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, *inset_owner_->getText(int(buf.getParFromID(owner_->id()).idx())),
1355                           buf.getParFromID(owner_->id()).pit(), os,
1356                           features.runparams(), string(), 0, -1, true);
1357                 if (os.length() > length)
1358                         features.addPreambleSnippet(os.release(), true);
1359         }
1360
1361         if (features.runparams().flavor == OutputParams::HTML
1362             && layout_->htmltitle()) {
1363                 features.setHTMLTitle(owner_->asString(AS_STR_INSETS | AS_STR_SKIPDELETE));
1364         }
1365
1366         // check the params.
1367         if (!params_.spacing().isDefault())
1368                 features.require("setspace");
1369
1370         // then the layouts
1371         features.useLayout(layout_->name());
1372
1373         // then the fonts
1374         fontlist_.validate(features);
1375
1376         // then the indentation
1377         if (!params_.leftIndent().zero())
1378                 features.require("ParagraphLeftIndent");
1379
1380         // then the insets
1381         InsetList::const_iterator icit = insetlist_.begin();
1382         InsetList::const_iterator iend = insetlist_.end();
1383         for (; icit != iend; ++icit) {
1384                 if (icit->inset) {
1385                         features.inDeletedInset(owner_->isDeleted(icit->pos));
1386                         if (icit->inset->lyxCode() == FOOT_CODE) {
1387                                 // FIXME: an item inset would make things much easier.
1388                                 if ((layout_->latextype == LATEX_LIST_ENVIRONMENT
1389                                      || (layout_->latextype == LATEX_ITEM_ENVIRONMENT
1390                                          && layout_->margintype == MARGIN_FIRST_DYNAMIC))
1391                                     && (icit->pos < begin_of_body_
1392                                         || (icit->pos == begin_of_body_
1393                                             && (icit->pos == 0 || text_[icit->pos - 1] != ' '))))
1394                                         features.saveNoteEnv("description");
1395                         }
1396                         icit->inset->validate(features);
1397                         features.inDeletedInset(false);
1398                         if (layout_->needprotect &&
1399                             icit->inset->lyxCode() == FOOT_CODE)
1400                                 features.require("NeedLyXFootnoteCode");
1401                 }
1402         }
1403
1404         // then the contents
1405         BufferParams const bp = features.runparams().is_child
1406                 ? features.buffer().masterParams() : features.buffer().params();
1407         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1408                 char_type c = text_[i];
1409                 CharInfo const & ci = Encodings::unicodeCharInfo(c);
1410                 if (c == 0x0022) {
1411                         if (features.runparams().isFullUnicode() && bp.useNonTeXFonts)
1412                                 features.require("textquotedblp");
1413                         else if (bp.main_font_encoding() != "T1"
1414                                  || ((&owner_->getFontSettings(bp, i))->language()->internalFontEncoding()))
1415                                 features.require("textquotedbl");
1416                 } else if (ci.textfeature() && contains(ci.textpreamble(), '=')) {
1417                         // features that depend on the font or input encoding
1418                         string feats = ci.textpreamble();
1419                         string fontenc = (&owner_->getFontSettings(bp, i))->language()->fontenc(bp);
1420                         if (fontenc.empty())
1421                                 fontenc = features.runparams().main_fontenc;
1422                         while (!feats.empty()) {
1423                                 string feat;
1424                                 feats = split(feats, feat, ',');
1425                                 if (contains(feat, "!=")) {
1426                                         // a feature that is required except for the spcified
1427                                         // font or input encodings
1428                                         string realfeature;
1429                                         string const contexts = ltrim(split(feat, realfeature, '!'), "=");
1430                                         // multiple encodings are separated by semicolon
1431                                         vector<string> context = getVectorFromString(contexts, ";");
1432                                         // require feature if the context matches neither current font
1433                                         // nor input encoding
1434                                         if (std::find(context.begin(), context.end(), fontenc) == context.end()
1435                                             && std::find(context.begin(), context.end(),
1436                                                          features.runparams().encoding->name()) == context.end())
1437                                                 features.require(realfeature);
1438                                 } else if (contains(feat, '=')) {
1439                                         // a feature that is required only for the spcified
1440                                         // font or input encodings
1441                                         string realfeature;
1442                                         string const contexts = split(feat, realfeature, '=');
1443                                         // multiple encodings are separated by semicolon
1444                                         vector<string> context = getVectorFromString(contexts, ";");
1445                                         // require feature if the context matches either current font
1446                                         // or input encoding
1447                                         if (std::find(context.begin(), context.end(), fontenc) != context.end()
1448                                             || std::find(context.begin(), context.end(),
1449                                                          features.runparams().encoding->name()) != context.end())
1450                                                 features.require(realfeature);
1451                                 }
1452                         }
1453                 } else if (!bp.use_dash_ligatures
1454                            && (c == 0x2013 || c == 0x2014)
1455                            && bp.useNonTeXFonts
1456                            && features.runparams().flavor == OutputParams::XETEX)
1457                         // XeTeX's dash behaviour is determined via a global setting
1458                         features.require("xetexdashbreakstate");
1459                 BufferEncodings::validate(c, features);
1460         }
1461 }
1462
1463 /////////////////////////////////////////////////////////////////////
1464 //
1465 // Paragraph
1466 //
1467 /////////////////////////////////////////////////////////////////////
1468
1469 namespace {
1470         Layout const emptyParagraphLayout;
1471 }
1472
1473 Paragraph::Paragraph()
1474         : d(new Paragraph::Private(this, emptyParagraphLayout))
1475 {
1476         itemdepth = 0;
1477         d->params_.clear();
1478 }
1479
1480
1481 Paragraph::Paragraph(Paragraph const & par)
1482         : itemdepth(par.itemdepth),
1483         d(new Paragraph::Private(*par.d, this))
1484 {
1485         registerWords();
1486 }
1487
1488
1489 Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
1490         : itemdepth(par.itemdepth),
1491         d(new Paragraph::Private(*par.d, this, beg, end))
1492 {
1493         registerWords();
1494 }
1495
1496
1497 Paragraph & Paragraph::operator=(Paragraph const & par)
1498 {
1499         // needed as we will destroy the private part before copying it
1500         if (&par != this) {
1501                 itemdepth = par.itemdepth;
1502
1503                 deregisterWords();
1504                 delete d;
1505                 d = new Private(*par.d, this);
1506                 registerWords();
1507         }
1508         return *this;
1509 }
1510
1511
1512 Paragraph::~Paragraph()
1513 {
1514         deregisterWords();
1515         delete d;
1516 }
1517
1518
1519 namespace {
1520
1521 // this shall be called just before every "os << ..." action.
1522 void flushString(ostream & os, docstring & s)
1523 {
1524         os << to_utf8(s);
1525         s.erase();
1526 }
1527
1528 } // namespace
1529
1530
1531 void Paragraph::write(ostream & os, BufferParams const & bparams,
1532         depth_type & dth) const
1533 {
1534         // The beginning or end of a deeper (i.e. nested) area?
1535         if (dth != d->params_.depth()) {
1536                 if (d->params_.depth() > dth) {
1537                         while (d->params_.depth() > dth) {
1538                                 os << "\n\\begin_deeper";
1539                                 ++dth;
1540                         }
1541                 } else {
1542                         while (d->params_.depth() < dth) {
1543                                 os << "\n\\end_deeper";
1544                                 --dth;
1545                         }
1546                 }
1547         }
1548
1549         // First write the layout
1550         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1551
1552         d->params_.write(os);
1553
1554         Font font1(inherit_font, bparams.language);
1555
1556         Change running_change = Change(Change::UNCHANGED);
1557
1558         // this string is used as a buffer to avoid repetitive calls
1559         // to to_utf8(), which turn out to be expensive (JMarc)
1560         docstring write_buffer;
1561
1562         int column = 0;
1563         for (pos_type i = 0; i <= size(); ++i) {
1564
1565                 Change const & change = lookupChange(i);
1566                 if (change != running_change)
1567                         flushString(os, write_buffer);
1568                 Changes::lyxMarkChange(os, bparams, column, running_change, change);
1569                 running_change = change;
1570
1571                 if (i == size())
1572                         break;
1573
1574                 // Write font changes
1575                 Font font2 = getFontSettings(bparams, i);
1576                 if (font2 != font1) {
1577                         flushString(os, write_buffer);
1578                         font2.lyxWriteChanges(font1, os);
1579                         column = 0;
1580                         font1 = font2;
1581                 }
1582
1583                 char_type const c = d->text_[i];
1584                 switch (c) {
1585                 case META_INSET:
1586                         if (Inset const * inset = getInset(i)) {
1587                                 flushString(os, write_buffer);
1588                                 if (inset->directWrite()) {
1589                                         // international char, let it write
1590                                         // code directly so it's shorter in
1591                                         // the file
1592                                         inset->write(os);
1593                                 } else {
1594                                         if (i)
1595                                                 os << '\n';
1596                                         os << "\\begin_inset ";
1597                                         inset->write(os);
1598                                         os << "\n\\end_inset\n\n";
1599                                         column = 0;
1600                                 }
1601                                 // FIXME This can be removed again once the mystery
1602                                 // crash has been resolved.
1603                                 os << flush;
1604                         }
1605                         break;
1606                 case '\\':
1607                         flushString(os, write_buffer);
1608                         os << "\n\\backslash\n";
1609                         column = 0;
1610                         break;
1611                 case '.':
1612                         flushString(os, write_buffer);
1613                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1614                                 os << ".\n";
1615                                 column = 0;
1616                         } else
1617                                 os << '.';
1618                         break;
1619                 default:
1620                         if ((column > 70 && c == ' ')
1621                             || column > 79) {
1622                                 flushString(os, write_buffer);
1623                                 os << '\n';
1624                                 column = 0;
1625                         }
1626                         // this check is to amend a bug. LyX sometimes
1627                         // inserts '\0' this could cause problems.
1628                         if (c != '\0')
1629                                 write_buffer.push_back(c);
1630                         else
1631                                 LYXERR0("NUL char in structure.");
1632                         ++column;
1633                         break;
1634                 }
1635         }
1636
1637         flushString(os, write_buffer);
1638         os << "\n\\end_layout\n";
1639         // FIXME This can be removed again once the mystery
1640         // crash has been resolved.
1641         os << flush;
1642 }
1643
1644
1645 void Paragraph::validate(LaTeXFeatures & features) const
1646 {
1647         d->validate(features);
1648         bool fragile = features.runparams().moving_arg;
1649         fragile |= layout().needprotect;
1650         if (needsCProtection(fragile))
1651                 features.require("cprotect");
1652 }
1653
1654
1655 void Paragraph::insert(pos_type start, docstring const & str,
1656                        Font const & font, Change const & change)
1657 {
1658         for (size_t i = 0, n = str.size(); i != n ; ++i)
1659                 insertChar(start + i, str[i], font, change);
1660 }
1661
1662
1663 void Paragraph::appendChar(char_type c, Font const & font,
1664                 Change const & change)
1665 {
1666         // track change
1667         d->changes_.insert(change, d->text_.size());
1668         // when appending characters, no need to update tables
1669         d->text_.push_back(c);
1670         setFont(d->text_.size() - 1, font);
1671         d->requestSpellCheck(d->text_.size() - 1);
1672 }
1673
1674
1675 void Paragraph::appendString(docstring const & s, Font const & font,
1676                 Change const & change)
1677 {
1678         pos_type end = s.size();
1679         size_t oldsize = d->text_.size();
1680         size_t newsize = oldsize + end;
1681         size_t capacity = d->text_.capacity();
1682         if (newsize >= capacity)
1683                 d->text_.reserve(max(capacity + 100, newsize));
1684
1685         // when appending characters, no need to update tables
1686         d->text_.append(s);
1687
1688         // FIXME: Optimize this!
1689         for (size_t i = oldsize; i != newsize; ++i) {
1690                 // track change
1691                 d->changes_.insert(change, i);
1692                 d->requestSpellCheck(i);
1693         }
1694         d->fontlist_.set(oldsize, font);
1695         d->fontlist_.set(newsize - 1, font);
1696 }
1697
1698
1699 void Paragraph::insertChar(pos_type pos, char_type c,
1700                            bool trackChanges)
1701 {
1702         d->insertChar(pos, c, Change(trackChanges ?
1703                            Change::INSERTED : Change::UNCHANGED));
1704 }
1705
1706
1707 void Paragraph::insertChar(pos_type pos, char_type c,
1708                            Font const & font, bool trackChanges)
1709 {
1710         d->insertChar(pos, c, Change(trackChanges ?
