bparams, rp, basefont,
2479                                                 basefont, needPar);
2480                                         open_font = false;
2481                                 }
2482                                 basefont = getLayoutFont(bparams, outerfont);
2483                                 running_font = basefont;
2484                                 column += Changes::latexMarkChange(os, bparams,
2485                                         Change(Change::INSERTED), change, rp);
2486                         }
2487                 }
2488
2489                 if (bparams.output_changes && runningChange != change) {
2490                         if (!alien_script.empty()) {
2491                                 column += 1;
2492                                 os << "}";
2493                                 alien_script.clear();
2494                         }
2495                         if (open_font) {
2496                                 bool needPar = false;
2497                                 column += running_font.latexWriteEndChanges(
2498                                                 os, bparams, runparams,
2499                                                 basefont, basefont, needPar);
2500                                 open_font = false;
2501                         }
2502                         basefont = getLayoutFont(bparams, outerfont);
2503                         running_font = basefont;
2504                         column += Changes::latexMarkChange(os, bparams, runningChange,
2505                                                            change, runparams);
2506                         runningChange = change;
2507                 }
2508
2509                 // do not output text which is marked deleted
2510                 // if change tracking output is disabled
2511                 if (!bparams.output_changes && change.deleted()) {
2512                         continue;
2513                 }
2514
2515                 ++column;
2516
2517                 // Fully instantiated font
2518                 Font const current_font = getFont(bparams, i, outerfont);
2519
2520                 Font const last_font = running_font;
2521
2522                 // Do we need to close the previous font?
2523                 if (open_font &&
2524                     (current_font != running_font ||
2525                      current_font.language() != running_font.language()))
2526                 {
2527                         // ensure there is no open script-wrapper
2528                         if (!alien_script.empty()) {
2529                                 column += 1;
2530                                 os << "}";
2531                                 alien_script.clear();
2532                         }
2533                         bool needPar = false;
2534                         column += running_font.latexWriteEndChanges(
2535                                     os, bparams, runparams, basefont,
2536                                     (i == body_pos-1) ? basefont : current_font,
2537                                     needPar);
2538                         running_font = basefont;
2539                         open_font = false;
2540                 }
2541
2542                 string const running_lang = runparams.use_polyglossia ?
2543                         running_font.language()->polyglossia() : running_font.language()->babel();
2544                 // close babel's font environment before opening CJK.
2545                 string const lang_end_command = runparams.use_polyglossia ?
2546                         "\\end{$$lang}" : lyxrc.language_command_end;
2547                 bool const using_begin_end = runparams.use_polyglossia ||
2548                                                 !lang_end_command.empty();
2549                 if (!running_lang.empty() &&
2550                     (!using_begin_end || running_lang == openLanguageName()) &&
2551                     current_font.language()->encoding()->package() == Encoding::CJK) {
2552                                 string end_tag = subst(lang_end_command,
2553                                                         "$$lang",
2554                                                         running_lang);
2555                                 os << from_ascii(end_tag);
2556                                 column += end_tag.length();
2557                                 if (using_begin_end)
2558                                         popLanguageName();
2559                 }
2560
2561                 // Switch file encoding if necessary (and allowed)
2562                 if (!runparams.pass_thru && !style.pass_thru &&
2563                     runparams.encoding->package() != Encoding::none &&
2564                     current_font.language()->encoding()->package() != Encoding::none) {
2565                         pair<bool, int> const enc_switch =
2566                                 switchEncoding(os.os(), bparams, runparams,
2567                                         *(current_font.language()->encoding()));
2568                         if (enc_switch.first) {
2569                                 column += enc_switch.second;
2570                                 runparams.encoding = current_font.language()->encoding();
2571                         }
2572                 }
2573
2574                 char_type const c = d->text_[i];
2575
2576                 // A display math inset inside an ulem command will be output
2577                 // as a box of width \linewidth, so we have to either disable
2578                 // indentation if the inset starts a paragraph, or start a new
2579                 // line to accommodate such box. This has to be done before
2580                 // writing any font changing commands.
2581                 if (runparams.inDisplayMath && !deleted_display_math
2582                     && runparams.inulemcmd) {
2583                         if (os.afterParbreak())
2584                                 os << "\\noindent";
2585                         else
2586                                 os << "\\\\\n";
2587                 }
2588
2589                 // Do we need to change font?
2590                 if ((current_font != running_font ||
2591                      current_font.language() != running_font.language()) &&
2592                         i != body_pos - 1)
2593                 {
2594                         odocstringstream ods;
2595                         column += current_font.latexWriteStartChanges(ods, bparams,
2596                                                               runparams, basefont,
2597                                                               last_font);
2598                         // Check again for display math in ulem commands as a
2599                         // font change may also occur just before a math inset.
2600                         if (runparams.inDisplayMath && !deleted_display_math
2601                             && runparams.inulemcmd) {
2602                                 if (os.afterParbreak())
2603                                         os << "\\noindent";
2604                                 else
2605                                         os << "\\\\\n";
2606                         }
2607                         running_font = current_font;
2608                         open_font = true;
2609                         docstring fontchange = ods.str();
2610                         // check whether the fontchange ends with a \\textcolor
2611                         // modifier and the text starts with a space (bug 4473)
2612                         docstring const last_modifier = rsplit(fontchange, '\\');
2613                         if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2614                                 os << fontchange << from_ascii("{}");
2615                         // check if the fontchange ends with a trailing blank
2616                         // (like "\small " (see bug 3382)
2617                         else if (suffixIs(fontchange, ' ') && c == ' ')
2618                                 os << fontchange.substr(0, fontchange.size() - 1)
2619                                    << from_ascii("{}");
2620                         else
2621                                 os << fontchange;
2622                 }
2623
2624                 // FIXME: think about end_pos implementation...
2625                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2626                         // FIXME: integrate this case in latexSpecialChar
2627                         // Do not print the separation of the optional argument
2628                         // if style.pass_thru is false. This works because
2629                         // latexSpecialChar ignores spaces if
2630                         // style.pass_thru is false.
2631                         if (i != body_pos - 1) {
2632                                 if (d->simpleTeXBlanks(bparams, runparams, os,
2633                                                 i, column, current_font, style)) {
2634                                         // A surrogate pair was output. We
2635                                         // must not call latexSpecialChar
2636                                         // in this iteration, since it would output
2637                                         // the combining character again.
2638                                         ++i;
2639                                         continue;
2640                                 }
2641                         }
2642                 }
2643
2644                 OutputParams rp = runparams;
2645                 rp.free_spacing = style.free_spacing;
2646                 rp.local_font = &current_font;
2647                 rp.intitle = style.intitle;
2648
2649                 // Two major modes:  LaTeX or plain
2650                 // Handle here those cases common to both modes
2651                 // and then split to handle the two modes separately.
2652                 if (c == META_INSET) {
2653                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2654                                 // Greyedout notes and, in general, all insets
2655                                 // with InsetLayout::isDisplay() == false,
2656                                 // are typeset inline with the text. So, we
2657                                 // can add a \par to the last paragraph of
2658                                 // such insets only if nothing else follows.
