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 (inInset().getLayout().isNeedProtect())
1651                 fragile = true;
1652         if (needsCProtection(fragile))
1653                 features.require("cprotect");
1654 }
1655
1656
1657 void Paragraph::insert(pos_type start, docstring const & str,
1658                        Font const & font, Change const & change)
1659 {
1660         for (size_t i = 0, n = str.size(); i != n ; ++i)
1661                 insertChar(start + i, str[i], font, change);
1662 }
1663
1664
1665 void Paragraph::appendChar(char_type c, Font const & font,
1666                 Change const & change)
1667 {
1668         // track change
1669         d->changes_.insert(change, d->text_.size());
1670         // when appending characters, no need to update tables
1671         d->text_.push_back(c);
1672         setFont(d->text_.size() - 1, font);
1673         d->requestSpellCheck(d->text_.size() - 1);
1674 }
1675
1676
1677 void Paragraph::appendString(docstring const & s, Font const & font,
1678                 Change const & change)
1679 {
1680         pos_type end = s.size();
1681         size_t oldsize = d->text_.size();
1682         size_t newsize = oldsize + end;
1683         size_t capacity = d->text_.capacity();
1684         if (newsize >= capacity)
1685                 d->text_.reserve(max(capacity + 100, newsize));
1686
1687         // when appending characters, no need to update tables
1688         d->text_.append(s);
1689
1690         // FIXME: Optimize this!
1691         for (size_t i = oldsize; i != newsize; ++i) {
1692                 // track change
1693                 d->changes_.insert(change, i);
1694                 d->requestSpellCheck(i);
1695         }
1696         d->fontlist_.set(oldsize, font);
1697         d->fontlist_.set(newsize - 1, font);
1698 }
1699
1700
1701 void Paragraph::insertChar(pos_type pos, char_type c,
1702                            bool trackChanges)
1703 {
1704         d->insertChar(pos, c, Change(trackChanges ?
1705                            Change::INSERTED : Change::UNCHANGED));
1706 }
1707
1708
1709 void Paragraph::insertChar(pos_type pos, char_type c,
1710                            Font const & font, bool trackChanges)
1711 {
1712         d->insertChar(pos, c, Change(trackChanges ?
1713                            Change::INSERTED : Change::UNCHANGED));
1714         setFont(pos, font);
1715 }
1716
1717
1718 void Paragraph::insertChar(pos_type pos, char_type c,
1719                            Font const & font, Change const & change)
1720 {
1721         d->insertChar(pos, c, change);
1722         setFont(pos, font);
1723 }
1724
1725
1726 void Paragraph::resetFonts(Font const & font)
1727 {
1728         d->fontlist_.clear();
1729         d->fontlist_.set(0, font);
1730         d->fontlist_.set(d->text_.size() - 1, font);
1731 }
1732
1733 // Gets uninstantiated font setting at position.
1734 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1735                                          pos_type pos) const
1736 {
1737         if (pos > size()) {
1738                 LYXERR0("pos: " << pos << " size: " << size());
1739                 LBUFERR(false);
1740         }
1741
1742         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1743         if (cit != d->fontlist_.end())
1744                 return cit->font();
1745
1746         if (pos == size() && !empty())
1747                 return getFontSettings(bparams, pos - 1);
1748
1749         // Optimisation: avoid a full font instantiation if there is no
1750         // language change from previous call.
1751         static Font previous_font;
1752         static Language const * previous_lang = 0;
1753         Language const * lang = getParLanguage(bparams);
1754         if (lang != previous_lang) {
1755                 previous_lang = lang;
1756                 previous_font = Font(inherit_font, lang);
1757         }
1758         return previous_font;
1759 }
1760
1761
1762 FontSpan Paragraph::fontSpan(pos_type pos) const
1763 {
1764         LBUFERR(pos <= size());
1765
1766         if (pos == size())
1767                 return FontSpan(pos, pos);
1768
1769         pos_type start = 0;
1770         FontList::const_iterator cit = d->fontlist_.begin();
1771         FontList::const_iterator end = d->fontlist_.end();
1772         for (; cit != end; ++cit) {
1773                 if (cit->pos() >= pos) {
1774                         if (pos >= beginOfBody())
1775                                 return FontSpan(max(start, beginOfBody()),
1776                                                 cit->pos());
1777                         else
1778                                 return FontSpan(start,
1779                                                 min(beginOfBody() - 1,
1780                                                          cit->pos()));
1781                 }
1782                 start = cit->pos() + 1;
1783         }
1784
1785         // This should not happen, but if so, we take no chances.
1786         LYXERR0("Paragraph::fontSpan: position not found in fontinfo table!");
1787         LASSERT(false, return FontSpan(pos, pos));
1788 }
1789
1790
1791 // Gets uninstantiated font setting at position 0
1792 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1793 {
1794         if (!empty() && !d->fontlist_.empty())
1795                 return d->fontlist_.begin()->font();
1796
1797         // Optimisation: avoid a full font instantiation if there is no
1798         // language change from previous call.
1799         static Font previous_font;
1800         static Language const * previous_lang = 0;
1801         if (bparams.language != previous_lang) {
1802                 previous_lang = bparams.language;
1803                 previous_font = Font(inherit_font, bparams.language);
1804         }
1805
1806         return previous_font;
1807 }
1808
1809
1810 // Gets the fully instantiated font at a given position in a paragraph
1811 // This is basically the same function as Text::GetFont() in text2.cpp.
1812 // The difference is that this one is used for generating the LaTeX file,
1813 // and thus cosmetic "improvements" are disallowed: This has to deliver
1814 // the true picture of the buffer. (Asger)
1815 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1816                                  Font const & outerfont) const
1817 {
1818         LBUFERR(pos >= 0);
1819
1820         Font font = getFontSettings(bparams, pos);
1821
1822         pos_type const body_pos = beginOfBody();
1823         FontInfo & fi = font.fontInfo();
1824         if (pos < body_pos)
1825                 fi.realize(d->layout_->labelfont);
1826         else
1827                 fi.realize(d->layout_->font);
1828
1829         fi.realize(outerfont.fontInfo());
1830         fi.realize(bparams.getFont().fontInfo());
1831
1832         return font;
1833 }
1834
1835
1836 Font const Paragraph::getLabelFont
1837         (BufferParams const & bparams, Font const & outerfont) const
1838 {
1839         FontInfo tmpfont = d->layout_->labelfont;
1840         tmpfont.realize(outerfont.fontInfo());
1841         tmpfont.realize(bparams.getFont().fontInfo());
1842         return Font(tmpfont, getParLanguage(bparams));
1843 }
1844
1845
1846 Font const Paragraph::getLayoutFont
1847         (BufferParams const & bparams, Font const & outerfont) const
1848 {
1849         FontInfo tmpfont = d->layout_->font;
1850         tmpfont.realize(outerfont.fontInfo());
1851         tmpfont.realize(bparams.getFont().fontInfo());
1852         return Font(tmpfont, getParLanguage(bparams));
1853 }
1854
1855
1856 char_type Paragraph::getUChar(BufferParams const & bparams,
1857                               OutputParams const & rp,
1858                               pos_type pos) const
1859 {
1860         char_type c = d->text_[pos];
1861
1862         // Return unchanged character in LTR languages
1863         // or if we use poylglossia/bidi.
1864         if (rp.use_polyglossia || !getFontSettings(bparams, pos).isRightToLeft())
1865                 return c;
1866
1867         // Without polyglossia/bidi, we need to account for some special cases.
1868         // FIXME This needs to be audited!
1869         // Check if:
1870         // * The input is as expected for all delimiters
1871         //   => checked for Hebrew!
1872         // * The output matches the display in the LyX workarea
1873         //   => checked for Hebrew!
1874         // * The special cases below are really necessary
1875         //   => checked for Hebrew!
1876         // * In arabic_arabi, brackets are transformed to Arabic
1877         //   Ornate Parentheses. Is this is really wanted?
1878
1879         string const & lang = getFontSettings(bparams, pos).language()->lang();
1880         char_type uc = c;
1881
1882         // 1. In the following languages, parentheses need to be reversed.
1883         bool const reverseparens = lang == "hebrew";
1884
1885         // 2. In the following languages, brackets don't need to be reversed.
1886         bool const reversebrackets = lang != "arabic_arabtex"
1887                         && lang != "arabic_arabi"
1888                         && lang != "farsi";
1889
1890         // Now swap delimiters if needed.