1711                            Change::INSERTED : Change::UNCHANGED));
1712         setFont(pos, font);
1713 }
1714
1715
1716 void Paragraph::insertChar(pos_type pos, char_type c,
1717                            Font const & font, Change const & change)
1718 {
1719         d->insertChar(pos, c, change);
1720         setFont(pos, font);
1721 }
1722
1723
1724 void Paragraph::resetFonts(Font const & font)
1725 {
1726         d->fontlist_.clear();
1727         d->fontlist_.set(0, font);
1728         d->fontlist_.set(d->text_.size() - 1, font);
1729 }
1730
1731 // Gets uninstantiated font setting at position.
1732 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1733                                          pos_type pos) const
1734 {
1735         if (pos > size()) {
1736                 LYXERR0("pos: " << pos << " size: " << size());
1737                 LBUFERR(false);
1738         }
1739
1740         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1741         if (cit != d->fontlist_.end())
1742                 return cit->font();
1743
1744         if (pos == size() && !empty())
1745                 return getFontSettings(bparams, pos - 1);
1746
1747         // Optimisation: avoid a full font instantiation if there is no
1748         // language change from previous call.
1749         static Font previous_font;
1750         static Language const * previous_lang = 0;
1751         Language const * lang = getParLanguage(bparams);
1752         if (lang != previous_lang) {
1753                 previous_lang = lang;
1754                 previous_font = Font(inherit_font, lang);
1755         }
1756         return previous_font;
1757 }
1758
1759
1760 FontSpan Paragraph::fontSpan(pos_type pos) const
1761 {
1762         LBUFERR(pos <= size());
1763
1764         if (pos == size())
1765                 return FontSpan(pos, pos);
1766
1767         pos_type start = 0;
1768         FontList::const_iterator cit = d->fontlist_.begin();
1769         FontList::const_iterator end = d->fontlist_.end();
1770         for (; cit != end; ++cit) {
1771                 if (cit->pos() >= pos) {
1772                         if (pos >= beginOfBody())
1773                                 return FontSpan(max(start, beginOfBody()),
1774                                                 cit->pos());
1775                         else
1776                                 return FontSpan(start,
1777                                                 min(beginOfBody() - 1,
1778                                                          cit->pos()));
1779                 }
1780                 start = cit->pos() + 1;
1781         }
1782
1783         // This should not happen, but if so, we take no chances.
1784         LYXERR0("Paragraph::fontSpan: position not found in fontinfo table!");
1785         LASSERT(false, return FontSpan(pos, pos));
1786 }
1787
1788
1789 // Gets uninstantiated font setting at position 0
1790 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1791 {
1792         if (!empty() && !d->fontlist_.empty())
1793                 return d->fontlist_.begin()->font();
1794
1795         // Optimisation: avoid a full font instantiation if there is no
1796         // language change from previous call.
1797         static Font previous_font;
1798         static Language const * previous_lang = 0;
1799         if (bparams.language != previous_lang) {
1800                 previous_lang = bparams.language;
1801                 previous_font = Font(inherit_font, bparams.language);
1802         }
1803
1804         return previous_font;
1805 }
1806
1807
1808 // Gets the fully instantiated font at a given position in a paragraph
1809 // This is basically the same function as Text::GetFont() in text2.cpp.
1810 // The difference is that this one is used for generating the LaTeX file,
1811 // and thus cosmetic "improvements" are disallowed: This has to deliver
1812 // the true picture of the buffer. (Asger)
1813 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1814                                  Font const & outerfont) const
1815 {
1816         LBUFERR(pos >= 0);
1817
1818         Font font = getFontSettings(bparams, pos);
1819
1820         pos_type const body_pos = beginOfBody();
1821         FontInfo & fi = font.fontInfo();
1822         if (pos < body_pos)
1823                 fi.realize(d->layout_->labelfont);
1824         else
1825                 fi.realize(d->layout_->font);
1826
1827         fi.realize(outerfont.fontInfo());
1828         fi.realize(bparams.getFont().fontInfo());
1829
1830         return font;
1831 }
1832
1833
1834 Font const Paragraph::getLabelFont
1835         (BufferParams const & bparams, Font const & outerfont) const
1836 {
1837         FontInfo tmpfont = d->layout_->labelfont;
1838         tmpfont.realize(outerfont.fontInfo());
1839         tmpfont.realize(bparams.getFont().fontInfo());
1840         return Font(tmpfont, getParLanguage(bparams));
1841 }
1842
1843
1844 Font const Paragraph::getLayoutFont
1845         (BufferParams const & bparams, Font const & outerfont) const
1846 {
1847         FontInfo tmpfont = d->layout_->font;
1848         tmpfont.realize(outerfont.fontInfo());
1849         tmpfont.realize(bparams.getFont().fontInfo());
1850         return Font(tmpfont, getParLanguage(bparams));
1851 }
1852
1853
1854 char_type Paragraph::getUChar(BufferParams const & bparams,
1855                               OutputParams const & rp,
1856                               pos_type pos) const
1857 {
1858         char_type c = d->text_[pos];
1859
1860         // Return unchanged character in LTR languages
1861         // or if we use poylglossia/bidi.
1862         if (rp.use_polyglossia || !getFontSettings(bparams, pos).isRightToLeft())
1863                 return c;
1864
1865         // Without polyglossia/bidi, we need to account for some special cases.
1866         // FIXME This needs to be audited!
1867         // Check if:
1868         // * The input is as expected for all delimiters
1869         //   => checked for Hebrew!
1870         // * The output matches the display in the LyX workarea
1871         //   => checked for Hebrew!
1872         // * The special cases below are really necessary
1873         //   => checked for Hebrew!
1874         // * In arabic_arabi, brackets are transformed to Arabic
1875         //   Ornate Parentheses. Is this is really wanted?
1876
1877         string const & lang = getFontSettings(bparams, pos).language()->lang();
1878         char_type uc = c;
1879
1880         // 1. In the following languages, parentheses need to be reversed.
1881         bool const reverseparens = lang == "hebrew";
1882
1883         // 2. In the following languages, brackets don't need to be reversed.
1884         bool const reversebrackets = lang != "arabic_arabtex"
1885                         && lang != "arabic_arabi"
1886                         && lang != "farsi";
1887
1888         // Now swap delimiters if needed.
1889         switch (c) {
1890         case '(':
1891                 if (reverseparens)
1892                         uc = ')';
1893                 break;
1894         case ')':
1895                 if (reverseparens)
1896                         uc = '(';
1897                 break;
1898         case '[':
1899                 if (reversebrackets)
1900                         uc = ']';
1901                 break;
1902         case ']':
1903                 if (reversebrackets)
1904                         uc = '[';
1905                 break;
1906         case '{':
1907                 uc = '}';
1908                 break;
1909         case '}':
1910                 uc = '{';
1911                 break;
1912         case '<':
1913                 uc = '>';
1914                 break;
1915         case '>':
1916                 uc = '<';
1917                 break;
1918         }
1919
1920         return uc;
1921 }
1922
1923
1924 void Paragraph::setFont(pos_type pos, Font const & font)
1925 {
1926         LASSERT(pos <= size(), return);
1927
1928         // First, reduce font against layout/label font
1929         // Update: The setCharFont() routine in text2.cpp already
1930         // reduces font, so we don't need to do that here. (Asger)
1931
1932         d->fontlist_.set(pos, font);
1933 }
1934
1935
1936 void Paragraph::makeSameLayout(Paragraph const & par)
1937 {
1938         d->layout_ = par.d->layout_;
1939         d->params_ = par.d->params_;
1940 }
1941
1942
1943 bool Paragraph::stripLeadingSpaces(bool trackChanges)
1944 {
1945         if (isFreeSpacing())
1946                 return false;
1947
1948         int pos = 0;
1949         int count = 0;
1950
1951         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
1952                 if (eraseChar(pos, trackChanges))
1953                         ++count;
1954                 else
1955                         ++pos;
1956         }
1957
1958         return count > 0 || pos > 0;
1959 }
1960
1961
1962 bool Paragraph::hasSameLayout(Paragraph const & par) const
1963 {
1964         return par.d->layout_ == d->layout_
1965                 && d->params_.sameLayout(par.d->params_);
1966 }
1967
1968
1969 depth_type Paragraph::getDepth() const
1970 {
1971         return d->params_.depth();
1972 }
1973
1974
1975 depth_type Paragraph::getMaxDepthAfter() const
1976 {
1977         if (d->layout_->isEnvironment())
1978                 return d->params_.depth() + 1;
1979         else
1980                 return d->params_.depth();
1981 }
1982
1983
1984 LyXAlignment Paragraph::getAlign() const
1985 {
1986         if (d->params_.align() == LYX_ALIGN_LAYOUT)
1987                 return d->layout_->align;
1988         else
1989                 return d->params_.align();
1990 }
1991
1992
1993 docstring const & Paragraph::labelString() const
1994 {
1995         return d->params_.labelString();
1996 }
1997
1998
1999 // the next two functions are for the manual labels
2000 docstring const Paragraph::getLabelWidthString() const
2001 {
2002         if (d->layout_->margintype == MARGIN_MANUAL
2003             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
2004                 return d->params_.labelWidthString();
2005         else
2006                 return _("Senseless with this layout!");
2007 }
2008
2009
2010 void Paragraph::setLabelWidthString(docstring const & s)
2011 {
2012         d->params_.labelWidthString(s);
2013 }
2014
2015
2016 docstring Paragraph::expandLabel(Layout const & layout,
2017                 BufferParams const & bparams) const
2018 {
2019         return expandParagraphLabel(layout, bparams, true);
2020 }
2021
2022
2023 docstring Paragraph::expandDocBookLabel(Layout const & layout,
2024                 BufferParams const & bparams) const
2025 {
2026         return expandParagraphLabel(layout, bparams, false);
2027 }
2028
2029
2030 docstring Paragraph::expandParagraphLabel(Layout const & layout,
2031                 BufferParams const & bparams, bool process_appendix) const
2032 {
2033         DocumentClass const & tclass = bparams.documentClass();
2034         string const & lang = getParLanguage(bparams)->code();
2035         bool const in_appendix = process_appendix && d->params_.appendix();
2036         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
2037
2038         if (fmt.empty() && !layout.counter.empty())
2039                 return tclass.counters().theCounter(layout.counter, lang);
2040
2041         // handle 'inherited level parts' in 'fmt',
2042         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
2043         size_t const i = fmt.find('@', 0);
2044         if (i != docstring::npos) {
2045                 size_t const j = fmt.find('@', i + 1);
2046                 if (j != docstring::npos) {
2047                         docstring parent(fmt, i + 1, j - i - 1);
2048                         docstring label = from_ascii("??");
2049                         if (tclass.hasLayout(parent))
2050                                 label = expandParagraphLabel(tclass[parent], bparams,
2051                                                       process_appendix);
2052                         fmt = docstring(fmt, 0, i) + label
2053                                 + docstring(fmt, j + 1, docstring::npos);
2054                 }
2055         }
2056
2057         return tclass.counters().counterLabel(fmt, lang);
2058 }
2059
2060
2061 void Paragraph::applyLayout(Layout const & new_layout)
2062 {
2063         d->layout_ = &new_layout;
2064         LyXAlignment const oldAlign = d->params_.align();
2065
2066         if (!(oldAlign & d->layout_->alignpossible)) {
2067                 frontend::Alert::warning(_("Alignment not permitted"),
2068                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
2069                 d->params_.align(LYX_ALIGN_LAYOUT);
2070         }
2071 }
2072
2073
2074 pos_type Paragraph::beginOfBody() const
2075 {
2076         return d->begin_of_body_;
2077 }
2078
2079
2080 void Paragraph::setBeginOfBody()
2081 {
2082         if (d->layout_->labeltype != LABEL_MANUAL) {
2083                 d->begin_of_body_ = 0;
2084                 return;
2085         }
2086
2087         // Unroll the first two cycles of the loop
2088         // and remember the previous character to
2089         // remove unnecessary getChar() calls
2090         pos_type i = 0;
2091         pos_type end = size();
2092         if (i < end && !(isNewline(i) || isEnvSeparator(i))) {
2093                 ++i;
2094                 if (i < end) {
2095                         char_type previous_char = d->text_[i];
2096                         if (!(isNewline(i) || isEnvSeparator(i))) {
2097                                 ++i;
2098                                 while (i < end && previous_char != ' ') {
2099                                         char_type temp = d->text_[i];
2100                                         if (isNewline(i) || isEnvSeparator(i))
2101                                                 break;
2102                                         ++i;
2103                                         previous_char = temp;
2104                                 }
2105                         }
2106                 }
2107         }
2108
2109         d->begin_of_body_ = i;
2110 }
2111
2112
2113 bool Paragraph::allowParagraphCustomization() const
2114 {
2115         return inInset().allowParagraphCustomization();
2116 }
2117
2118
2119 bool Paragraph::usePlainLayout() const
2120 {
2121         return inInset().usePlainLayout();
2122 }
2123
2124
2125 bool Paragraph::isPassThru() const
2126 {
2127         return inInset().isPassThru() || d->layout_->pass_thru;
2128 }
2129
2130 namespace {
2131
2132 // paragraphs inside floats need different alignment tags to avoid
2133 // unwanted space
2134
2135 bool noTrivlistCentering(InsetCode code)
2136 {
2137         return code == FLOAT_CODE
2138                || code == WRAP_CODE
2139                || code == CELL_CODE;
2140 }
2141
2142
2143 string correction(string const & orig)
2144 {
2145         if (orig == "flushleft")
2146                 return "raggedright";
2147         if (orig == "flushright")
2148                 return "raggedleft";
2149         if (orig == "center")
2150                 return "centering";
2151         return orig;
2152 }
2153
2154
2155 bool corrected_env(otexstream & os, string const & suffix, string const & env,
2156         InsetCode code, bool const lastpar, int & col)
2157 {
2158         string macro = suffix + "{";
2159         if (noTrivlistCentering(code)) {
2160                 if (lastpar) {
2161                         // the last paragraph in non-trivlist-aligned
2162                         // context is special (to avoid unwanted whitespace)
2163                         if (suffix == "\\begin") {
2164                                 macro = "\\" + correction(env) + "{}";
2165                                 os << from_ascii(macro);
2166                                 col += macro.size();
2167                                 return true;
2168                         }
2169                         return false;
2170                 }
2171                 macro += correction(env);
2172         } else
2173                 macro += env;
2174         macro += "}";
2175         if (suffix == "\\par\\end") {
2176                 os << breakln;
2177                 col = 0;
2178         }
2179         os << from_ascii(macro);
2180         col += macro.size();
2181         if (suffix == "\\begin") {
2182                 os << breakln;
2183                 col = 0;
2184         }
2185         return true;
2186 }
2187
2188 } // namespace
2189
2190
2191 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
2192                         otexstream & os, OutputParams const & runparams) const
2193 {
2194         int column = 0;
2195
2196         bool canindent =
2197                 (bparams.paragraph_separation == BufferParams::ParagraphIndentSeparation) ?