2659                                 bool incremented = false;
2660                                 Inset const * inset = getInset(i);
2661                                 InsetText const * textinset = inset
2662                                                         ? inset->asInsetText()
2663                                                         : 0;
2664                                 if (i + 1 == size() && textinset
2665                                     && !inset->getLayout().isDisplay()) {
2666                                         ParagraphList const & pars =
2667                                                 textinset->text().paragraphs();
2668                                         pit_type const pit = pars.size() - 1;
2669                                         Font const lastfont =
2670                                                 pit < 0 || pars[pit].empty()
2671                                                 ? pars[pit].getLayoutFont(
2672                                                                 bparams,
2673                                                                 outerfont)
2674                                                 : pars[pit].getFont(bparams,
2675                                                         pars[pit].size() - 1,
2676                                                         outerfont);
2677                                         if (lastfont.fontInfo().size() !=
2678                                             basefont.fontInfo().size()) {
2679                                                 ++parInline;
2680                                                 incremented = true;
2681                                         }
2682                                 }
2683                                 d->latexInset(bparams, os, rp, running_font,
2684                                                 basefont, outerfont, open_font,
2685                                                 runningChange, style, i, column);
2686                                 if (incremented)
2687                                         --parInline;
2688                         }
2689                 } else if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2690                         if (!bparams.useNonTeXFonts)
2691                           script = Encodings::isKnownScriptChar(c);
2692                         if (script != alien_script) {
2693                                 if (!alien_script.empty()) {
2694                                         os << "}";
2695                                         alien_script.clear();
2696                                 }
2697                                 string fontenc = running_font.language()->fontenc(bparams);
2698                                 if (!script.empty()
2699                                         && !Encodings::fontencSupportsScript(fontenc, script)) {
2700                                         column += script.length() + 2;
2701                                         os << "\\" << script << "{";
2702                                         alien_script = script;
2703                                 }
2704                         }
2705                         try {
2706                                 d->latexSpecialChar(os, bparams, rp, running_font,
2707                                                                         alien_script, style, i, end_pos, column);
2708                         } catch (EncodingException & e) {
2709                                 if (runparams.dryrun) {
2710                                         os << "<" << _("LyX Warning: ")
2711                                            << _("uncodable character") << " '";
2712                                         os.put(c);
2713                                         os << "'>";
2714                                 } else {
2715                                         // add location information and throw again.
2716                                         e.par_id = id();
2717                                         e.pos = i;
2718                                         throw(e);
2719                                 }
2720                         }
2721                 }
2722
2723                 // Set the encoding to that returned from latexSpecialChar (see
2724                 // comment for encoding member in OutputParams.h)
2725                 runparams.encoding = rp.encoding;
2726
2727                 // Also carry on the info on a closed ulem command for insets
2728                 // such as Note that do not produce any output, so that no
2729                 // command is ever executed but its opening was recorded.
2730                 runparams.inulemcmd = rp.inulemcmd;
2731
2732                 // And finally, pass the post_macros upstream
2733                 runparams.post_macro = rp.post_macro;
2734         }
2735
2736     // Close wrapper for alien script
2737         if (!alien_script.empty()) {
2738                 os << "}";
2739                 alien_script.clear();
2740         }
2741
2742         // If we have an open font definition, we have to close it
2743         if (open_font) {
2744                 // Make sure that \\par is done with the font of the last
2745                 // character if this has another size as the default.
2746                 // This is necessary because LaTeX (and LyX on the screen)
2747                 // calculates the space between the baselines according
2748                 // to this font. (Matthias)
2749                 //
2750                 // We must not change the font for the last paragraph
2751                 // of non-multipar insets, tabular cells or commands,
2752                 // since this produces unwanted whitespace.
2753
2754                 Font const font = empty()
2755                         ? getLayoutFont(bparams, outerfont)
2756                         : getFont(bparams, size() - 1, outerfont);
2757
2758                 InsetText const * textinset = inInset().asInsetText();
2759
2760                 bool const maintext = textinset
2761                         ? textinset->text().isMainText()
2762                         : false;
2763
2764                 size_t const numpars = textinset
2765                         ? textinset->text().paragraphs().size()
2766                         : 0;
2767
2768                 bool needPar = false;
2769
2770                 if (style.resfont.size() != font.fontInfo().size()
2771                     && (!runparams.isLastPar || maintext
2772                         || (numpars > 1 && d->ownerCode() != CELL_CODE
2773                             && (inInset().getLayout().isDisplay()
2774                                 || parInline)))
2775                     && !style.isCommand()) {
2776                         needPar = true;
2777                 }
2778 #ifdef FIXED_LANGUAGE_END_DETECTION
2779                 if (next_) {
2780                         running_font.latexWriteEndChanges(os, bparams,
2781                                         runparams, basefont,
2782                                         next_->getFont(bparams, 0, outerfont),
2783                                                        needPar);
2784                 } else {
2785                         running_font.latexWriteEndChanges(os, bparams,
2786                                         runparams, basefont, basefont, needPar);
2787                 }
2788 #else
2789 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2790 //FIXME: there as we start another \selectlanguage with the next paragraph if
2791 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2792                 running_font.latexWriteEndChanges(os, bparams, runparams,
2793                                 basefont, basefont, needPar);
2794 #endif
2795                 if (needPar) {
2796                         // The \par could not be inserted at the same nesting
2797                         // level of the font size change, so do it now.
2798                         os << "{\\" << font.latexSize() << "\\par}";
2799                 }
2800         }
2801
2802         column += Changes::latexMarkChange(os, bparams, runningChange,
2803                                            Change(Change::UNCHANGED), runparams);
2804
2805         // Needed if there is an optional argument but no contents.
2806         if (body_pos > 0 && body_pos == size()) {
2807                 os << "}]~";
2808         }
2809
2810         if (!style.rightdelim().empty()) {
2811                 os << style.rightdelim();
2812                 column += style.rightdelim().size();
2813         }
2814
2815         if (allowcust && d->endTeXParParams(bparams, os, runparams)
2816             && runparams.encoding != prev_encoding) {
2817                 runparams.encoding = prev_encoding;
2818                 os << setEncoding(prev_encoding->iconvName());
2819         }
2820
2821         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2822 }
2823
2824
2825 bool Paragraph::emptyTag() const
2826 {
2827         for (pos_type i = 0; i < size(); ++i) {
2828                 if (Inset const * inset = getInset(i)) {
2829                         InsetCode lyx_code = inset->lyxCode();
2830                         // FIXME testing like that is wrong. What is
2831                         // the intent?
2832                         if (lyx_code != TOC_CODE &&
2833                             lyx_code != INCLUDE_CODE &&
2834                             lyx_code != GRAPHICS_CODE &&
2835                             lyx_code != ERT_CODE &&
2836                             lyx_code != LISTINGS_CODE &&
2837                             lyx_code != FLOAT_CODE &&
2838                             lyx_code != TABULAR_CODE) {
2839                                 return false;
2840                         }
2841                 } else {
2842                         char_type c = d->text_[i];
2843                         if (c != ' ' && c != '\t')
2844                                 return false;
2845                 }
2846         }
2847         return true;
2848 }
2849
2850
2851 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
2852         const
2853 {
2854         for (pos_type i = 0; i < size(); ++i) {
2855                 if (Inset const * inset = getInset(i)) {
2856                         InsetCode lyx_code = inset->lyxCode();
2857                         if (lyx_code == LABEL_CODE) {
2858                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2859                                 docstring const & id = il->getParam("name");
2860                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2861                         }
2862                 }
2863         }
2864         return string();
2865 }
2866
2867
2868 pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runparams)
2869         const
2870 {
2871         pos_type i;
2872         for (i = 0; i < size(); ++i) {
2873                 if (Inset const * inset = getInset(i)) {
2874                         inset->docbook(os, runparams);
2875                 } else {
2876                         char_type c = d->text_[i];
2877                         if (c == ' ')
2878                                 break;
2879                         os << sgml::escapeChar(c);
2880                 }
2881         }
2882         return i;
2883 }
2884
2885
2886 pos_type Paragraph::firstWordLyXHTML(XHTMLStream & xs, OutputParams const & runparams)
2887         const
2888 {
2889         pos_type i;
2890         for (i = 0; i < size(); ++i) {
2891                 if (Inset const * inset = getInset(i)) {
2892                         inset->xhtml(xs, runparams);
2893                 } else {
2894                         char_type c = d->text_[i];
2895                         if (c == ' ')