1891         switch (c) {
1892         case '(':
1893                 if (reverseparens)
1894                         uc = ')';
1895                 break;
1896         case ')':
1897                 if (reverseparens)
1898                         uc = '(';
1899                 break;
1900         case '[':
1901                 if (reversebrackets)
1902                         uc = ']';
1903                 break;
1904         case ']':
1905                 if (reversebrackets)
1906                         uc = '[';
1907                 break;
1908         case '{':
1909                 uc = '}';
1910                 break;
1911         case '}':
1912                 uc = '{';
1913                 break;
1914         case '<':
1915                 uc = '>';
1916                 break;
1917         case '>':
1918                 uc = '<';
1919                 break;
1920         }
1921
1922         return uc;
1923 }
1924
1925
1926 void Paragraph::setFont(pos_type pos, Font const & font)
1927 {
1928         LASSERT(pos <= size(), return);
1929
1930         // First, reduce font against layout/label font
1931         // Update: The setCharFont() routine in text2.cpp already
1932         // reduces font, so we don't need to do that here. (Asger)
1933
1934         d->fontlist_.set(pos, font);
1935 }
1936
1937
1938 void Paragraph::makeSameLayout(Paragraph const & par)
1939 {
1940         d->layout_ = par.d->layout_;
1941         d->params_ = par.d->params_;
1942 }
1943
1944
1945 bool Paragraph::stripLeadingSpaces(bool trackChanges)
1946 {
1947         if (isFreeSpacing())
1948                 return false;
1949
1950         int pos = 0;
1951         int count = 0;
1952
1953         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
1954                 if (eraseChar(pos, trackChanges))
1955                         ++count;
1956                 else
1957                         ++pos;
1958         }
1959
1960         return count > 0 || pos > 0;
1961 }
1962
1963
1964 bool Paragraph::hasSameLayout(Paragraph const & par) const
1965 {
1966         return par.d->layout_ == d->layout_
1967                 && d->params_.sameLayout(par.d->params_);
1968 }
1969
1970
1971 depth_type Paragraph::getDepth() const
1972 {
1973         return d->params_.depth();
1974 }
1975
1976
1977 depth_type Paragraph::getMaxDepthAfter() const
1978 {
1979         if (d->layout_->isEnvironment())
1980                 return d->params_.depth() + 1;
1981         else
1982                 return d->params_.depth();
1983 }
1984
1985
1986 LyXAlignment Paragraph::getAlign() const
1987 {
1988         if (d->params_.align() == LYX_ALIGN_LAYOUT)
1989                 return d->layout_->align;
1990         else
1991                 return d->params_.align();
1992 }
1993
1994
1995 docstring const & Paragraph::labelString() const
1996 {
1997         return d->params_.labelString();
1998 }
1999
2000
2001 // the next two functions are for the manual labels
2002 docstring const Paragraph::getLabelWidthString() const
2003 {
2004         if (d->layout_->margintype == MARGIN_MANUAL
2005             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
2006                 return d->params_.labelWidthString();
2007         else
2008                 return _("Senseless with this layout!");
2009 }
2010
2011
2012 void Paragraph::setLabelWidthString(docstring const & s)
2013 {
2014         d->params_.labelWidthString(s);
2015 }
2016
2017
2018 docstring Paragraph::expandLabel(Layout const & layout,
2019                 BufferParams const & bparams) const
2020 {
2021         return expandParagraphLabel(layout, bparams, true);
2022 }
2023
2024
2025 docstring Paragraph::expandDocBookLabel(Layout const & layout,
2026                 BufferParams const & bparams) const
2027 {
2028         return expandParagraphLabel(layout, bparams, false);
2029 }
2030
2031
2032 docstring Paragraph::expandParagraphLabel(Layout const & layout,
2033                 BufferParams const & bparams, bool process_appendix) const
2034 {
2035         DocumentClass const & tclass = bparams.documentClass();
2036         string const & lang = getParLanguage(bparams)->code();
2037         bool const in_appendix = process_appendix && d->params_.appendix();
2038         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
2039
2040         if (fmt.empty() && !layout.counter.empty())
2041                 return tclass.counters().theCounter(layout.counter, lang);
2042
2043         // handle 'inherited level parts' in 'fmt',
2044         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
2045         size_t const i = fmt.find('@', 0);
2046         if (i != docstring::npos) {
2047                 size_t const j = fmt.find('@', i + 1);
2048                 if (j != docstring::npos) {
2049                         docstring parent(fmt, i + 1, j - i - 1);
2050                         docstring label = from_ascii("??");
2051                         if (tclass.hasLayout(parent))
2052                                 label = expandParagraphLabel(tclass[parent], bparams,
2053                                                       process_appendix);
2054                         fmt = docstring(fmt, 0, i) + label
2055                                 + docstring(fmt, j + 1, docstring::npos);
2056                 }
2057         }
2058
2059         return tclass.counters().counterLabel(fmt, lang);
2060 }
2061
2062
2063 void Paragraph::applyLayout(Layout const & new_layout)
2064 {
2065         d->layout_ = &new_layout;
2066         LyXAlignment const oldAlign = d->params_.align();
2067
2068         if (!(oldAlign & d->layout_->alignpossible)) {
2069                 frontend::Alert::warning(_("Alignment not permitted"),
2070                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
2071                 d->params_.align(LYX_ALIGN_LAYOUT);
2072         }
2073 }
2074
2075
2076 pos_type Paragraph::beginOfBody() const
2077 {
2078         return d->begin_of_body_;
2079 }
2080
2081
2082 void Paragraph::setBeginOfBody()
2083 {
2084         if (d->layout_->labeltype != LABEL_MANUAL) {
2085                 d->begin_of_body_ = 0;
2086                 return;
2087         }
2088
2089         // Unroll the first two cycles of the loop
2090         // and remember the previous character to
2091         // remove unnecessary getChar() calls
2092         pos_type i = 0;
2093         pos_type end = size();
2094         if (i < end && !(isNewline(i) || isEnvSeparator(i))) {
2095                 ++i;
2096                 if (i < end) {
2097                         char_type previous_char = d->text_[i];
2098                         if (!(isNewline(i) || isEnvSeparator(i))) {
2099                                 ++i;
2100                                 while (i < end && previous_char != ' ') {
2101                                         char_type temp = d->text_[i];
2102                                         if (isNewline(i) || isEnvSeparator(i))
2103                                                 break;
2104                                         ++i;
2105                                         previous_char = temp;
2106                                 }
2107                         }
2108                 }
2109         }
2110
2111         d->begin_of_body_ = i;
2112 }
2113
2114
2115 bool Paragraph::allowParagraphCustomization() const
2116 {
2117         return inInset().allowParagraphCustomization();
2118 }
2119
2120
2121 bool Paragraph::usePlainLayout() const
2122 {
2123         return inInset().usePlainLayout();
2124 }
2125
2126
2127 bool Paragraph::isPassThru() const
2128 {
2129         return inInset().isPassThru() || d->layout_->pass_thru;
2130 }
2131
2132 namespace {
2133
2134 // paragraphs inside floats need different alignment tags to avoid
2135 // unwanted space
2136
2137 bool noTrivlistCentering(InsetCode code)
2138 {
2139         return code == FLOAT_CODE
2140                || code == WRAP_CODE
2141                || code == CELL_CODE;
2142 }
2143
2144
2145 string correction(string const & orig)
2146 {
2147         if (orig == "flushleft")
2148                 return "raggedright";
2149         if (orig == "flushright")
2150                 return "raggedleft";
2151         if (orig == "center")
2152                 return "centering";
2153         return orig;
2154 }
2155
2156
2157 bool corrected_env(otexstream & os, string const & suffix, string const & env,
2158         InsetCode code, bool const lastpar, int & col)
2159 {
2160         string macro = suffix + "{";
2161         if (noTrivlistCentering(code)) {
2162                 if (lastpar) {
2163                         // the last paragraph in non-trivlist-aligned
2164                         // context is special (to avoid unwanted whitespace)
2165                         if (suffix == "\\begin") {
2166                                 macro = "\\" + correction(env) + "{}";
2167                                 os << from_ascii(macro);
2168                                 col += macro.size();
2169                                 return true;
2170                         }
2171                         return false;
2172                 }
2173                 macro += correction(env);
2174         } else
2175                 macro += env;
2176         macro += "}";
2177         if (suffix == "\\par\\end") {
2178                 os << breakln;
2179                 col = 0;
2180         }
2181         os << from_ascii(macro);
2182         col += macro.size();
2183         if (suffix == "\\begin") {
2184                 os << breakln;
2185                 col = 0;
2186         }
2187         return true;
2188 }
2189
2190 } // namespace
2191
2192
2193 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
2194                         otexstream & os, OutputParams const & runparams) const
2195 {
2196         int column = 0;
2197
2198         bool canindent =
2199                 (bparams.paragraph_separation == BufferParams::ParagraphIndentSeparation) ?
2200                         (layout_->toggle_indent != ITOGGLE_NEVER) :
2201                         (layout_->toggle_indent == ITOGGLE_ALWAYS);
2202
2203         if (canindent && params_.noindent() && !layout_->pass_thru) {
2204                 os << "\\noindent ";
2205                 column += 10;
2206         }
2207
2208         LyXAlignment const curAlign = params_.align();
2209
2210         if (curAlign == layout_->align)
2211                 return column;
2212
2213         switch (curAlign) {
2214         case LYX_ALIGN_NONE:
2215         case LYX_ALIGN_BLOCK:
2216         case LYX_ALIGN_LAYOUT:
2217         case LYX_ALIGN_SPECIAL:
2218         case LYX_ALIGN_DECIMAL:
2219                 break;
2220         case LYX_ALIGN_LEFT:
2221         case LYX_ALIGN_RIGHT:
2222         case LYX_ALIGN_CENTER:
2223                 if (runparams.moving_arg) {
2224                         os << "\\protect";
2225                         column += 8;
2226                 }
2227                 break;
2228         }
2229
2230         string const begin_tag = "\\begin";
2231         InsetCode code = ownerCode();
2232         bool const lastpar = runparams.isLastPar;
2233         // RTL in classic (PDF)LaTeX (without the Bidi package)
2234         bool const rtl_classic = owner_->getParLanguage(bparams)->rightToLeft()
2235                 && !runparams.use_polyglossia;
2236
2237         switch (curAlign) {
2238         case LYX_ALIGN_NONE:
2239         case LYX_ALIGN_BLOCK:
2240         case LYX_ALIGN_LAYOUT:
2241         case LYX_ALIGN_SPECIAL:
2242         case LYX_ALIGN_DECIMAL:
2243                 break;
2244         case LYX_ALIGN_LEFT: {
2245                 if (rtl_classic)
2246                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2247                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2248                 else
2249                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2250                 break;
2251         } case LYX_ALIGN_RIGHT: {
2252                 if (rtl_classic)
2253                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2254                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2255                 else
2256                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2257                 break;
2258         } case LYX_ALIGN_CENTER: {
2259                 corrected_env(os, begin_tag, "center", code, lastpar, column);
2260                 break;
2261         }
2262         }
2263
2264         return column;
2265 }
2266
2267
2268 bool Paragraph::Private::endTeXParParams(BufferParams const & bparams,
2269                         otexstream & os, OutputParams const & runparams) const
2270 {
2271         LyXAlignment const curAlign = params_.align();
2272
2273         if (curAlign == layout_->align)
2274                 return false;
2275
2276         switch (curAlign) {
2277         case LYX_ALIGN_NONE:
2278         case LYX_ALIGN_BLOCK:
2279         case LYX_ALIGN_LAYOUT:
2280         case LYX_ALIGN_SPECIAL:
2281         case LYX_ALIGN_DECIMAL:
2282                 break;
2283         case LYX_ALIGN_LEFT:
2284         case LYX_ALIGN_RIGHT:
2285         case LYX_ALIGN_CENTER:
2286                 if (runparams.moving_arg)
2287                         os << "\\protect";
2288                 break;
2289         }
2290
2291         bool output = false;
2292         int col = 0;
2293         string const end_tag = "\\par\\end";
2294         InsetCode code = ownerCode();
2295         bool const lastpar = runparams.isLastPar;
2296         // RTL in classic (PDF)LaTeX (without the Bidi package)
2297         bool const rtl_classic = owner_->getParLanguage(bparams)->rightToLeft()
2298                 && !runparams.use_polyglossia;
2299
2300         switch (curAlign) {
2301         case LYX_ALIGN_NONE:
2302         case LYX_ALIGN_BLOCK:
2303         case LYX_ALIGN_LAYOUT:
2304         case LYX_ALIGN_SPECIAL:
2305         case LYX_ALIGN_DECIMAL:
2306                 break;
2307         case LYX_ALIGN_LEFT: {
2308                 if (rtl_classic)
2309                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2310                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2311                 else
2312                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2313                 break;
2314         } case LYX_ALIGN_RIGHT: {
2315                 if (rtl_classic)
2316                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2317                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2318                 else
2319                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2320                 break;
2321         } case LYX_ALIGN_CENTER: {
2322                 corrected_env(os, end_tag, "center", code, lastpar, col);
2323                 break;
2324         }
2325         }
2326
2327         return output || lastpar;
2328 }
2329
2330
2331 // This one spits out the text of the paragraph
2332 void Paragraph::latex(BufferParams const & bparams,
2333         Font const & outerfont,
2334         otexstream & os,
2335         OutputParams const & runparams,
2336         int start_pos, int end_pos, bool force) const
2337 {
2338         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
2339
2340         // FIXME This check should not be needed. Perhaps issue an
2341         // error if it triggers.
2342         Layout const & style = inInset().forcePlainLayout() ?
2343                 bparams.documentClass().plainLayout() : *d->layout_;
2344
2345         if (!force && style.inpreamble)
2346                 return;
2347
2348         bool const allowcust = allowParagraphCustomization();
2349
2350         // Current base font for all inherited font changes, without any
2351         // change caused by an individual character, except for the language:
2352         // It is set to the language of the first character.
2353         // As long as we are in the label, this font is the base font of the
2354         // label. Before the first body character it is set to the base font
2355         // of the body.
2356         Font basefont;
2357
2358         // If there is an open font-encoding changing command (script wrapper),
2359         // alien_script is set to its name
2360         string alien_script;
2361         string script;
2362
2363         // Maybe we have to create a optional argument.
2364         pos_type body_pos = beginOfBody();
2365         unsigned int column = 0;
2366
2367         if (body_pos > 0) {
2368                 // the optional argument is kept in curly brackets in
2369                 // case it contains a ']'
2370                 // This is not strictly needed, but if this is changed it
2371                 // would be a file format change, and tex2lyx would need
2372                 // to be adjusted, since it unconditionally removes the
2373                 // braces when it parses \item.
