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 Change Paragraph::parEndChange() const
 625 {
 626         return d->changes_.lookup(size());
 627 }
 628
 629
 630 void Paragraph::setChange(Change const & change)
 631 {
 632         // beware of the imaginary end-of-par character!
 633         d->changes_.set(change, 0, size() + 1);
 634
 635         /*
 636          * Propagate the change recursively - but not in case of DELETED!
 637          *
 638          * Imagine that your co-author makes changes in an existing inset. He
 639          * sends your document to you and you come to the conclusion that the
 640          * inset should go completely. If you erase it, LyX must not delete all
 641          * text within the inset. Otherwise, the change tracked insertions of
 642          * your co-author get lost and there is no way to restore them later.
 643          *
 644          * Conclusion: An inset's content should remain untouched if you delete it
 645          */
 646
 647         if (!change.deleted()) {
 648                 for (pos_type pos = 0; pos < size(); ++pos) {
 649                         if (Inset * inset = getInset(pos))
 650                                 inset->setChange(change);
 651                 }
 652         }
 653 }
 654
 655
 656 void Paragraph::setChange(pos_type pos, Change const & change)
 657 {
 658         LASSERT(pos >= 0 && pos <= size(), return);
 659         d->changes_.set(change, pos);
 660
 661         // see comment in setChange(Change const &) above
 662         if (!change.deleted() && pos < size())
 663                         if (Inset * inset = getInset(pos))
 664                                 inset->setChange(change);
 665 }
 666
 667
 668 Change const & Paragraph::lookupChange(pos_type pos) const
 669 {
 670         LBUFERR(pos >= 0 && pos <= size());
 671         return d->changes_.lookup(pos);
 672 }
 673
 674
 675 void Paragraph::acceptChanges(pos_type start, pos_type end)
 676 {
 677         LASSERT(start >= 0 && start <= size(), return);
 678         LASSERT(end > start && end <= size() + 1, return);
 679
 680         for (pos_type pos = start; pos < end; ++pos) {
 681                 switch (lookupChange(pos).type) {
 682                         case Change::UNCHANGED:
 683                                 // accept changes in nested inset
 684                                 if (Inset * inset = getInset(pos))
 685                                         inset->acceptChanges();
 686                                 break;
 687
 688                         case Change::INSERTED:
 689                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 690                                 // also accept changes in nested inset
 691                                 if (Inset * inset = getInset(pos))
 692                                         inset->acceptChanges();
 693                                 break;
 694
 695                         case Change::DELETED:
 696                                 // Suppress access to non-existent
 697                                 // "end-of-paragraph char"
 698                                 if (pos < size()) {
 699                                         eraseChar(pos, false);
 700                                         --end;
 701                                         --pos;
 702                                 }
 703                                 break;
 704                 }
 705
 706         }
 707 }
 708
 709
 710 void Paragraph::rejectChanges(pos_type start, pos_type end)
 711 {
 712         LASSERT(start >= 0 && start <= size(), return);
 713         LASSERT(end > start && end <= size() + 1, return);
 714
 715         for (pos_type pos = start; pos < end; ++pos) {
 716                 switch (lookupChange(pos).type) {
 717                         case Change::UNCHANGED:
 718                                 // reject changes in nested inset
 719                                 if (Inset * inset = getInset(pos))
 720                                                 inset->rejectChanges();
 721                                 break;
 722
 723                         case Change::INSERTED:
 724                                 // Suppress access to non-existent
 725                                 // "end-of-paragraph char"
 726                                 if (pos < size()) {
 727                                         eraseChar(pos, false);
 728                                         --end;
 729                                         --pos;
 730                                 }
 731                                 break;
 732
 733                         case Change::DELETED:
 734                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 735
 736                                 // Do NOT reject changes within a deleted inset!
 737                                 // There may be insertions of a co-author inside of it!
 738
 739                                 break;
 740                 }
 741         }
 742 }
 743
 744
 745 void Paragraph::Private::insertChar(pos_type pos, char_type c,
 746                 Change const & change)
 747 {
 748         LASSERT(pos >= 0 && pos <= int(text_.size()), return);
 749
 750         // track change
 751         changes_.insert(change, pos);
 752
 753         // This is actually very common when parsing buffers (and
 754         // maybe inserting ascii text)
 755         if (pos == pos_type(text_.size())) {
 756                 // when appending characters, no need to update tables
 757                 text_.push_back(c);
 758                 // but we want spell checking
 759                 requestSpellCheck(pos);
 760                 return;
 761         }
 762
 763         text_.insert(text_.begin() + pos, c);
 764
 765         // Update the font table.
 766         fontlist_.increasePosAfterPos(pos);
 767
 768         // Update the insets
 769         insetlist_.increasePosAfterPos(pos);
 770
 771         // Update list of misspelled positions
 772         speller_state_.increasePosAfterPos(pos);
 773 }
 774
 775
 776 bool Paragraph::insertInset(pos_type pos, Inset * inset,
 777                                    Font const & font, Change const & change)
 778 {
 779         LASSERT(inset, return false);
 780         LASSERT(pos >= 0 && pos <= size(), return false);
 781
 782         // Paragraph::insertInset() can be used in cut/copy/paste operation where
 783         // d->inset_owner_ is not set yet.
 784         if (d->inset_owner_ && !d->inset_owner_->insetAllowed(inset->lyxCode()))
 785                 return false;
 786
 787         d->insertChar(pos, META_INSET, change);
 788         LASSERT(d->text_[pos] == META_INSET, return false);
 789
 790         // Add a new entry in the insetlist_.
 791         d->insetlist_.insert(inset, pos);
 792
 793         // Some insets require run of spell checker
 794         requestSpellCheck(pos);
 795         setFont(pos, font);
 796         return true;
 797 }
 798
 799
 800 bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
 801 {
 802         LASSERT(pos >= 0 && pos <= size(), return false);
 803
 804         // keep the logic here in sync with the logic of isMergedOnEndOfParDeletion()
 805
 806         if (trackChanges) {
 807                 Change change = d->changes_.lookup(pos);
 808
 809                 // set the character to DELETED if
 810                 //  a) it was previously unchanged or
 811                 //  b) it was inserted by a co-author
 812
 813                 if (!change.changed() ||
 814                       (change.inserted() && !change.currentAuthor())) {
 815                         setChange(pos, Change(Change::DELETED));
 816                         // request run of spell checker
 817                         requestSpellCheck(pos);
 818                         return false;
 819                 }
 820
 821                 if (change.deleted())
 822                         return false;
 823         }
 824
 825         // Don't physically access the imaginary end-of-paragraph character.
 826         // eraseChar() can only mark it as DELETED. A physical deletion of
 827         // end-of-par must be handled externally.
 828         if (pos == size()) {
 829                 return false;
 830         }
 831
 832         // track change
 833         d->changes_.erase(pos);
 834
 835         // if it is an inset, delete the inset entry
 836         if (d->text_[pos] == META_INSET)
 837                 d->insetlist_.erase(pos);
 838
 839         d->text_.erase(d->text_.begin() + pos);
 840
 841         // Update the fontlist_
 842         d->fontlist_.erase(pos);
 843
 844         // Update the insetlist_
 845         d->insetlist_.decreasePosAfterPos(pos);
 846
 847         // Update list of misspelled positions
 848         d->speller_state_.decreasePosAfterPos(pos);
 849         d->speller_state_.refreshLast(size());
 850
 851         return true;
 852 }
 853
 854
 855 int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
 856 {
 857         LASSERT(start >= 0 && start <= size(), return 0);
 858         LASSERT(end >= start && end <= size() + 1, return 0);
 859
 860         pos_type i = start;
 861         for (pos_type count = end - start; count; --count) {
 862                 if (!eraseChar(i, trackChanges))
 863                         ++i;
 864         }
 865         return end - i;
 866 }
 867
 868 // Handle combining characters
 869 int Paragraph::Private::latexSurrogatePair(BufferParams const & bparams,
 870                 otexstream & os, char_type c, char_type next,
 871                 OutputParams const & runparams)
 872 {
 873         // Writing next here may circumvent a possible font change between
 874         // c and next. Since next is only output if it forms a surrogate pair
 875         // with c we can ignore this:
 876         // A font change inside a surrogate pair does not make sense and is
 877         // hopefully impossible to input.
 878         // FIXME: change tracking
 879         // Is this correct WRT change tracking?
 880         Encoding const & encoding = *(runparams.encoding);
 881         docstring latex1 = encoding.latexChar(next).first;
 882         if (runparams.inIPA) {
 883                 string const tipashortcut = Encodings::TIPAShortcut(next);
 884                 if (!tipashortcut.empty()) {
 885                         latex1 = from_ascii(tipashortcut);
 886                 }
 887         }
 888         docstring latex2 = encoding.latexChar(c).first;
 889
 890         if (bparams.useNonTeXFonts || docstring(1, next) == latex1) {
 891                 // Encoding supports the combination:
 892                 // output as is (combining char after base char).
 893                 os << latex2 << latex1;
 894                 return latex1.length() + latex2.length();
 895         }
 896
 897         os << latex1 << "{" << latex2 << "}";
 898         return latex1.length() + latex2.length() + 2;
 899 }
 900
 901
 902 bool Paragraph::Private::simpleTeXBlanks(BufferParams const & bparams,
 903                                        OutputParams const & runparams,
 904                                        otexstream & os,
 905                                        pos_type i,
 906                                        unsigned int & column,
 907                                        Font const & font,
 908                                        Layout const & style)
 909 {
 910         if (style.pass_thru || runparams.pass_thru)
 911                 return false;
 912
 913         if (i + 1 < int(text_.size())) {
 914                 char_type next = text_[i + 1];
 915                 if (Encodings::isCombiningChar(next)) {
 916                         // This space has an accent, so we must always output it.
 917                         column += latexSurrogatePair(bparams, os, ' ', next, runparams) - 1;
 918                         return true;
 919                 }
 920         }
 921
 922         if (runparams.linelen > 0
 923             && column > runparams.linelen
 924             && i
 925             && text_[i - 1] != ' '
 926             && (i + 1 < int(text_.size()))
 927             // same in FreeSpacing mode
 928             && !owner_->isFreeSpacing()
 929             // In typewriter mode, we want to avoid
 930             // ! . ? : at the end of a line
 931             && !(font.fontInfo().family() == TYPEWRITER_FAMILY
 932                  && (text_[i - 1] == '.'
 933                      || text_[i - 1] == '?'