2896                                 break;
2897                         xs << c;
2898                 }
2899         }
2900         return i;
2901 }
2902
2903
2904 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2905 {
2906         Font font_old;
2907         pos_type size = text_.size();
2908         for (pos_type i = initial; i < size; ++i) {
2909                 Font font = owner_->getFont(buf.params(), i, outerfont);
2910                 if (text_[i] == META_INSET)
2911                         return false;
2912                 if (i != initial && font != font_old)
2913                         return false;
2914                 font_old = font;
2915         }
2916
2917         return true;
2918 }
2919
2920
2921 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
2922                                     odocstream & os,
2923                                     OutputParams const & runparams,
2924                                     Font const & outerfont,
2925                                     pos_type initial) const
2926 {
2927         bool emph_flag = false;
2928
2929         Layout const & style = *d->layout_;
2930         FontInfo font_old =
2931                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2932
2933         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2934                 os << "]]>";
2935
2936         // parsing main loop
2937         for (pos_type i = initial; i < size(); ++i) {
2938                 Font font = getFont(buf.params(), i, outerfont);
2939
2940                 // handle <emphasis> tag
2941                 if (font_old.emph() != font.fontInfo().emph()) {
2942                         if (font.fontInfo().emph() == FONT_ON) {
2943                                 os << "<emphasis>";
2944                                 emph_flag = true;
2945                         } else if (i != initial) {
2946                                 os << "</emphasis>";
2947                                 emph_flag = false;
2948                         }
2949                 }
2950
2951                 if (Inset const * inset = getInset(i)) {
2952                         inset->docbook(os, runparams);
2953                 } else {
2954                         char_type c = d->text_[i];
2955
2956                         if (style.pass_thru)
2957                                 os.put(c);
2958                         else
2959                                 os << sgml::escapeChar(c);
2960                 }
2961                 font_old = font.fontInfo();
2962         }
2963
2964         if (emph_flag) {
2965                 os << "</emphasis>";
2966         }
2967
2968         if (style.free_spacing)
2969                 os << '\n';
2970         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2971                 os << "<![CDATA[";
2972 }
2973
2974
2975 namespace {
2976 void doFontSwitch(vector<html::FontTag> & tagsToOpen,
2977                   vector<html::EndFontTag> & tagsToClose,
2978                   bool & flag, FontState curstate, html::FontTypes type)
2979 {
2980         if (curstate == FONT_ON) {
2981                 tagsToOpen.push_back(html::FontTag(type));
2982                 flag = true;
2983         } else if (flag) {
2984                 tagsToClose.push_back(html::EndFontTag(type));
2985                 flag = false;
2986         }
2987 }
2988 } // namespace
2989
2990
2991 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
2992                                     XHTMLStream & xs,
2993                                     OutputParams const & runparams,
2994                                     Font const & outerfont,
2995                                     bool start_paragraph, bool close_paragraph,
2996                                     pos_type initial) const
2997 {
2998         docstring retval;
2999
3000         // track whether we have opened these tags
3001         bool emph_flag = false;
3002         bool bold_flag = false;
3003         bool noun_flag = false;
3004         bool ubar_flag = false;
3005         bool dbar_flag = false;
3006         bool sout_flag = false;
3007         bool xout_flag = false;
3008         bool wave_flag = false;
3009         // shape tags
3010         bool shap_flag = false;
3011         // family tags
3012         bool faml_flag = false;
3013         // size tags
3014         bool size_flag = false;
3015
3016         Layout const & style = *d->layout_;
3017
3018         if (start_paragraph)
3019                 xs.startDivision(allowEmpty());
3020
3021         FontInfo font_old =
3022                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
3023
3024         FontShape  curr_fs   = INHERIT_SHAPE;
3025         FontFamily curr_fam  = INHERIT_FAMILY;
3026         FontSize   curr_size = FONT_SIZE_INHERIT;
3027
3028         string const default_family =
3029                 buf.masterBuffer()->params().fonts_default_family;
3030
3031         vector<html::FontTag> tagsToOpen;
3032         vector<html::EndFontTag> tagsToClose;
3033
3034         // parsing main loop
3035         for (pos_type i = initial; i < size(); ++i) {
3036                 // let's not show deleted material in the output
3037                 if (isDeleted(i))
3038                         continue;
3039
3040                 Font const font = getFont(buf.masterBuffer()->params(), i, outerfont);
3041
3042                 // emphasis
3043                 FontState curstate = font.fontInfo().emph();
3044                 if (font_old.emph() != curstate)
3045                         doFontSwitch(tagsToOpen, tagsToClose, emph_flag, curstate, html::FT_EMPH);
3046
3047                 // noun
3048                 curstate = font.fontInfo().noun();
3049                 if (font_old.noun() != curstate)
3050                         doFontSwitch(tagsToOpen, tagsToClose, noun_flag, curstate, html::FT_NOUN);
3051
3052                 // underbar
3053                 curstate = font.fontInfo().underbar();
3054                 if (font_old.underbar() != curstate)
3055                         doFontSwitch(tagsToOpen, tagsToClose, ubar_flag, curstate, html::FT_UBAR);
3056
3057                 // strikeout
3058                 curstate = font.fontInfo().strikeout();
3059                 if (font_old.strikeout() != curstate)
3060                         doFontSwitch(tagsToOpen, tagsToClose, sout_flag, curstate, html::FT_SOUT);
3061
3062                 // xout
3063                 curstate = font.fontInfo().xout();
3064                 if (font_old.xout() != curstate)
3065                         doFontSwitch(tagsToOpen, tagsToClose, xout_flag, curstate, html::FT_XOUT);
3066
3067                 // double underbar
3068                 curstate = font.fontInfo().uuline();
3069                 if (font_old.uuline() != curstate)
3070                         doFontSwitch(tagsToOpen, tagsToClose, dbar_flag, curstate, html::FT_DBAR);
3071
3072                 // wavy line
3073                 curstate = font.fontInfo().uwave();
3074                 if (font_old.uwave() != curstate)
3075                         doFontSwitch(tagsToOpen, tagsToClose, wave_flag, curstate, html::FT_WAVE);
3076
3077                 // bold
3078                 // a little hackish, but allows us to reuse what we have.
3079                 curstate = (font.fontInfo().series() == BOLD_SERIES ? FONT_ON : FONT_OFF);
3080                 if (font_old.series() != font.fontInfo().series())
3081                         doFontSwitch(tagsToOpen, tagsToClose, bold_flag, curstate, html::FT_BOLD);
3082
3083                 // Font shape
3084                 curr_fs = font.fontInfo().shape();
3085                 FontShape old_fs = font_old.shape();
3086                 if (old_fs != curr_fs) {
3087                         if (shap_flag) {
3088                                 switch (old_fs) {
3089                                 case ITALIC_SHAPE:
3090                                         tagsToClose.push_back(html::EndFontTag(html::FT_ITALIC));
3091                                         break;
3092                                 case SLANTED_SHAPE:
3093                                         tagsToClose.push_back(html::EndFontTag(html::FT_SLANTED));
3094                                         break;
3095                                 case SMALLCAPS_SHAPE:
3096                                         tagsToClose.push_back(html::EndFontTag(html::FT_SMALLCAPS));
3097                                         break;
3098                                 case UP_SHAPE:
3099                                 case INHERIT_SHAPE:
3100                                         break;
3101                                 default:
3102                                         // the other tags are for internal use
3103                                         LATTEST(false);
3104                                         break;
3105                                 }
3106                                 shap_flag = false;
3107                         }
3108                         switch (curr_fs) {
3109                         case ITALIC_SHAPE:
3110                                 tagsToOpen.push_back(html::FontTag(html::FT_ITALIC));
3111                                 shap_flag = true;
3112                                 break;
3113                         case SLANTED_SHAPE:
3114                                 tagsToOpen.push_back(html::FontTag(html::FT_SLANTED));
3115                                 shap_flag = true;
3116                                 break;
3117                         case SMALLCAPS_SHAPE:
3118                                 tagsToOpen.push_back(html::FontTag(html::FT_SMALLCAPS));
3119                                 shap_flag = true;
3120                                 break;
3121                         case UP_SHAPE:
3122                         case INHERIT_SHAPE:
3123                                 break;
3124                         default:
3125                                 // the other tags are for internal use
3126                                 LATTEST(false);
3127                                 break;
3128                         }
3129                 }
3130
3131                 // Font family
3132                 curr_fam = font.fontInfo().family();
3133                 FontFamily old_fam = font_old.family();
3134                 if (old_fam != curr_fam) {
3135                         if (faml_flag) {
3136                                 switch (old_fam) {
3137                                 case ROMAN_FAMILY:
3138                                         tagsToClose.push_back(html::EndFontTag(html::FT_ROMAN));
3139                                         break;
3140                                 case SANS_FAMILY:
3141                                         tagsToClose.push_back(html::EndFontTag(html::FT_SANS));
3142                                         break;
3143                                 case TYPEWRITER_FAMILY:
3144                                         tagsToClose.push_back(html::EndFontTag(html::FT_TYPE));
3145                                         break;
3146                                 case INHERIT_FAMILY:
3147                                         break;
3148                                 default:
3149                                         // the other tags are for internal use
3150                                         LATTEST(false);
3151                                         break;
3152                                 }
3153                                 faml_flag = false;
3154                         }
3155                         switch (curr_fam) {
3156                         case ROMAN_FAMILY:
3157                                 // we will treat a "default" font family as roman, since we have
3158                                 // no other idea what to do.