2374                 os << "[{";
2375                 column += 2;
2376                 basefont = getLabelFont(bparams, outerfont);
2377         } else {
2378                 basefont = getLayoutFont(bparams, outerfont);
2379         }
2380
2381         // Which font is currently active?
2382         Font running_font(basefont);
2383         // Do we have an open font change?
2384         bool open_font = false;
2385
2386         Change runningChange = Change(Change::UNCHANGED);
2387
2388         Encoding const * const prev_encoding = runparams.encoding;
2389
2390         os.texrow().start(id(), 0);
2391
2392         // if the paragraph is empty, the loop will not be entered at all
2393         if (empty()) {
2394                 // For InTitle commands, we have already opened a group
2395                 // in output_latex::TeXOnePar.
2396                 if (style.isCommand() && !style.intitle) {
2397                         os << '{';
2398                         ++column;
2399                 }
2400                 if (!style.leftdelim().empty()) {
2401                         os << style.leftdelim();
2402                         column += style.leftdelim().size();
2403                 }
2404                 if (allowcust)
2405                         column += d->startTeXParParams(bparams, os, runparams);
2406         }
2407
2408         // Whether a \par can be issued for insets typeset inline with text.
2409         // Yes if greater than 0. This has to be static.
2410         THREAD_LOCAL_STATIC int parInline = 0;
2411
2412         for (pos_type i = 0; i < size(); ++i) {
2413                 // First char in paragraph or after label?
2414                 if (i == body_pos) {
2415                         if (body_pos > 0) {
2416                                 if (open_font) {
2417                                         bool needPar = false;
2418                                         column += running_font.latexWriteEndChanges(
2419                                                 os, bparams, runparams,
2420                                                 basefont, basefont, needPar);
2421                                         open_font = false;
2422                                 }
2423                                 basefont = getLayoutFont(bparams, outerfont);
2424                                 running_font = basefont;
2425
2426                                 column += Changes::latexMarkChange(os, bparams,
2427                                                 runningChange, Change(Change::UNCHANGED),
2428                                                 runparams);
2429                                 runningChange = Change(Change::UNCHANGED);
2430
2431                                 os << "}] ";
2432                                 column +=3;
2433                         }
2434                         // For InTitle commands, we have already opened a group
2435                         // in output_latex::TeXOnePar.
2436                         if (style.isCommand() && !style.intitle) {
2437                                 os << '{';
2438                                 ++column;
2439                         }
2440
2441                         if (!style.leftdelim().empty()) {
2442                                 os << style.leftdelim();
2443                                 column += style.leftdelim().size();
2444                         }
2445
2446                         if (allowcust)
2447                                 column += d->startTeXParParams(bparams, os,
2448                                                             runparams);
2449                 }
2450
2451                 runparams.wasDisplayMath = runparams.inDisplayMath;
2452                 runparams.inDisplayMath = false;
2453                 bool deleted_display_math = false;
2454                 Change const & change = runparams.inDeletedInset
2455                         ? runparams.changeOfDeletedInset : lookupChange(i);
2456
2457                 // Check whether a display math inset follows
2458                 if (d->text_[i] == META_INSET
2459                     && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2460                         InsetMath const * im = getInset(i)->asInsetMath();
2461                         if (im && im->asHullInset()
2462                             && im->asHullInset()->outerDisplay()) {
2463                                 runparams.inDisplayMath = true;
2464                                 // runparams.inDeletedInset will be set by
2465                                 // latexInset later, but we need this info
2466                                 // before it is called. On the other hand, we
2467                                 // cannot set it here because it is a counter.
2468                                 deleted_display_math = isDeleted(i);
2469                         }
2470                         if (bparams.output_changes && deleted_display_math
2471                             && runningChange == change
2472                             && change.type == Change::DELETED
2473                             && !os.afterParbreak()) {
2474                                 // A display math in the same paragraph follows.
2475                                 // We have to close and then reopen \lyxdeleted,
2476                                 // otherwise the math will be shifted up.
2477                                 OutputParams rp = runparams;
2478                                 if (open_font) {
2479                                         bool needPar = false;
2480                                         column += running_font.latexWriteEndChanges(
2481                                                 os, bparams, rp, basefont,
2482                                                 basefont, needPar);
2483                                         open_font = false;
2484                                 }
2485                                 basefont = getLayoutFont(bparams, outerfont);
2486                                 running_font = basefont;
2487                                 column += Changes::latexMarkChange(os, bparams,
2488                                         Change(Change::INSERTED), change, rp);
2489                         }
2490                 }
2491
2492                 if (bparams.output_changes && runningChange != change) {
2493                         if (!alien_script.empty()) {
2494                                 column += 1;
2495                                 os << "}";
2496                                 alien_script.clear();
2497                         }
2498                         if (open_font) {
2499                                 bool needPar = false;
2500                                 column += running_font.latexWriteEndChanges(
2501                                                 os, bparams, runparams,
2502                                                 basefont, basefont, needPar);
2503                                 open_font = false;
2504                         }
2505                         basefont = getLayoutFont(bparams, outerfont);
2506                         running_font = basefont;
2507                         column += Changes::latexMarkChange(os, bparams, runningChange,
2508                                                            change, runparams);
2509                         runningChange = change;
2510                 }
2511
2512                 // do not output text which is marked deleted
2513                 // if change tracking output is disabled
2514                 if (!bparams.output_changes && change.deleted()) {
2515                         continue;
2516                 }
2517
2518                 ++column;
2519
2520                 // Fully instantiated font
2521                 Font const current_font = getFont(bparams, i, outerfont);
2522
2523                 Font const last_font = running_font;
2524
2525                 // Do we need to close the previous font?
2526                 if (open_font &&
2527                     (current_font != running_font ||
2528                      current_font.language() != running_font.language()))
2529                 {
2530                         // ensure there is no open script-wrapper
2531                         if (!alien_script.empty()) {
2532                                 column += 1;
2533                                 os << "}";
2534                                 alien_script.clear();
2535                         }
2536                         bool needPar = false;
2537                         column += running_font.latexWriteEndChanges(
2538                                     os, bparams, runparams, basefont,
2539                                     (i == body_pos-1) ? basefont : current_font,
2540                                     needPar);
2541                         running_font = basefont;
2542                         open_font = false;
2543                 }
2544
2545                 string const running_lang = runparams.use_polyglossia ?
2546                         running_font.language()->polyglossia() : running_font.language()->babel();
2547                 // close babel's font environment before opening CJK.
2548                 string const lang_end_command = runparams.use_polyglossia ?
2549                         "\\end{$$lang}" : lyxrc.language_command_end;
2550                 bool const using_begin_end = runparams.use_polyglossia ||
2551                                                 !lang_end_command.empty();
2552                 if (!running_lang.empty() &&
2553                     (!using_begin_end || running_lang == openLanguageName()) &&
2554                     current_font.language()->encoding()->package() == Encoding::CJK) {
2555                                 string end_tag = subst(lang_end_command,
2556                                                         "$$lang",
2557                                                         running_lang);
2558                                 os << from_ascii(end_tag);
2559                                 column += end_tag.length();
2560                                 if (using_begin_end)
2561                                         popLanguageName();
2562                 }
2563
2564                 // Switch file encoding if necessary (and allowed)
2565                 if (!runparams.pass_thru && !style.pass_thru &&
2566                     runparams.encoding->package() != Encoding::none &&
2567                     current_font.language()->encoding()->package() != Encoding::none) {
2568                         pair<bool, int> const enc_switch =
2569                                 switchEncoding(os.os(), bparams, runparams,
2570                                         *(current_font.language()->encoding()));
2571                         if (enc_switch.first) {
2572                                 column += enc_switch.second;
2573                                 runparams.encoding = current_font.language()->encoding();
2574                         }
2575                 }
2576
2577                 char_type const c = d->text_[i];
2578
2579                 // A display math inset inside an ulem command will be output
2580                 // as a box of width \linewidth, so we have to either disable
2581                 // indentation if the inset starts a paragraph, or start a new
2582                 // line to accommodate such box. This has to be done before
2583                 // writing any font changing commands.
2584                 if (runparams.inDisplayMath && !deleted_display_math
2585                     && runparams.inulemcmd) {
2586                         if (os.afterParbreak())
2587                                 os << "\\noindent";
2588                         else
2589                                 os << "\\\\\n";
2590                 }
2591
2592                 // Do we need to change font?
2593                 if ((current_font != running_font ||
2594                      current_font.language() != running_font.language()) &&
2595                         i != body_pos - 1)
2596                 {
2597                         odocstringstream ods;
2598                         column += current_font.latexWriteStartChanges(ods, bparams,
2599                                                               runparams, basefont,
2600                                                               last_font);
2601                         // Check again for display math in ulem commands as a
2602                         // font change may also occur just before a math inset.
2603                         if (runparams.inDisplayMath && !deleted_display_math
2604                             && runparams.inulemcmd) {
2605                                 if (os.afterParbreak())
2606                                         os << "\\noindent";
2607                                 else
2608                                         os << "\\\\\n";
2609                         }
2610                         running_font = current_font;
2611                         open_font = true;
2612                         docstring fontchange = ods.str();
2613                         // check whether the fontchange ends with a \\textcolor
2614                         // modifier and the text starts with a space (bug 4473)
2615                         docstring const last_modifier = rsplit(fontchange, '\\');
2616                         if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2617                                 os << fontchange << from_ascii("{}");
2618                         // check if the fontchange ends with a trailing blank
2619                         // (like "\small " (see bug 3382)
2620                         else if (suffixIs(fontchange, ' ') && c == ' ')
2621                                 os << fontchange.substr(0, fontchange.size() - 1)
2622                                    << from_ascii("{}");
2623                         else
2624                                 os << fontchange;
2625                 }
2626
2627                 // FIXME: think about end_pos implementation...
2628                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2629                         // FIXME: integrate this case in latexSpecialChar
2630                         // Do not print the separation of the optional argument
2631                         // if style.pass_thru is false. This works because
2632                         // latexSpecialChar ignores spaces if
2633                         // style.pass_thru is false.
2634                         if (i != body_pos - 1) {
2635                                 if (d->simpleTeXBlanks(bparams, runparams, os,
2636                                                 i, column, current_font, style)) {
2637                                         // A surrogate pair was output. We
2638                                         // must not call latexSpecialChar
2639                                         // in this iteration, since it would output
2640                                         // the combining character again.
2641                                         ++i;
2642                                         continue;
2643                                 }
2644                         }
2645                 }
2646
2647                 OutputParams rp = runparams;
2648                 rp.free_spacing = style.free_spacing;
2649                 rp.local_font = &current_font;
2650                 rp.intitle = style.intitle;
2651
2652                 // Two major modes:  LaTeX or plain
2653                 // Handle here those cases common to both modes
2654                 // and then split to handle the two modes separately.
2655                 if (c == META_INSET) {
2656                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2657                                 // Greyedout notes and, in general, all insets
2658                                 // with InsetLayout::isDisplay() == false,
2659                                 // are typeset inline with the text. So, we
2660                                 // can add a \par to the last paragraph of
2661                                 // such insets only if nothing else follows.