 934                      || text_[i - 1] == ':'
 935                      || text_[i - 1] == '!'))) {
 936                 os << '\n';
 937                 os.texrow().start(owner_->id(), i + 1);
 938                 column = 0;
 939         } else if (style.free_spacing) {
 940                 os << '~';
 941         } else {
 942                 os << ' ';
 943         }
 944         return false;
 945 }
 946
 947
 948 void Paragraph::Private::latexInset(BufferParams const & bparams,
 949                                     otexstream & os,
 950                                     OutputParams & runparams,
 951                                     Font & running_font,
 952                                     Font & basefont,
 953                                     Font const & outerfont,
 954                                     bool & open_font,
 955                                     Change & running_change,
 956                                     Layout const & style,
 957                                     pos_type & i,
 958                                     unsigned int & column)
 959 {
 960         Inset * inset = owner_->getInset(i);
 961         LBUFERR(inset);
 962
 963         if (style.pass_thru) {
 964                 odocstringstream ods;
 965                 inset->plaintext(ods, runparams);
 966                 os << ods.str();
 967                 return;
 968         }
 969
 970         // FIXME: move this to InsetNewline::latex
 971         if (inset->lyxCode() == NEWLINE_CODE || inset->lyxCode() == SEPARATOR_CODE) {
 972                 // newlines are handled differently here than
 973                 // the default in simpleTeXSpecialChars().
 974                 if (!style.newline_allowed) {
 975                         os << '\n';
 976                 } else {
 977                         if (open_font) {
 978                                 bool needPar = false;
 979                                 column += running_font.latexWriteEndChanges(
 980                                         os, bparams, runparams,
 981                                         basefont, basefont, needPar);
 982                                 open_font = false;
 983                         }
 984
 985                         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY)
 986                                 os << '~';
 987
 988                         basefont = owner_->getLayoutFont(bparams, outerfont);
 989                         running_font = basefont;
 990
 991                         if (runparams.moving_arg)
 992                                 os << "\\protect ";
 993
 994                 }
 995                 os.texrow().start(owner_->id(), i + 1);
 996                 column = 0;
 997         }
 998
 999         if (owner_->isDeleted(i)) {
1000                 if( ++runparams.inDeletedInset == 1)
1001                         runparams.changeOfDeletedInset = owner_->lookupChange(i);
1002         }
1003
1004         if (inset->canTrackChanges()) {
1005                 column += Changes::latexMarkChange(os, bparams, running_change,
1006                         Change(Change::UNCHANGED), runparams);
1007                 running_change = Change(Change::UNCHANGED);
1008         }
1009
1010         bool close = false;
1011         odocstream::pos_type const len = os.os().tellp();
1012
1013         if (inset->forceLTR(runparams)
1014             && running_font.isRightToLeft()
1015             // ERT is an exception, it should be output with no
1016             // decorations at all
1017             && inset->lyxCode() != ERT_CODE) {
1018                 if (runparams.use_polyglossia) {
1019                         os << "\\LRE{";
1020                 } else if (running_font.language()->lang() == "farsi"
1021                            || running_font.language()->lang() == "arabic_arabi")
1022                         os << "\\textLR{" << termcmd;
1023                 else
1024                         os << "\\L{";
1025                 close = true;
1026         }
1027
1028         // FIXME: Bug: we can have an empty font change here!
1029         // if there has just been a font change, we are going to close it
1030         // right now, which means stupid latex code like \textsf{}. AFAIK,
1031         // this does not harm dvi output. A minor bug, thus (JMarc)
1032
1033         // Some insets cannot be inside a font change command.
1034         // However, even such insets *can* be placed in \L or \R
1035         // or their equivalents (for RTL language switches), so we don't
1036         // close the language in those cases.
1037         // ArabTeX, though, cannot handle this special behavior, it seems.
1038         bool arabtex = basefont.language()->lang() == "arabic_arabtex"
1039                 || running_font.language()->lang() == "arabic_arabtex";
1040         if (open_font && !inset->inheritFont()) {
1041                 bool needPar = false;
1042                 bool closeLanguage = arabtex
1043                         || basefont.isRightToLeft() == running_font.isRightToLeft();
1044                 unsigned int count = running_font.latexWriteEndChanges(os,
1045                                         bparams, runparams, basefont, basefont,
1046                                         needPar, closeLanguage);
1047                 column += count;
1048                 // if any font properties were closed, update the running_font,
1049                 // making sure, however, to leave the language as it was
1050                 if (count > 0) {
1051                         // FIXME: probably a better way to keep track of the old
1052                         // language, than copying the entire font?
1053                         Font const copy_font(running_font);
1054                         basefont = owner_->getLayoutFont(bparams, outerfont);
1055                         running_font = basefont;
1056                         if (!closeLanguage)
1057                                 running_font.setLanguage(copy_font.language());
1058                         // leave font open if language is still open
1059                         open_font = (running_font.language() == basefont.language());
1060                         if (closeLanguage)
1061                                 runparams.local_font = &basefont;
1062                 }
1063         }
1064
1065         size_t const previous_row_count = os.texrow().rows();
1066
1067         try {
1068                 runparams.lastid = id_;
1069                 runparams.lastpos = i;
1070                 inset->latex(os, runparams);
1071         } catch (EncodingException & e) {
1072                 // add location information and throw again.
1073                 e.par_id = id_;
1074                 e.pos = i;
1075                 throw(e);
1076         }
1077
1078         if (close)
1079                 os << '}';
1080
1081         if (os.texrow().rows() > previous_row_count) {
1082                 os.texrow().start(owner_->id(), i + 1);
1083                 column = 0;
1084         } else {
1085                 column += (unsigned int)(os.os().tellp() - len);
1086         }
1087
1088         if (owner_->isDeleted(i))
1089                 --runparams.inDeletedInset;
1090 }
1091
1092
1093 void Paragraph::Private::latexSpecialChar(otexstream & os,
1094                                           BufferParams const & bparams,
1095                                           OutputParams const & runparams,
1096                                           Font const & running_font,
1097                                           string & alien_script,
1098                                           Layout const & style,
1099                                           pos_type & i,
1100                                           pos_type end_pos,
1101                                           unsigned int & column)
1102 {
1103         char_type const c = owner_->getUChar(bparams, runparams, i);
1104
1105         if (style.pass_thru || runparams.pass_thru
1106             || contains(style.pass_thru_chars, c)
1107             || contains(runparams.pass_thru_chars, c)) {
1108                 if (c != '\0') {
1109                         Encoding const * const enc = runparams.encoding;
1110                         if (enc && !enc->encodable(c))
1111                                 throw EncodingException(c);
1112                         os.put(c);
1113                 }
1114                 return;
1115         }
1116
1117         // TIPA uses its own T3 encoding
1118         if (runparams.inIPA && latexSpecialT3(c, os, i, column))
1119                 return;
1120         // If T1 font encoding is used, use the special
1121         // characters it provides.
1122         // NOTE: Some languages reset the font encoding internally to a
1123         //       non-standard font encoding. If we are using such a language,
1124         //       we do not output special T1 chars.
1125         if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1126             && !runparams.isFullUnicode() && bparams.main_font_encoding() == "T1"
1127             && latexSpecialT1(c, os, i, column))
1128                 return;
1129         // NOTE: "fontspec" (non-TeX fonts) sets the font encoding to "TU" (untill 2017 "EU1" or "EU2")
1130         else if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1131                  && runparams.isFullUnicode() && latexSpecialTU(c, os, i, column))
1132                      return;
1133
1134         // Otherwise, we use what LaTeX provides us.
1135         switch (c) {
1136         case '\\':
1137                 os << "\\textbackslash" << termcmd;
1138                 column += 15;
1139                 break;
1140         case '<':
1141                 os << "\\textless" << termcmd;
1142                 column += 10;
1143                 break;
1144         case '>':
1145                 os << "\\textgreater" << termcmd;
1146                 column += 13;
1147                 break;
1148         case '|':
1149                 os << "\\textbar" << termcmd;
1150                 column += 9;
1151                 break;
1152         case '-':
1153                 os << '-';
1154                 if (i + 1 < static_cast<pos_type>(text_.size()) &&
1155                     (end_pos == -1 || i + 1 < end_pos) &&
1156                     text_[i+1] == '-') {
1157                         // Prevent "--" becoming an en dash and "---" an em dash.
1158                         // (Within \ttfamily, "---" is merged to en dash + hyphen.)
1159                         os << "{}";
1160                         column += 2;
1161                 }
1162                 break;
1163         case '\"':
1164                 os << "\\textquotedbl" << termcmd;
1165                 column += 14;
1166                 break;
1167
1168         case '$': case '&':
1169         case '%': case '#': case '{':
1170         case '}': case '_':
1171                 os << '\\';
1172                 os.put(c);
1173                 column += 1;
1174                 break;
1175
1176         case '~':
1177                 os << "\\textasciitilde" << termcmd;
1178                 column += 16;
1179                 break;
1180
1181         case '^':
1182                 os << "\\textasciicircum" << termcmd;
1183                 column += 17;
1184                 break;
1185
1186         case '*':
1187         case '[':
1188         case ']':
1189                 // avoid being mistaken for optional arguments
1190                 os << '{';
1191                 os.put(c);
1192                 os << '}';
1193                 column += 2;
1194                 break;
1195
1196         case ' ':
1197                 // Blanks are printed before font switching.