3159                                 if (default_family != "rmdefault" && default_family != "default") {
3160                                         tagsToOpen.push_back(html::FontTag(html::FT_ROMAN));
3161                                         faml_flag = true;
3162                                 }
3163                                 break;
3164                         case SANS_FAMILY:
3165                                 if (default_family != "sfdefault") {
3166                                         tagsToOpen.push_back(html::FontTag(html::FT_SANS));
3167                                         faml_flag = true;
3168                                 }
3169                                 break;
3170                         case TYPEWRITER_FAMILY:
3171                                 if (default_family != "ttdefault") {
3172                                         tagsToOpen.push_back(html::FontTag(html::FT_TYPE));
3173                                         faml_flag = true;
3174                                 }
3175                                 break;
3176                         case INHERIT_FAMILY:
3177                                 break;
3178                         default:
3179                                 // the other tags are for internal use
3180                                 LATTEST(false);
3181                                 break;
3182                         }
3183                 }
3184
3185                 // Font size
3186                 curr_size = font.fontInfo().size();
3187                 FontSize old_size = font_old.size();
3188                 if (old_size != curr_size) {
3189                         if (size_flag) {
3190                                 switch (old_size) {
3191                                 case FONT_SIZE_TINY:
3192                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_TINY));
3193                                         break;
3194                                 case FONT_SIZE_SCRIPT:
3195                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SCRIPT));
3196                                         break;
3197                                 case FONT_SIZE_FOOTNOTE:
3198                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_FOOTNOTE));
3199                                         break;
3200                                 case FONT_SIZE_SMALL:
3201                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SMALL));
3202                                         break;
3203                                 case FONT_SIZE_LARGE:
3204                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGE));
3205                                         break;
3206                                 case FONT_SIZE_LARGER:
3207                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGER));
3208                                         break;
3209                                 case FONT_SIZE_LARGEST:
3210                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGEST));
3211                                         break;
3212                                 case FONT_SIZE_HUGE:
3213                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGE));
3214                                         break;
3215                                 case FONT_SIZE_HUGER:
3216                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGER));
3217                                         break;
3218                                 case FONT_SIZE_INCREASE:
3219                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_INCREASE));
3220                                         break;
3221                                 case FONT_SIZE_DECREASE:
3222                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_DECREASE));
3223                                         break;
3224                                 case FONT_SIZE_INHERIT:
3225                                 case FONT_SIZE_NORMAL:
3226                                         break;
3227                                 default:
3228                                         // the other tags are for internal use
3229                                         LATTEST(false);
3230                                         break;
3231                                 }
3232                                 size_flag = false;
3233                         }
3234                         switch (curr_size) {
3235                         case FONT_SIZE_TINY:
3236                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_TINY));
3237                                 size_flag = true;
3238                                 break;
3239                         case FONT_SIZE_SCRIPT:
3240                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SCRIPT));
3241                                 size_flag = true;
3242                                 break;
3243                         case FONT_SIZE_FOOTNOTE:
3244                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_FOOTNOTE));
3245                                 size_flag = true;
3246                                 break;
3247                         case FONT_SIZE_SMALL:
3248                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SMALL));
3249                                 size_flag = true;
3250                                 break;
3251                         case FONT_SIZE_LARGE:
3252                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGE));
3253                                 size_flag = true;
3254                                 break;
3255                         case FONT_SIZE_LARGER:
3256                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGER));
3257                                 size_flag = true;
3258                                 break;
3259                         case FONT_SIZE_LARGEST:
3260                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGEST));
3261                                 size_flag = true;
3262                                 break;
3263                         case FONT_SIZE_HUGE:
3264                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGE));
3265                                 size_flag = true;
3266                                 break;
3267                         case FONT_SIZE_HUGER:
3268                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGER));
3269                                 size_flag = true;
3270                                 break;
3271                         case FONT_SIZE_INCREASE:
3272                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_INCREASE));
3273                                 size_flag = true;
3274                                 break;
3275                         case FONT_SIZE_DECREASE:
3276                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_DECREASE));
3277                                 size_flag = true;
3278                                 break;
3279                         case FONT_SIZE_NORMAL:
3280                         case FONT_SIZE_INHERIT:
3281                                 break;
3282                         default:
3283                                 // the other tags are for internal use
3284                                 LATTEST(false);
3285                                 break;
3286                         }
3287                 }
3288
3289                 // FIXME XHTML
3290                 // Other such tags? What about the other text ranges?
3291
3292                 vector<html::EndFontTag>::const_iterator cit = tagsToClose.begin();
3293                 vector<html::EndFontTag>::const_iterator cen = tagsToClose.end();
3294                 for (; cit != cen; ++cit)
3295                         xs << *cit;
3296
3297                 vector<html::FontTag>::const_iterator sit = tagsToOpen.begin();
3298                 vector<html::FontTag>::const_iterator sen = tagsToOpen.end();
3299                 for (; sit != sen; ++sit)
3300                         xs << *sit;
3301
3302                 tagsToClose.clear();
3303                 tagsToOpen.clear();
3304
3305                 Inset const * inset = getInset(i);
3306                 if (inset) {
3307                         if (!runparams.for_toc || inset->isInToc()) {
3308                                 OutputParams np = runparams;
3309                                 np.local_font = &font;
3310                                 // If the paragraph has size 1, then we are in the "special
3311                                 // case" where we do not output the containing paragraph info
3312                                 if (!inset->getLayout().htmlisblock() && size() != 1)
3313                                         np.html_in_par = true;
3314                                 retval += inset->xhtml(xs, np);
3315                         }
3316                 } else {
3317                         char_type c = getUChar(buf.masterBuffer()->params(),
3318                                                runparams, i);
3319                         if (c == ' ' && (style.free_spacing || runparams.free_spacing))
3320                                 xs << XHTMLStream::ESCAPE_NONE << "&nbsp;";
3321                         else
3322                                 xs << c;
3323                 }
3324                 font_old = font.fontInfo();
3325         }
3326
3327         // FIXME XHTML
3328         // I'm worried about what happens if a branch, say, is itself
3329         // wrapped in some font stuff. I think that will not work.
3330         xs.closeFontTags();
3331         if (close_paragraph)
3332                 xs.endDivision();
3333
3334         return retval;
3335 }
3336
3337
3338 bool Paragraph::isHfill(pos_type pos) const
3339 {
3340         Inset const * inset = getInset(pos);
3341         return inset && inset->isHfill();
3342 }
3343
3344
3345 bool Paragraph::isNewline(pos_type pos) const
3346 {
3347         // U+2028 LINE SEPARATOR
3348         // U+2029 PARAGRAPH SEPARATOR
3349         char_type const c = d->text_[pos];
3350         if (c == 0x2028 || c == 0x2029)
3351                 return true;
3352         Inset const * inset = getInset(pos);
3353         return inset && inset->lyxCode() == NEWLINE_CODE;
3354 }
3355
3356
3357 bool Paragraph::isEnvSeparator(pos_type pos) const
3358 {
3359         Inset const * inset = getInset(pos);
3360         return inset && inset->lyxCode() == SEPARATOR_CODE;
3361 }
3362
3363
3364 bool Paragraph::isLineSeparator(pos_type pos) const
3365 {
3366         char_type const c = d->text_[pos];
3367         if (isLineSeparatorChar(c))
3368                 return true;
3369         Inset const * inset = getInset(pos);
3370         return inset && inset->isLineSeparator();
3371 }
3372
3373
3374 bool Paragraph::isWordSeparator(pos_type pos) const
3375 {
3376         if (pos == size())
3377                 return true;
3378         if (Inset const * inset = getInset(pos))
3379                 return !inset->isLetter();
3380         // if we have a hard hyphen (no en- or emdash) or apostrophe
3381         // we pass this to the spell checker
3382         // FIXME: this method is subject to change, visit
3383         // https://bugzilla.mozilla.org/show_bug.cgi?id=355178
3384         // to get an impression how complex this is.