2662                                 bool incremented = false;
2663                                 Inset const * inset = getInset(i);
2664                                 InsetText const * textinset = inset
2665                                                         ? inset->asInsetText()
2666                                                         : 0;
2667                                 if (i + 1 == size() && textinset
2668                                     && !inset->getLayout().isDisplay()) {
2669                                         ParagraphList const & pars =
2670                                                 textinset->text().paragraphs();
2671                                         pit_type const pit = pars.size() - 1;
2672                                         Font const lastfont =
2673                                                 pit < 0 || pars[pit].empty()
2674                                                 ? pars[pit].getLayoutFont(
2675                                                                 bparams,
2676                                                                 outerfont)
2677                                                 : pars[pit].getFont(bparams,
2678                                                         pars[pit].size() - 1,
2679                                                         outerfont);
2680                                         if (lastfont.fontInfo().size() !=
2681                                             basefont.fontInfo().size()) {
2682                                                 ++parInline;
2683                                                 incremented = true;
2684                                         }
2685                                 }
2686                                 d->latexInset(bparams, os, rp, running_font,
2687                                                 basefont, outerfont, open_font,
2688                                                 runningChange, style, i, column);
2689                                 if (incremented)
2690                                         --parInline;
2691                         }
2692                 } else if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2693                         if (!bparams.useNonTeXFonts)
2694                           script = Encodings::isKnownScriptChar(c);
2695                         if (script != alien_script) {
2696                                 if (!alien_script.empty()) {
2697                                         os << "}";
2698                                         alien_script.clear();
2699                                 }
2700                                 string fontenc = running_font.language()->fontenc(bparams);
2701                                 if (!script.empty()
2702                                         && !Encodings::fontencSupportsScript(fontenc, script)) {
2703                                         column += script.length() + 2;
2704                                         os << "\\" << script << "{";
2705                                         alien_script = script;
2706                                 }
2707                         }
2708                         try {
2709                                 d->latexSpecialChar(os, bparams, rp, running_font,
2710                                                                         alien_script, style, i, end_pos, column);
2711                         } catch (EncodingException & e) {
2712                                 if (runparams.dryrun) {
2713                                         os << "<" << _("LyX Warning: ")
2714                                            << _("uncodable character") << " '";
2715                                         os.put(c);
2716                                         os << "'>";
2717                                 } else {
2718                                         // add location information and throw again.
2719                                         e.par_id = id();
2720                                         e.pos = i;
2721                                         throw(e);
2722                                 }
2723                         }
2724                 }
2725
2726                 // Set the encoding to that returned from latexSpecialChar (see
2727                 // comment for encoding member in OutputParams.h)
2728                 runparams.encoding = rp.encoding;
2729
2730                 // Also carry on the info on a closed ulem command for insets
2731                 // such as Note that do not produce any output, so that no
2732                 // command is ever executed but its opening was recorded.
2733                 runparams.inulemcmd = rp.inulemcmd;
2734
2735                 // And finally, pass the post_macros upstream
2736                 runparams.post_macro = rp.post_macro;
2737         }
2738
2739     // Close wrapper for alien script
2740         if (!alien_script.empty()) {
2741                 os << "}";
2742                 alien_script.clear();
2743         }
2744
2745         // If we have an open font definition, we have to close it
2746         if (open_font) {
2747                 // Make sure that \\par is done with the font of the last
2748                 // character if this has another size as the default.
2749                 // This is necessary because LaTeX (and LyX on the screen)
2750                 // calculates the space between the baselines according
2751                 // to this font. (Matthias)
2752                 //
2753                 // We must not change the font for the last paragraph
2754                 // of non-multipar insets, tabular cells or commands,
2755                 // since this produces unwanted whitespace.
2756
2757                 Font const font = empty()
2758                         ? getLayoutFont(bparams, outerfont)
2759                         : getFont(bparams, size() - 1, outerfont);
2760
2761                 InsetText const * textinset = inInset().asInsetText();
2762
2763                 bool const maintext = textinset
2764                         ? textinset->text().isMainText()
2765                         : false;
2766
2767                 size_t const numpars = textinset
2768                         ? textinset->text().paragraphs().size()
2769                         : 0;
2770
2771                 bool needPar = false;
2772
2773                 if (style.resfont.size() != font.fontInfo().size()
2774                     && (!runparams.isLastPar || maintext
2775                         || (numpars > 1 && d->ownerCode() != CELL_CODE
2776                             && (inInset().getLayout().isDisplay()
2777                                 || parInline)))
2778                     && !style.isCommand()) {
2779                         needPar = true;
2780                 }
2781 #ifdef FIXED_LANGUAGE_END_DETECTION
2782                 if (next_) {
2783                         running_font.latexWriteEndChanges(os, bparams,
2784                                         runparams, basefont,
2785                                         next_->getFont(bparams, 0, outerfont),
2786                                                        needPar);
2787                 } else {
2788                         running_font.latexWriteEndChanges(os, bparams,
2789                                         runparams, basefont, basefont, needPar);
2790                 }
2791 #else
2792 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2793 //FIXME: there as we start another \selectlanguage with the next paragraph if
2794 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2795                 running_font.latexWriteEndChanges(os, bparams, runparams,
2796                                 basefont, basefont, needPar);
2797 #endif
2798                 if (needPar) {
2799                         // The \par could not be inserted at the same nesting
2800                         // level of the font size change, so do it now.
2801                         os << "{\\" << font.latexSize() << "\\par}";
2802                 }
2803         }
2804
2805         column += Changes::latexMarkChange(os, bparams, runningChange,
2806                                            Change(Change::UNCHANGED), runparams);
2807
2808         // Needed if there is an optional argument but no contents.
2809         if (body_pos > 0 && body_pos == size()) {
2810                 os << "}]~";
2811         }
2812
2813         if (!style.rightdelim().empty()) {
2814                 os << style.rightdelim();
2815                 column += style.rightdelim().size();
2816         }
2817
2818         if (allowcust && d->endTeXParParams(bparams, os, runparams)
2819             && runparams.encoding != prev_encoding) {
2820                 runparams.encoding = prev_encoding;
2821                 os << setEncoding(prev_encoding->iconvName());
2822         }
2823
2824         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2825 }
2826
2827
2828 bool Paragraph::emptyTag() const
2829 {
2830         for (pos_type i = 0; i < size(); ++i) {
2831                 if (Inset const * inset = getInset(i)) {
2832                         InsetCode lyx_code = inset->lyxCode();
2833                         // FIXME testing like that is wrong. What is
2834                         // the intent?
2835                         if (lyx_code != TOC_CODE &&
2836                             lyx_code != INCLUDE_CODE &&
2837                             lyx_code != GRAPHICS_CODE &&
2838                             lyx_code != ERT_CODE &&
2839                             lyx_code != LISTINGS_CODE &&
2840                             lyx_code != FLOAT_CODE &&
2841                             lyx_code != TABULAR_CODE) {
2842                                 return false;
2843                         }
2844                 } else {
2845                         char_type c = d->text_[i];
2846                         if (c != ' ' && c != '\t')
2847                                 return false;
2848                 }
2849         }
2850         return true;
2851 }
2852
2853
2854 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
2855         const
2856 {
2857         for (pos_type i = 0; i < size(); ++i) {
2858                 if (Inset const * inset = getInset(i)) {
2859                         InsetCode lyx_code = inset->lyxCode();
2860                         if (lyx_code == LABEL_CODE) {
2861                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2862                                 docstring const & id = il->getParam("name");
2863                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2864                         }
2865                 }
2866         }
2867         return string();
2868 }
2869
2870
2871 pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runparams)
2872         const
2873 {
2874         pos_type i;
2875         for (i = 0; i < size(); ++i) {
2876                 if (Inset const * inset = getInset(i)) {
2877                         inset->docbook(os, runparams);
2878                 } else {
2879                         char_type c = d->text_[i];
2880                         if (c == ' ')
2881                                 break;
2882                         os << sgml::escapeChar(c);
2883                 }
2884         }
2885         return i;
2886 }
2887
2888
2889 pos_type Paragraph::firstWordLyXHTML(XHTMLStream & xs, OutputParams const & runparams)
2890         const
2891 {
2892         pos_type i;
2893         for (i = 0; i < size(); ++i) {
2894                 if (Inset const * inset = getInset(i)) {
2895                         inset->xhtml(xs, runparams);
2896                 } else {
2897                         char_type c = d->text_[i];
2898                         if (c == ' ')
2899                                 break;
2900                         xs << c;
2901                 }
2902         }
2903         return i;
2904 }
2905
2906
2907 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2908 {
2909         Font font_old;
2910         pos_type size = text_.size();
2911         for (pos_type i = initial; i < size; ++i) {
2912                 Font font = owner_->getFont(buf.params(), i, outerfont);
2913                 if (text_[i] == META_INSET)
2914                         return false;
2915                 if (i != initial && font != font_old)
2916                         return false;
2917                 font_old = font;
2918         }
2919
2920         return true;
2921 }
2922
2923
2924 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
2925                                     odocstream & os,
2926                                     OutputParams const & runparams,
2927                                     Font const & outerfont,
2928                                     pos_type initial) const
2929 {
2930         bool emph_flag = false;
2931
2932         Layout const & style = *d->layout_;
2933         FontInfo font_old =
2934                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2935
2936         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2937                 os << "]]>";
2938
2939         // parsing main loop
2940         for (pos_type i = initial; i < size(); ++i) {
2941                 Font font = getFont(buf.params(), i, outerfont);
2942
2943                 // handle <emphasis> tag
2944                 if (font_old.emph() != font.fontInfo().emph()) {
2945                         if (font.fontInfo().emph() == FONT_ON) {
2946                                 os << "<emphasis>";
2947                                 emph_flag = true;
2948                         } else if (i != initial) {
2949                                 os << "</emphasis>";
2950                                 emph_flag = false;
2951                         }
2952                 }
2953
2954                 if (Inset const * inset = getInset(i)) {
2955                         inset->docbook(os, runparams);
2956                 } else {
2957                         char_type c = d->text_[i];
2958
2959                         if (style.pass_thru)
2960                                 os.put(c);
2961                         else
2962                                 os << sgml::escapeChar(c);
2963                 }
2964                 font_old = font.fontInfo();
2965         }
2966
2967         if (emph_flag) {
2968                 os << "</emphasis>";
2969         }
2970
2971         if (style.free_spacing)
2972                 os << '\n';
2973         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2974                 os << "<![CDATA[";
2975 }
2976
2977
2978 namespace {
2979 void doFontSwitch(vector<html::FontTag> & tagsToOpen,
2980                   vector<html::EndFontTag> & tagsToClose,
2981                   bool & flag, FontState curstate, html::FontTypes type)
2982 {
2983         if (curstate == FONT_ON) {
2984                 tagsToOpen.push_back(html::FontTag(type));
2985                 flag = true;
2986         } else if (flag) {
2987                 tagsToClose.push_back(html::EndFontTag(type));
2988                 flag = false;
2989         }
2990 }
2991 } // namespace
2992
2993
2994 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
2995                                     XHTMLStream & xs,
2996                                     OutputParams const & runparams,
2997                                     Font const & outerfont,
2998                                     bool start_paragraph, bool close_paragraph,
2999                                     pos_type initial) const
3000 {
3001         docstring retval;
3002
3003         // track whether we have opened these tags
3004         bool emph_flag = false;
3005         bool bold_flag = false;
3006         bool noun_flag = false;
3007         bool ubar_flag = false;
3008         bool dbar_flag = false;
3009         bool sout_flag = false;
3010         bool xout_flag = false;
3011         bool wave_flag = false;
3012         // shape tags
3013         bool shap_flag = false;
3014         // family tags
3015         bool faml_flag = false;
3016         // size tags
3017         bool size_flag = false;
3018
3019         Layout const & style = *d->layout_;
3020
3021         if (start_paragraph)
3022                 xs.startDivision(allowEmpty());
3023
3024         FontInfo font_old =
3025                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
3026
3027         FontShape  curr_fs   = INHERIT_SHAPE;
3028         FontFamily curr_fam  = INHERIT_FAMILY;
3029         FontSize   curr_size = FONT_SIZE_INHERIT;
3030
3031         string const default_family =
3032                 buf.masterBuffer()->params().fonts_default_family;
3033
3034         vector<html::FontTag> tagsToOpen;
3035         vector<html::EndFontTag> tagsToClose;
3036
3037         // parsing main loop
3038         for (pos_type i = initial; i < size(); ++i) {
3039                 // let's not show deleted material in the output
3040                 if (isDeleted(i))
3041                         continue;
3042
3043                 Font const font = getFont(buf.masterBuffer()->params(), i, outerfont);
3044
3045                 // emphasis
3046                 FontState curstate = font.fontInfo().emph();
3047                 if (font_old.emph() != curstate)
3048                         doFontSwitch(tagsToOpen, tagsToClose, emph_flag, curstate, html::FT_EMPH);
3049
3050                 // noun
3051                 curstate = font.fontInfo().noun();
3052                 if (font_old.noun() != curstate)
3053                         doFontSwitch(tagsToOpen, tagsToClose, noun_flag, curstate, html::FT_NOUN);
3054
3055                 // underbar
3056                 curstate = font.fontInfo().underbar();
3057                 if (font_old.underbar() != curstate)
3058                         doFontSwitch(tagsToOpen, tagsToClose, ubar_flag, curstate, html::FT_UBAR);
3059
3060                 // strikeout
3061                 curstate = font.fontInfo().strikeout();
3062                 if (font_old.strikeout() != curstate)
3063                         doFontSwitch(tagsToOpen, tagsToClose, sout_flag, curstate, html::FT_SOUT);
3064
3065                 // xout
3066                 curstate = font.fontInfo().xout();
3067                 if (font_old.xout() != curstate)
3068                         doFontSwitch(tagsToOpen, tagsToClose, xout_flag, curstate, html::FT_XOUT);
3069
3070                 // double underbar
3071                 curstate = font.fontInfo().uuline();
3072                 if (font_old.uuline() != curstate)
3073                         doFontSwitch(tagsToOpen, tagsToClose, dbar_flag, curstate, html::FT_DBAR);
3074
3075                 // wavy line
3076                 curstate = font.fontInfo().uwave();
3077                 if (font_old.uwave() != curstate)
3078                         doFontSwitch(tagsToOpen, tagsToClose, wave_flag, curstate, html::FT_WAVE);
3079
3080                 // bold
3081                 // a little hackish, but allows us to reuse what we have.