1198                 // Sure? I am not! (try nice-latex)
1199                 // I am sure it's correct. LyX might be smarter
1200                 // in the future, but for now, nothing wrong is
1201                 // written. (Asger)
1202                 break;
1203
1204         case 0x2013:
1205         case 0x2014:
1206                 // XeTeX's dash behaviour is determined via a global setting
1207                 if (bparams.use_dash_ligatures
1208                     && owner_->getFontSettings(bparams, i).fontInfo().family() != TYPEWRITER_FAMILY
1209                     && !runparams.inIPA
1210                         // TODO #10961: && not in inset Flex Code
1211                         // TODO #10961: && not in layout LyXCode
1212                     && (!bparams.useNonTeXFonts || runparams.flavor != OutputParams::XETEX)) {
1213                         if (c == 0x2013) {
1214                                 // en-dash
1215                                 os << "--";
1216                                 column +=2;
1217                         } else {
1218                                 // em-dash
1219                                 os << "---";
1220                                 column +=3;
1221                         }
1222                         break;
1223                 }
1224                 // fall through
1225         default:
1226                 if (c == '\0')
1227                         return;
1228
1229                 Encoding const & encoding = *(runparams.encoding);
1230                 char_type next = '\0';
1231                 if (i + 1 < int(text_.size())) {
1232                         next = text_[i + 1];
1233                         if (Encodings::isCombiningChar(next)) {
1234                                 column += latexSurrogatePair(bparams, os, c, next, runparams) - 1;
1235                                 ++i;
1236                                 break;
1237                         }
1238                 }
1239                 pair<docstring, bool> latex = encoding.latexChar(c);
1240                 docstring nextlatex;
1241                 bool nexttipas = false;
1242                 string nexttipashortcut;
1243                 if (next != '\0' && next != META_INSET && !encoding.encodable(next)) {
1244                         nextlatex = encoding.latexChar(next).first;
1245                         if (runparams.inIPA) {
1246                                 nexttipashortcut = Encodings::TIPAShortcut(next);
1247                                 nexttipas = !nexttipashortcut.empty();
1248                         }
1249                 }
1250                 bool tipas = false;
1251                 if (runparams.inIPA) {
1252                         string const tipashortcut = Encodings::TIPAShortcut(c);
1253                         if (!tipashortcut.empty()) {
1254                                 latex.first = from_ascii(tipashortcut);
1255                                 latex.second = false;
1256                                 tipas = true;
1257                         }
1258                 }
1259                 // eventually close "script wrapper" command (see `Paragraph::latex`)
1260                 if (!alien_script.empty()
1261                         && alien_script != Encodings::isKnownScriptChar(next)) {
1262                         column += latex.first.length();
1263                         alien_script.clear();
1264                         os << latex.first << "}";
1265                         break;
1266                 }
1267                 if (latex.second
1268                          && ((!prefixIs(nextlatex, '\\')
1269                                && !prefixIs(nextlatex, '{')
1270                                && !prefixIs(nextlatex, '}'))
1271                              || (nexttipas
1272                                  && !prefixIs(from_ascii(nexttipashortcut), '\\')))
1273                          && !tipas) {
1274                         // Prevent eating of a following space or command corruption by
1275                         // following characters
1276                         if (next == ' ' || next == '\0') {
1277                                 column += latex.first.length() + 1;
1278                                 os << latex.first << "{}";
1279                         } else {
1280                                 column += latex.first.length();
1281                                 os << latex.first << " ";
1282                         }
1283                 } else {
1284                         column += latex.first.length() - 1;
1285                         os << latex.first;
1286                 }
1287                 break;
1288         }
1289 }
1290
1291
1292 bool Paragraph::Private::latexSpecialT1(char_type const c, otexstream & os,
1293         pos_type i, unsigned int & column)
1294 {
1295         switch (c) {
1296         case '>':
1297         case '<':
1298                 os.put(c);
1299                 // In T1 encoding, these characters exist
1300                 // but we should avoid ligatures
1301                 if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
1302                         return true;
1303                 os << "\\textcompwordmark" << termcmd;
1304                 column += 19;
1305                 return true;
1306         case '|':
1307                 os.put(c);
1308                 return true;
1309         case '\"':
1310                 // soul.sty breaks with \char`\"
1311                 os << "\\textquotedbl" << termcmd;
1312                 column += 14;
1313                 return true;
1314         default:
1315                 return false;
1316         }
1317 }
1318
1319
1320 bool Paragraph::Private::latexSpecialTU(char_type const c, otexstream & os,
1321         pos_type i, unsigned int & column)
1322 {
1323         // TU encoding is currently on par with T1.
1324         return latexSpecialT1(c, os, i, column);
1325 }
1326
1327
1328 bool Paragraph::Private::latexSpecialT3(char_type const c, otexstream & os,
1329         pos_type /*i*/, unsigned int & column)
1330 {
1331         switch (c) {
1332         case '*':
1333         case '[':
1334         case ']':
1335         case '\"':
1336                 os.put(c);
1337                 return true;
1338         case '|':
1339                 os << "\\textvertline" << termcmd;
1340                 column += 14;
1341                 return true;
1342         default:
1343                 return false;
1344         }
1345 }
1346
1347
1348 void Paragraph::Private::validate(LaTeXFeatures & features) const
1349 {
1350         if (layout_->inpreamble && inset_owner_) {
1351                 // FIXME: Using a string stream here circumvents the encoding
1352                 // switching machinery of odocstream. Therefore the
1353                 // output is wrong if this paragraph contains content
1354                 // that needs to switch encoding.
1355                 Buffer const & buf = inset_owner_->buffer();
1356                 otexstringstream os;
1357                 os << layout_->preamble();
1358                 size_t const length = os.length();
1359                 TeXOnePar(buf, *inset_owner_->getText(int(buf.getParFromID(owner_->id()).idx())),
1360                           buf.getParFromID(owner_->id()).pit(), os,
1361                           features.runparams(), string(), 0, -1, true);
1362                 if (os.length() > length)
1363                         features.addPreambleSnippet(os.release(), true);
1364         }
1365
1366         if (features.runparams().flavor == OutputParams::HTML
1367             && layout_->htmltitle()) {
1368                 features.setHTMLTitle(owner_->asString(AS_STR_INSETS | AS_STR_SKIPDELETE));
1369         }
1370
1371         // check the params.
1372         if (!params_.spacing().isDefault())
1373                 features.require("setspace");
1374
1375         // then the layouts
1376         features.useLayout(layout_->name());
1377
1378         // then the fonts
1379         fontlist_.validate(features);
1380
1381         // then the indentation
1382         if (!params_.leftIndent().zero())
1383                 features.require("ParagraphLeftIndent");
1384
1385         // then the insets
1386         InsetList::const_iterator icit = insetlist_.begin();
1387         InsetList::const_iterator iend = insetlist_.end();
1388         for (; icit != iend; ++icit) {
1389                 if (icit->inset) {
1390                         features.inDeletedInset(owner_->isDeleted(icit->pos));
1391                         if (icit->inset->lyxCode() == FOOT_CODE) {
1392                                 // FIXME: an item inset would make things much easier.
1393                                 if ((layout_->latextype == LATEX_LIST_ENVIRONMENT
1394                                      || (layout_->latextype == LATEX_ITEM_ENVIRONMENT
1395                                          && layout_->margintype == MARGIN_FIRST_DYNAMIC))
1396                                     && (icit->pos < begin_of_body_
1397                                         || (icit->pos == begin_of_body_
1398                                             && (icit->pos == 0 || text_[icit->pos - 1] != ' '))))
1399                                         features.saveNoteEnv("description");
1400                         }
1401                         icit->inset->validate(features);
1402                         features.inDeletedInset(false);
1403                         if (layout_->needprotect &&
1404                             icit->inset->lyxCode() == FOOT_CODE)
1405                                 features.require("NeedLyXFootnoteCode");
1406                 }
1407         }
1408
1409         // then the contents
1410         BufferParams const bp = features.runparams().is_child
1411                 ? features.buffer().masterParams() : features.buffer().params();
1412         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1413                 char_type c = text_[i];
1414                 CharInfo const & ci = Encodings::unicodeCharInfo(c);
1415                 if (c == 0x0022) {
1416                         if (features.runparams().isFullUnicode() && bp.useNonTeXFonts)
1417                                 features.require("textquotedblp");
1418                         else if (bp.main_font_encoding() != "T1"
1419                                  || ((&owner_->getFontSettings(bp, i))->language()->internalFontEncoding()))
1420                                 features.require("textquotedbl");
1421                 } else if (ci.textfeature() && contains(ci.textpreamble(), '=')) {
1422                         // features that depend on the font or input encoding
1423                         string feats = ci.textpreamble();
1424                         string fontenc = (&owner_->getFontSettings(bp, i))->language()->fontenc(bp);
1425                         if (fontenc.empty())
1426                                 fontenc = features.runparams().main_fontenc;
1427                         while (!feats.empty()) {
1428                                 string feat;
1429                                 feats = split(feats, feat, ',');
1430                                 if (contains(feat, "!=")) {
1431                                         // a feature that is required except for the spcified
1432                                         // font or input encodings
1433                                         string realfeature;
1434                                         string const contexts = ltrim(split(feat, realfeature, '!'), "=");
1435                                         // multiple encodings are separated by semicolon
1436                                         vector<string> context = getVectorFromString(contexts, ";");
1437                                         // require feature if the context matches neither current font
1438                                         // nor input encoding
1439                                         if (std::find(context.begin(), context.end(), fontenc) == context.end()
1440                                             && std::find(context.begin(), context.end(),
1441                                                          features.runparams().encoding->name()) == context.end())
1442                                                 features.require(realfeature);
1443                                 } else if (contains(feat, '=')) {
1444                                         // a feature that is required only for the spcified
1445                                         // font or input encodings
1446                                         string realfeature;
1447                                         string const contexts = split(feat, realfeature, '=');
1448                                         // multiple encodings are separated by semicolon
1449                                         vector<string> context = getVectorFromString(contexts, ";");
1450                                         // require feature if the context matches either current font
1451                                         // or input encoding
1452                                         if (std::find(context.begin(), context.end(), fontenc) != context.end()
1453                                             || std::find(context.begin(), context.end(),
1454                                                          features.runparams().encoding->name()) != context.end())
1455                                                 features.require(realfeature);
1456                                 }
1457                         }
1458                 } else if (!bp.use_dash_ligatures
1459                            && (c == 0x2013 || c == 0x2014)
1460                            && bp.useNonTeXFonts
1461                            && features.runparams().flavor == OutputParams::XETEX)
1462                         // XeTeX's dash behaviour is determined via a global setting
1463                         features.require("xetexdashbreakstate");
1464                 BufferEncodings::validate(c, features);
1465         }
1466 }
1467
1468 /////////////////////////////////////////////////////////////////////
1469 //
1470 // Paragraph
1471 //
1472 /////////////////////////////////////////////////////////////////////
1473
1474 namespace {
1475         Layout const emptyParagraphLayout;
1476 }
1477
1478 Paragraph::Paragraph()
1479         : d(new Paragraph::Private(this, emptyParagraphLayout))
1480 {
1481         itemdepth = 0;
1482         d->params_.clear();
1483 }
1484
1485
1486 Paragraph::Paragraph(Paragraph const & par)
1487         : itemdepth(par.itemdepth),
1488         d(new Paragraph::Private(*par.d, this))
1489 {
1490         registerWords();
1491 }
1492
1493
1494 Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
1495         : itemdepth(par.itemdepth),
1496         d(new Paragraph::Private(*par.d, this, beg, end))
1497 {
1498         registerWords();
1499 }
1500
1501
1502 Paragraph & Paragraph::operator=(Paragraph const & par)
1503 {
1504         // needed as we will destroy the private part before copying it
1505         if (&par != this) {
1506                 itemdepth = par.itemdepth;
1507
1508                 deregisterWords();
1509                 delete d;
1510                 d = new Private(*par.d, this);
1511                 registerWords();
1512         }
1513         return *this;
1514 }
1515
1516
1517 Paragraph::~Paragraph()
1518 {
1519         deregisterWords();
1520         delete d;
1521 }
1522
1523
1524 namespace {
1525
1526 // this shall be called just before every "os << ..." action.