3385         if (isHardHyphenOrApostrophe(pos))
3386                 return false;
3387         char_type const c = d->text_[pos];
3388         // We want to pass the escape chars to the spellchecker
3389         docstring const escape_chars = from_utf8(lyxrc.spellchecker_esc_chars);
3390         return !isLetterChar(c) && !isDigitASCII(c) && !contains(escape_chars, c);
3391 }
3392
3393
3394 bool Paragraph::isHardHyphenOrApostrophe(pos_type pos) const
3395 {
3396         pos_type const psize = size();
3397         if (pos >= psize)
3398                 return false;
3399         char_type const c = d->text_[pos];
3400         if (c != '-' && c != '\'')
3401                 return false;
3402         int nextpos = pos + 1;
3403         int prevpos = pos > 0 ? pos - 1 : 0;
3404         if ((nextpos == psize || isSpace(nextpos))
3405                 && (pos == 0 || isSpace(prevpos)))
3406                 return false;
3407         return true;
3408 }
3409
3410
3411 bool Paragraph::needsCProtection(bool const fragile) const
3412 {
3413         // first check the layout of the paragraph, but only in insets
3414         InsetText const * textinset = inInset().asInsetText();
3415         bool const maintext = textinset
3416                 ? textinset->text().isMainText()
3417                 : false;
3418
3419         if (!maintext && layout().needcprotect) {
3420                 // Environments need cprotection regardless the content
3421                 if (layout().latextype == LATEX_ENVIRONMENT)
3422                         return true;
3423
3424                 // Commands need cprotection if they contain specific chars
3425                 int const nchars_escape = 9;
3426                 static char_type const chars_escape[nchars_escape] = {
3427                         '&', '_', '$', '%', '#', '^', '{', '}', '\\'};
3428
3429                 docstring const pars = asString();
3430                 for (int k = 0; k < nchars_escape; k++) {
3431                         if (contains(pars, chars_escape[k]))
3432                                 return true;
3433                 }
3434         }
3435
3436         // now check whether we have insets that need cprotection
3437         pos_type size = d->text_.size();
3438         for (pos_type i = 0; i < size; ++i)
3439                 if (isInset(i) && getInset(i)->needsCProtection(maintext, fragile))
3440                         return true;
3441
3442         return false;
3443 }
3444
3445
3446 FontSpan const & Paragraph::getSpellRange(pos_type pos) const
3447 {
3448         return d->speller_state_.getRange(pos);
3449 }
3450
3451
3452 bool Paragraph::isChar(pos_type pos) const
3453 {
3454         if (Inset const * inset = getInset(pos))
3455                 return inset->isChar();
3456         char_type const c = d->text_[pos];
3457         return !isLetterChar(c) && !isDigitASCII(c) && !lyx::isSpace(c);
3458 }
3459
3460
3461 bool Paragraph::isSpace(pos_type pos) const
3462 {
3463         if (Inset const * inset = getInset(pos))
3464                 return inset->isSpace();
3465         char_type const c = d->text_[pos];
3466         return lyx::isSpace(c);
3467 }
3468
3469
3470 Language const *
3471 Paragraph::getParLanguage(BufferParams const & bparams) const
3472 {
3473         if (!empty())
3474                 return getFirstFontSettings(bparams).language();
3475         // FIXME: we should check the prev par as well (Lgb)
3476         return bparams.language;
3477 }
3478
3479
3480 bool Paragraph::isRTL(BufferParams const & bparams) const
3481 {
3482         return getParLanguage(bparams)->rightToLeft()
3483                 && !inInset().getLayout().forceLTR();
3484 }
3485
3486
3487 void Paragraph::changeLanguage(BufferParams const & bparams,
3488                                Language const * from, Language const * to)
3489 {
3490         // change language including dummy font change at the end
3491         for (pos_type i = 0; i <= size(); ++i) {
3492                 Font font = getFontSettings(bparams, i);
3493                 if (font.language() == from) {
3494                         font.setLanguage(to);
3495                         setFont(i, font);
3496                         d->requestSpellCheck(i);
3497                 }
3498         }
3499 }
3500
3501
3502 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
3503 {
3504         Language const * doc_language = bparams.language;
3505         for (auto const & f : d->fontlist_)
3506                 if (f.font().language() != ignore_language &&
3507                     f.font().language() != latex_language &&
3508                     f.font().language() != doc_language)
3509                         return true;
3510         return false;
3511 }
3512
3513
3514 void Paragraph::getLanguages(std::set<Language const *> & langs) const
3515 {
3516         for (auto const & f : d->fontlist_) {
3517                 Language const * lang = f.font().language();
3518                 if (lang != ignore_language &&
3519                     lang != latex_language)
3520                         langs.insert(lang);
3521         }
3522 }
3523
3524
3525 docstring Paragraph::asString(int options) const
3526 {
3527         return asString(0, size(), options);
3528 }
3529
3530
3531 docstring Paragraph::asString(pos_type beg, pos_type end, int options, const OutputParams *runparams) const
3532 {
3533         odocstringstream os;
3534
3535         if (beg == 0
3536             && options & AS_STR_LABEL
3537             && !d->params_.labelString().empty())
3538                 os << d->params_.labelString() << ' ';
3539
3540         for (pos_type i = beg; i < end; ++i) {
3541                 if ((options & AS_STR_SKIPDELETE) && isDeleted(i))
3542                         continue;
3543                 char_type const c = d->text_[i];
3544                 if (isPrintable(c) || c == '\t'
3545                     || (c == '\n' && (options & AS_STR_NEWLINES)))
3546                         os.put(c);
3547                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
3548                         if (c == META_INSET && (options & AS_STR_PLAINTEXT)) {
3549                                 LASSERT(runparams != 0, return docstring());
3550                                 getInset(i)->plaintext(os, *runparams);
3551                         } else {
3552                                 getInset(i)->toString(os);
3553                         }
3554                 }
3555         }
3556
3557         return os.str();
3558 }
3559
3560
3561 void Paragraph::forOutliner(docstring & os, size_t const maxlen,
3562                             bool const shorten, bool const label) const
3563 {
3564         size_t tmplen = shorten ? maxlen + 1 : maxlen;
3565         if (label && !labelString().empty())
3566                 os += labelString() + ' ';
3567         if (!layout().isTocCaption())
3568                 return;
3569         for (pos_type i = 0; i < size() && os.length() < tmplen; ++i) {
3570                 if (isDeleted(i))
3571                         continue;
3572                 char_type const c = d->text_[i];
3573                 if (isPrintable(c))
3574                         os += c;
3575                 else if (c == META_INSET)
3576                         getInset(i)->forOutliner(os, tmplen, false);
3577         }
3578         if (shorten)
3579                 Text::shortenForOutliner(os, maxlen);
3580 }
3581
3582
3583 void Paragraph::setInsetOwner(Inset const * inset)
3584 {
3585         d->inset_owner_ = inset;
3586 }
3587
3588
3589 int Paragraph::id() const
3590 {
3591         return d->id_;
3592 }
3593
3594
3595 void Paragraph::setId(int id)
3596 {
3597         d->id_ = id;
3598 }
3599
3600
3601 Layout const & Paragraph::layout() const
3602 {
3603         return *d->layout_;
3604 }
3605
3606
3607 void Paragraph::setLayout(Layout const & layout)
3608 {
3609         d->layout_ = &layout;
3610 }
3611
3612
3613 void Paragraph::setDefaultLayout(DocumentClass const & tc)
3614 {
3615         setLayout(tc.defaultLayout());
3616 }
3617
3618
3619 void Paragraph::setPlainLayout(DocumentClass const & tc)
3620 {
3621         setLayout(tc.plainLayout());
3622 }
3623
3624
3625 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
3626 {
3627         if (usePlainLayout())
3628                 setPlainLayout(tclass);
3629         else
3630                 setDefaultLayout(tclass);
3631 }
3632
3633
3634 Inset const & Paragraph::inInset() const
3635 {
3636         LBUFERR(d->inset_owner_);
3637         return *d->inset_owner_;
3638 }
3639
3640
3641 ParagraphParameters & Paragraph::params()
3642 {
3643         return d->params_;
3644 }
3645
3646
3647 ParagraphParameters const & Paragraph::params() const
3648 {
3649         return d->params_;
3650 }
3651
3652
3653 bool Paragraph::isFreeSpacing() const
3654 {
3655         if (d->layout_->free_spacing)
3656                 return true;
3657         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
3658 }
3659
3660
3661 bool Paragraph::allowEmpty() const
3662 {
3663         if (d->layout_->keepempty)
3664                 return true;
3665         return d->inset_owner_ && d->inset_owner_->allowEmpty();
3666 }
3667
3668
3669 bool Paragraph::brokenBiblio() const
3670 {
3671         // there is a problem if there is no bibitem at position 0 or
3672         // if there is another bibitem in the paragraph.