2198                         (layout_->toggle_indent != ITOGGLE_NEVER) :
2199                         (layout_->toggle_indent == ITOGGLE_ALWAYS);
2200
2201         if (canindent && params_.noindent() && !layout_->pass_thru) {
2202                 os << "\\noindent ";
2203                 column += 10;
2204         }
2205
2206         LyXAlignment const curAlign = params_.align();
2207
2208         if (curAlign == layout_->align)
2209                 return column;
2210
2211         switch (curAlign) {
2212         case LYX_ALIGN_NONE:
2213         case LYX_ALIGN_BLOCK:
2214         case LYX_ALIGN_LAYOUT:
2215         case LYX_ALIGN_SPECIAL:
2216         case LYX_ALIGN_DECIMAL:
2217                 break;
2218         case LYX_ALIGN_LEFT:
2219         case LYX_ALIGN_RIGHT:
2220         case LYX_ALIGN_CENTER:
2221                 if (runparams.moving_arg) {
2222                         os << "\\protect";
2223                         column += 8;
2224                 }
2225                 break;
2226         }
2227
2228         string const begin_tag = "\\begin";
2229         InsetCode code = ownerCode();
2230         bool const lastpar = runparams.isLastPar;
2231         // RTL in classic (PDF)LaTeX (without the Bidi package)
2232         bool const rtl_classic = owner_->getParLanguage(bparams)->rightToLeft()
2233                 && !runparams.use_polyglossia;
2234
2235         switch (curAlign) {
2236         case LYX_ALIGN_NONE:
2237         case LYX_ALIGN_BLOCK:
2238         case LYX_ALIGN_LAYOUT:
2239         case LYX_ALIGN_SPECIAL:
2240         case LYX_ALIGN_DECIMAL:
2241                 break;
2242         case LYX_ALIGN_LEFT: {
2243                 if (rtl_classic)
2244                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2245                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2246                 else
2247                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2248                 break;
2249         } case LYX_ALIGN_RIGHT: {
2250                 if (rtl_classic)
2251                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2252                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2253                 else
2254                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2255                 break;
2256         } case LYX_ALIGN_CENTER: {
2257                 corrected_env(os, begin_tag, "center", code, lastpar, column);
2258                 break;
2259         }
2260         }
2261
2262         return column;
2263 }
2264
2265
2266 bool Paragraph::Private::endTeXParParams(BufferParams const & bparams,
2267                         otexstream & os, OutputParams const & runparams) const
2268 {
2269         LyXAlignment const curAlign = params_.align();
2270
2271         if (curAlign == layout_->align)
2272                 return false;
2273
2274         switch (curAlign) {
2275         case LYX_ALIGN_NONE:
2276         case LYX_ALIGN_BLOCK:
2277         case LYX_ALIGN_LAYOUT:
2278         case LYX_ALIGN_SPECIAL:
2279         case LYX_ALIGN_DECIMAL:
2280                 break;
2281         case LYX_ALIGN_LEFT:
2282         case LYX_ALIGN_RIGHT:
2283         case LYX_ALIGN_CENTER:
2284                 if (runparams.moving_arg)
2285                         os << "\\protect";
2286                 break;
2287         }
2288
2289         bool output = false;
2290         int col = 0;
2291         string const end_tag = "\\par\\end";
2292         InsetCode code = ownerCode();
2293         bool const lastpar = runparams.isLastPar;
2294         // RTL in classic (PDF)LaTeX (without the Bidi package)
2295         bool const rtl_classic = owner_->getParLanguage(bparams)->rightToLeft()
2296                 && !runparams.use_polyglossia;
2297
2298         switch (curAlign) {
2299         case LYX_ALIGN_NONE:
2300         case LYX_ALIGN_BLOCK:
2301         case LYX_ALIGN_LAYOUT:
2302         case LYX_ALIGN_SPECIAL:
2303         case LYX_ALIGN_DECIMAL:
2304                 break;
2305         case LYX_ALIGN_LEFT: {
2306                 if (rtl_classic)
2307                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2308                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2309                 else
2310                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2311                 break;
2312         } case LYX_ALIGN_RIGHT: {
2313                 if (rtl_classic)
2314                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2315                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2316                 else
2317                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2318                 break;
2319         } case LYX_ALIGN_CENTER: {
2320                 corrected_env(os, end_tag, "center", code, lastpar, col);
2321                 break;
2322         }
2323         }
2324
2325         return output || lastpar;
2326 }
2327
2328
2329 // This one spits out the text of the paragraph
2330 void Paragraph::latex(BufferParams const & bparams,
2331         Font const & outerfont,
2332         otexstream & os,
2333         OutputParams const & runparams,
2334         int start_pos, int end_pos, bool force) const
2335 {
2336         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
2337
2338         // FIXME This check should not be needed. Perhaps issue an
2339         // error if it triggers.
2340         Layout const & style = inInset().forcePlainLayout() ?
2341                 bparams.documentClass().plainLayout() : *d->layout_;
2342
2343         if (!force && style.inpreamble)
2344                 return;
2345
2346         bool const allowcust = allowParagraphCustomization();
2347
2348         // Current base font for all inherited font changes, without any
2349         // change caused by an individual character, except for the language:
2350         // It is set to the language of the first character.
2351         // As long as we are in the label, this font is the base font of the
2352         // label. Before the first body character it is set to the base font
2353         // of the body.
2354         Font basefont;
2355
2356         // If there is an open font-encoding changing command (script wrapper),
2357         // alien_script is set to its name
2358         string alien_script;
2359         string script;
2360
2361         // Maybe we have to create a optional argument.
2362         pos_type body_pos = beginOfBody();
2363         unsigned int column = 0;
2364
2365         if (body_pos > 0) {
2366                 // the optional argument is kept in curly brackets in
2367                 // case it contains a ']'
2368                 // This is not strictly needed, but if this is changed it
2369                 // would be a file format change, and tex2lyx would need
2370                 // to be adjusted, since it unconditionally removes the
2371                 // braces when it parses \item.
2372                 os << "[{";
2373                 column += 2;
2374                 basefont = getLabelFont(bparams, outerfont);
2375         } else {
2376                 basefont = getLayoutFont(bparams, outerfont);
2377         }
2378
2379         // Which font is currently active?
2380         Font running_font(basefont);
2381         // Do we have an open font change?
2382         bool open_font = false;
2383
2384         Change runningChange = Change(Change::UNCHANGED);
2385
2386         Encoding const * const prev_encoding = runparams.encoding;
2387
2388         os.texrow().start(id(), 0);
2389
2390         // if the paragraph is empty, the loop will not be entered at all
2391         if (empty()) {
2392                 // For InTitle commands, we have already opened a group
2393                 // in output_latex::TeXOnePar.
2394                 if (style.isCommand() && !style.intitle) {
2395                         os << '{';
2396                         ++column;
2397                 }
2398                 if (!style.leftdelim().empty()) {
2399                         os << style.leftdelim();
2400                         column += style.leftdelim().size();
2401                 }
2402                 if (allowcust)
2403                         column += d->startTeXParParams(bparams, os, runparams);
2404         }
2405
2406         // Whether a \par can be issued for insets typeset inline with text.
2407         // Yes if greater than 0. This has to be static.
2408         THREAD_LOCAL_STATIC int parInline = 0;
2409
2410         for (pos_type i = 0; i < size(); ++i) {
2411                 // First char in paragraph or after label?
2412                 if (i == body_pos) {
2413                         if (body_pos > 0) {
2414                                 if (open_font) {
2415                                         bool needPar = false;
2416                                         column += running_font.latexWriteEndChanges(
2417                                                 os, bparams, runparams,
2418                                                 basefont, basefont, needPar);
2419                                         open_font = false;
2420                                 }
2421                                 basefont = getLayoutFont(bparams, outerfont);
2422                                 running_font = basefont;
2423
2424                                 column += Changes::latexMarkChange(os, bparams,
2425                                                 runningChange, Change(Change::UNCHANGED),
2426                                                 runparams);
2427                                 runningChange = Change(Change::UNCHANGED);
2428
2429                                 os << "}] ";
2430                                 column +=3;
2431                         }
2432                         // For InTitle commands, we have already opened a group
2433                         // in output_latex::TeXOnePar.
2434                         if (style.isCommand() && !style.intitle) {
2435                                 os << '{';
2436                                 ++column;
2437                         }
2438
2439                         if (!style.leftdelim().empty()) {
2440                                 os << style.leftdelim();
2441                                 column += style.leftdelim().size();
2442                         }
2443
2444                         if (allowcust)
2445                                 column += d->startTeXParParams(bparams, os,
2446                                                             runparams);
2447                 }
2448
2449                 runparams.wasDisplayMath = runparams.inDisplayMath;
2450                 runparams.inDisplayMath = false;
2451                 bool deleted_display_math = false;
2452                 Change const & change = runparams.inDeletedInset
2453                         ? runparams.changeOfDeletedInset : lookupChange(i);
2454
2455                 // Check whether a display math inset follows
2456                 if (d->text_[i] == META_INSET
2457                     && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2458                         InsetMath const * im = getInset(i)->asInsetMath();
2459                         if (im && im->asHullInset()
2460                             && im->asHullInset()->outerDisplay()) {
2461                                 runparams.inDisplayMath = true;
2462                                 // runparams.inDeletedInset will be set by
2463                                 // latexInset later, but we need this info
2464                                 // before it is called. On the other hand, we
2465                                 // cannot set it here because it is a counter.
2466                                 deleted_display_math = isDeleted(i);
2467                         }
2468                         if (bparams.output_changes && deleted_display_math
2469                             && runningChange == change
2470                             && change.type == Change::DELETED
2471                             && !os.afterParbreak()) {
2472                                 // A display math in the same paragraph follows.
2473                                 // We have to close and then reopen \lyxdeleted,
2474                                 // otherwise the math will be shifted up.