1527 void flushString(ostream & os, docstring & s)
1528 {
1529         os << to_utf8(s);
1530         s.erase();
1531 }
1532
1533 } // namespace
1534
1535
1536 void Paragraph::write(ostream & os, BufferParams const & bparams,
1537         depth_type & dth) const
1538 {
1539         // The beginning or end of a deeper (i.e. nested) area?
1540         if (dth != d->params_.depth()) {
1541                 if (d->params_.depth() > dth) {
1542                         while (d->params_.depth() > dth) {
1543                                 os << "\n\\begin_deeper";
1544                                 ++dth;
1545                         }
1546                 } else {
1547                         while (d->params_.depth() < dth) {
1548                                 os << "\n\\end_deeper";
1549                                 --dth;
1550                         }
1551                 }
1552         }
1553
1554         // First write the layout
1555         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1556
1557         d->params_.write(os);
1558
1559         Font font1(inherit_font, bparams.language);
1560
1561         Change running_change = Change(Change::UNCHANGED);
1562
1563         // this string is used as a buffer to avoid repetitive calls
1564         // to to_utf8(), which turn out to be expensive (JMarc)
1565         docstring write_buffer;
1566
1567         int column = 0;
1568         for (pos_type i = 0; i <= size(); ++i) {
1569
1570                 Change const & change = lookupChange(i);
1571                 if (change != running_change)
1572                         flushString(os, write_buffer);
1573                 Changes::lyxMarkChange(os, bparams, column, running_change, change);
1574                 running_change = change;
1575
1576                 if (i == size())
1577                         break;
1578
1579                 // Write font changes
1580                 Font font2 = getFontSettings(bparams, i);
1581                 if (font2 != font1) {
1582                         flushString(os, write_buffer);
1583                         font2.lyxWriteChanges(font1, os);
1584                         column = 0;
1585                         font1 = font2;
1586                 }
1587
1588                 char_type const c = d->text_[i];
1589                 switch (c) {
1590                 case META_INSET:
1591                         if (Inset const * inset = getInset(i)) {
1592                                 flushString(os, write_buffer);
1593                                 if (inset->directWrite()) {
1594                                         // international char, let it write
1595                                         // code directly so it's shorter in
1596                                         // the file
1597                                         inset->write(os);
1598                                 } else {
1599                                         if (i)
1600                                                 os << '\n';
1601                                         os << "\\begin_inset ";
1602                                         inset->write(os);
1603                                         os << "\n\\end_inset\n\n";
1604                                         column = 0;
1605                                 }
1606                                 // FIXME This can be removed again once the mystery
1607                                 // crash has been resolved.
1608                                 os << flush;
1609                         }
1610                         break;
1611                 case '\\':
1612                         flushString(os, write_buffer);
1613                         os << "\n\\backslash\n";
1614                         column = 0;
1615                         break;
1616                 case '.':
1617                         flushString(os, write_buffer);
1618                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1619                                 os << ".\n";
1620                                 column = 0;
1621                         } else
1622                                 os << '.';
1623                         break;
1624                 default:
1625                         if ((column > 70 && c == ' ')
1626                             || column > 79) {
1627                                 flushString(os, write_buffer);
1628                                 os << '\n';
1629                                 column = 0;
1630                         }
1631                         // this check is to amend a bug. LyX sometimes
1632                         // inserts '\0' this could cause problems.
1633                         if (c != '\0')
1634                                 write_buffer.push_back(c);
1635                         else
1636                                 LYXERR0("NUL char in structure.");
1637                         ++column;
1638                         break;
1639                 }
1640         }
1641
1642         flushString(os, write_buffer);
1643         os << "\n\\end_layout\n";
1644         // FIXME This can be removed again once the mystery
1645         // crash has been resolved.
1646         os << flush;
1647 }
1648
1649
1650 void Paragraph::validate(LaTeXFeatures & features) const
1651 {
1652         d->validate(features);
1653         bool fragile = features.runparams().moving_arg;
1654         fragile |= layout().needprotect;
1655         if (inInset().getLayout().isNeedProtect())
1656                 fragile = true;
1657         if (needsCProtection(fragile))
1658                 features.require("cprotect");
1659 }
1660
1661
1662 void Paragraph::insert(pos_type start, docstring const & str,
1663                        Font const & font, Change const & change)
1664 {
1665         for (size_t i = 0, n = str.size(); i != n ; ++i)
1666                 insertChar(start + i, str[i], font, change);
1667 }
1668
1669
1670 void Paragraph::appendChar(char_type c, Font const & font,
1671                 Change const & change)
1672 {
1673         // track change
1674         d->changes_.insert(change, d->text_.size());
1675         // when appending characters, no need to update tables
1676         d->text_.push_back(c);
1677         setFont(d->text_.size() - 1, font);
1678         d->requestSpellCheck(d->text_.size() - 1);
1679 }
1680
1681
1682 void Paragraph::appendString(docstring const & s, Font const & font,
1683                 Change const & change)
1684 {
1685         pos_type end = s.size();
1686         size_t oldsize = d->text_.size();
1687         size_t newsize = oldsize + end;
1688         size_t capacity = d->text_.capacity();
1689         if (newsize >= capacity)
1690                 d->text_.reserve(max(capacity + 100, newsize));
1691
1692         // when appending characters, no need to update tables
1693         d->text_.append(s);
1694
1695         // FIXME: Optimize this!
1696         for (size_t i = oldsize; i != newsize; ++i) {
1697                 // track change
1698                 d->changes_.insert(change, i);
1699                 d->requestSpellCheck(i);
1700         }
1701         d->fontlist_.set(oldsize, font);
1702         d->fontlist_.set(newsize - 1, font);
1703 }
1704
1705
1706 void Paragraph::insertChar(pos_type pos, char_type c,
1707                            bool trackChanges)
1708 {
1709         d->insertChar(pos, c, Change(trackChanges ?
1710                            Change::INSERTED : Change::UNCHANGED));
1711 }
1712
1713
1714 void Paragraph::insertChar(pos_type pos, char_type c,
1715                            Font const & font, bool trackChanges)
1716 {
1717         d->insertChar(pos, c, Change(trackChanges ?
1718                            Change::INSERTED : Change::UNCHANGED));
1719         setFont(pos, font);
1720 }
1721
1722
1723 void Paragraph::insertChar(pos_type pos, char_type c,
1724                            Font const & font, Change const & change)
1725 {
1726         d->insertChar(pos, c, change);
1727         setFont(pos, font);
1728 }
1729
1730
1731 void Paragraph::resetFonts(Font const & font)
1732 {
1733         d->fontlist_.clear();
1734         d->fontlist_.set(0, font);
1735         d->fontlist_.set(d->text_.size() - 1, font);
1736 }
1737
1738 // Gets uninstantiated font setting at position.
1739 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1740                                          pos_type pos) const
1741 {
1742         if (pos > size()) {
1743                 LYXERR0("pos: " << pos << " size: " << size());
1744                 LBUFERR(false);
1745         }
1746
1747         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1748         if (cit != d->fontlist_.end())
1749                 return cit->font();
1750
1751         if (pos == size() && !empty())
1752                 return getFontSettings(bparams, pos - 1);
1753
1754         // Optimisation: avoid a full font instantiation if there is no
1755         // language change from previous call.
1756         static Font previous_font;
1757         static Language const * previous_lang = 0;
1758         Language const * lang = getParLanguage(bparams);
1759         if (lang != previous_lang) {
1760                 previous_lang = lang;
1761                 previous_font = Font(inherit_font, lang);
1762         }
1763         return previous_font;
1764 }
1765
1766
1767 FontSpan Paragraph::fontSpan(pos_type pos) const
1768 {
1769         LBUFERR(pos <= size());
1770
1771         if (pos == size())
1772                 return FontSpan(pos, pos);
1773
1774         pos_type start = 0;
1775         FontList::const_iterator cit = d->fontlist_.begin();
1776         FontList::const_iterator end = d->fontlist_.end();
1777         for (; cit != end; ++cit) {
1778                 if (cit->pos() >= pos) {
1779                         if (pos >= beginOfBody())
1780                                 return FontSpan(max(start, beginOfBody()),
1781                                                 cit->pos());
1782                         else
1783                                 return FontSpan(start,
1784                                                 min(beginOfBody() - 1,
1785                                                          cit->pos()));
1786                 }
1787                 start = cit->pos() + 1;
1788         }
1789
1790         // This should not happen, but if so, we take no chances.
1791         LYXERR0("Paragraph::fontSpan: position not found in fontinfo table!");
1792         LASSERT(false, return FontSpan(pos, pos));
1793 }
1794
1795
1796 // Gets uninstantiated font setting at position 0
1797 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1798 {
1799         if (!empty() && !d->fontlist_.empty())
1800                 return d->fontlist_.begin()->font();
1801
1802         // Optimisation: avoid a full font instantiation if there is no
1803         // language change from previous call.
1804         static Font previous_font;
1805         static Language const * previous_lang = 0;
1806         if (bparams.language != previous_lang) {
1807                 previous_lang = bparams.language;
1808                 previous_font = Font(inherit_font, bparams.language);
1809         }
1810
1811         return previous_font;
1812 }
1813
1814
1815 // Gets the fully instantiated font at a given position in a paragraph
1816 // This is basically the same function as Text::GetFont() in text2.cpp.
1817 // The difference is that this one is used for generating the LaTeX file,
1818 // and thus cosmetic "improvements" are disallowed: This has to deliver
1819 // the true picture of the buffer. (Asger)
1820 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1821                                  Font const & outerfont) const
1822 {
1823         LBUFERR(pos >= 0);
1824
1825         Font font = getFontSettings(bparams, pos);
1826
1827         pos_type const body_pos = beginOfBody();
1828         FontInfo & fi = font.fontInfo();
1829         if (pos < body_pos)
1830                 fi.realize(d->layout_->labelfont);
1831         else
1832                 fi.realize(d->layout_->font);
1833
1834         fi.realize(outerfont.fontInfo());
1835         fi.realize(bparams.getFont().fontInfo());
1836
1837         return font;
1838 }
1839
1840
1841 Font const Paragraph::getLabelFont
1842         (BufferParams const & bparams, Font const & outerfont) const
1843 {
1844         FontInfo tmpfont = d->layout_->labelfont;
1845         tmpfont.realize(outerfont.fontInfo());
1846         tmpfont.realize(bparams.getFont().fontInfo());
1847         return Font(tmpfont, getParLanguage(bparams));
1848 }
1849
1850
1851 Font const Paragraph::getLayoutFont
1852         (BufferParams const & bparams, Font const & outerfont) const
1853 {
1854         FontInfo tmpfont = d->layout_->font;
1855         tmpfont.realize(outerfont.fontInfo());
1856         tmpfont.realize(bparams.getFont().fontInfo());
1857         return Font(tmpfont, getParLanguage(bparams));
1858 }
1859
1860
1861 char_type Paragraph::getUChar(BufferParams const & bparams,
1862                               OutputParams const & rp,
1863                               pos_type pos) const
1864 {
1865         char_type c = d->text_[pos];
1866
1867         // Return unchanged character in LTR languages
1868         // or if we use poylglossia/bidi (XeTeX).