3082                 curstate = (font.fontInfo().series() == BOLD_SERIES ? FONT_ON : FONT_OFF);
3083                 if (font_old.series() != font.fontInfo().series())
3084                         doFontSwitch(tagsToOpen, tagsToClose, bold_flag, curstate, html::FT_BOLD);
3085
3086                 // Font shape
3087                 curr_fs = font.fontInfo().shape();
3088                 FontShape old_fs = font_old.shape();
3089                 if (old_fs != curr_fs) {
3090                         if (shap_flag) {
3091                                 switch (old_fs) {
3092                                 case ITALIC_SHAPE:
3093                                         tagsToClose.push_back(html::EndFontTag(html::FT_ITALIC));
3094                                         break;
3095                                 case SLANTED_SHAPE:
3096                                         tagsToClose.push_back(html::EndFontTag(html::FT_SLANTED));
3097                                         break;
3098                                 case SMALLCAPS_SHAPE:
3099                                         tagsToClose.push_back(html::EndFontTag(html::FT_SMALLCAPS));
3100                                         break;
3101                                 case UP_SHAPE:
3102                                 case INHERIT_SHAPE:
3103                                         break;
3104                                 default:
3105                                         // the other tags are for internal use
3106                                         LATTEST(false);
3107                                         break;
3108                                 }
3109                                 shap_flag = false;
3110                         }
3111                         switch (curr_fs) {
3112                         case ITALIC_SHAPE:
3113                                 tagsToOpen.push_back(html::FontTag(html::FT_ITALIC));
3114                                 shap_flag = true;
3115                                 break;
3116                         case SLANTED_SHAPE:
3117                                 tagsToOpen.push_back(html::FontTag(html::FT_SLANTED));
3118                                 shap_flag = true;
3119                                 break;
3120                         case SMALLCAPS_SHAPE:
3121                                 tagsToOpen.push_back(html::FontTag(html::FT_SMALLCAPS));
3122                                 shap_flag = true;
3123                                 break;
3124                         case UP_SHAPE:
3125                         case INHERIT_SHAPE:
3126                                 break;
3127                         default:
3128                                 // the other tags are for internal use
3129                                 LATTEST(false);
3130                                 break;
3131                         }
3132                 }
3133
3134                 // Font family
3135                 curr_fam = font.fontInfo().family();
3136                 FontFamily old_fam = font_old.family();
3137                 if (old_fam != curr_fam) {
3138                         if (faml_flag) {
3139                                 switch (old_fam) {
3140                                 case ROMAN_FAMILY:
3141                                         tagsToClose.push_back(html::EndFontTag(html::FT_ROMAN));
3142                                         break;
3143                                 case SANS_FAMILY:
3144                                         tagsToClose.push_back(html::EndFontTag(html::FT_SANS));
3145                                         break;
3146                                 case TYPEWRITER_FAMILY:
3147                                         tagsToClose.push_back(html::EndFontTag(html::FT_TYPE));
3148                                         break;
3149                                 case INHERIT_FAMILY:
3150                                         break;
3151                                 default:
3152                                         // the other tags are for internal use
3153                                         LATTEST(false);
3154                                         break;
3155                                 }
3156                                 faml_flag = false;
3157                         }
3158                         switch (curr_fam) {
3159                         case ROMAN_FAMILY:
3160                                 // we will treat a "default" font family as roman, since we have
3161                                 // no other idea what to do.
3162                                 if (default_family != "rmdefault" && default_family != "default") {
3163                                         tagsToOpen.push_back(html::FontTag(html::FT_ROMAN));
3164                                         faml_flag = true;
3165                                 }
3166                                 break;
3167                         case SANS_FAMILY:
3168                                 if (default_family != "sfdefault") {
3169                                         tagsToOpen.push_back(html::FontTag(html::FT_SANS));
3170                                         faml_flag = true;
3171                                 }
3172                                 break;
3173                         case TYPEWRITER_FAMILY:
3174                                 if (default_family != "ttdefault") {
3175                                         tagsToOpen.push_back(html::FontTag(html::FT_TYPE));
3176                                         faml_flag = true;
3177                                 }
3178                                 break;
3179                         case INHERIT_FAMILY:
3180                                 break;
3181                         default:
3182                                 // the other tags are for internal use
3183                                 LATTEST(false);
3184                                 break;
3185                         }
3186                 }
3187
3188                 // Font size
3189                 curr_size = font.fontInfo().size();
3190                 FontSize old_size = font_old.size();
3191                 if (old_size != curr_size) {
3192                         if (size_flag) {
3193                                 switch (old_size) {
3194                                 case FONT_SIZE_TINY:
3195                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_TINY));
3196                                         break;
3197                                 case FONT_SIZE_SCRIPT:
3198                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SCRIPT));
3199                                         break;
3200                                 case FONT_SIZE_FOOTNOTE:
3201                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_FOOTNOTE));
3202                                         break;
3203                                 case FONT_SIZE_SMALL:
3204                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SMALL));
3205                                         break;
3206                                 case FONT_SIZE_LARGE:
3207                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGE));
3208                                         break;
3209                                 case FONT_SIZE_LARGER:
3210                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGER));
3211                                         break;
3212                                 case FONT_SIZE_LARGEST:
3213                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGEST));
3214                                         break;
3215                                 case FONT_SIZE_HUGE:
3216                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGE));
3217                                         break;
3218                                 case FONT_SIZE_HUGER:
3219                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGER));
3220                                         break;
3221                                 case FONT_SIZE_INCREASE:
3222                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_INCREASE));
3223                                         break;
3224                                 case FONT_SIZE_DECREASE:
3225                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_DECREASE));
3226                                         break;
3227                                 case FONT_SIZE_INHERIT:
3228                                 case FONT_SIZE_NORMAL:
3229                                         break;
3230                                 default:
3231                                         // the other tags are for internal use
3232                                         LATTEST(false);
3233                                         break;
3234                                 }
3235                                 size_flag = false;
3236                         }
3237                         switch (curr_size) {
3238                         case FONT_SIZE_TINY:
3239                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_TINY));
3240                                 size_flag = true;
3241                                 break;
3242                         case FONT_SIZE_SCRIPT:
3243                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SCRIPT));
3244                                 size_flag = true;
3245                                 break;
3246                         case FONT_SIZE_FOOTNOTE:
3247                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_FOOTNOTE));
3248                                 size_flag = true;
3249                                 break;
3250                         case FONT_SIZE_SMALL:
3251                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SMALL));
3252                                 size_flag = true;
3253                                 break;
3254                         case FONT_SIZE_LARGE:
3255                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGE));
3256                                 size_flag = true;
3257                                 break;
3258                         case FONT_SIZE_LARGER:
3259                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGER));
3260                                 size_flag = true;
3261                                 break;
3262                         case FONT_SIZE_LARGEST:
3263                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGEST));
3264                                 size_flag = true;
3265                                 break;
3266                         case FONT_SIZE_HUGE:
3267                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGE));
3268                                 size_flag = true;
3269                                 break;
3270                         case FONT_SIZE_HUGER:
3271                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGER));
3272                                 size_flag = true;
3273                                 break;
3274                         case FONT_SIZE_INCREASE:
3275                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_INCREASE));
3276                                 size_flag = true;
3277                                 break;
3278                         case FONT_SIZE_DECREASE:
3279                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_DECREASE));
3280                                 size_flag = true;
3281                                 break;
3282                         case FONT_SIZE_NORMAL:
3283                         case FONT_SIZE_INHERIT:
3284                                 break;
3285                         default:
3286                                 // the other tags are for internal use
3287                                 LATTEST(false);
3288                                 break;
3289                         }
3290                 }
3291
3292                 // FIXME XHTML
3293                 // Other such tags? What about the other text ranges?
3294
3295                 vector<html::EndFontTag>::const_iterator cit = tagsToClose.begin();
3296                 vector<html::EndFontTag>::const_iterator cen = tagsToClose.end();
3297                 for (; cit != cen; ++cit)
3298                         xs << *cit;
3299
3300                 vector<html::FontTag>::const_iterator sit = tagsToOpen.begin();
3301                 vector<html::FontTag>::const_iterator sen = tagsToOpen.end();
3302                 for (; sit != sen; ++sit)
3303                         xs << *sit;
3304
3305                 tagsToClose.clear();
3306                 tagsToOpen.clear();
3307
3308                 Inset const * inset = getInset(i);
3309                 if (inset) {
3310                         if (!runparams.for_toc || inset->isInToc()) {
3311                                 OutputParams np = runparams;
3312                                 np.local_font = &font;
3313                                 // If the paragraph has size 1, then we are in the "special
3314                                 // case" where we do not output the containing paragraph info
3315                                 if (!inset->getLayout().htmlisblock() && size() != 1)
3316                                         np.html_in_par = true;
3317                                 retval += inset->xhtml(xs, np);
3318                         }
3319                 } else {
3320                         char_type c = getUChar(buf.masterBuffer()->params(),
3321                                                runparams, i);
3322                         if (c == ' ' && (style.free_spacing || runparams.free_spacing))
3323                                 xs << XHTMLStream::ESCAPE_NONE << "&nbsp;";
3324                         else
3325                                 xs << c;
3326                 }
3327                 font_old = font.fontInfo();
3328         }
3329
3330         // FIXME XHTML
3331         // I'm worried about what happens if a branch, say, is itself
3332         // wrapped in some font stuff. I think that will not work.