3673         return d->layout_->labeltype == LABEL_BIBLIO
3674                 && (d->insetlist_.find(BIBITEM_CODE) != 0
3675                     || d->insetlist_.find(BIBITEM_CODE, 1) > 0);
3676 }
3677
3678
3679 int Paragraph::fixBiblio(Buffer const & buffer)
3680 {
3681         // FIXME: What about the case where paragraph is not BIBLIO
3682         // but there is an InsetBibitem?
3683         // FIXME: when there was already an inset at 0, the return value is 1,
3684         // which does not tell whether another inset has been remove; the
3685         // cursor cannot be correctly updated.
3686
3687         if (d->layout_->labeltype != LABEL_BIBLIO)
3688                 return 0;
3689
3690         bool const track_changes = buffer.params().track_changes;
3691         int bibitem_pos = d->insetlist_.find(BIBITEM_CODE);
3692         bool const hasbibitem0 = bibitem_pos == 0;
3693
3694         if (hasbibitem0) {
3695                 bibitem_pos = d->insetlist_.find(BIBITEM_CODE, 1);
3696                 // There was an InsetBibitem at pos 0, and no other one => OK
3697                 if (bibitem_pos == -1)
3698                         return 0;
3699                 // there is a bibitem at the 0 position, but since
3700                 // there is a second one, we copy the second on the
3701                 // first. We're assuming there are at most two of
3702                 // these, which there should be.
3703                 // FIXME: why does it make sense to do that rather
3704                 // than keep the first? (JMarc)
3705                 Inset * inset = releaseInset(bibitem_pos);
3706                 d->insetlist_.begin()->inset = inset;
3707                 return -bibitem_pos;
3708         }
3709
3710         // We need to create an inset at the beginning
3711         Inset * inset = 0;
3712         if (bibitem_pos > 0) {
3713                 // there was one somewhere in the paragraph, let's move it
3714                 inset = d->insetlist_.release(bibitem_pos);
3715                 eraseChar(bibitem_pos, track_changes);
3716         } else
3717                 // make a fresh one
3718                 inset = new InsetBibitem(const_cast<Buffer *>(&buffer),
3719                                          InsetCommandParams(BIBITEM_CODE));
3720
3721         Font font(inherit_font, buffer.params().language);
3722         insertInset(0, inset, font, Change(track_changes ? Change::INSERTED
3723                                                    : Change::UNCHANGED));
3724
3725         return 1;
3726 }
3727
3728
3729 void Paragraph::checkAuthors(AuthorList const & authorList)
3730 {
3731         d->changes_.checkAuthors(authorList);
3732 }
3733
3734
3735 bool Paragraph::isChanged(pos_type pos) const
3736 {
3737         return lookupChange(pos).changed();
3738 }
3739
3740
3741 bool Paragraph::isInserted(pos_type pos) const
3742 {
3743         return lookupChange(pos).inserted();
3744 }
3745
3746
3747 bool Paragraph::isDeleted(pos_type pos) const
3748 {
3749         return lookupChange(pos).deleted();
3750 }
3751
3752
3753 InsetList const & Paragraph::insetList() const
3754 {
3755         return d->insetlist_;
3756 }
3757
3758
3759 void Paragraph::setInsetBuffers(Buffer & b)
3760 {
3761         d->insetlist_.setBuffer(b);
3762 }
3763
3764
3765 void Paragraph::resetBuffer()
3766 {
3767         d->insetlist_.resetBuffer();
3768 }
3769
3770
3771 Inset * Paragraph::releaseInset(pos_type pos)
3772 {
3773         Inset * inset = d->insetlist_.release(pos);
3774         /// does not honour change tracking!
3775         eraseChar(pos, false);
3776         return inset;
3777 }
3778
3779
3780 Inset * Paragraph::getInset(pos_type pos)
3781 {
3782         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3783                  ? d->insetlist_.get(pos) : 0;
3784 }
3785
3786
3787 Inset const * Paragraph::getInset(pos_type pos) const
3788 {
3789         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3790                  ? d->insetlist_.get(pos) : 0;
3791 }
3792
3793
3794 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
3795                 pos_type & right, TextCase action)
3796 {
3797         // process sequences of modified characters; in change
3798         // tracking mode, this approach results in much better
3799         // usability than changing case on a char-by-char basis
3800         // We also need to track the current font, since font
3801         // changes within sequences can occur.
3802         vector<pair<char_type, Font> > changes;
3803
3804         bool const trackChanges = bparams.track_changes;
3805
3806         bool capitalize = true;
3807
3808         for (; pos < right; ++pos) {
3809                 char_type oldChar = d->text_[pos];
3810                 char_type newChar = oldChar;
3811
3812                 // ignore insets and don't play with deleted text!
3813                 if (oldChar != META_INSET && !isDeleted(pos)) {
3814                         switch (action) {
3815                                 case text_lowercase:
3816                                         newChar = lowercase(oldChar);
3817                                         break;
3818                                 case text_capitalization:
3819                                         if (capitalize) {
3820                                                 newChar = uppercase(oldChar);
3821                                                 capitalize = false;
3822                                         }
3823                                         break;
3824                                 case text_uppercase:
3825                                         newChar = uppercase(oldChar);
3826                                         break;
3827                         }
3828                 }
3829
3830                 if (isWordSeparator(pos) || isDeleted(pos)) {
3831                         // permit capitalization again
3832                         capitalize = true;
3833                 }
3834
3835                 if (oldChar != newChar) {
3836                         changes.push_back(make_pair(newChar, getFontSettings(bparams, pos)));
3837                         if (pos != right - 1)
3838                                 continue;
3839                         // step behind the changing area
3840                         pos++;
3841                 }
3842
3843                 int erasePos = pos - changes.size();
3844                 for (size_t i = 0; i < changes.size(); i++) {
3845                         insertChar(pos, changes[i].first,
3846                                    changes[i].second,
3847                                    trackChanges);
3848                         if (!eraseChar(erasePos, trackChanges)) {
3849                                 ++erasePos;
3850                                 ++pos; // advance
3851                                 ++right; // expand selection
3852                         }
3853                 }
3854                 changes.clear();
3855         }
3856 }
3857
3858
3859 int Paragraph::find(docstring const & str, bool cs, bool mw,
3860                 pos_type start_pos, bool del) const
3861 {
3862         pos_type pos = start_pos;
3863         int const strsize = str.length();
3864         int i = 0;
3865         pos_type const parsize = d->text_.size();
3866         for (i = 0; i < strsize && pos < parsize; ++i, ++pos) {
3867                 // Ignore "invisible" letters such as ligature breaks
3868                 // and hyphenation chars while searching
3869                 while (pos < parsize - 1 && isInset(pos)) {
3870                         odocstringstream os;
3871                         getInset(pos)->toString(os);
3872                         if (!getInset(pos)->isLetter() || !os.str().empty())
3873                                 break;
3874                         pos++;
3875                 }
3876                 if (cs && str[i] != d->text_[pos])
3877                         break;
3878                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos]))
3879                         break;
3880                 if (!del && isDeleted(pos))
3881                         break;
3882         }
3883
3884         if (i != strsize)
3885                 return 0;
3886
3887         // if necessary, check whether string matches word
3888         if (mw) {
3889                 if (start_pos > 0 && !isWordSeparator(start_pos - 1))
3890                         return 0;
3891                 if (pos < parsize
3892                         && !isWordSeparator(pos))
3893                         return 0;
3894         }
3895
3896         return pos - start_pos;
3897 }
3898
3899
3900 char_type Paragraph::getChar(pos_type pos) const
3901 {
3902         return d->text_[pos];
3903 }
3904
3905
3906 pos_type Paragraph::size() const
3907 {
3908         return d->text_.size();
3909 }
3910
3911
3912 bool Paragraph::empty() const
3913 {
3914         return d->text_.empty();
3915 }
3916
3917
3918 bool Paragraph::isInset(pos_type pos) const
3919 {
3920         return d->text_[pos] == META_INSET;
3921 }
3922
3923
3924 bool Paragraph::isSeparator(pos_type pos) const
3925 {
3926         //FIXME: Are we sure this can be the only separator?