2475                                 OutputParams rp = runparams;
2476                                 if (open_font) {
2477                                         bool needPar = false;
2478                                         column += running_font.latexWriteEndChanges(
2479                                                 os, bparams, rp, basefont,
2480                                                 basefont, needPar);
2481                                         open_font = false;
2482                                 }
2483                                 basefont = getLayoutFont(bparams, outerfont);
2484                                 running_font = basefont;
2485                                 column += Changes::latexMarkChange(os, bparams,
2486                                         Change(Change::INSERTED), change, rp);
2487                         }
2488                 }
2489
2490                 if (bparams.output_changes && runningChange != change) {
2491                         if (!alien_script.empty()) {
2492                                 column += 1;
2493                                 os << "}";
2494                                 alien_script.clear();
2495                         }
2496                         if (open_font) {
2497                                 bool needPar = false;
2498                                 column += running_font.latexWriteEndChanges(
2499                                                 os, bparams, runparams,
2500                                                 basefont, basefont, needPar);
2501                                 open_font = false;
2502                         }
2503                         basefont = getLayoutFont(bparams, outerfont);
2504                         running_font = basefont;
2505                         column += Changes::latexMarkChange(os, bparams, runningChange,
2506                                                            change, runparams);
2507                         runningChange = change;
2508                 }
2509
2510                 // do not output text which is marked deleted
2511                 // if change tracking output is disabled
2512                 if (!bparams.output_changes && change.deleted()) {
2513                         continue;
2514                 }
2515
2516                 ++column;
2517
2518                 // Fully instantiated font
2519                 Font const current_font = getFont(bparams, i, outerfont);
2520
2521                 Font const last_font = running_font;
2522
2523                 // Do we need to close the previous font?
2524                 if (open_font &&
2525                     (current_font != running_font ||
2526                      current_font.language() != running_font.language()))
2527                 {
2528                         // ensure there is no open script-wrapper
2529                         if (!alien_script.empty()) {
2530                                 column += 1;
2531                                 os << "}";
2532                                 alien_script.clear();
2533                         }
2534                         bool needPar = false;
2535                         column += running_font.latexWriteEndChanges(
2536                                     os, bparams, runparams, basefont,
2537                                     (i == body_pos-1) ? basefont : current_font,
2538                                     needPar);
2539                         running_font = basefont;
2540                         open_font = false;
2541                 }
2542
2543                 string const running_lang = runparams.use_polyglossia ?
2544                         running_font.language()->polyglossia() : running_font.language()->babel();
2545                 // close babel's font environment before opening CJK.
2546                 string const lang_end_command = runparams.use_polyglossia ?
2547                         "\\end{$$lang}" : lyxrc.language_command_end;
2548                 bool const using_begin_end = runparams.use_polyglossia ||
2549                                                 !lang_end_command.empty();
2550                 if (!running_lang.empty() &&
2551                     (!using_begin_end || running_lang == openLanguageName()) &&
2552                     current_font.language()->encoding()->package() == Encoding::CJK) {
2553                                 string end_tag = subst(lang_end_command,
2554                                                         "$$lang",
2555                                                         running_lang);
2556                                 os << from_ascii(end_tag);
2557                                 column += end_tag.length();
2558                                 if (using_begin_end)
2559                                         popLanguageName();
2560                 }
2561
2562                 // Switch file encoding if necessary (and allowed)
2563                 if (!runparams.pass_thru && !style.pass_thru &&
2564                     runparams.encoding->package() != Encoding::none &&
2565                     current_font.language()->encoding()->package() != Encoding::none) {
2566                         pair<bool, int> const enc_switch =
2567                                 switchEncoding(os.os(), bparams, runparams,
2568                                         *(current_font.language()->encoding()));
2569                         if (enc_switch.first) {
2570                                 column += enc_switch.second;
2571                                 runparams.encoding = current_font.language()->encoding();
2572                         }
2573                 }
2574
2575                 char_type const c = d->text_[i];
2576
2577                 // A display math inset inside an ulem command will be output
2578                 // as a box of width \linewidth, so we have to either disable
2579                 // indentation if the inset starts a paragraph, or start a new
2580                 // line to accommodate such box. This has to be done before
2581                 // writing any font changing commands.
2582                 if (runparams.inDisplayMath && !deleted_display_math
2583                     && runparams.inulemcmd) {
2584                         if (os.afterParbreak())
2585                                 os << "\\noindent";
2586                         else
2587                                 os << "\\\\\n";
2588                 }
2589
2590                 // Do we need to change font?
2591                 if ((current_font != running_font ||
2592                      current_font.language() != running_font.language()) &&
2593                         i != body_pos - 1)
2594                 {
2595                         odocstringstream ods;
2596                         column += current_font.latexWriteStartChanges(ods, bparams,
2597                                                               runparams, basefont,
2598                                                               last_font);
2599                         // Check again for display math in ulem commands as a
2600                         // font change may also occur just before a math inset.
2601                         if (runparams.inDisplayMath && !deleted_display_math
2602                             && runparams.inulemcmd) {
2603                                 if (os.afterParbreak())
2604                                         os << "\\noindent";
2605                                 else
2606                                         os << "\\\\\n";
2607                         }
2608                         running_font = current_font;
2609                         open_font = true;
2610                         docstring fontchange = ods.str();
2611                         // check whether the fontchange ends with a \\textcolor
2612                         // modifier and the text starts with a space (bug 4473)
2613                         docstring const last_modifier = rsplit(fontchange, '\\');
2614                         if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2615                                 os << fontchange << from_ascii("{}");
2616                         // check if the fontchange ends with a trailing blank
2617                         // (like "\small " (see bug 3382)
2618                         else if (suffixIs(fontchange, ' ') && c == ' ')
2619                                 os << fontchange.substr(0, fontchange.size() - 1)
2620                                    << from_ascii("{}");
2621                         else
2622                                 os << fontchange;
2623                 }
2624
2625                 // FIXME: think about end_pos implementation...
2626                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2627                         // FIXME: integrate this case in latexSpecialChar
2628                         // Do not print the separation of the optional argument
2629                         // if style.pass_thru is false. This works because
2630                         // latexSpecialChar ignores spaces if
2631                         // style.pass_thru is false.
2632                         if (i != body_pos - 1) {
2633                                 if (d->simpleTeXBlanks(bparams, runparams, os,
2634                                                 i, column, current_font, style)) {
2635                                         // A surrogate pair was output. We
2636                                         // must not call latexSpecialChar
2637                                         // in this iteration, since it would output
2638                                         // the combining character again.
2639                                         ++i;
2640                                         continue;
2641                                 }
2642                         }
2643                 }
2644
2645                 OutputParams rp = runparams;
2646                 rp.free_spacing = style.free_spacing;
2647                 rp.local_font = &current_font;
2648                 rp.intitle = style.intitle;
2649
2650                 // Two major modes:  LaTeX or plain
2651                 // Handle here those cases common to both modes
2652                 // and then split to handle the two modes separately.
2653                 if (c == META_INSET) {
2654                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2655                                 // Greyedout notes and, in general, all insets
2656                                 // with InsetLayout::isDisplay() == false,
2657                                 // are typeset inline with the text. So, we
2658                                 // can add a \par to the last paragraph of
2659                                 // such insets only if nothing else follows.
2660                                 bool incremented = false;
2661                                 Inset const * inset = getInset(i);
2662                                 InsetText const * textinset = inset
2663                                                         ? inset->asInsetText()
2664                                                         : 0;
2665                                 if (i + 1 == size() && textinset
2666                                     && !inset->getLayout().isDisplay()) {
2667                                         ParagraphList const & pars =
2668                                                 textinset->text().paragraphs();
2669                                         pit_type const pit = pars.size() - 1;
2670                                         Font const lastfont =
2671                                                 pit < 0 || pars[pit].empty()
2672                                                 ? pars[pit].getLayoutFont(
2673                                                                 bparams,
2674                                                                 outerfont)
2675                                                 : pars[pit].getFont(bparams,
2676                                                         pars[pit].size() - 1,
2677                                                         outerfont);
2678                                         if (lastfont.fontInfo().size() !=
2679                                             basefont.fontInfo().size()) {
2680                                                 ++parInline;
2681                                                 incremented = true;
2682                                         }
2683                                 }
2684                                 d->latexInset(bparams, os, rp, running_font,
2685                                                 basefont, outerfont, open_font,
2686                                                 runningChange, style, i, column);
2687                                 if (incremented)
2688                                         --parInline;
2689                         }
2690                 } else if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2691                         if (!bparams.useNonTeXFonts)
2692                           script = Encodings::isKnownScriptChar(c);
2693                         if (script != alien_script) {
2694                                 if (!alien_script.empty()) {
2695                                         os << "}";
2696                                         alien_script.clear();
2697                                 }
2698                                 string fontenc = running_font.language()->fontenc(bparams);
2699                                 if (!script.empty()
2700                                         && !Encodings::fontencSupportsScript(fontenc, script)) {
2701                                         column += script.length() + 2;
2702                                         os << "\\" << script << "{";
2703                                         alien_script = script;
2704                                 }
2705                         }
2706                         try {
2707                                 d->latexSpecialChar(os, bparams, rp, running_font,
2708                                                                         alien_script, style, i, end_pos, column);
2709                         } catch (EncodingException & e) {
2710                                 if (runparams.dryrun) {
2711                                         os << "<" << _("LyX Warning: ")
2712                                            << _("uncodable character") << " '";
2713                                         os.put(c);
2714                                         os << "'>";
2715                                 } else {
2716                                         // add location information and throw again.
2717                                         e.par_id = id();
2718                                         e.pos = i;
2719                                         throw(e);
2720                                 }
2721                         }
2722                 }
2723
2724                 // Set the encoding to that returned from latexSpecialChar (see
2725                 // comment for encoding member in OutputParams.h)
2726                 runparams.encoding = rp.encoding;
2727
2728                 // Also carry on the info on a closed ulem command for insets
2729                 // such as Note that do not produce any output, so that no
2730                 // command is ever executed but its opening was recorded.
2731                 runparams.inulemcmd = rp.inulemcmd;
2732
2733                 // And finally, pass the post_macros upstream
2734                 runparams.post_macro = rp.post_macro;
2735         }
2736
2737     // Close wrapper for alien script
2738         if (!alien_script.empty()) {
2739                 os << "}";
2740                 alien_script.clear();
2741         }
2742
2743         // If we have an open font definition, we have to close it
2744         if (open_font) {
2745                 // Make sure that \\par is done with the font of the last
2746                 // character if this has another size as the default.
2747                 // This is necessary because LaTeX (and LyX on the screen)
2748                 // calculates the space between the baselines according
2749                 // to this font. (Matthias)
2750                 //
2751                 // We must not change the font for the last paragraph
2752                 // of non-multipar insets, tabular cells or commands,
2753                 // since this produces unwanted whitespace.
2754
2755                 Font const font = empty()
2756                         ? getLayoutFont(bparams, outerfont)
2757                         : getFont(bparams, size() - 1, outerfont);
2758
2759                 InsetText const * textinset = inInset().asInsetText();
2760
2761                 bool const maintext = textinset
2762                         ? textinset->text().isMainText()
2763                         : false;
2764
2765                 size_t const numpars = textinset
2766                         ? textinset->text().paragraphs().size()
2767                         : 0;
2768
2769                 bool needPar = false;
2770
2771                 if (style.resfont.size() != font.fontInfo().size()
2772                     && (!runparams.isLastPar || maintext
2773                         || (numpars > 1 && d->ownerCode() != CELL_CODE
2774                             && (inInset().getLayout().isDisplay()
2775                                 || parInline)))
2776                     && !style.isCommand()) {
2777                         needPar = true;
2778                 }
2779 #ifdef FIXED_LANGUAGE_END_DETECTION
2780                 if (next_) {
2781                         running_font.latexWriteEndChanges(os, bparams,
2782                                         runparams, basefont,
2783                                         next_->getFont(bparams, 0, outerfont),
2784                                                        needPar);
2785                 } else {
2786                         running_font.latexWriteEndChanges(os, bparams,
2787                                         runparams, basefont, basefont, needPar);
2788                 }
2789 #else
2790 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2791 //FIXME: there as we start another \selectlanguage with the next paragraph if
2792 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2793                 running_font.latexWriteEndChanges(os, bparams, runparams,
2794                                 basefont, basefont, needPar);
2795 #endif
2796                 if (needPar) {
2797                         // The \par could not be inserted at the same nesting
2798                         // level of the font size change, so do it now.
2799                         os << "{\\" << font.latexSize() << "\\par}";
2800                 }
2801         }
2802
2803         column += Changes::latexMarkChange(os, bparams, runningChange,
2804                                            Change(Change::UNCHANGED), runparams);
2805
2806         // Needed if there is an optional argument but no contents.
2807         if (body_pos > 0 && body_pos == size()) {
2808                 os << "}]~";
2809         }
2810
2811         if (!style.rightdelim().empty()) {
2812                 os << style.rightdelim();
2813                 column += style.rightdelim().size();
2814         }
2815
2816         if (allowcust && d->endTeXParParams(bparams, os, runparams)
2817             && runparams.encoding != prev_encoding) {
2818                 runparams.encoding = prev_encoding;
2819                 os << setEncoding(prev_encoding->iconvName());
2820         }
2821
2822         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2823 }
2824
2825
2826 bool Paragraph::emptyTag() const
2827 {
2828         for (pos_type i = 0; i < size(); ++i) {
2829                 if (Inset const * inset = getInset(i)) {
2830                         InsetCode lyx_code = inset->lyxCode();
2831                         // FIXME testing like that is wrong. What is
2832                         // the intent?