3333         xs.closeFontTags();
3334         if (close_paragraph)
3335                 xs.endDivision();
3336
3337         return retval;
3338 }
3339
3340
3341 bool Paragraph::isHfill(pos_type pos) const
3342 {
3343         Inset const * inset = getInset(pos);
3344         return inset && inset->isHfill();
3345 }
3346
3347
3348 bool Paragraph::isNewline(pos_type pos) const
3349 {
3350         // U+2028 LINE SEPARATOR
3351         // U+2029 PARAGRAPH SEPARATOR
3352         char_type const c = d->text_[pos];
3353         if (c == 0x2028 || c == 0x2029)
3354                 return true;
3355         Inset const * inset = getInset(pos);
3356         return inset && inset->lyxCode() == NEWLINE_CODE;
3357 }
3358
3359
3360 bool Paragraph::isEnvSeparator(pos_type pos) const
3361 {
3362         Inset const * inset = getInset(pos);
3363         return inset && inset->lyxCode() == SEPARATOR_CODE;
3364 }
3365
3366
3367 bool Paragraph::isLineSeparator(pos_type pos) const
3368 {
3369         char_type const c = d->text_[pos];
3370         if (isLineSeparatorChar(c))
3371                 return true;
3372         Inset const * inset = getInset(pos);
3373         return inset && inset->isLineSeparator();
3374 }
3375
3376
3377 bool Paragraph::isWordSeparator(pos_type pos) const
3378 {
3379         if (pos == size())
3380                 return true;
3381         if (Inset const * inset = getInset(pos))
3382                 return !inset->isLetter();
3383         // if we have a hard hyphen (no en- or emdash) or apostrophe
3384         // we pass this to the spell checker
3385         // FIXME: this method is subject to change, visit
3386         // https://bugzilla.mozilla.org/show_bug.cgi?id=355178
3387         // to get an impression how complex this is.
3388         if (isHardHyphenOrApostrophe(pos))
3389                 return false;
3390         char_type const c = d->text_[pos];
3391         // We want to pass the escape chars to the spellchecker
3392         docstring const escape_chars = from_utf8(lyxrc.spellchecker_esc_chars);
3393         return !isLetterChar(c) && !isDigitASCII(c) && !contains(escape_chars, c);
3394 }
3395
3396
3397 bool Paragraph::isHardHyphenOrApostrophe(pos_type pos) const
3398 {
3399         pos_type const psize = size();
3400         if (pos >= psize)
3401                 return false;
3402         char_type const c = d->text_[pos];
3403         if (c != '-' && c != '\'')
3404                 return false;
3405         int nextpos = pos + 1;
3406         int prevpos = pos > 0 ? pos - 1 : 0;
3407         if ((nextpos == psize || isSpace(nextpos))
3408                 && (pos == 0 || isSpace(prevpos)))
3409                 return false;
3410         return true;
3411 }
3412
3413
3414 bool Paragraph::needsCProtection(bool const fragile) const
3415 {
3416         // first check the layout of the paragraph, but only in insets
3417         InsetText const * textinset = inInset().asInsetText();
3418         bool const maintext = textinset
3419                 ? textinset->text().isMainText()
3420                 : false;
3421
3422         if (!maintext && layout().needcprotect) {
3423                 // Environments need cprotection regardless the content
3424                 if (layout().latextype == LATEX_ENVIRONMENT)
3425                         return true;
3426
3427                 // Commands need cprotection if they contain specific chars
3428                 int const nchars_escape = 9;
3429                 static char_type const chars_escape[nchars_escape] = {
3430                         '&', '_', '$', '%', '#', '^', '{', '}', '\\'};
3431
3432                 docstring const pars = asString();
3433                 for (int k = 0; k < nchars_escape; k++) {
3434                         if (contains(pars, chars_escape[k]))
3435                                 return true;
3436                 }
3437         }
3438
3439         // now check whether we have insets that need cprotection
3440         pos_type size = pos_type(d->text_.size());
3441         for (pos_type i = 0; i < size; ++i) {
3442                 if (!isInset(i))
3443                         continue;
3444                 Inset const * ins = getInset(i);
3445                 if (ins->needsCProtection(maintext, fragile))
3446                         return true;
3447                 if (ins->getLayout().latextype() == InsetLayout::ENVIRONMENT)
3448                         // Environments need cprotection regardless the content
3449                         return true;
3450                 // Now check math environments
3451                 InsetMath const * im = getInset(i)->asInsetMath();
3452                 if (!im || im->cell(0).empty())
3453                         continue;
3454                 switch(im->cell(0)[0]->lyxCode()) {
3455                 case MATH_AMSARRAY_CODE:
3456                 case MATH_SUBSTACK_CODE:
3457                 case MATH_ENV_CODE:
3458                 case MATH_XYMATRIX_CODE:
3459                         // these need cprotection
3460                         return true;
3461                 default:
3462                         break;
3463                 }
3464         }
3465
3466         return false;
3467 }
3468
3469
3470 FontSpan const & Paragraph::getSpellRange(pos_type pos) const
3471 {
3472         return d->speller_state_.getRange(pos);
3473 }
3474
3475
3476 bool Paragraph::isChar(pos_type pos) const
3477 {
3478         if (Inset const * inset = getInset(pos))
3479                 return inset->isChar();
3480         char_type const c = d->text_[pos];
3481         return !isLetterChar(c) && !isDigitASCII(c) && !lyx::isSpace(c);
3482 }
3483
3484
3485 bool Paragraph::isSpace(pos_type pos) const
3486 {
3487         if (Inset const * inset = getInset(pos))
3488                 return inset->isSpace();
3489         char_type const c = d->text_[pos];
3490         return lyx::isSpace(c);
3491 }
3492
3493
3494 Language const *
3495 Paragraph::getParLanguage(BufferParams const & bparams) const
3496 {
3497         if (!empty())
3498                 return getFirstFontSettings(bparams).language();
3499         // FIXME: we should check the prev par as well (Lgb)
3500         return bparams.language;
3501 }
3502
3503
3504 bool Paragraph::isRTL(BufferParams const & bparams) const
3505 {
3506         return getParLanguage(bparams)->rightToLeft()
3507                 && !inInset().getLayout().forceLTR();
3508 }
3509
3510
3511 void Paragraph::changeLanguage(BufferParams const & bparams,
3512                                Language const * from, Language const * to)
3513 {
3514         // change language including dummy font change at the end
3515         for (pos_type i = 0; i <= size(); ++i) {
3516                 Font font = getFontSettings(bparams, i);
3517                 if (font.language() == from) {
3518                         font.setLanguage(to);
3519                         setFont(i, font);
3520                         d->requestSpellCheck(i);
3521                 }
3522         }
3523 }
3524
3525
3526 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
3527 {
3528         Language const * doc_language = bparams.language;
3529         for (auto const & f : d->fontlist_)
3530                 if (f.font().language() != ignore_language &&
3531                     f.font().language() != latex_language &&
3532                     f.font().language() != doc_language)
3533                         return true;
3534         return false;
3535 }
3536
3537
3538 void Paragraph::getLanguages(std::set<Language const *> & langs) const
3539 {
3540         for (auto const & f : d->fontlist_) {
3541                 Language const * lang = f.font().language();
3542                 if (lang != ignore_language &&
3543                     lang != latex_language)
3544                         langs.insert(lang);
3545         }
3546 }
3547
3548
3549 docstring Paragraph::asString(int options) const
3550 {
3551         return asString(0, size(), options);
3552 }
3553
3554
3555 docstring Paragraph::asString(pos_type beg, pos_type end, int options, const OutputParams *runparams) const
3556 {
3557         odocstringstream os;
3558
3559         if (beg == 0
3560             && options & AS_STR_LABEL
3561             && !d->params_.labelString().empty())
3562                 os << d->params_.labelString() << ' ';
3563
3564         for (pos_type i = beg; i < end; ++i) {
3565                 if ((options & AS_STR_SKIPDELETE) && isDeleted(i))
3566                         continue;
3567                 char_type const c = d->text_[i];
3568                 if (isPrintable(c) || c == '\t'
3569                     || (c == '\n' && (options & AS_STR_NEWLINES)))
3570                         os.put(c);
3571                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
3572                         if (c == META_INSET && (options & AS_STR_PLAINTEXT)) {
3573                                 LASSERT(runparams != 0, return docstring());
3574                                 getInset(i)->plaintext(os, *runparams);
3575                         } else {
3576                                 getInset(i)->toString(os);
3577                         }
3578                 }
3579         }
3580
3581         return os.str();
3582 }
3583
3584
3585 void Paragraph::forOutliner(docstring & os, size_t const maxlen,
3586                             bool const shorten, bool const label) const
3587 {
3588         size_t tmplen = shorten ? maxlen + 1 : maxlen;
3589         if (label && !labelString().empty())
3590                 os += labelString() + ' ';
3591         if (!layout().isTocCaption())
3592                 return;
3593         for (pos_type i = 0; i < size() && os.length() < tmplen; ++i) {
3594                 if (isDeleted(i))
3595                         continue;
3596                 char_type const c = d->text_[i];
3597                 if (isPrintable(c))
3598                         os += c;
3599                 else if (c == META_INSET)
3600                         getInset(i)->forOutliner(os, tmplen, false);
3601         }
3602         if (shorten)
3603                 Text::shortenForOutliner(os, maxlen);
3604 }
3605
3606
3607 void Paragraph::setInsetOwner(Inset const * inset)
3608 {
3609         d->inset_owner_ = inset;
3610 }
3611
3612
3613 int Paragraph::id() const
3614 {
3615         return d->id_;
3616 }
3617
3618
3619 void Paragraph::setId(int id)
3620 {
3621         d->id_ = id;
3622 }
3623
3624
3625 Layout const & Paragraph::layout() const
3626 {
3627         return *d->layout_;
3628 }
3629
3630
3631 void Paragraph::setLayout(Layout const & layout)
3632 {
3633         d->layout_ = &layout;
3634 }
3635
3636
3637 void Paragraph::setDefaultLayout(DocumentClass const & tc)
3638 {
3639         setLayout(tc.defaultLayout());
3640 }
3641
3642
3643 void Paragraph::setPlainLayout(DocumentClass const & tc)
3644 {
3645         setLayout(tc.plainLayout());
3646 }
3647
3648
3649 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
3650 {
3651         if (usePlainLayout())
3652                 setPlainLayout(tclass);
3653         else
3654                 setDefaultLayout(tclass);
3655 }
3656
3657
3658 Inset const & Paragraph::inInset() const
3659 {
3660         LBUFERR(d->inset_owner_);
3661         return *d->inset_owner_;
3662 }
3663
3664
3665 ParagraphParameters & Paragraph::params()
3666 {
3667         return d->params_;
3668 }
3669
3670
3671 ParagraphParameters const & Paragraph::params() const
3672 {
3673         return d->params_;
3674 }
3675
3676
3677 bool Paragraph::isFreeSpacing() const
3678 {
3679         if (d->layout_->free_spacing)
3680                 return true;
3681         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
3682 }
3683
3684
3685 bool Paragraph::allowEmpty() const
3686 {
3687         if (d->layout_->keepempty)
3688                 return true;
3689         return d->inset_owner_ && d->inset_owner_->allowEmpty();
3690 }
3691
3692
3693 bool Paragraph::brokenBiblio() const
3694 {
3695         // There is a problem if there is no bibitem at position 0 in
3696         // paragraphs that need one, if there is another bibitem in the
3697         // paragraph or if this paragraph is not supposed to have
3698         // a bibitem inset at all.