3927         return d->text_[pos] == ' ';
3928 }
3929
3930
3931 void Paragraph::deregisterWords()
3932 {
3933         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3934         Private::LangWordsMap::const_iterator ite = d->words_.end();
3935         for (; itl != ite; ++itl) {
3936                 WordList & wl = theWordList(itl->first);
3937                 Private::Words::const_iterator it = (itl->second).begin();
3938                 Private::Words::const_iterator et = (itl->second).end();
3939                 for (; it != et; ++it)
3940                         wl.remove(*it);
3941         }
3942         d->words_.clear();
3943 }
3944
3945
3946 void Paragraph::locateWord(pos_type & from, pos_type & to,
3947         word_location const loc) const
3948 {
3949         switch (loc) {
3950         case WHOLE_WORD_STRICT:
3951                 if (from == 0 || from == size()
3952                     || isWordSeparator(from)
3953                     || isWordSeparator(from - 1)) {
3954                         to = from;
3955                         return;
3956                 }
3957                 // fall through
3958
3959         case WHOLE_WORD:
3960                 // If we are already at the beginning of a word, do nothing
3961                 if (!from || isWordSeparator(from - 1))
3962                         break;
3963                 // fall through
3964
3965         case PREVIOUS_WORD:
3966                 // always move the cursor to the beginning of previous word
3967                 while (from && !isWordSeparator(from - 1))
3968                         --from;
3969                 break;
3970         case NEXT_WORD:
3971                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
3972                 break;
3973         case PARTIAL_WORD:
3974                 // no need to move the 'from' cursor
3975                 break;
3976         }
3977         to = from;
3978         while (to < size() && !isWordSeparator(to))
3979                 ++to;
3980 }
3981
3982
3983 void Paragraph::collectWords()
3984 {
3985         for (pos_type pos = 0; pos < size(); ++pos) {
3986                 if (isWordSeparator(pos))
3987                         continue;
3988                 pos_type from = pos;
3989                 locateWord(from, pos, WHOLE_WORD);
3990                 // Work around MSVC warning: The statement
3991                 // if (pos < from + lyxrc.completion_minlength)
3992                 // triggers a signed vs. unsigned warning.
3993                 // I don't know why this happens, it could be a MSVC bug, or
3994                 // related to LLP64 (windows) vs. LP64 (unix) programming
3995                 // model, or the C++ standard might be ambigous in the section
3996                 // defining the "usual arithmetic conversions". However, using
3997                 // a temporary variable is safe and works on all compilers.
3998                 pos_type const endpos = from + lyxrc.completion_minlength;
3999                 if (pos < endpos)
4000                         continue;
4001                 FontList::const_iterator cit = d->fontlist_.fontIterator(from);
4002                 if (cit == d->fontlist_.end())
4003                         return;
4004                 Language const * lang = cit->font().language();
4005                 docstring const word = asString(from, pos, AS_STR_NONE);
4006                 d->words_[lang->lang()].insert(word);
4007         }
4008 }
4009
4010
4011 void Paragraph::registerWords()
4012 {
4013         Private::LangWordsMap::const_iterator itl = d->words_.begin();
4014         Private::LangWordsMap::const_iterator ite = d->words_.end();
4015         for (; itl != ite; ++itl) {
4016                 WordList & wl = theWordList(itl->first);
4017                 Private::Words::const_iterator it = (itl->second).begin();
4018                 Private::Words::const_iterator et = (itl->second).end();
4019                 for (; it != et; ++it)
4020                         wl.insert(*it);
4021         }
4022 }
4023
4024
4025 void Paragraph::updateWords()
4026 {
4027         deregisterWords();
4028         collectWords();
4029         registerWords();
4030 }
4031
4032
4033 void Paragraph::Private::appendSkipPosition(SkipPositions & skips, pos_type const pos) const
4034 {
4035         SkipPositionsIterator begin = skips.begin();
4036         SkipPositions::iterator end = skips.end();
4037         if (pos > 0 && begin < end) {
4038                 --end;
4039                 if (end->last == pos - 1) {
4040                         end->last = pos;
4041                         return;
4042                 }
4043         }
4044         skips.insert(end, FontSpan(pos, pos));
4045 }
4046
4047
4048 Language * Paragraph::Private::locateSpellRange(
4049         pos_type & from, pos_type & to,
4050         SkipPositions & skips) const
4051 {
4052         // skip leading white space
4053         while (from < to && owner_->isWordSeparator(from))
4054                 ++from;
4055         // don't check empty range
4056         if (from >= to)
4057                 return 0;
4058         // get current language
4059         Language * lang = getSpellLanguage(from);
4060         pos_type last = from;
4061         bool samelang = true;
4062         bool sameinset = true;
4063         while (last < to && samelang && sameinset) {
4064                 // hop to end of word
4065                 while (last < to && !owner_->isWordSeparator(last)) {
4066                         if (owner_->getInset(last)) {
4067                                 appendSkipPosition(skips, last);
4068                         } else if (owner_->isDeleted(last)) {
4069                                 appendSkipPosition(skips, last);
4070                         }
4071                         ++last;
4072                 }
4073                 // hop to next word while checking for insets
4074                 while (sameinset && last < to && owner_->isWordSeparator(last)) {
4075                         if (Inset const * inset = owner_->getInset(last))
4076                                 sameinset = inset->isChar() && inset->isLetter();
4077                         if (sameinset && owner_->isDeleted(last)) {
4078                                 appendSkipPosition(skips, last);
4079                         }
4080                         if (sameinset)
4081                                 last++;
4082                 }
4083                 if (sameinset && last < to) {
4084                         // now check for language change
4085                         samelang = lang == getSpellLanguage(last);
4086                 }
4087         }
4088         // if language change detected backstep is needed
4089         if (!samelang)
4090                 --last;
4091         to = last;
4092         return lang;
4093 }
4094
4095
4096 Language * Paragraph::Private::getSpellLanguage(pos_type const from) const
4097 {
4098         Language * lang =
4099                 const_cast<Language *>(owner_->getFontSettings(
4100                         inset_owner_->buffer().params(), from).language());
4101         if (lang == inset_owner_->buffer().params().language
4102                 && !lyxrc.spellchecker_alt_lang.empty()) {
4103                 string lang_code;
4104                 string const lang_variety =
4105                         split(lyxrc.spellchecker_alt_lang, lang_code, '-');
4106                 lang->setCode(lang_code);
4107                 lang->setVariety(lang_variety);
4108         }
4109         return lang;
4110 }
4111
4112
4113 void Paragraph::requestSpellCheck(pos_type pos)
4114 {
4115         d->requestSpellCheck(pos);
4116 }
4117
4118
4119 bool Paragraph::needsSpellCheck() const
4120 {
4121         SpellChecker::ChangeNumber speller_change_number = 0;
4122         if (theSpellChecker())
4123                 speller_change_number = theSpellChecker()->changeNumber();
4124         if (speller_change_number > d->speller_state_.currentChangeNumber()) {
4125                 d->speller_state_.needsCompleteRefresh(speller_change_number);
4126         }
4127         return d->needsSpellCheck();
4128 }
4129
4130
4131 bool Paragraph::Private::ignoreWord(docstring const & word) const
4132 {
4133         // Ignore words with digits
4134         // FIXME: make this customizable
4135         // (note that some checkers ignore words with digits by default)
4136         docstring::const_iterator cit = word.begin();