1869         if (rp.useBidiPackage()
1870             || !getFontSettings(bparams, pos).isRightToLeft())
1871                 return c;
1872
1873         // Without polyglossia/bidi, we need to account for some special cases.
1874         // FIXME This needs to be audited!
1875         // Check if:
1876         // * The input is as expected for all delimiters
1877         //   => checked for Hebrew!
1878         // * The output matches the display in the LyX workarea
1879         //   => checked for Hebrew!
1880         // * The special cases below are really necessary
1881         //   => checked for Hebrew!
1882         // * In arabic_arabi, brackets are transformed to Arabic
1883         //   Ornate Parentheses. Is this is really wanted?
1884
1885         string const & lang = getFontSettings(bparams, pos).language()->lang();
1886         char_type uc = c;
1887
1888         // 1. In the following languages, parentheses need to be reversed.
1889         //    Also with polyglodia/luabidi
1890         bool const reverseparens = (lang == "hebrew" || rp.use_polyglossia);
1891
1892         // 2. In the following languages, brackets don't need to be reversed.
1893         bool const reversebrackets = lang != "arabic_arabtex"
1894                         && lang != "arabic_arabi"
1895                         && lang != "farsi";
1896
1897         // Now swap delimiters if needed.
1898         switch (c) {
1899         case '(':
1900                 if (reverseparens)
1901                         uc = ')';
1902                 break;
1903         case ')':
1904                 if (reverseparens)
1905                         uc = '(';
1906                 break;
1907         case '[':
1908                 if (reversebrackets)
1909                         uc = ']';
1910                 break;
1911         case ']':
1912                 if (reversebrackets)
1913                         uc = '[';
1914                 break;
1915         case '{':
1916                 uc = '}';
1917                 break;
1918         case '}':
1919                 uc = '{';
1920                 break;
1921         case '<':
1922                 uc = '>';
1923                 break;
1924         case '>':
1925                 uc = '<';
1926                 break;
1927         }
1928
1929         return uc;
1930 }
1931
1932
1933 void Paragraph::setFont(pos_type pos, Font const & font)
1934 {
1935         LASSERT(pos <= size(), return);
1936
1937         // First, reduce font against layout/label font
1938         // Update: The setCharFont() routine in text2.cpp already
1939         // reduces font, so we don't need to do that here. (Asger)
1940
1941         d->fontlist_.set(pos, font);
1942 }
1943
1944
1945 void Paragraph::makeSameLayout(Paragraph const & par)
1946 {
1947         d->layout_ = par.d->layout_;
1948         d->params_ = par.d->params_;
1949 }
1950
1951
1952 bool Paragraph::stripLeadingSpaces(bool trackChanges)
1953 {
1954         if (isFreeSpacing())
1955                 return false;
1956
1957         int pos = 0;
1958         int count = 0;
1959
1960         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
1961                 if (eraseChar(pos, trackChanges))
1962                         ++count;
1963                 else
1964                         ++pos;
1965         }
1966
1967         return count > 0 || pos > 0;
1968 }
1969
1970
1971 bool Paragraph::hasSameLayout(Paragraph const & par) const
1972 {
1973         return par.d->layout_ == d->layout_
1974                 && d->params_.sameLayout(par.d->params_);
1975 }
1976
1977
1978 depth_type Paragraph::getDepth() const
1979 {
1980         return d->params_.depth();
1981 }
1982
1983
1984 depth_type Paragraph::getMaxDepthAfter() const
1985 {
1986         if (d->layout_->isEnvironment())
1987                 return d->params_.depth() + 1;
1988         else
1989                 return d->params_.depth();
1990 }
1991
1992
1993 LyXAlignment Paragraph::getAlign(BufferParams const & bparams) const
1994 {
1995         if (d->params_.align() == LYX_ALIGN_LAYOUT)
1996                 return getDefaultAlign(bparams);
1997         else
1998                 return d->params_.align();
1999 }
2000
2001
2002 LyXAlignment Paragraph::getDefaultAlign(BufferParams const & bparams) const
2003 {
2004         LyXAlignment res = layout().align;
2005         if (isRTL(bparams)) {
2006                 // Swap sides
2007                 if (res == LYX_ALIGN_LEFT)
2008                         res = LYX_ALIGN_RIGHT;
2009                 else if  (res == LYX_ALIGN_RIGHT)
2010                         res = LYX_ALIGN_LEFT;
2011         }
2012         return res;
2013 }
2014
2015
2016 docstring const & Paragraph::labelString() const
2017 {
2018         return d->params_.labelString();
2019 }
2020
2021
2022 // the next two functions are for the manual labels
2023 docstring const Paragraph::getLabelWidthString() const
2024 {
2025         if (d->layout_->margintype == MARGIN_MANUAL
2026             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
2027                 return d->params_.labelWidthString();
2028         else
2029                 return _("Senseless with this layout!");
2030 }
2031
2032
2033 void Paragraph::setLabelWidthString(docstring const & s)
2034 {
2035         d->params_.labelWidthString(s);
2036 }
2037
2038
2039 docstring Paragraph::expandLabel(Layout const & layout,
2040                 BufferParams const & bparams) const
2041 {
2042         return expandParagraphLabel(layout, bparams, true);
2043 }
2044
2045
2046 docstring Paragraph::expandDocBookLabel(Layout const & layout,
2047                 BufferParams const & bparams) const
2048 {
2049         return expandParagraphLabel(layout, bparams, false);
2050 }
2051
2052
2053 docstring Paragraph::expandParagraphLabel(Layout const & layout,
2054                 BufferParams const & bparams, bool process_appendix) const
2055 {
2056         DocumentClass const & tclass = bparams.documentClass();
2057         string const & lang = getParLanguage(bparams)->code();
2058         bool const in_appendix = process_appendix && d->params_.appendix();
2059         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
2060
2061         if (fmt.empty() && !layout.counter.empty())
2062                 return tclass.counters().theCounter(layout.counter, lang);
2063
2064         // handle 'inherited level parts' in 'fmt',
2065         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
2066         size_t const i = fmt.find('@', 0);
2067         if (i != docstring::npos) {
2068                 size_t const j = fmt.find('@', i + 1);
2069                 if (j != docstring::npos) {
2070                         docstring parent(fmt, i + 1, j - i - 1);
2071                         docstring label = from_ascii("??");
2072                         if (tclass.hasLayout(parent))
2073                                 label = expandParagraphLabel(tclass[parent], bparams,
2074                                                       process_appendix);
2075                         fmt = docstring(fmt, 0, i) + label
2076                                 + docstring(fmt, j + 1, docstring::npos);
2077                 }
2078         }
2079
2080         return tclass.counters().counterLabel(fmt, lang);
2081 }
2082
2083
2084 void Paragraph::applyLayout(Layout const & new_layout)
2085 {
2086         d->layout_ = &new_layout;
2087         LyXAlignment const oldAlign = d->params_.align();
2088
2089         if (!(oldAlign & d->layout_->alignpossible)) {
2090                 frontend::Alert::warning(_("Alignment not permitted"),
2091                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
2092                 d->params_.align(LYX_ALIGN_LAYOUT);
2093         }
2094 }
2095
2096
2097 pos_type Paragraph::beginOfBody() const
2098 {
2099         return d->begin_of_body_;
2100 }
2101
2102
2103 void Paragraph::setBeginOfBody()
2104 {
2105         if (d->layout_->labeltype != LABEL_MANUAL) {
2106                 d->begin_of_body_ = 0;
2107                 return;
2108         }
2109
2110         // Unroll the first two cycles of the loop
2111         // and remember the previous character to
2112         // remove unnecessary getChar() calls
2113         pos_type i = 0;
2114         pos_type end = size();
2115         if (i < end && !(isNewline(i) || isEnvSeparator(i))) {
2116                 ++i;
2117                 if (i < end) {
2118                         char_type previous_char = d->text_[i];
2119                         if (!(isNewline(i) || isEnvSeparator(i))) {
2120                                 ++i;
2121                                 while (i < end && previous_char != ' ') {
2122                                         char_type temp = d->text_[i];
2123                                         if (isNewline(i) || isEnvSeparator(i))
2124                                                 break;
2125                                         ++i;
2126                                         previous_char = temp;
2127                                 }
2128                         }
2129                 }
2130         }
2131
2132         d->begin_of_body_ = i;
2133 }
2134
2135
2136 bool Paragraph::allowParagraphCustomization() const
2137 {
2138         return inInset().allowParagraphCustomization();
2139 }
2140
2141
2142 bool Paragraph::usePlainLayout() const
2143 {
2144         return inInset().usePlainLayout();
2145 }
2146
2147
2148 bool Paragraph::isPassThru() const
2149 {
2150         return inInset().isPassThru() || d->layout_->pass_thru;
2151 }
2152
2153 namespace {
2154
2155 // paragraphs inside floats need different alignment tags to avoid
2156 // unwanted space
2157
2158 bool noTrivlistCentering(InsetCode code)
2159 {
2160         return code == FLOAT_CODE
2161                || code == WRAP_CODE
2162                || code == CELL_CODE;
2163 }
2164
2165
2166 string correction(string const & orig)
2167 {
2168         if (orig == "flushleft")
2169                 return "raggedright";
2170         if (orig == "flushright")
2171                 return "raggedleft";
2172         if (orig == "center")
2173                 return "centering";
2174         return orig;
2175 }
2176
2177
2178 bool corrected_env(otexstream & os, string const & suffix, string const & env,
2179         InsetCode code, bool const lastpar, int & col)
2180 {
2181         string macro = suffix + "{";
2182         if (noTrivlistCentering(code)) {
2183                 if (lastpar) {
2184                         // the last paragraph in non-trivlist-aligned
2185                         // context is special (to avoid unwanted whitespace)
2186                         if (suffix == "\\begin") {
2187                                 macro = "\\" + correction(env) + "{}";
2188                                 os << from_ascii(macro);
2189                                 col += macro.size();
2190                                 return true;
2191                         }
2192                         return false;
2193                 }
2194                 macro += correction(env);
2195         } else
2196                 macro += env;
2197         macro += "}";
2198         if (suffix == "\\par\\end") {
2199                 os << breakln;
2200                 col = 0;
2201         }
2202         os << from_ascii(macro);
2203         col += macro.size();
2204         if (suffix == "\\begin") {
2205                 os << breakln;
2206                 col = 0;
2207         }
2208         return true;
2209 }
2210
2211 } // namespace
2212
2213
2214 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
2215                         otexstream & os, OutputParams const & runparams) const
2216 {
2217         int column = 0;
2218
2219         bool canindent =
2220                 (bparams.paragraph_separation == BufferParams::ParagraphIndentSeparation) ?