3699         return ((d->layout_->labeltype == LABEL_BIBLIO
3700                 && (d->insetlist_.find(BIBITEM_CODE) != 0
3701                     || d->insetlist_.find(BIBITEM_CODE, 1) > 0))
3702                 || (d->layout_->labeltype != LABEL_BIBLIO
3703                     && d->insetlist_.find(BIBITEM_CODE) != -1));
3704 }
3705
3706
3707 int Paragraph::fixBiblio(Buffer const & buffer)
3708 {
3709         // FIXME: when there was already an inset at 0, the return value is 1,
3710         // which does not tell whether another inset has been remove; the
3711         // cursor cannot be correctly updated.
3712
3713         bool const track_changes = buffer.params().track_changes;
3714         int bibitem_pos = d->insetlist_.find(BIBITEM_CODE);
3715
3716         // The case where paragraph is not BIBLIO
3717         if (d->layout_->labeltype != LABEL_BIBLIO) {
3718                 if (bibitem_pos == -1)
3719                         // No InsetBibitem => OK
3720                         return 0;
3721                 // There is an InsetBibitem: remove it!
3722                 d->insetlist_.release(bibitem_pos);
3723                 eraseChar(bibitem_pos, track_changes);
3724                 return (bibitem_pos == 0) ? -1 : -bibitem_pos;
3725         }
3726
3727         bool const hasbibitem0 = bibitem_pos == 0;
3728         if (hasbibitem0) {
3729                 bibitem_pos = d->insetlist_.find(BIBITEM_CODE, 1);
3730                 // There was an InsetBibitem at pos 0,
3731                 // and no other one => OK
3732                 if (bibitem_pos == -1)
3733                         return 0;
3734                 // there is a bibitem at the 0 position, but since
3735                 // there is a second one, we copy the second on the
3736                 // first. We're assuming there are at most two of
3737                 // these, which there should be.
3738                 // FIXME: why does it make sense to do that rather
3739                 // than keep the first? (JMarc)
3740                 Inset * inset = releaseInset(bibitem_pos);
3741                 d->insetlist_.begin()->inset = inset;
3742                 return -bibitem_pos;
3743         }
3744
3745         // We need to create an inset at the beginning
3746         Inset * inset = nullptr;
3747         if (bibitem_pos > 0) {
3748                 // there was one somewhere in the paragraph, let's move it
3749                 inset = d->insetlist_.release(bibitem_pos);
3750                 eraseChar(bibitem_pos, track_changes);
3751         } else
3752                 // make a fresh one
3753                 inset = new InsetBibitem(const_cast<Buffer *>(&buffer),
3754                                          InsetCommandParams(BIBITEM_CODE));
3755
3756         Font font(inherit_font, buffer.params().language);
3757         insertInset(0, inset, font, Change(track_changes ? Change::INSERTED
3758                                                    : Change::UNCHANGED));
3759
3760         // This is needed to get the counters right
3761         buffer.updateBuffer();
3762         return 1;
3763 }
3764
3765
3766 void Paragraph::checkAuthors(AuthorList const & authorList)
3767 {
3768         d->changes_.checkAuthors(authorList);
3769 }
3770
3771
3772 bool Paragraph::isChanged(pos_type pos) const
3773 {
3774         return lookupChange(pos).changed();
3775 }
3776
3777
3778 bool Paragraph::isInserted(pos_type pos) const
3779 {
3780         return lookupChange(pos).inserted();
3781 }
3782
3783
3784 bool Paragraph::isDeleted(pos_type pos) const
3785 {
3786         return lookupChange(pos).deleted();
3787 }
3788
3789
3790 InsetList const & Paragraph::insetList() const
3791 {
3792         return d->insetlist_;
3793 }
3794
3795
3796 void Paragraph::setInsetBuffers(Buffer & b)
3797 {
3798         d->insetlist_.setBuffer(b);
3799 }
3800
3801
3802 void Paragraph::resetBuffer()
3803 {
3804         d->insetlist_.resetBuffer();
3805 }
3806
3807
3808 Inset * Paragraph::releaseInset(pos_type pos)
3809 {
3810         Inset * inset = d->insetlist_.release(pos);
3811         /// does not honour change tracking!
3812         eraseChar(pos, false);
3813         return inset;
3814 }
3815
3816
3817 Inset * Paragraph::getInset(pos_type pos)
3818 {
3819         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3820                  ? d->insetlist_.get(pos) : 0;
3821 }
3822
3823
3824 Inset const * Paragraph::getInset(pos_type pos) const
3825 {
3826         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3827                  ? d->insetlist_.get(pos) : 0;
3828 }
3829
3830
3831 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
3832                 pos_type & right, TextCase action)
3833 {
3834         // process sequences of modified characters; in change
3835         // tracking mode, this approach results in much better
3836         // usability than changing case on a char-by-char basis
3837         // We also need to track the current font, since font
3838         // changes within sequences can occur.
3839         vector<pair<char_type, Font> > changes;
3840
3841         bool const trackChanges = bparams.track_changes;
3842
3843         bool capitalize = true;
3844
3845         for (; pos < right; ++pos) {
3846                 char_type oldChar = d->text_[pos];
3847                 char_type newChar = oldChar;
3848
3849                 // ignore insets and don't play with deleted text!
3850                 if (oldChar != META_INSET && !isDeleted(pos)) {
3851                         switch (action) {
3852                                 case text_lowercase:
3853                                         newChar = lowercase(oldChar);
3854                                         break;
3855                                 case text_capitalization:
3856                                         if (capitalize) {
3857                                                 newChar = uppercase(oldChar);
3858                                                 capitalize = false;
3859                                         }
3860                                         break;
3861                                 case text_uppercase:
3862                                         newChar = uppercase(oldChar);
3863                                         break;
3864                         }
3865                 }
3866
3867                 if (isWordSeparator(pos) || isDeleted(pos)) {
3868                         // permit capitalization again
3869                         capitalize = true;
3870                 }
3871
3872                 if (oldChar != newChar) {
3873                         changes.push_back(make_pair(newChar, getFontSettings(bparams, pos)));
3874                         if (pos != right - 1)
3875                                 continue;
3876                         // step behind the changing area
3877                         pos++;
3878                 }
3879
3880                 int erasePos = pos - changes.size();
3881                 for (size_t i = 0; i < changes.size(); i++) {
3882                         insertChar(pos, changes[i].first,
3883                                    changes[i].second,
3884                                    trackChanges);
3885                         if (!eraseChar(erasePos, trackChanges)) {
3886                                 ++erasePos;
3887                                 ++pos; // advance
3888                                 ++right; // expand selection
3889                         }
3890                 }
3891                 changes.clear();
3892         }
3893 }
3894
3895
3896 int Paragraph::find(docstring const & str, bool cs, bool mw,
3897                 pos_type start_pos, bool del) const
3898 {
3899         pos_type pos = start_pos;
3900         int const strsize = str.length();
3901         int i = 0;
3902         pos_type const parsize = d->text_.size();
3903         for (i = 0; i < strsize && pos < parsize; ++i, ++pos) {
3904                 // Ignore "invisible" letters such as ligature breaks
3905                 // and hyphenation chars while searching
3906                 while (pos < parsize - 1 && isInset(pos)) {
3907                         odocstringstream os;
3908                         getInset(pos)->toString(os);
3909                         if (!getInset(pos)->isLetter() || !os.str().empty())
3910                                 break;
3911                         pos++;
3912                 }
3913                 if (cs && str[i] != d->text_[pos])
3914                         break;
3915                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos]))
3916                         break;
3917                 if (!del && isDeleted(pos))
3918                         break;
3919         }
3920
3921         if (i != strsize)
3922                 return 0;
3923
3924         // if necessary, check whether string matches word
3925         if (mw) {
3926                 if (start_pos > 0 && !isWordSeparator(start_pos - 1))
3927                         return 0;
3928                 if (pos < parsize
3929                         && !isWordSeparator(pos))
3930                         return 0;
3931         }
3932
3933         return pos - start_pos;
3934 }
3935
3936
3937 char_type Paragraph::getChar(pos_type pos) const
3938 {
3939         return d->text_[pos];
3940 }
3941
3942
3943 pos_type Paragraph::size() const
3944 {
3945         return d->text_.size();
3946 }
3947
3948
3949 bool Paragraph::empty() const
3950 {
3951         return d->text_.empty();
3952 }
3953
3954
3955 bool Paragraph::isInset(pos_type pos) const
3956 {
3957         return d->text_[pos] == META_INSET;
3958 }
3959
3960
3961 bool Paragraph::isSeparator(pos_type pos) const
3962 {
3963         //FIXME: Are we sure this can be the only separator?
3964         return d->text_[pos] == ' ';
3965 }
3966
3967
3968 void Paragraph::deregisterWords()
3969 {
3970         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3971         Private::LangWordsMap::const_iterator ite = d->words_.end();
3972         for (; itl != ite; ++itl) {
3973                 WordList & wl = theWordList(itl->first);
3974                 Private::Words::const_iterator it = (itl->second).begin();
3975                 Private::Words::const_iterator et = (itl->second).end();
3976                 for (; it != et; ++it)
3977                         wl.remove(*it);
3978         }
3979         d->words_.clear();
3980 }
3981
3982
3983 void Paragraph::locateWord(pos_type & from, pos_type & to,
3984         word_location const loc) const
3985 {
3986         switch (loc) {
3987         case WHOLE_WORD_STRICT:
3988                 if (from == 0 || from == size()
3989                     || isWordSeparator(from)
3990                     || isWordSeparator(from - 1)) {
3991                         to = from;
3992                         return;
3993                 }
3994                 // fall through
3995
3996         case WHOLE_WORD:
3997                 // If we are already at the beginning of a word, do nothing
3998                 if (!from || isWordSeparator(from - 1))
3999                         break;
4000                 // fall through
4001
4002         case PREVIOUS_WORD:
4003                 // always move the cursor to the beginning of previous word
4004                 while (from && !isWordSeparator(from - 1))
4005                         --from;
4006                 break;
4007         case NEXT_WORD:
4008                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
4009                 break;
4010         case PARTIAL_WORD:
4011                 // no need to move the 'from' cursor
4012                 break;
4013         }
4014         to = from;
4015         while (to < size() && !isWordSeparator(to))
4016                 ++to;
4017 }
4018
4019
4020 void Paragraph::collectWords()
4021 {
4022         for (pos_type pos = 0; pos < size(); ++pos) {
4023                 if (isWordSeparator(pos))
4024                         continue;
4025                 pos_type from = pos;
4026                 locateWord(from, pos, WHOLE_WORD);
4027                 // Work around MSVC warning: The statement
4028                 // if (pos < from + lyxrc.completion_minlength)
4029                 // triggers a signed vs. unsigned warning.
4030                 // I don't know why this happens, it could be a MSVC bug, or
4031                 // related to LLP64 (windows) vs. LP64 (unix) programming
4032                 // model, or the C++ standard might be ambigous in the section
4033                 // defining the "usual arithmetic conversions". However, using
4034                 // a temporary variable is safe and works on all compilers.