4137         docstring::const_iterator const end = word.end();
4138         for (; cit != end; ++cit) {
4139                 if (isNumber((*cit)))
4140                         return true;
4141         }
4142         return false;
4143 }
4144
4145
4146 SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
4147         WordLangTuple & wl, docstring_list & suggestions,
4148         bool do_suggestion, bool check_learned) const
4149 {
4150         SpellChecker::Result result = SpellChecker::WORD_OK;
4151         SpellChecker * speller = theSpellChecker();
4152         if (!speller)
4153                 return result;
4154
4155         if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
4156                 return result;
4157
4158         locateWord(from, to, WHOLE_WORD);
4159         if (from == to || from >= size())
4160                 return result;
4161
4162         docstring word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
4163         Language * lang = d->getSpellLanguage(from);
4164
4165         if (getFontSettings(d->inset_owner_->buffer().params(), from).fontInfo().nospellcheck() == FONT_ON)
4166                 return result;
4167
4168         wl = WordLangTuple(word, lang);
4169
4170         if (word.empty())
4171                 return result;
4172
4173         if (needsSpellCheck() || check_learned) {
4174                 pos_type end = to;
4175                 if (!d->ignoreWord(word)) {
4176                         bool const trailing_dot = to < size() && d->text_[to] == '.';
4177                         result = speller->check(wl);
4178                         if (SpellChecker::misspelled(result) && trailing_dot) {
4179                                 wl = WordLangTuple(word.append(from_ascii(".")), lang);
4180                                 result = speller->check(wl);
4181                                 if (!SpellChecker::misspelled(result)) {
4182                                         LYXERR(Debug::GUI, "misspelled word is correct with dot: \"" <<
4183                                            word << "\" [" <<
4184                                            from << ".." << to << "]");
4185                                 } else {
4186                                         // spell check with dot appended failed too
4187                                         // restore original word/lang value
4188                                         word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
4189                                         wl = WordLangTuple(word, lang);
4190                                 }
4191                         }
4192                 }
4193                 if (!SpellChecker::misspelled(result)) {
4194                         // area up to the begin of the next word is not misspelled
4195                         while (end < size() && isWordSeparator(end))
4196                                 ++end;
4197                 }
4198                 d->setMisspelled(from, end, result);
4199         } else {
4200                 result = d->speller_state_.getState(from);
4201         }
4202
4203         if (do_suggestion)
4204                 suggestions.clear();
4205
4206         if (SpellChecker::misspelled(result)) {
4207                 LYXERR(Debug::GUI, "misspelled word: \"" <<
4208                            word << "\" [" <<
4209                            from << ".." << to << "]");
4210                 if (do_suggestion)
4211                         speller->suggest(wl, suggestions);
4212         }
4213         return result;
4214 }
4215
4216
4217 void Paragraph::anonymize()
4218 {
4219         // This is a very crude anonymization for now
4220         for (char_type & c : d->text_)
4221                 if (isLetterChar(c) || isNumber(c))
4222                         c = 'a';
4223 }
4224
4225
4226 void Paragraph::Private::markMisspelledWords(
4227         pos_type const & first, pos_type const & last,
4228         SpellChecker::Result result,
4229         docstring const & word,
4230         SkipPositions const & skips)
4231 {
4232         if (!SpellChecker::misspelled(result)) {
4233                 setMisspelled(first, last, SpellChecker::WORD_OK);
4234                 return;
4235         }
4236         int snext = first;
4237         SpellChecker * speller = theSpellChecker();
4238         // locate and enumerate the error positions
4239         int nerrors = speller->numMisspelledWords();
4240         int numskipped = 0;
4241         SkipPositionsIterator it = skips.begin();
4242         SkipPositionsIterator et = skips.end();
4243         for (int index = 0; index < nerrors; ++index) {
4244                 int wstart;
4245                 int wlen = 0;
4246                 speller->misspelledWord(index, wstart, wlen);
4247                 /// should not happen if speller supports range checks
4248                 if (!wlen) continue;
4249                 docstring const misspelled = word.substr(wstart, wlen);
4250                 wstart += first + numskipped;
4251                 if (snext < wstart) {
4252                         /// mark the range of correct spelling
4253                         numskipped += countSkips(it, et, wstart);
4254                         setMisspelled(snext,
4255                                 wstart - 1, SpellChecker::WORD_OK);
4256                 }
4257                 snext = wstart + wlen;
4258                 numskipped += countSkips(it, et, snext);
4259                 /// mark the range of misspelling
4260                 setMisspelled(wstart, snext, result);
4261                 LYXERR(Debug::GUI, "misspelled word: \"" <<
4262                            misspelled << "\" [" <<
4263                            wstart << ".." << (snext-1) << "]");
4264                 ++snext;
4265         }
4266         if (snext <= last) {
4267                 /// mark the range of correct spelling at end
4268                 setMisspelled(snext, last, SpellChecker::WORD_OK);
4269         }
4270 }
4271
4272
4273 void Paragraph::spellCheck() const
4274 {
4275         SpellChecker * speller = theSpellChecker();
4276         if (!speller || empty() ||!needsSpellCheck())
4277                 return;
4278         pos_type start;
4279         pos_type endpos;
4280         d->rangeOfSpellCheck(start, endpos);
4281         if (speller->canCheckParagraph()) {
4282                 // loop until we leave the range
4283                 for (pos_type first = start; first < endpos; ) {
4284                         pos_type last = endpos;
4285                         Private::SkipPositions skips;
4286                         Language * lang = d->locateSpellRange(first, last, skips);
4287                         if (first >= endpos)
4288                                 break;
4289                         // start the spell checker on the unit of meaning
4290                         docstring word = asString(first, last, AS_STR_INSETS + AS_STR_SKIPDELETE);
4291                         WordLangTuple wl = WordLangTuple(word, lang);
4292                         SpellChecker::Result result = word.size() ?
4293                                 speller->check(wl) : SpellChecker::WORD_OK;
4294                         d->markMisspelledWords(first, last, result, word, skips);
4295                         first = ++last;
4296                 }
4297         } else {
4298                 static docstring_list suggestions;
4299                 pos_type to = endpos;
4300                 while (start < endpos) {
4301                         WordLangTuple wl;
4302                         spellCheck(start, to, wl, suggestions, false);
4303                         start = to + 1;
4304                 }
4305         }
4306         d->readySpellCheck();
4307 }
4308
4309
4310 bool Paragraph::isMisspelled(pos_type pos, bool check_boundary) const
4311 {
4312         bool result = SpellChecker::misspelled(d->speller_state_.getState(pos));
4313         if (result || pos <= 0 || pos > size())
4314                 return result;
4315         if (check_boundary && (pos == size() || isWordSeparator(pos)))
4316                 result = SpellChecker::misspelled(d->speller_state_.getState(pos - 1));
4317         return result;
4318 }
4319
4320
4321 string Paragraph::magicLabel() const
4322 {
4323         stringstream ss;
4324         ss << "magicparlabel-" << id();
4325         return ss.str();
4326 }
4327
4328
4329 } // namespace lyx