2221                         (layout_->toggle_indent != ITOGGLE_NEVER) :
2222                         (layout_->toggle_indent == ITOGGLE_ALWAYS);
2223
2224         if (canindent && params_.noindent() && !layout_->pass_thru) {
2225                 os << "\\noindent ";
2226                 column += 10;
2227         }
2228
2229         LyXAlignment const curAlign = params_.align();
2230
2231         if (curAlign == layout_->align)
2232                 return column;
2233
2234         switch (curAlign) {
2235         case LYX_ALIGN_NONE:
2236         case LYX_ALIGN_BLOCK:
2237         case LYX_ALIGN_LAYOUT:
2238         case LYX_ALIGN_SPECIAL:
2239         case LYX_ALIGN_DECIMAL:
2240                 break;
2241         case LYX_ALIGN_LEFT:
2242         case LYX_ALIGN_RIGHT:
2243         case LYX_ALIGN_CENTER:
2244                 if (runparams.moving_arg) {
2245                         os << "\\protect";
2246                         column += 8;
2247                 }
2248                 break;
2249         }
2250
2251         string const begin_tag = "\\begin";
2252         InsetCode code = ownerCode();
2253         bool const lastpar = runparams.isLastPar;
2254         // RTL in classic (PDF)LaTeX (without the Bidi package)
2255         // Luabibdi (used by LuaTeX) behaves like classic
2256         bool const rtl_classic = owner_->getParLanguage(bparams)->rightToLeft()
2257                 && !runparams.useBidiPackage();
2258
2259         switch (curAlign) {
2260         case LYX_ALIGN_NONE:
2261         case LYX_ALIGN_BLOCK:
2262         case LYX_ALIGN_LAYOUT:
2263         case LYX_ALIGN_SPECIAL:
2264         case LYX_ALIGN_DECIMAL:
2265                 break;
2266         case LYX_ALIGN_LEFT: {
2267                 if (rtl_classic)
2268                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2269                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2270                 else
2271                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2272                 break;
2273         } case LYX_ALIGN_RIGHT: {
2274                 if (rtl_classic)
2275                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2276                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2277                 else
2278                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2279                 break;
2280         } case LYX_ALIGN_CENTER: {
2281                 corrected_env(os, begin_tag, "center", code, lastpar, column);
2282                 break;
2283         }
2284         }
2285
2286         return column;
2287 }
2288
2289
2290 bool Paragraph::Private::endTeXParParams(BufferParams const & bparams,
2291                         otexstream & os, OutputParams const & runparams) const
2292 {
2293         LyXAlignment const curAlign = params_.align();
2294
2295         if (curAlign == layout_->align)
2296                 return false;
2297
2298         switch (curAlign) {
2299         case LYX_ALIGN_NONE:
2300         case LYX_ALIGN_BLOCK:
2301         case LYX_ALIGN_LAYOUT:
2302         case LYX_ALIGN_SPECIAL:
2303         case LYX_ALIGN_DECIMAL:
2304                 break;
2305         case LYX_ALIGN_LEFT:
2306         case LYX_ALIGN_RIGHT:
2307         case LYX_ALIGN_CENTER:
2308                 if (runparams.moving_arg)
2309                         os << "\\protect";
2310                 break;
2311         }
2312
2313         bool output = false;
2314         int col = 0;
2315         string const end_tag = "\\par\\end";
2316         InsetCode code = ownerCode();
2317         bool const lastpar = runparams.isLastPar;
2318         // RTL in classic (PDF)LaTeX (without the Bidi package)
2319         // Luabibdi (used by LuaTeX) behaves like classic
2320         bool const rtl_classic = owner_->getParLanguage(bparams)->rightToLeft()
2321                 && !runparams.useBidiPackage();
2322
2323         switch (curAlign) {
2324         case LYX_ALIGN_NONE:
2325         case LYX_ALIGN_BLOCK:
2326         case LYX_ALIGN_LAYOUT:
2327         case LYX_ALIGN_SPECIAL:
2328         case LYX_ALIGN_DECIMAL:
2329                 break;
2330         case LYX_ALIGN_LEFT: {
2331                 if (rtl_classic)
2332                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2333                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2334                 else
2335                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2336                 break;
2337         } case LYX_ALIGN_RIGHT: {
2338                 if (rtl_classic)
2339                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2340                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2341                 else
2342                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2343                 break;
2344         } case LYX_ALIGN_CENTER: {
2345                 corrected_env(os, end_tag, "center", code, lastpar, col);
2346                 break;
2347         }
2348         }
2349
2350         return output || lastpar;
2351 }
2352
2353
2354 // This one spits out the text of the paragraph
2355 void Paragraph::latex(BufferParams const & bparams,
2356         Font const & outerfont,
2357         otexstream & os,
2358         OutputParams const & runparams,
2359         int start_pos, int end_pos, bool force) const
2360 {
2361         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
2362
2363         // FIXME This check should not be needed. Perhaps issue an
2364         // error if it triggers.
2365         Layout const & style = inInset().forcePlainLayout() ?
2366                 bparams.documentClass().plainLayout() : *d->layout_;
2367
2368         if (!force && style.inpreamble)
2369                 return;
2370
2371         bool const allowcust = allowParagraphCustomization();
2372
2373         // Current base font for all inherited font changes, without any
2374         // change caused by an individual character, except for the language:
2375         // It is set to the language of the first character.
2376         // As long as we are in the label, this font is the base font of the
2377         // label. Before the first body character it is set to the base font
2378         // of the body.
2379         Font basefont;
2380
2381         // If there is an open font-encoding changing command (script wrapper),
2382         // alien_script is set to its name
2383         string alien_script;
2384         string script;
2385
2386         // Maybe we have to create a optional argument.
2387         pos_type body_pos = beginOfBody();
2388         unsigned int column = 0;
2389
2390         if (body_pos > 0) {
2391                 // the optional argument is kept in curly brackets in
2392                 // case it contains a ']'
2393                 // This is not strictly needed, but if this is changed it
2394                 // would be a file format change, and tex2lyx would need
2395                 // to be adjusted, since it unconditionally removes the
2396                 // braces when it parses \item.
2397                 os << "[{";
2398                 column += 2;
2399                 basefont = getLabelFont(bparams, outerfont);
2400         } else {
2401                 basefont = getLayoutFont(bparams, outerfont);
2402         }
2403
2404         // Which font is currently active?
2405         Font running_font(basefont);
2406         // Do we have an open font change?
2407         bool open_font = false;
2408
2409         Change runningChange = Change(Change::UNCHANGED);
2410
2411         Encoding const * const prev_encoding = runparams.encoding;
2412
2413         os.texrow().start(id(), 0);
2414
2415         // if the paragraph is empty, the loop will not be entered at all
2416         if (empty()) {
2417                 // For InTitle commands, we have already opened a group
2418                 // in output_latex::TeXOnePar.
2419                 if (style.isCommand() && !style.intitle) {
2420                         os << '{';
2421                         ++column;
2422                 }
2423                 if (!style.leftdelim().empty()) {
2424                         os << style.leftdelim();
2425                         column += style.leftdelim().size();
2426                 }
2427                 if (allowcust)
2428                         column += d->startTeXParParams(bparams, os, runparams);
2429         }
2430
2431         // Whether a \par can be issued for insets typeset inline with text.
2432         // Yes if greater than 0. This has to be static.
2433         THREAD_LOCAL_STATIC int parInline = 0;
2434
2435         for (pos_type i = 0; i < size(); ++i) {
2436                 // First char in paragraph or after label?
2437                 if (i == body_pos) {
2438                         if (body_pos > 0) {
2439                                 if (open_font) {
2440                                         bool needPar = false;
2441                                         column += running_font.latexWriteEndChanges(
2442                                                 os, bparams, runparams,
2443                                                 basefont, basefont, needPar);
2444                                         open_font = false;
2445                                 }
2446                                 basefont = getLayoutFont(bparams, outerfont);
2447                                 running_font = basefont;
2448
2449                                 column += Changes::latexMarkChange(os, bparams,
2450                                                 runningChange, Change(Change::UNCHANGED),
2451                                                 runparams);
2452                                 runningChange = Change(Change::UNCHANGED);
2453
2454                                 os << "}] ";
2455                                 column +=3;
2456                         }
2457                         // For InTitle commands, we have already opened a group
2458                         // in output_latex::TeXOnePar.
2459                         if (style.isCommand() && !style.intitle) {
2460                                 os << '{';
2461                                 ++column;
2462                         }
2463
2464                         if (!style.leftdelim().empty()) {
2465                                 os << style.leftdelim();
2466                                 column += style.leftdelim().size();
2467                         }
2468
2469                         if (allowcust)
2470                                 column += d->startTeXParParams(bparams, os,
2471                                                             runparams);
2472                 }
2473
2474                 runparams.wasDisplayMath = runparams.inDisplayMath;
2475                 runparams.inDisplayMath = false;
2476                 bool deleted_display_math = false;
2477                 Change const & change = runparams.inDeletedInset
2478                         ? runparams.changeOfDeletedInset : lookupChange(i);
2479
2480                 // Check whether a display math inset follows
2481                 if (d->text_[i] == META_INSET
2482                     && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2483                         InsetMath const * im = getInset(i)->asInsetMath();
2484                         if (im && im->asHullInset()
2485                             && im->asHullInset()->outerDisplay()) {
2486                                 runparams.inDisplayMath = true;
2487                                 // runparams.inDeletedInset will be set by
2488                                 // latexInset later, but we need this info
2489                                 // before it is called. On the other hand, we
2490                                 // cannot set it here because it is a counter.
2491                                 deleted_display_math = isDeleted(i);
2492                         }
2493                         if (bparams.output_changes && deleted_display_math
2494                             && runningChange == change
2495                             && change.type == Change::DELETED
2496                             && !os.afterParbreak()) {
2497                                 // A display math in the same paragraph follows.
2498                                 // We have to close and then reopen \lyxdeleted,
2499                                 // otherwise the math will be shifted up.