4035                 pos_type const endpos = from + lyxrc.completion_minlength;
4036                 if (pos < endpos)
4037                         continue;
4038                 FontList::const_iterator cit = d->fontlist_.fontIterator(from);
4039                 if (cit == d->fontlist_.end())
4040                         return;
4041                 Language const * lang = cit->font().language();
4042                 docstring const word = asString(from, pos, AS_STR_NONE);
4043                 d->words_[lang->lang()].insert(word);
4044         }
4045 }
4046
4047
4048 void Paragraph::registerWords()
4049 {
4050         Private::LangWordsMap::const_iterator itl = d->words_.begin();
4051         Private::LangWordsMap::const_iterator ite = d->words_.end();
4052         for (; itl != ite; ++itl) {
4053                 WordList & wl = theWordList(itl->first);
4054                 Private::Words::const_iterator it = (itl->second).begin();
4055                 Private::Words::const_iterator et = (itl->second).end();
4056                 for (; it != et; ++it)
4057                         wl.insert(*it);
4058         }
4059 }
4060
4061
4062 void Paragraph::updateWords()
4063 {
4064         deregisterWords();
4065         collectWords();
4066         registerWords();
4067 }
4068
4069
4070 void Paragraph::Private::appendSkipPosition(SkipPositions & skips, pos_type const pos) const
4071 {
4072         SkipPositionsIterator begin = skips.begin();
4073         SkipPositions::iterator end = skips.end();
4074         if (pos > 0 && begin < end) {
4075                 --end;
4076                 if (end->last == pos - 1) {
4077                         end->last = pos;
4078                         return;
4079                 }
4080         }
4081         skips.insert(end, FontSpan(pos, pos));
4082 }
4083
4084
4085 Language * Paragraph::Private::locateSpellRange(
4086         pos_type & from, pos_type & to,
4087         SkipPositions & skips) const
4088 {
4089         // skip leading white space
4090         while (from < to && owner_->isWordSeparator(from))
4091                 ++from;
4092         // don't check empty range
4093         if (from >= to)
4094                 return 0;
4095         // get current language
4096         Language * lang = getSpellLanguage(from);
4097         pos_type last = from;
4098         bool samelang = true;
4099         bool sameinset = true;
4100         while (last < to && samelang && sameinset) {
4101                 // hop to end of word
4102                 while (last < to && !owner_->isWordSeparator(last)) {
4103                         if (owner_->getInset(last)) {
4104                                 appendSkipPosition(skips, last);
4105                         } else if (owner_->isDeleted(last)) {
4106                                 appendSkipPosition(skips, last);
4107                         }
4108                         ++last;
4109                 }
4110                 // hop to next word while checking for insets
4111                 while (sameinset && last < to && owner_->isWordSeparator(last)) {
4112                         if (Inset const * inset = owner_->getInset(last))
4113                                 sameinset = inset->isChar() && inset->isLetter();
4114                         if (sameinset && owner_->isDeleted(last)) {
4115                                 appendSkipPosition(skips, last);
4116                         }
4117                         if (sameinset)
4118                                 last++;
4119                 }
4120                 if (sameinset && last < to) {
4121                         // now check for language change
4122                         samelang = lang == getSpellLanguage(last);
4123                 }
4124         }
4125         // if language change detected backstep is needed
4126         if (!samelang)
4127                 --last;
4128         to = last;
4129         return lang;
4130 }
4131
4132
4133 Language * Paragraph::Private::getSpellLanguage(pos_type const from) const
4134 {
4135         Language * lang =
4136                 const_cast<Language *>(owner_->getFontSettings(
4137                         inset_owner_->buffer().params(), from).language());
4138         if (lang == inset_owner_->buffer().params().language
4139                 && !lyxrc.spellchecker_alt_lang.empty()) {
4140                 string lang_code;
4141                 string const lang_variety =
4142                         split(lyxrc.spellchecker_alt_lang, lang_code, '-');
4143                 lang->setCode(lang_code);
4144                 lang->setVariety(lang_variety);
4145         }
4146         return lang;
4147 }
4148
4149
4150 void Paragraph::requestSpellCheck(pos_type pos)
4151 {
4152         d->requestSpellCheck(pos);
4153 }
4154
4155
4156 bool Paragraph::needsSpellCheck() const
4157 {
4158         SpellChecker::ChangeNumber speller_change_number = 0;
4159         if (theSpellChecker())
4160                 speller_change_number = theSpellChecker()->changeNumber();
4161         if (speller_change_number > d->speller_state_.currentChangeNumber()) {
4162                 d->speller_state_.needsCompleteRefresh(speller_change_number);
4163         }
4164         return d->needsSpellCheck();
4165 }
4166
4167
4168 bool Paragraph::Private::ignoreWord(docstring const & word) const
4169 {
4170         // Ignore words with digits
4171         // FIXME: make this customizable
4172         // (note that some checkers ignore words with digits by default)
4173         docstring::const_iterator cit = word.begin();
4174         docstring::const_iterator const end = word.end();
4175         for (; cit != end; ++cit) {
4176                 if (isNumber((*cit)))
4177                         return true;
4178         }
4179         return false;
4180 }
4181
4182
4183 SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
4184         WordLangTuple & wl, docstring_list & suggestions,
4185         bool do_suggestion, bool check_learned) const
4186 {
4187         SpellChecker::Result result = SpellChecker::WORD_OK;
4188         SpellChecker * speller = theSpellChecker();
4189         if (!speller)
4190                 return result;
4191
4192         if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
4193                 return result;
4194
4195         locateWord(from, to, WHOLE_WORD);
4196         if (from == to || from >= size())
4197                 return result;
4198
4199         docstring word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
4200         Language * lang = d->getSpellLanguage(from);
4201
4202         if (getFontSettings(d->inset_owner_->buffer().params(), from).fontInfo().nospellcheck() == FONT_ON)
4203                 return result;
4204
4205         wl = WordLangTuple(word, lang);
4206
4207         if (word.empty())
4208                 return result;
4209
4210         if (needsSpellCheck() || check_learned) {
4211                 pos_type end = to;
4212                 if (!d->ignoreWord(word)) {
4213                         bool const trailing_dot = to < size() && d->text_[to] == '.';
4214                         result = speller->check(wl);
4215                         if (SpellChecker::misspelled(result) && trailing_dot) {
4216                                 wl = WordLangTuple(word.append(from_ascii(".")), lang);
4217                                 result = speller->check(wl);
4218                                 if (!SpellChecker::misspelled(result)) {
4219                                         LYXERR(Debug::GUI, "misspelled word is correct with dot: \"" <<
4220                                            word << "\" [" <<
4221                                            from << ".." << to << "]");
4222                                 } else {
4223                                         // spell check with dot appended failed too
4224                                         // restore original word/lang value
4225                                         word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
4226                                         wl = WordLangTuple(word, lang);
4227                                 }
4228                         }
4229                 }
4230                 if (!SpellChecker::misspelled(result)) {
4231                         // area up to the begin of the next word is not misspelled
4232                         while (end < size() && isWordSeparator(end))
4233                                 ++end;
4234                 }
4235                 d->setMisspelled(from, end, result);
4236         } else {
4237                 result = d->speller_state_.getState(from);
4238         }
4239
4240         if (do_suggestion)
4241                 suggestions.clear();
4242
4243         if (SpellChecker::misspelled(result)) {
4244                 LYXERR(Debug::GUI, "misspelled word: \"" <<
4245                            word << "\" [" <<
4246                            from << ".." << to << "]");
4247                 if (do_suggestion)
4248                         speller->suggest(wl, suggestions);
4249         }
4250         return result;
4251 }
4252
4253
4254 void Paragraph::anonymize()
4255 {
4256         // This is a very crude anonymization for now
4257         for (char_type & c : d->text_)
4258                 if (isLetterChar(c) || isNumber(c))
4259                         c = 'a';
4260 }
4261
4262
4263 void Paragraph::Private::markMisspelledWords(
4264         pos_type const & first, pos_type const & last,
4265         SpellChecker::Result result,
4266         docstring const & word,
4267         SkipPositions const & skips)
4268 {
4269         if (!SpellChecker::misspelled(result)) {
4270                 setMisspelled(first, last, SpellChecker::WORD_OK);
4271                 return;
4272         }
4273         int snext = first;
4274         SpellChecker * speller = theSpellChecker();
4275         // locate and enumerate the error positions
4276         int nerrors = speller->numMisspelledWords();
4277         int numskipped = 0;
4278         SkipPositionsIterator it = skips.begin();
4279         SkipPositionsIterator et = skips.end();
4280         for (int index = 0; index < nerrors; ++index) {
4281                 int wstart;
4282                 int wlen = 0;
4283                 speller->misspelledWord(index, wstart, wlen);
4284                 /// should not happen if speller supports range checks
4285                 if (!wlen) continue;
4286                 docstring const misspelled = word.substr(wstart, wlen);
4287                 wstart += first + numskipped;
4288                 if (snext < wstart) {
4289                         /// mark the range of correct spelling
4290                         numskipped += countSkips(it, et, wstart);
4291                         setMisspelled(snext,
4292                                 wstart - 1, SpellChecker::WORD_OK);
4293                 }
4294                 snext = wstart + wlen;
4295                 numskipped += countSkips(it, et, snext);
4296                 /// mark the range of misspelling
4297                 setMisspelled(wstart, snext, result);
4298                 LYXERR(Debug::GUI, "misspelled word: \"" <<
4299                            misspelled << "\" [" <<
4300                            wstart << ".." << (snext-1) << "]");
4301                 ++snext;
4302         }
4303         if (snext <= last) {
4304                 /// mark the range of correct spelling at end
4305                 setMisspelled(snext, last, SpellChecker::WORD_OK);
4306         }
4307 }
4308
4309
4310 void Paragraph::spellCheck() const
4311 {
4312         SpellChecker * speller = theSpellChecker();
4313         if (!speller || empty() ||!needsSpellCheck())
4314                 return;
4315         pos_type start;
4316         pos_type endpos;
4317         d->rangeOfSpellCheck(start, endpos);
4318         if (speller->canCheckParagraph()) {
4319                 // loop until we leave the range
4320                 for (pos_type first = start; first < endpos; ) {
4321                         pos_type last = endpos;
4322                         Private::SkipPositions skips;
4323                         Language * lang = d->locateSpellRange(first, last, skips);
4324                         if (first >= endpos)
4325                                 break;
4326                         // start the spell checker on the unit of meaning
4327                         docstring word = asString(first, last, AS_STR_INSETS + AS_STR_SKIPDELETE);
4328                         WordLangTuple wl = WordLangTuple(word, lang);
4329                         SpellChecker::Result result = word.size() ?
4330                                 speller->check(wl) : SpellChecker::WORD_OK;
4331                         d->markMisspelledWords(first, last, result, word, skips);
4332                         first = ++last;
4333                 }
4334         } else {
4335                 static docstring_list suggestions;
4336                 pos_type to = endpos;
4337                 while (start < endpos) {
4338                         WordLangTuple wl;
4339                         spellCheck(start, to, wl, suggestions, false);
4340                         start = to + 1;
4341                 }
4342         }
4343         d->readySpellCheck();
4344 }
4345
4346
4347 bool Paragraph::isMisspelled(pos_type pos, bool check_boundary) const
4348 {
4349         bool result = SpellChecker::misspelled(d->speller_state_.getState(pos));
4350         if (result || pos <= 0 || pos > size())
4351                 return result;
4352         if (check_boundary && (pos == size() || isWordSeparator(pos)))
4353                 result = SpellChecker::misspelled(d->speller_state_.getState(pos - 1));
4354         return result;
4355 }
4356
4357
4358 string Paragraph::magicLabel() const
4359 {
4360         stringstream ss;
4361         ss << "magicparlabel-" << id();
4362         return ss.str();
4363 }
4364
4365
4366 } // namespace lyx