2833                         if (lyx_code != TOC_CODE &&
2834                             lyx_code != INCLUDE_CODE &&
2835                             lyx_code != GRAPHICS_CODE &&
2836                             lyx_code != ERT_CODE &&
2837                             lyx_code != LISTINGS_CODE &&
2838                             lyx_code != FLOAT_CODE &&
2839                             lyx_code != TABULAR_CODE) {
2840                                 return false;
2841                         }
2842                 } else {
2843                         char_type c = d->text_[i];
2844                         if (c != ' ' && c != '\t')
2845                                 return false;
2846                 }
2847         }
2848         return true;
2849 }
2850
2851
2852 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
2853         const
2854 {
2855         for (pos_type i = 0; i < size(); ++i) {
2856                 if (Inset const * inset = getInset(i)) {
2857                         InsetCode lyx_code = inset->lyxCode();
2858                         if (lyx_code == LABEL_CODE) {
2859                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2860                                 docstring const & id = il->getParam("name");
2861                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2862                         }
2863                 }
2864         }
2865         return string();
2866 }
2867
2868
2869 pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runparams)
2870         const
2871 {
2872         pos_type i;
2873         for (i = 0; i < size(); ++i) {
2874                 if (Inset const * inset = getInset(i)) {
2875                         inset->docbook(os, runparams);
2876                 } else {
2877                         char_type c = d->text_[i];
2878                         if (c == ' ')
2879                                 break;
2880                         os << sgml::escapeChar(c);
2881                 }
2882         }
2883         return i;
2884 }
2885
2886
2887 pos_type Paragraph::firstWordLyXHTML(XHTMLStream & xs, OutputParams const & runparams)
2888         const
2889 {
2890         pos_type i;
2891         for (i = 0; i < size(); ++i) {
2892                 if (Inset const * inset = getInset(i)) {
2893                         inset->xhtml(xs, runparams);
2894                 } else {
2895                         char_type c = d->text_[i];
2896                         if (c == ' ')
2897                                 break;
2898                         xs << c;
2899                 }
2900         }
2901         return i;
2902 }
2903
2904
2905 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2906 {
2907         Font font_old;
2908         pos_type size = text_.size();
2909         for (pos_type i = initial; i < size; ++i) {
2910                 Font font = owner_->getFont(buf.params(), i, outerfont);
2911                 if (text_[i] == META_INSET)
2912                         return false;
2913                 if (i != initial && font != font_old)
2914                         return false;
2915                 font_old = font;
2916         }
2917
2918         return true;
2919 }
2920
2921
2922 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
2923                                     odocstream & os,
2924                                     OutputParams const & runparams,
2925                                     Font const & outerfont,
2926                                     pos_type initial) const
2927 {
2928         bool emph_flag = false;
2929
2930         Layout const & style = *d->layout_;
2931         FontInfo font_old =
2932                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2933
2934         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2935                 os << "]]>";
2936
2937         // parsing main loop
2938         for (pos_type i = initial; i < size(); ++i) {
2939                 Font font = getFont(buf.params(), i, outerfont);
2940
2941                 // handle <emphasis> tag
2942                 if (font_old.emph() != font.fontInfo().emph()) {
2943                         if (font.fontInfo().emph() == FONT_ON) {
2944                                 os << "<emphasis>";
2945                                 emph_flag = true;
2946                         } else if (i != initial) {
2947                                 os << "</emphasis>";
2948                                 emph_flag = false;
2949                         }
2950                 }
2951
2952                 if (Inset const * inset = getInset(i)) {
2953                         inset->docbook(os, runparams);
2954                 } else {
2955                         char_type c = d->text_[i];
2956
2957                         if (style.pass_thru)
2958                                 os.put(c);
2959                         else
2960                                 os << sgml::escapeChar(c);
2961                 }
2962                 font_old = font.fontInfo();
2963         }
2964
2965         if (emph_flag) {
2966                 os << "</emphasis>";
2967         }
2968
2969         if (style.free_spacing)
2970                 os << '\n';
2971         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2972                 os << "<![CDATA[";
2973 }
2974
2975
2976 namespace {
2977 void doFontSwitch(vector<html::FontTag> & tagsToOpen,
2978                   vector<html::EndFontTag> & tagsToClose,
2979                   bool & flag, FontState curstate, html::FontTypes type)
2980 {
2981         if (curstate == FONT_ON) {
2982                 tagsToOpen.push_back(html::FontTag(type));
2983                 flag = true;
2984         } else if (flag) {
2985                 tagsToClose.push_back(html::EndFontTag(type));
2986                 flag = false;
2987         }
2988 }
2989 } // namespace
2990
2991
2992 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
2993                                     XHTMLStream & xs,
2994                                     OutputParams const & runparams,
2995                                     Font const & outerfont,
2996                                     bool start_paragraph, bool close_paragraph,
2997                                     pos_type initial) const
2998 {
2999         docstring retval;
3000
3001         // track whether we have opened these tags
3002         bool emph_flag = false;
3003         bool bold_flag = false;
3004         bool noun_flag = false;
3005         bool ubar_flag = false;
3006         bool dbar_flag = false;
3007         bool sout_flag = false;
3008         bool xout_flag = false;
3009         bool wave_flag = false;
3010         // shape tags
3011         bool shap_flag = false;
3012         // family tags
3013         bool faml_flag = false;
3014         // size tags
3015         bool size_flag = false;
3016
3017         Layout const & style = *d->layout_;
3018
3019         if (start_paragraph)
3020                 xs.startDivision(allowEmpty());
3021
3022         FontInfo font_old =
3023                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
3024
3025         FontShape  curr_fs   = INHERIT_SHAPE;
3026         FontFamily curr_fam  = INHERIT_FAMILY;
3027         FontSize   curr_size = FONT_SIZE_INHERIT;
3028
3029         string const default_family =
3030                 buf.masterBuffer()->params().fonts_default_family;
3031
3032         vector<html::FontTag> tagsToOpen;
3033         vector<html::EndFontTag> tagsToClose;
3034
3035         // parsing main loop
3036         for (pos_type i = initial; i < size(); ++i) {
3037                 // let's not show deleted material in the output
3038                 if (isDeleted(i))
3039                         continue;
3040
3041                 Font const font = getFont(buf.masterBuffer()->params(), i, outerfont);
3042
3043                 // emphasis
3044                 FontState curstate = font.fontInfo().emph();
3045                 if (font_old.emph() != curstate)
3046                         doFontSwitch(tagsToOpen, tagsToClose, emph_flag, curstate, html::FT_EMPH);
3047
3048                 // noun
3049                 curstate = font.fontInfo().noun();
3050                 if (font_old.noun() != curstate)
3051                         doFontSwitch(tagsToOpen, tagsToClose, noun_flag, curstate, html::FT_NOUN);
3052
3053                 // underbar
3054                 curstate = font.fontInfo().underbar();
3055                 if (font_old.underbar() != curstate)
3056                         doFontSwitch(tagsToOpen, tagsToClose, ubar_flag, curstate, html::FT_UBAR);
3057
3058                 // strikeout
3059                 curstate = font.fontInfo().strikeout();
3060                 if (font_old.strikeout() != curstate)
3061                         doFontSwitch(tagsToOpen, tagsToClose, sout_flag, curstate, html::FT_SOUT);
3062
3063                 // xout
3064                 curstate = font.fontInfo().xout();
3065                 if (font_old.xout() != curstate)
3066                         doFontSwitch(tagsToOpen, tagsToClose, xout_flag, curstate, html::FT_XOUT);
3067
3068                 // double underbar
3069                 curstate = font.fontInfo().uuline();
3070                 if (font_old.uuline() != curstate)
3071                         doFontSwitch(tagsToOpen, tagsToClose, dbar_flag, curstate, html::FT_DBAR);
3072
3073                 // wavy line
3074                 curstate = font.fontInfo().uwave();
3075                 if (font_old.uwave() != curstate)
3076                         doFontSwitch(tagsToOpen, tagsToClose, wave_flag, curstate, html::FT_WAVE);
3077
3078                 // bold
3079                 // a little hackish, but allows us to reuse what we have.
3080                 curstate = (font.fontInfo().series() == BOLD_SERIES ? FONT_ON : FONT_OFF);
3081                 if (font_old.series() != font.fontInfo().series())
3082                         doFontSwitch(tagsToOpen, tagsToClose, bold_flag, curstate, html::FT_BOLD);
3083
3084                 // Font shape
3085                 curr_fs = font.fontInfo().shape();
3086                 FontShape old_fs = font_old.shape();
3087                 if (old_fs != curr_fs) {
3088                         if (shap_flag) {
3089                                 switch (old_fs) {
3090                                 case ITALIC_SHAPE:
3091                                         tagsToClose.push_back(html::EndFontTag(html::FT_ITALIC));
3092                                         break;
3093                                 case SLANTED_SHAPE:
3094                                         tagsToClose.push_back(html::EndFontTag(html::FT_SLANTED));
3095                                         break;
3096                                 case SMALLCAPS_SHAPE:
3097                                         tagsToClose.push_back(html::EndFontTag(html::FT_SMALLCAPS));
3098                                         break;
3099                                 case UP_SHAPE:
3100                                 case INHERIT_SHAPE:
3101                                         break;
3102                                 default:
3103                                         // the other tags are for internal use
3104                                         LATTEST(false);
3105                                         break;
3106                                 }
3107                                 shap_flag = false;
3108                         }
3109                         switch (curr_fs) {
3110                         case ITALIC_SHAPE:
3111                                 tagsToOpen.push_back(html::FontTag(html::FT_ITALIC));
3112                                 shap_flag = true;
3113                                 break;
3114                         case SLANTED_SHAPE:
3115                                 tagsToOpen.push_back(html::FontTag(html::FT_SLANTED));
3116                                 shap_flag = true;
3117                                 break;
3118                         case SMALLCAPS_SHAPE:
3119                                 tagsToOpen.push_back(html::FontTag(html::FT_SMALLCAPS));
3120                                 shap_flag = true;
3121                                 break;
3122                         case UP_SHAPE:
3123                         case INHERIT_SHAPE:
3124                                 break;
3125                         default:
3126                                 // the other tags are for internal use
3127                                 LATTEST(false);
3128                                 break;
3129                         }
3130                 }
3131
3132                 // Font family
3133                 curr_fam = font.fontInfo().family();
3134                 FontFamily old_fam = font_old.family();
3135                 if (old_fam != curr_fam) {
3136                         if (faml_flag) {
3137                                 switch (old_fam) {
3138                                 case ROMAN_FAMILY:
3139                                         tagsToClose.push_back(html::EndFontTag(html::FT_ROMAN));
3140                                         break;
3141                                 case SANS_FAMILY:
3142                                         tagsToClose.push_back(html::EndFontTag(html::FT_SANS));
3143                                         break;
3144                                 case TYPEWRITER_FAMILY:
3145                                         tagsToClose.push_back(html::EndFontTag(html::FT_TYPE));
3146                                         break;
3147                                 case INHERIT_FAMILY:
3148                                         break;
3149                                 default:
3150                                         // the other tags are for internal use
3151                                         LATTEST(false);
3152                                         break;
3153                                 }
3154                                 faml_flag = false;
3155                         }
3156                         switch (curr_fam) {
3157                         case ROMAN_FAMILY:
3158                                 // we will treat a "default" font family as roman, since we have
3159                                 // no other idea what to do.