2500                                 OutputParams rp = runparams;
2501                                 if (open_font) {
2502                                         bool needPar = false;
2503                                         column += running_font.latexWriteEndChanges(
2504                                                 os, bparams, rp, basefont,
2505                                                 basefont, needPar);
2506                                         open_font = false;
2507                                 }
2508                                 basefont = (body_pos > i) ? getLabelFont(bparams, outerfont)
2509                                                           : getLayoutFont(bparams, outerfont);
2510                                 running_font = basefont;
2511                                 column += Changes::latexMarkChange(os, bparams,
2512                                         Change(Change::INSERTED), change, rp);
2513                         }
2514                 }
2515
2516                 if (bparams.output_changes && runningChange != change) {
2517                         if (!alien_script.empty()) {
2518                                 column += 1;
2519                                 os << "}";
2520                                 alien_script.clear();
2521                         }
2522                         if (open_font) {
2523                                 bool needPar = false;
2524                                 column += running_font.latexWriteEndChanges(
2525                                                 os, bparams, runparams,
2526                                                 basefont, basefont, needPar);
2527                                 open_font = false;
2528                         }
2529                         basefont = (body_pos > i) ? getLabelFont(bparams, outerfont)
2530                                                   : getLayoutFont(bparams, outerfont);
2531                         running_font = basefont;
2532                         column += Changes::latexMarkChange(os, bparams, runningChange,
2533                                                            change, runparams);
2534                         runningChange = change;
2535                 }
2536
2537                 // do not output text which is marked deleted
2538                 // if change tracking output is disabled
2539                 if (!bparams.output_changes && change.deleted()) {
2540                         continue;
2541                 }
2542
2543                 ++column;
2544
2545                 // Fully instantiated font
2546                 Font const current_font = getFont(bparams, i, outerfont);
2547
2548                 Font const last_font = running_font;
2549
2550                 // Do we need to close the previous font?
2551                 if (open_font &&
2552                     (current_font != running_font ||
2553                      current_font.language() != running_font.language()))
2554                 {
2555                         // ensure there is no open script-wrapper
2556                         if (!alien_script.empty()) {
2557                                 column += 1;
2558                                 os << "}";
2559                                 alien_script.clear();
2560                         }
2561                         bool needPar = false;
2562                         column += running_font.latexWriteEndChanges(
2563                                     os, bparams, runparams, basefont,
2564                                     (i == body_pos-1) ? basefont : current_font,
2565                                     needPar);
2566                         running_font = basefont;
2567                         open_font = false;
2568                 }
2569
2570                 // if necessary, close language environment before opening CJK
2571                 string const running_lang = running_font.language()->babel();
2572                 string const lang_end_command = lyxrc.language_command_end;
2573                 if (!lang_end_command.empty() && !bparams.useNonTeXFonts
2574                         && !running_lang.empty()
2575                         && running_lang == openLanguageName()
2576                         && current_font.language()->encoding()->package() == Encoding::CJK) {
2577                         string end_tag = subst(lang_end_command, "$$lang", running_lang);
2578                         os << from_ascii(end_tag);
2579                         column += end_tag.length();
2580                         popLanguageName();
2581                 }
2582
2583                 // Switch file encoding if necessary (and allowed)
2584                 if (!runparams.pass_thru && !style.pass_thru &&
2585                     runparams.encoding->package() != Encoding::none &&
2586                     current_font.language()->encoding()->package() != Encoding::none) {
2587                         pair<bool, int> const enc_switch =
2588                                 switchEncoding(os.os(), bparams, runparams,
2589                                         *(current_font.language()->encoding()));
2590                         if (enc_switch.first) {
2591                                 column += enc_switch.second;
2592                                 runparams.encoding = current_font.language()->encoding();
2593                         }
2594                 }
2595
2596                 char_type const c = d->text_[i];
2597
2598                 // A display math inset inside an ulem command will be output
2599                 // as a box of width \linewidth, so we have to either disable
2600                 // indentation if the inset starts a paragraph, or start a new
2601                 // line to accommodate such box. This has to be done before
2602                 // writing any font changing commands.
2603                 if (runparams.inDisplayMath && !deleted_display_math
2604                     && runparams.inulemcmd) {
2605                         if (os.afterParbreak())
2606                                 os << "\\noindent";
2607                         else
2608                                 os << "\\\\\n";
2609                 }
2610
2611                 // Do we need to change font?
2612                 if ((current_font != running_font ||
2613                      current_font.language() != running_font.language()) &&
2614                         i != body_pos - 1)
2615                 {
2616                         odocstringstream ods;
2617                         column += current_font.latexWriteStartChanges(ods, bparams,
2618                                                               runparams, basefont,
2619                                                               last_font);
2620                         // Check again for display math in ulem commands as a
2621                         // font change may also occur just before a math inset.
2622                         if (runparams.inDisplayMath && !deleted_display_math
2623                             && runparams.inulemcmd) {
2624                                 if (os.afterParbreak())
2625                                         os << "\\noindent";
2626                                 else
2627                                         os << "\\\\\n";
2628                         }
2629                         running_font = current_font;
2630                         open_font = true;
2631                         docstring fontchange = ods.str();
2632                         // check whether the fontchange ends with a \\textcolor
2633                         // modifier and the text starts with a space (bug 4473)
2634                         docstring const last_modifier = rsplit(fontchange, '\\');
2635                         if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2636                                 os << fontchange << from_ascii("{}");
2637                         // check if the fontchange ends with a trailing blank
2638                         // (like "\small " (see bug 3382)
2639                         else if (suffixIs(fontchange, ' ') && c == ' ')
2640                                 os << fontchange.substr(0, fontchange.size() - 1)
2641                                    << from_ascii("{}");
2642                         else
2643                                 os << fontchange;
2644                 }
2645
2646                 // FIXME: think about end_pos implementation...
2647                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2648                         // FIXME: integrate this case in latexSpecialChar
2649                         // Do not print the separation of the optional argument
2650                         // if style.pass_thru is false. This works because
2651                         // latexSpecialChar ignores spaces if
2652                         // style.pass_thru is false.
2653                         if (i != body_pos - 1) {
2654                                 if (d->simpleTeXBlanks(bparams, runparams, os,
2655                                                 i, column, current_font, style)) {
2656                                         // A surrogate pair was output. We
2657                                         // must not call latexSpecialChar
2658                                         // in this iteration, since it would output
2659                                         // the combining character again.
2660                                         ++i;
2661                                         continue;
2662                                 }
2663                         }
2664                 }
2665
2666                 OutputParams rp = runparams;
2667                 rp.free_spacing = style.free_spacing;
2668                 rp.local_font = &current_font;
2669                 rp.intitle = style.intitle;
2670
2671                 // Two major modes:  LaTeX or plain
2672                 // Handle here those cases common to both modes
2673                 // and then split to handle the two modes separately.
2674                 if (c == META_INSET) {
2675                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2676                                 // Greyedout notes and, in general, all insets
2677                                 // with InsetLayout::isDisplay() == false,
2678                                 // are typeset inline with the text. So, we
2679                                 // can add a \par to the last paragraph of
2680                                 // such insets only if nothing else follows.
2681                                 bool incremented = false;
2682                                 Inset const * inset = getInset(i);
2683                                 InsetText const * textinset = inset
2684                                                         ? inset->asInsetText()
2685                                                         : 0;
2686                                 if (i + 1 == size() && textinset
2687                                     && !inset->getLayout().isDisplay()) {
2688                                         ParagraphList const & pars =
2689                                                 textinset->text().paragraphs();
2690                                         pit_type const pit = pars.size() - 1;
2691                                         Font const lastfont =
2692                                                 pit < 0 || pars[pit].empty()
2693                                                 ? pars[pit].getLayoutFont(
2694                                                                 bparams,
2695                                                                 outerfont)
2696                                                 : pars[pit].getFont(bparams,
2697                                                         pars[pit].size() - 1,
2698                                                         outerfont);
2699                                         if (lastfont.fontInfo().size() !=
2700                                             basefont.fontInfo().size()) {
2701                                                 ++parInline;
2702                                                 incremented = true;
2703                                         }
2704                                 }
2705                                 d->latexInset(bparams, os, rp, running_font,
2706                                                 basefont, outerfont, open_font,
2707                                                 runningChange, style, i, column);
2708                                 if (incremented)
2709                                         --parInline;
2710                         }
2711                 } else if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2712                         if (!bparams.useNonTeXFonts)
2713                           script = Encodings::isKnownScriptChar(c);
2714                         if (script != alien_script) {
2715                                 if (!alien_script.empty()) {
2716                                         os << "}";
2717                                         alien_script.clear();
2718                                 }
2719                                 string fontenc = running_font.language()->fontenc(bparams);
2720                                 if (!script.empty()
2721                                         && !Encodings::fontencSupportsScript(fontenc, script)) {
2722                                         column += script.length() + 2;
2723                                         os << "\\" << script << "{";
2724                                         alien_script = script;
2725                                 }
2726                         }
2727                         try {
2728                                 d->latexSpecialChar(os, bparams, rp, running_font,
2729                                                                         alien_script, style, i, end_pos, column);
2730                         } catch (EncodingException & e) {
2731                                 if (runparams.dryrun) {
2732                                         os << "<" << _("LyX Warning: ")
2733                                            << _("uncodable character") << " '";
2734                                         os.put(c);
2735                                         os << "'>";
2736                                 } else {
2737                                         // add location information and throw again.
2738                                         e.par_id = id();
2739                                         e.pos = i;
2740                                         throw(e);
2741                                 }
2742                         }
2743                 }
2744
2745                 // Set the encoding to that returned from latexSpecialChar (see
2746                 // comment for encoding member in OutputParams.h)
2747                 runparams.encoding = rp.encoding;
2748
2749                 // Also carry on the info on a closed ulem command for insets
2750                 // such as Note that do not produce any output, so that no
2751                 // command is ever executed but its opening was recorded.
2752                 runparams.inulemcmd = rp.inulemcmd;
2753
2754                 // And finally, pass the post_macros upstream
2755                 runparams.post_macro = rp.post_macro;
2756         }
2757
2758         // Close wrapper for alien script
2759         if (!alien_script.empty()) {
2760                 os << "}";
2761                 alien_script.clear();
2762         }
2763
2764         // If we have an open font definition, we have to close it
2765         if (open_font) {
2766                 // Make sure that \\par is done with the font of the last
2767                 // character if this has another size as the default.
2768                 // This is necessary because LaTeX (and LyX on the screen)
2769                 // calculates the space between the baselines according
2770                 // to this font. (Matthias)
2771                 //
2772                 // We must not change the font for the last paragraph
2773                 // of non-multipar insets, tabular cells or commands,
2774                 // since this produces unwanted whitespace.