3160                                 if (default_family != "rmdefault" && default_family != "default") {
3161                                         tagsToOpen.push_back(html::FontTag(html::FT_ROMAN));
3162                                         faml_flag = true;
3163                                 }
3164                                 break;
3165                         case SANS_FAMILY:
3166                                 if (default_family != "sfdefault") {
3167                                         tagsToOpen.push_back(html::FontTag(html::FT_SANS));
3168                                         faml_flag = true;
3169                                 }
3170                                 break;
3171                         case TYPEWRITER_FAMILY:
3172                                 if (default_family != "ttdefault") {
3173                                         tagsToOpen.push_back(html::FontTag(html::FT_TYPE));
3174                                         faml_flag = true;
3175                                 }
3176                                 break;
3177                         case INHERIT_FAMILY:
3178                                 break;
3179                         default:
3180                                 // the other tags are for internal use
3181                                 LATTEST(false);
3182                                 break;
3183                         }
3184                 }
3185
3186                 // Font size
3187                 curr_size = font.fontInfo().size();
3188                 FontSize old_size = font_old.size();
3189                 if (old_size != curr_size) {
3190                         if (size_flag) {
3191                                 switch (old_size) {
3192                                 case FONT_SIZE_TINY:
3193                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_TINY));
3194                                         break;
3195                                 case FONT_SIZE_SCRIPT:
3196                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SCRIPT));
3197                                         break;
3198                                 case FONT_SIZE_FOOTNOTE:
3199                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_FOOTNOTE));
3200                                         break;
3201                                 case FONT_SIZE_SMALL:
3202                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SMALL));
3203                                         break;
3204                                 case FONT_SIZE_LARGE:
3205                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGE));
3206                                         break;
3207                                 case FONT_SIZE_LARGER:
3208                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGER));
3209                                         break;
3210                                 case FONT_SIZE_LARGEST:
3211                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGEST));
3212                                         break;
3213                                 case FONT_SIZE_HUGE:
3214                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGE));
3215                                         break;
3216                                 case FONT_SIZE_HUGER:
3217                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGER));
3218                                         break;
3219                                 case FONT_SIZE_INCREASE:
3220                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_INCREASE));
3221                                         break;
3222                                 case FONT_SIZE_DECREASE:
3223                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_DECREASE));
3224                                         break;
3225                                 case FONT_SIZE_INHERIT:
3226                                 case FONT_SIZE_NORMAL:
3227                                         break;
3228                                 default:
3229                                         // the other tags are for internal use
3230                                         LATTEST(false);
3231                                         break;
3232                                 }
3233                                 size_flag = false;
3234                         }
3235                         switch (curr_size) {
3236                         case FONT_SIZE_TINY:
3237                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_TINY));
3238                                 size_flag = true;
3239                                 break;
3240                         case FONT_SIZE_SCRIPT:
3241                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SCRIPT));
3242                                 size_flag = true;
3243                                 break;
3244                         case FONT_SIZE_FOOTNOTE:
3245                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_FOOTNOTE));
3246                                 size_flag = true;
3247                                 break;
3248                         case FONT_SIZE_SMALL:
3249                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SMALL));
3250                                 size_flag = true;
3251                                 break;
3252                         case FONT_SIZE_LARGE:
3253                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGE));
3254                                 size_flag = true;
3255                                 break;
3256                         case FONT_SIZE_LARGER:
3257                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGER));
3258                                 size_flag = true;
3259                                 break;
3260                         case FONT_SIZE_LARGEST:
3261                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGEST));
3262                                 size_flag = true;
3263                                 break;
3264                         case FONT_SIZE_HUGE:
3265                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGE));
3266                                 size_flag = true;
3267                                 break;
3268                         case FONT_SIZE_HUGER:
3269                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGER));
3270                                 size_flag = true;
3271                                 break;
3272                         case FONT_SIZE_INCREASE:
3273                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_INCREASE));
3274                                 size_flag = true;
3275                                 break;
3276                         case FONT_SIZE_DECREASE:
3277                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_DECREASE));
3278                                 size_flag = true;
3279                                 break;
3280                         case FONT_SIZE_NORMAL:
3281                         case FONT_SIZE_INHERIT:
3282                                 break;
3283                         default:
3284                                 // the other tags are for internal use
3285                                 LATTEST(false);
3286                                 break;
3287                         }
3288                 }
3289
3290                 // FIXME XHTML
3291                 // Other such tags? What about the other text ranges?
3292
3293                 vector<html::EndFontTag>::const_iterator cit = tagsToClose.begin();
3294                 vector<html::EndFontTag>::const_iterator cen = tagsToClose.end();
3295                 for (; cit != cen; ++cit)
3296                         xs << *cit;
3297
3298                 vector<html::FontTag>::const_iterator sit = tagsToOpen.begin();
3299                 vector<html::FontTag>::const_iterator sen = tagsToOpen.end();
3300                 for (; sit != sen; ++sit)
3301                         xs << *sit;
3302
3303                 tagsToClose.clear();
3304                 tagsToOpen.clear();
3305
3306                 Inset const * inset = getInset(i);
3307                 if (inset) {
3308                         if (!runparams.for_toc || inset->isInToc()) {
3309                                 OutputParams np = runparams;
3310                                 np.local_font = &font;
3311                                 // If the paragraph has size 1, then we are in the "special
3312                                 // case" where we do not output the containing paragraph info
3313                                 if (!inset->getLayout().htmlisblock() && size() != 1)
3314                                         np.html_in_par = true;
3315                                 retval += inset->xhtml(xs, np);
3316                         }
3317                 } else {
3318                         char_type c = getUChar(buf.masterBuffer()->params(),
3319                                                runparams, i);
3320                         if (c == ' ' && (style.free_spacing || runparams.free_spacing))
3321                                 xs << XHTMLStream::ESCAPE_NONE << "&nbsp;";
3322                         else
3323                                 xs << c;
3324                 }
3325                 font_old = font.fontInfo();
3326         }
3327
3328         // FIXME XHTML
3329         // I'm worried about what happens if a branch, say, is itself
3330         // wrapped in some font stuff. I think that will not work.
3331         xs.closeFontTags();
3332         if (close_paragraph)
3333                 xs.endDivision();
3334
3335         return retval;
3336 }
3337
3338
3339 bool Paragraph::isHfill(pos_type pos) const
3340 {
3341         Inset const * inset = getInset(pos);
3342         return inset && inset->isHfill();
3343 }
3344
3345
3346 bool Paragraph::isNewline(pos_type pos) const
3347 {
3348         // U+2028 LINE SEPARATOR
3349         // U+2029 PARAGRAPH SEPARATOR
3350         char_type const c = d->text_[pos];
3351         if (c == 0x2028 || c == 0x2029)
3352                 return true;
3353         Inset const * inset = getInset(pos);
3354         return inset && inset->lyxCode() == NEWLINE_CODE;
3355 }
3356
3357
3358 bool Paragraph::isEnvSeparator(pos_type pos) const
3359 {
3360         Inset const * inset = getInset(pos);
3361         return inset && inset->lyxCode() == SEPARATOR_CODE;
3362 }
3363
3364
3365 bool Paragraph::isLineSeparator(pos_type pos) const
3366 {
3367         char_type const c = d->text_[pos];
3368         if (isLineSeparatorChar(c))
3369                 return true;
3370         Inset const * inset = getInset(pos);
3371         return inset && inset->isLineSeparator();
3372 }
3373
3374
3375 bool Paragraph::isWordSeparator(pos_type pos) const
3376 {
3377         if (pos == size())
3378                 return true;
3379         if (Inset const * inset = getInset(pos))
3380                 return !inset->isLetter();
3381         // if we have a hard hyphen (no en- or emdash) or apostrophe
3382         // we pass this to the spell checker
3383         // FIXME: this method is subject to change, visit
3384         // https://bugzilla.mozilla.org/show_bug.cgi?id=355178
3385         // to get an impression how complex this is.
3386         if (isHardHyphenOrApostrophe(pos))
3387                 return false;
3388         char_type const c = d->text_[pos];
3389         // We want to pass the escape chars to the spellchecker
3390         docstring const escape_chars = from_utf8(lyxrc.spellchecker_esc_chars);
3391         return !isLetterChar(c) && !isDigitASCII(c) && !contains(escape_chars, c);
3392 }
3393
3394
3395 bool Paragraph::isHardHyphenOrApostrophe(pos_type pos) const
3396 {
3397         pos_type const psize = size();
3398         if (pos >= psize)
3399                 return false;
3400         char_type const c = d->text_[pos];
3401         if (c != '-' && c != '\'')
3402                 return false;
3403         int nextpos = pos + 1;
3404         int prevpos = pos > 0 ? pos - 1 : 0;
3405         if ((nextpos == psize || isSpace(nextpos))
3406                 && (pos == 0 || isSpace(prevpos)))
3407                 return false;
3408         return true;
3409 }
3410
3411
3412 bool Paragraph::needsCProtection(bool const fragile) const
3413 {
3414         // first check the layout of the paragraph, but only in insets
3415         InsetText const * textinset = inInset().asInsetText();
3416         bool const maintext = textinset
3417                 ? textinset->text().isMainText()
3418                 : false;
3419
3420         if (!maintext && layout().needcprotect) {
3421                 // Environments need cprotection regardless the content
3422                 if (layout().latextype == LATEX_ENVIRONMENT)
3423                         return true;
3424
3425                 // Commands need cprotection if they contain specific chars
3426                 int const nchars_escape = 9;
3427                 static char_type const chars_escape[nchars_escape] = {
3428                         '&', '_', '$', '%', '#', '^', '{', '}', '\\'};
3429
3430                 docstring const pars = asString();
3431                 for (int k = 0; k < nchars_escape; k++) {
3432                         if (contains(pars, chars_escape[k]))
3433                                 return true;
3434                 }
3435         }
3436
3437         // now check whether we have insets that need cprotection
3438         pos_type size = d->text_.size();
3439         for (pos_type i = 0; i < size; ++i)
3440                 if (isInset(i) && getInset(i)->needsCProtection(maintext, fragile))
3441                         return true;
3442
3443         return false;
3444 }
3445
3446
3447 FontSpan const & Paragraph::getSpellRange(pos_type pos) const
3448 {
3449         return d->speller_state_.getRange(pos);
3450 }
3451
3452
3453 bool Paragraph::isChar(pos_type pos) const
3454 {
3455         if (Inset const * inset = getInset(pos))
3456                 return inset->isChar();
3457         char_type const c = d->text_[pos];
3458         return !isLetterChar(c) && !isDigitASCII(c) && !lyx::isSpace(c);
3459 }
3460
3461
3462 bool Paragraph::isSpace(pos_type pos) const
3463 {
3464         if (Inset const * inset = getInset(pos))
3465                 return inset->isSpace();
3466         char_type const c = d->text_[pos];
3467         return lyx::isSpace(c);
3468 }
3469
3470
3471 Language const *
3472 Paragraph::getParLanguage(BufferParams const & bparams) const
3473 {
3474         if (!empty())
3475                 return getFirstFontSettings(bparams).language();
3476         // FIXME: we should check the prev par as well (Lgb)
3477         return bparams.language;
3478 }
3479
3480
3481 bool Paragraph::isRTL(BufferParams const & bparams) const
3482 {
3483         return getParLanguage(bparams)->rightToLeft()
3484                 && !inInset().getLayout().forceLTR();
3485 }
3486
3487
3488 void Paragraph::changeLanguage(BufferParams const & bparams,
3489                                Language const * from, Language const * to)
3490 {
3491         // change language including dummy font change at the end
3492         for (pos_type i = 0; i <= size(); ++i) {
3493                 Font font = getFontSettings(bparams, i);
3494                 if (font.language() == from) {
3495                         font.setLanguage(to);
3496                         setFont(i, font);
3497                         d->requestSpellCheck(i);
3498                 }
3499         }
3500 }
3501
3502
3503 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
3504 {
3505         Language const * doc_language = bparams.language;
3506         for (auto const & f : d->fontlist_)
3507                 if (f.font().language() != ignore_language &&
3508                     f.font().language() != latex_language &&
3509                     f.font().language() != doc_language)
3510                         return true;
3511         return false;
3512 }
3513
3514
3515 void Paragraph::getLanguages(std::set<Language const *> & langs) const
3516 {
3517         for (auto const & f : d->fontlist_) {
3518                 Language const * lang = f.font().language();
3519                 if (lang != ignore_language &&
3520                     lang != latex_language)
3521                         langs.insert(lang);
3522         }
3523 }
3524
3525
3526 docstring Paragraph::asString(int options) const
3527 {
3528         return asString(0, size(), options);
3529 }
3530
3531
3532 docstring Paragraph::asString(pos_type beg, pos_type end, int options, const OutputParams *runparams) const
3533 {
3534         odocstringstream os;
3535
3536         if (beg == 0
3537             && options & AS_STR_LABEL
3538             && !d->params_.labelString().empty())
3539                 os << d->params_.labelString() << ' ';
3540
3541         for (pos_type i = beg; i < end; ++i) {
3542                 if ((options & AS_STR_SKIPDELETE) && isDeleted(i))
3543                         continue;
3544                 char_type const c = d->text_[i];
3545                 if (isPrintable(c) || c == '\t'
3546                     || (c == '\n' && (options & AS_STR_NEWLINES)))
3547                         os.put(c);
3548                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
3549                         if (c == META_INSET && (options & AS_STR_PLAINTEXT)) {
3550                                 LASSERT(runparams != 0, return docstring());
3551                                 getInset(i)->plaintext(os, *runparams);
3552                         } else {
3553                                 getInset(i)->toString(os);
3554                         }
3555                 }
3556         }
3557
3558         return os.str();
3559 }
3560
3561
3562 void Paragraph::forOutliner(docstring & os, size_t const maxlen,
3563                             bool const shorten, bool const label) const
3564 {
3565         size_t tmplen = shorten ? maxlen + 1 : maxlen;
3566         if (label && !labelString().empty())
3567                 os += labelString() + ' ';
3568         if (!layout().isTocCaption())
3569                 return;
3570         for (pos_type i = 0; i < size() && os.length() < tmplen; ++i) {
3571                 if (isDeleted(i))
3572                         continue;
3573                 char_type const c = d->text_[i];
3574                 if (isPrintable(c))
3575                         os += c;
3576                 else if (c == META_INSET)
3577                         getInset(i)->forOutliner(os, tmplen, false);
3578         }
3579         if (shorten)
3580                 Text::shortenForOutliner(os, maxlen);
3581 }
3582
3583
3584 void Paragraph::setInsetOwner(Inset const * inset)
3585 {
3586         d->inset_owner_ = inset;
3587 }
3588
3589
3590 int Paragraph::id() const
3591 {
3592         return d->id_;
3593 }
3594
3595
3596 void Paragraph::setId(int id)
3597 {
3598         d->id_ = id;
3599 }
3600
3601
3602 Layout const & Paragraph::layout() const
3603 {
3604         return *d->layout_;
3605 }
3606
3607
3608 void Paragraph::setLayout(Layout const & layout)
3609 {
3610         d->layout_ = &layout;
3611 }
3612
3613
3614 void Paragraph::setDefaultLayout(DocumentClass const & tc)
3615 {
3616         setLayout(tc.defaultLayout());
3617 }
3618
3619
3620 void Paragraph::setPlainLayout(DocumentClass const & tc)
3621 {
3622         setLayout(tc.plainLayout());
3623 }
3624
3625
3626 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
3627 {
3628         if (usePlainLayout())
3629                 setPlainLayout(tclass);
3630         else
3631                 setDefaultLayout(tclass);
3632 }
3633
3634
3635 Inset const & Paragraph::inInset() const
3636 {
3637         LBUFERR(d->inset_owner_);
3638         return *d->inset_owner_;
3639 }
3640
3641
3642 ParagraphParameters & Paragraph::params()
3643 {
3644         return d->params_;
3645 }
3646
3647
3648 ParagraphParameters const & Paragraph::params() const
3649 {
3650         return d->params_;
3651 }
3652
3653
3654 bool Paragraph::isFreeSpacing() const
3655 {
3656         if (d->layout_->free_spacing)
3657                 return true;
3658         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
3659 }
3660
3661
3662 bool Paragraph::allowEmpty() const
3663 {
3664         if (d->layout_->keepempty)
3665                 return true;
3666         return d->inset_owner_ && d->inset_owner_->allowEmpty();
3667 }
3668
3669
3670 bool Paragraph::brokenBiblio() const
3671 {
3672         // there is a problem if there is no bibitem at position 0 or
3673         // if there is another bibitem in the paragraph.
3674         return d->layout_->labeltype == LABEL_BIBLIO
3675                 && (d->insetlist_.find(BIBITEM_CODE) != 0
3676                     || d->insetlist_.find(BIBITEM_CODE, 1) > 0);
3677 }
3678
3679
3680 int Paragraph::fixBiblio(Buffer const & buffer)
3681 {
3682         // FIXME: What about the case where paragraph is not BIBLIO
3683         // but there is an InsetBibitem?
3684         // FIXME: when there was already an inset at 0, the return value is 1,
3685         // which does not tell whether another inset has been remove; the
3686         // cursor cannot be correctly updated.
3687
3688         if (d->layout_->labeltype != LABEL_BIBLIO)
3689                 return 0;
3690
3691         bool const track_changes = buffer.params().track_changes;
3692         int bibitem_pos = d->insetlist_.find(BIBITEM_CODE);
3693         bool const hasbibitem0 = bibitem_pos == 0;
3694
3695         if (hasbibitem0) {
3696                 bibitem_pos = d->insetlist_.find(BIBITEM_CODE, 1);
3697                 // There was an InsetBibitem at pos 0, and no other one => OK
3698                 if (bibitem_pos == -1)
3699                         return 0;
3700                 // there is a bibitem at the 0 position, but since
3701                 // there is a second one, we copy the second on the
3702                 // first. We're assuming there are at most two of
3703                 // these, which there should be.
3704                 // FIXME: why does it make sense to do that rather
3705                 // than keep the first? (JMarc)
3706                 Inset * inset = releaseInset(bibitem_pos);
3707                 d->insetlist_.begin()->inset = inset;
3708                 return -bibitem_pos;
3709         }
3710
3711         // We need to create an inset at the beginning
3712         Inset * inset = 0;
3713         if (bibitem_pos > 0) {
3714                 // there was one somewhere in the paragraph, let's move it
3715                 inset = d->insetlist_.release(bibitem_pos);
3716                 eraseChar(bibitem_pos, track_changes);
3717         } else
3718                 // make a fresh one
3719                 inset = new InsetBibitem(const_cast<Buffer *>(&buffer),
3720                                          InsetCommandParams(BIBITEM_CODE));
3721
3722         Font font(inherit_font, buffer.params().language);
3723         insertInset(0, inset, font, Change(track_changes ? Change::INSERTED
3724                                                    : Change::UNCHANGED));
3725
3726         // This is needed to get the counters right
3727         buffer.updateBuffer();
3728         return 1;
3729 }
3730
3731
3732 void Paragraph::checkAuthors(AuthorList const & authorList)
3733 {
3734         d->changes_.checkAuthors(authorList);
3735 }
3736
3737
3738 bool Paragraph::isChanged(pos_type pos) const
3739 {
3740         return lookupChange(pos).changed();
3741 }
3742
3743
3744 bool Paragraph::isInserted(pos_type pos) const
3745 {
3746         return lookupChange(pos).inserted();
3747 }
3748
3749
3750 bool Paragraph::isDeleted(pos_type pos) const
3751 {
3752         return lookupChange(pos).deleted();
3753 }
3754
3755
3756 InsetList const & Paragraph::insetList() const
3757 {
3758         return d->insetlist_;
3759 }
3760
3761
3762 void Paragraph::setInsetBuffers(Buffer & b)
3763 {
3764         d->insetlist_.setBuffer(b);
3765 }
3766
3767
3768 void Paragraph::resetBuffer()
3769 {
3770         d->insetlist_.resetBuffer();
3771 }
3772
3773
3774 Inset * Paragraph::releaseInset(pos_type pos)
3775 {
3776         Inset * inset = d->insetlist_.release(pos);
3777         /// does not honour change tracking!
3778         eraseChar(pos, false);
3779         return inset;
3780 }
3781
3782
3783 Inset * Paragraph::getInset(pos_type pos)
3784 {
3785         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3786                  ? d->insetlist_.get(pos) : 0;
3787 }
3788
3789
3790 Inset const * Paragraph::getInset(pos_type pos) const
3791 {
3792         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3793                  ? d->insetlist_.get(pos) : 0;
3794 }
3795
3796
3797 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
3798                 pos_type & right, TextCase action)
3799 {
3800         // process sequences of modified characters; in change
3801         // tracking mode, this approach results in much better
3802         // usability than changing case on a char-by-char basis
3803         // We also need to track the current font, since font
3804         // changes within sequences can occur.
3805         vector<pair<char_type, Font> > changes;
3806
3807         bool const trackChanges = bparams.track_changes;
3808
3809         bool capitalize = true;
3810
3811         for (; pos < right; ++pos) {
3812                 char_type oldChar = d->text_[pos];
3813                 char_type newChar = oldChar;
3814
3815                 // ignore insets and don't play with deleted text!
3816                 if (oldChar != META_INSET && !isDeleted(pos)) {
3817                         switch (action) {
3818                                 case text_lowercase:
3819                                         newChar = lowercase(oldChar);
3820                                         break;
3821                                 case text_capitalization:
3822                                         if (capitalize) {
3823                                                 newChar = uppercase(oldChar);
3824                                                 capitalize = false;
3825                                         }
3826                                         break;
3827                                 case text_uppercase:
3828                                         newChar = uppercase(oldChar);
3829                                         break;
3830                         }
3831                 }
3832
3833                 if (isWordSeparator(pos) || isDeleted(pos)) {
3834                         // permit capitalization again
3835                         capitalize = true;
3836                 }
3837
3838                 if (oldChar != newChar) {
3839                         changes.push_back(make_pair(newChar, getFontSettings(bparams, pos)));
3840                         if (pos != right - 1)
3841                                 continue;
3842                         // step behind the changing area
3843                         pos++;
3844                 }
3845
3846                 int erasePos = pos - changes.size();
3847                 for (size_t i = 0; i < changes.size(); i++) {
3848                         insertChar(pos, changes[i].first,
3849                                    changes[i].second,
3850                                    trackChanges);
3851                         if (!eraseChar(erasePos, trackChanges)) {
3852                                 ++erasePos;
3853                                 ++pos; // advance
3854                                 ++right; // expand selection
3855                         }
3856                 }
3857                 changes.clear();
3858         }
3859 }
3860
3861
3862 int Paragraph::find(docstring const & str, bool cs, bool mw,
3863                 pos_type start_pos, bool del) const
3864 {
3865         pos_type pos = start_pos;
3866         int const strsize = str.length();
3867         int i = 0;
3868         pos_type const parsize = d->text_.size();
3869         for (i = 0; i < strsize && pos < parsize; ++i, ++pos) {
3870                 // Ignore "invisible" letters such as ligature breaks
3871                 // and hyphenation chars while searching
3872                 while (pos < parsize - 1 && isInset(pos)) {
3873                         odocstringstream os;
3874                         getInset(pos)->toString(os);
3875                         if (!getInset(pos)->isLetter() || !os.str().empty())
3876                                 break;
3877                         pos++;
3878                 }
3879                 if (cs && str[i] != d->text_[pos])
3880                         break;
3881                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos]))
3882                         break;
3883                 if (!del && isDeleted(pos))
3884                         break;
3885         }
3886
3887         if (i != strsize)
3888                 return 0;
3889
3890         // if necessary, check whether string matches word
3891         if (mw) {
3892                 if (start_pos > 0 && !isWordSeparator(start_pos - 1))
3893                         return 0;
3894                 if (pos < parsize
3895                         && !isWordSeparator(pos))
3896                         return 0;
3897         }
3898
3899         return pos - start_pos;
3900 }
3901
3902
3903 char_type Paragraph::getChar(pos_type pos) const
3904 {
3905         return d->text_[pos];
3906 }
3907
3908
3909 pos_type Paragraph::size() const
3910 {
3911         return d->text_.size();
3912 }
3913
3914
3915 bool Paragraph::empty() const
3916 {
3917         return d->text_.empty();
3918 }
3919
3920
3921 bool Paragraph::isInset(pos_type pos) const
3922 {
3923         return d->text_[pos] == META_INSET;
3924 }
3925
3926
3927 bool Paragraph::isSeparator(pos_type pos) const
3928 {
3929         //FIXME: Are we sure this can be the only separator?
3930         return d->text_[pos] == ' ';
3931 }
3932
3933
3934 void Paragraph::deregisterWords()
3935 {
3936         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3937         Private::LangWordsMap::const_iterator ite = d->words_.end();
3938         for (; itl != ite; ++itl) {
3939                 WordList & wl = theWordList(itl->first);
3940                 Private::Words::const_iterator it = (itl->second).begin();
3941                 Private::Words::const_iterator et = (itl->second).end();
3942                 for (; it != et; ++it)
3943                         wl.remove(*it);
3944         }
3945         d->words_.clear();
3946 }
3947
3948
3949 void Paragraph::locateWord(pos_type & from, pos_type & to,
3950         word_location const loc) const
3951 {
3952         switch (loc) {
3953         case WHOLE_WORD_STRICT:
3954                 if (from == 0 || from == size()
3955                     || isWordSeparator(from)
3956                     || isWordSeparator(from - 1)) {
3957                         to = from;
3958                         return;
3959                 }
3960                 // fall through
3961
3962         case WHOLE_WORD:
3963                 // If we are already at the beginning of a word, do nothing
3964                 if (!from || isWordSeparator(from - 1))
3965                         break;
3966                 // fall through
3967
3968         case PREVIOUS_WORD:
3969                 // always move the cursor to the beginning of previous word
3970                 while (from && !isWordSeparator(from - 1))
3971                         --from;
3972                 break;
3973         case NEXT_WORD:
3974                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
3975                 break;
3976         case PARTIAL_WORD:
3977                 // no need to move the 'from' cursor
3978                 break;
3979         }
3980         to = from;
3981         while (to < size() && !isWordSeparator(to))
3982                 ++to;
3983 }
3984
3985
3986 void Paragraph::collectWords()
3987 {
3988         for (pos_type pos = 0; pos < size(); ++pos) {
3989                 if (isWordSeparator(pos))
3990                         continue;
3991                 pos_type from = pos;
3992                 locateWord(from, pos, WHOLE_WORD);
3993                 // Work around MSVC warning: The statement
3994                 // if (pos < from + lyxrc.completion_minlength)
3995                 // triggers a signed vs. unsigned warning.
3996                 // I don't know why this happens, it could be a MSVC bug, or
3997                 // related to LLP64 (windows) vs. LP64 (unix) programming
3998                 // model, or the C++ standard might be ambigous in the section
3999                 // defining the "usual arithmetic conversions". However, using
4000                 // a temporary variable is safe and works on all compilers.
4001                 pos_type const endpos = from + lyxrc.completion_minlength;
4002                 if (pos < endpos)
4003                         continue;
4004                 FontList::const_iterator cit = d->fontlist_.fontIterator(from);
4005                 if (cit == d->fontlist_.end())
4006                         return;
4007                 Language const * lang = cit->font().language();
4008                 docstring const word = asString(from, pos, AS_STR_NONE);
4009                 d->words_[lang->lang()].insert(word);
4010         }
4011 }
4012
4013
4014 void Paragraph::registerWords()
4015 {
4016         Private::LangWordsMap::const_iterator itl = d->words_.begin();
4017         Private::LangWordsMap::const_iterator ite = d->words_.end();
4018         for (; itl != ite; ++itl) {
4019                 WordList & wl = theWordList(itl->first);
4020                 Private::Words::const_iterator it = (itl->second).begin();
4021                 Private::Words::const_iterator et = (itl->second).end();
4022                 for (; it != et; ++it)
4023                         wl.insert(*it);
4024         }
4025 }
4026
4027
4028 void Paragraph::updateWords()
4029 {
4030         deregisterWords();
4031         collectWords();
4032         registerWords();
4033 }
4034
4035
4036 void Paragraph::Private::appendSkipPosition(SkipPositions & skips, pos_type const pos) const
4037 {
4038         SkipPositionsIterator begin = skips.begin();
4039         SkipPositions::iterator end = skips.end();
4040         if (pos > 0 && begin < end) {
4041                 --end;
4042                 if (end->last == pos - 1) {
4043                         end->last = pos;
4044                         return;
4045                 }
4046         }
4047         skips.insert(end, FontSpan(pos, pos));
4048 }
4049
4050
4051 Language * Paragraph::Private::locateSpellRange(
4052         pos_type & from, pos_type & to,
4053         SkipPositions & skips) const
4054 {
4055         // skip leading white space
4056         while (from < to && owner_->isWordSeparator(from))
4057                 ++from;
4058         // don't check empty range
4059         if (from >= to)
4060                 return 0;
4061         // get current language
4062         Language * lang = getSpellLanguage(from);
4063         pos_type last = from;
4064         bool samelang = true;
4065         bool sameinset = true;
4066         while (last < to && samelang && sameinset) {
4067                 // hop to end of word
4068                 while (last < to && !owner_->isWordSeparator(last)) {
4069                         if (owner_->getInset(last)) {
4070                                 appendSkipPosition(skips, last);
4071                         } else if (owner_->isDeleted(last)) {
4072                                 appendSkipPosition(skips, last);
4073                         }
4074                         ++last;
4075                 }
4076                 // hop to next word while checking for insets
4077                 while (sameinset && last < to && owner_->isWordSeparator(last)) {
4078                         if (Inset const * inset = owner_->getInset(last))
4079                                 sameinset = inset->isChar() && inset->isLetter();
4080                         if (sameinset && owner_->isDeleted(last)) {
4081                                 appendSkipPosition(skips, last);
4082                         }
4083                         if (sameinset)
4084                                 last++;
4085                 }
4086                 if (sameinset && last < to) {
4087                         // now check for language change
4088                         samelang = lang == getSpellLanguage(last);
4089                 }
4090         }
4091         // if language change detected backstep is needed
4092         if (!samelang)
4093                 --last;
4094         to = last;
4095         return lang;
4096 }
4097
4098
4099 Language * Paragraph::Private::getSpellLanguage(pos_type const from) const
4100 {
4101         Language * lang =
4102                 const_cast<Language *>(owner_->getFontSettings(
4103                         inset_owner_->buffer().params(), from).language());
4104         if (lang == inset_owner_->buffer().params().language
4105                 && !lyxrc.spellchecker_alt_lang.empty()) {
4106                 string lang_code;
4107                 string const lang_variety =
4108                         split(lyxrc.spellchecker_alt_lang, lang_code, '-');
4109                 lang->setCode(lang_code);
4110                 lang->setVariety(lang_variety);
4111         }
4112         return lang;
4113 }
4114
4115
4116 void Paragraph::requestSpellCheck(pos_type pos)
4117 {
4118         d->requestSpellCheck(pos);
4119 }
4120
4121
4122 bool Paragraph::needsSpellCheck() const
4123 {
4124         SpellChecker::ChangeNumber speller_change_number = 0;
4125         if (theSpellChecker())
4126                 speller_change_number = theSpellChecker()->changeNumber();
4127         if (speller_change_number > d->speller_state_.currentChangeNumber()) {
4128                 d->speller_state_.needsCompleteRefresh(speller_change_number);
4129         }
4130         return d->needsSpellCheck();
4131 }
4132
4133
4134 bool Paragraph::Private::ignoreWord(docstring const & word) const
4135 {
4136         // Ignore words with digits
4137         // FIXME: make this customizable
4138         // (note that some checkers ignore words with digits by default)
4139         docstring::const_iterator cit = word.begin();
4140         docstring::const_iterator const end = word.end();
4141         for (; cit != end; ++cit) {
4142                 if (isNumber((*cit)))
4143                         return true;
4144         }
4145         return false;
4146 }
4147
4148
4149 SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
4150         WordLangTuple & wl, docstring_list & suggestions,
4151         bool do_suggestion, bool check_learned) const
4152 {
4153         SpellChecker::Result result = SpellChecker::WORD_OK;
4154         SpellChecker * speller = theSpellChecker();
4155         if (!speller)
4156                 return result;
4157
4158         if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
4159                 return result;
4160
4161         locateWord(from, to, WHOLE_WORD);
4162         if (from == to || from >= size())
4163                 return result;
4164
4165         docstring word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
4166         Language * lang = d->getSpellLanguage(from);
4167
4168         if (getFontSettings(d->inset_owner_->buffer().params(), from).fontInfo().nospellcheck() == FONT_ON)
4169                 return result;
4170
4171         wl = WordLangTuple(word, lang);
4172
4173         if (word.empty())
4174                 return result;
4175
4176         if (needsSpellCheck() || check_learned) {
4177                 pos_type end = to;
4178                 if (!d->ignoreWord(word)) {
4179                         bool const trailing_dot = to < size() && d->text_[to] == '.';
4180                         result = speller->check(wl);
4181                         if (SpellChecker::misspelled(result) && trailing_dot) {
4182                                 wl = WordLangTuple(word.append(from_ascii(".")), lang);
4183                                 result = speller->check(wl);
4184                                 if (!SpellChecker::misspelled(result)) {
4185                                         LYXERR(Debug::GUI, "misspelled word is correct with dot: \"" <<
4186                                            word << "\" [" <<
4187                                            from << ".." << to << "]");
4188                                 } else {
4189                                         // spell check with dot appended failed too
4190                                         // restore original word/lang value
4191                                         word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
4192                                         wl = WordLangTuple(word, lang);
4193                                 }
4194                         }
4195                 }
4196                 if (!SpellChecker::misspelled(result)) {
4197                         // area up to the begin of the next word is not misspelled
4198                         while (end < size() && isWordSeparator(end))
4199                                 ++end;
4200                 }
4201                 d->setMisspelled(from, end, result);
4202         } else {
4203                 result = d->speller_state_.getState(from);
4204         }
4205
4206         if (do_suggestion)
4207                 suggestions.clear();
4208
4209         if (SpellChecker::misspelled(result)) {
4210                 LYXERR(Debug::GUI, "misspelled word: \"" <<
4211                            word << "\" [" <<
4212                            from << ".." << to << "]");
4213                 if (do_suggestion)
4214                         speller->suggest(wl, suggestions);
4215         }
4216         return result;
4217 }
4218
4219
4220 void Paragraph::anonymize()
4221 {
4222         // This is a very crude anonymization for now
4223         for (char_type & c : d->text_)
4224                 if (isLetterChar(c) || isNumber(c))
4225                         c = 'a';
4226 }
4227
4228
4229 void Paragraph::Private::markMisspelledWords(
4230         pos_type const & first, pos_type const & last,
4231         SpellChecker::Result result,
4232         docstring const & word,
4233         SkipPositions const & skips)
4234 {
4235         if (!SpellChecker::misspelled(result)) {
4236                 setMisspelled(first, last, SpellChecker::WORD_OK);
4237                 return;
4238         }
4239         int snext = first;
4240         SpellChecker * speller = theSpellChecker();
4241         // locate and enumerate the error positions
4242         int nerrors = speller->numMisspelledWords();
4243         int numskipped = 0;
4244         SkipPositionsIterator it = skips.begin();
4245         SkipPositionsIterator et = skips.end();
4246         for (int index = 0; index < nerrors; ++index) {
4247                 int wstart;
4248                 int wlen = 0;
4249                 speller->misspelledWord(index, wstart, wlen);
4250                 /// should not happen if speller supports range checks
4251                 if (!wlen) continue;
4252                 docstring const misspelled = word.substr(wstart, wlen);
4253                 wstart += first + numskipped;
4254                 if (snext < wstart) {
4255                         /// mark the range of correct spelling
4256                         numskipped += countSkips(it, et, wstart);
4257                         setMisspelled(snext,
4258                                 wstart - 1, SpellChecker::WORD_OK);
4259                 }
4260                 snext = wstart + wlen;
4261                 numskipped += countSkips(it, et, snext);
4262                 /// mark the range of misspelling
4263                 setMisspelled(wstart, snext, result);
4264                 LYXERR(Debug::GUI, "misspelled word: \"" <<
4265                            misspelled << "\" [" <<
4266                            wstart << ".." << (snext-1) << "]");
4267                 ++snext;
4268         }
4269         if (snext <= last) {
4270                 /// mark the range of correct spelling at end
4271                 setMisspelled(snext, last, SpellChecker::WORD_OK);
4272         }
4273 }
4274
4275
4276 void Paragraph::spellCheck() const
4277 {
4278         SpellChecker * speller = theSpellChecker();
4279         if (!speller || empty() ||!needsSpellCheck())
4280                 return;
4281         pos_type start;
4282         pos_type endpos;
4283         d->rangeOfSpellCheck(start, endpos);
4284         if (speller->canCheckParagraph()) {
4285                 // loop until we leave the range
4286                 for (pos_type first = start; first < endpos; ) {
4287                         pos_type last = endpos;
4288                         Private::SkipPositions skips;
4289                         Language * lang = d->locateSpellRange(first, last, skips);
4290                         if (first >= endpos)
4291                                 break;
4292                         // start the spell checker on the unit of meaning
4293                         docstring word = asString(first, last, AS_STR_INSETS + AS_STR_SKIPDELETE);
4294                         WordLangTuple wl = WordLangTuple(word, lang);
4295                         SpellChecker::Result result = word.size() ?
4296                                 speller->check(wl) : SpellChecker::WORD_OK;
4297                         d->markMisspelledWords(first, last, result, word, skips);
4298                         first = ++last;
4299                 }
4300         } else {
4301                 static docstring_list suggestions;
4302                 pos_type to = endpos;
4303                 while (start < endpos) {
4304                         WordLangTuple wl;
4305                         spellCheck(start, to, wl, suggestions, false);
4306                         start = to + 1;
4307                 }
4308         }
4309         d->readySpellCheck();
4310 }
4311
4312
4313 bool Paragraph::isMisspelled(pos_type pos, bool check_boundary) const
4314 {
4315         bool result = SpellChecker::misspelled(d->speller_state_.getState(pos));
4316         if (result || pos <= 0 || pos > size())
4317                 return result;
4318         if (check_boundary && (pos == size() || isWordSeparator(pos)))
4319                 result = SpellChecker::misspelled(d->speller_state_.getState(pos - 1));
4320         return result;
4321 }
4322
4323
4324 string Paragraph::magicLabel() const
4325 {
4326         stringstream ss;
4327         ss << "magicparlabel-" << id();
4328         return ss.str();
4329 }
4330
4331
4332 } // namespace lyx