2775
2776                 Font const font = empty()
2777                         ? getLayoutFont(bparams, outerfont)
2778                         : getFont(bparams, size() - 1, outerfont);
2779
2780                 InsetText const * textinset = inInset().asInsetText();
2781
2782                 bool const maintext = textinset
2783                         ? textinset->text().isMainText()
2784                         : false;
2785
2786                 size_t const numpars = textinset
2787                         ? textinset->text().paragraphs().size()
2788                         : 0;
2789
2790                 bool needPar = false;
2791
2792                 if (style.resfont.size() != font.fontInfo().size()
2793                     && (!runparams.isLastPar || maintext
2794                         || (numpars > 1 && d->ownerCode() != CELL_CODE
2795                             && (inInset().getLayout().isDisplay()
2796                                 || parInline)))
2797                     && !style.isCommand()) {
2798                         needPar = true;
2799                 }
2800 #ifdef FIXED_LANGUAGE_END_DETECTION
2801                 if (next_) {
2802                         running_font.latexWriteEndChanges(os, bparams,
2803                                         runparams, basefont,
2804                                         next_->getFont(bparams, 0, outerfont),
2805                                                        needPar);
2806                 } else {
2807                         running_font.latexWriteEndChanges(os, bparams,
2808                                         runparams, basefont, basefont, needPar);
2809                 }
2810 #else
2811 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2812 //FIXME: there as we start another \selectlanguage with the next paragraph if
2813 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2814                 running_font.latexWriteEndChanges(os, bparams, runparams,
2815                                 basefont, basefont, needPar);
2816 #endif
2817                 if (needPar) {
2818                         // The \par could not be inserted at the same nesting
2819                         // level of the font size change, so do it now.
2820                         os << "{\\" << font.latexSize() << "\\par}";
2821                 }
2822         }
2823
2824         column += Changes::latexMarkChange(os, bparams, runningChange,
2825                                            Change(Change::UNCHANGED), runparams);
2826
2827         // Needed if there is an optional argument but no contents.
2828         if (body_pos > 0 && body_pos == size()) {
2829                 os << "}]~";
2830         }
2831
2832         if (!style.rightdelim().empty()) {
2833                 os << style.rightdelim();
2834                 column += style.rightdelim().size();
2835         }
2836
2837         if (allowcust && d->endTeXParParams(bparams, os, runparams)
2838             && runparams.encoding != prev_encoding) {
2839                 runparams.encoding = prev_encoding;
2840                 os << setEncoding(prev_encoding->iconvName());
2841         }
2842
2843         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2844 }
2845
2846
2847 bool Paragraph::emptyTag() const
2848 {
2849         for (pos_type i = 0; i < size(); ++i) {
2850                 if (Inset const * inset = getInset(i)) {
2851                         InsetCode lyx_code = inset->lyxCode();
2852                         // FIXME testing like that is wrong. What is
2853                         // the intent?
2854                         if (lyx_code != TOC_CODE &&
2855                             lyx_code != INCLUDE_CODE &&
2856                             lyx_code != GRAPHICS_CODE &&
2857                             lyx_code != ERT_CODE &&
2858                             lyx_code != LISTINGS_CODE &&
2859                             lyx_code != FLOAT_CODE &&
2860                             lyx_code != TABULAR_CODE) {
2861                                 return false;
2862                         }
2863                 } else {
2864                         char_type c = d->text_[i];
2865                         if (c != ' ' && c != '\t')
2866                                 return false;
2867                 }
2868         }
2869         return true;
2870 }
2871
2872
2873 string Paragraph::getID(Buffer const & buf, OutputParams const & runparams)
2874         const
2875 {
2876         for (pos_type i = 0; i < size(); ++i) {
2877                 if (Inset const * inset = getInset(i)) {
2878                         InsetCode lyx_code = inset->lyxCode();
2879                         if (lyx_code == LABEL_CODE) {
2880                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2881                                 docstring const & id = il->getParam("name");
2882                                 return "id='" + to_utf8(sgml::cleanID(buf, runparams, id)) + "'";
2883                         }
2884                 }
2885         }
2886         return string();
2887 }
2888
2889
2890 pos_type Paragraph::firstWordDocBook(odocstream & os, OutputParams const & runparams)
2891         const
2892 {
2893         pos_type i;
2894         for (i = 0; i < size(); ++i) {
2895                 if (Inset const * inset = getInset(i)) {
2896                         inset->docbook(os, runparams);
2897                 } else {
2898                         char_type c = d->text_[i];
2899                         if (c == ' ')
2900                                 break;
2901                         os << sgml::escapeChar(c);
2902                 }
2903         }
2904         return i;
2905 }
2906
2907
2908 pos_type Paragraph::firstWordLyXHTML(XHTMLStream & xs, OutputParams const & runparams)
2909         const
2910 {
2911         pos_type i;
2912         for (i = 0; i < size(); ++i) {
2913                 if (Inset const * inset = getInset(i)) {
2914                         inset->xhtml(xs, runparams);
2915                 } else {
2916                         char_type c = d->text_[i];
2917                         if (c == ' ')
2918                                 break;
2919                         xs << c;
2920                 }
2921         }
2922         return i;
2923 }
2924
2925
2926 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2927 {
2928         Font font_old;
2929         pos_type size = text_.size();
2930         for (pos_type i = initial; i < size; ++i) {
2931                 Font font = owner_->getFont(buf.params(), i, outerfont);
2932                 if (text_[i] == META_INSET)
2933                         return false;
2934                 if (i != initial && font != font_old)
2935                         return false;
2936                 font_old = font;
2937         }
2938
2939         return true;
2940 }
2941
2942
2943 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
2944                                     odocstream & os,
2945                                     OutputParams const & runparams,
2946                                     Font const & outerfont,
2947                                     pos_type initial) const
2948 {
2949         bool emph_flag = false;
2950
2951         Layout const & style = *d->layout_;
2952         FontInfo font_old =
2953                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
2954
2955         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2956                 os << "]]>";
2957
2958         // parsing main loop
2959         for (pos_type i = initial; i < size(); ++i) {
2960                 Font font = getFont(buf.params(), i, outerfont);
2961
2962                 // handle <emphasis> tag
2963                 if (font_old.emph() != font.fontInfo().emph()) {
2964                         if (font.fontInfo().emph() == FONT_ON) {
2965                                 os << "<emphasis>";
2966                                 emph_flag = true;
2967                         } else if (i != initial) {
2968                                 os << "</emphasis>";
2969                                 emph_flag = false;
2970                         }
2971                 }
2972
2973                 if (Inset const * inset = getInset(i)) {
2974                         inset->docbook(os, runparams);
2975                 } else {
2976                         char_type c = d->text_[i];
2977
2978                         if (style.pass_thru)
2979                                 os.put(c);
2980                         else
2981                                 os << sgml::escapeChar(c);
2982                 }
2983                 font_old = font.fontInfo();
2984         }
2985
2986         if (emph_flag) {
2987                 os << "</emphasis>";
2988         }
2989
2990         if (style.free_spacing)
2991                 os << '\n';
2992         if (style.pass_thru && !d->onlyText(buf, outerfont, initial))
2993                 os << "<![CDATA[";
2994 }
2995
2996
2997 namespace {
2998 void doFontSwitch(vector<html::FontTag> & tagsToOpen,
2999                   vector<html::EndFontTag> & tagsToClose,
3000                   bool & flag, FontState curstate, html::FontTypes type)
3001 {
3002         if (curstate == FONT_ON) {
3003                 tagsToOpen.push_back(html::FontTag(type));
3004                 flag = true;
3005         } else if (flag) {
3006                 tagsToClose.push_back(html::EndFontTag(type));
3007                 flag = false;
3008         }
3009 }
3010 } // namespace
3011
3012
3013 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
3014                                     XHTMLStream & xs,
3015                                     OutputParams const & runparams,
3016                                     Font const & outerfont,
3017                                     bool start_paragraph, bool close_paragraph,
3018                                     pos_type initial) const
3019 {
3020         docstring retval;
3021
3022         // track whether we have opened these tags
3023         bool emph_flag = false;
3024         bool bold_flag = false;
3025         bool noun_flag = false;
3026         bool ubar_flag = false;
3027         bool dbar_flag = false;
3028         bool sout_flag = false;
3029         bool xout_flag = false;
3030         bool wave_flag = false;
3031         // shape tags
3032         bool shap_flag = false;
3033         // family tags
3034         bool faml_flag = false;
3035         // size tags
3036         bool size_flag = false;
3037
3038         Layout const & style = *d->layout_;
3039
3040         if (start_paragraph)
3041                 xs.startDivision(allowEmpty());
3042
3043         FontInfo font_old =
3044                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
3045
3046         FontShape  curr_fs   = INHERIT_SHAPE;
3047         FontFamily curr_fam  = INHERIT_FAMILY;
3048         FontSize   curr_size = INHERIT_SIZE;
3049
3050         string const default_family =
3051                 buf.masterBuffer()->params().fonts_default_family;
3052
3053         vector<html::FontTag> tagsToOpen;
3054         vector<html::EndFontTag> tagsToClose;
3055
3056         // parsing main loop
3057         for (pos_type i = initial; i < size(); ++i) {
3058                 // let's not show deleted material in the output
3059                 if (isDeleted(i))
3060                         continue;
3061
3062                 Font const font = getFont(buf.masterBuffer()->params(), i, outerfont);
3063
3064                 // emphasis
3065                 FontState curstate = font.fontInfo().emph();
3066                 if (font_old.emph() != curstate)
3067                         doFontSwitch(tagsToOpen, tagsToClose, emph_flag, curstate, html::FT_EMPH);
3068
3069                 // noun
3070                 curstate = font.fontInfo().noun();
3071                 if (font_old.noun() != curstate)
3072                         doFontSwitch(tagsToOpen, tagsToClose, noun_flag, curstate, html::FT_NOUN);
3073
3074                 // underbar
3075                 curstate = font.fontInfo().underbar();
3076                 if (font_old.underbar() != curstate)
3077                         doFontSwitch(tagsToOpen, tagsToClose, ubar_flag, curstate, html::FT_UBAR);
3078
3079                 // strikeout
3080                 curstate = font.fontInfo().strikeout();
3081                 if (font_old.strikeout() != curstate)
3082                         doFontSwitch(tagsToOpen, tagsToClose, sout_flag, curstate, html::FT_SOUT);
3083
3084                 // xout
3085                 curstate = font.fontInfo().xout();
3086                 if (font_old.xout() != curstate)
3087                         doFontSwitch(tagsToOpen, tagsToClose, xout_flag, curstate, html::FT_XOUT);
3088
3089                 // double underbar
3090                 curstate = font.fontInfo().uuline();
3091                 if (font_old.uuline() != curstate)
3092                         doFontSwitch(tagsToOpen, tagsToClose, dbar_flag, curstate, html::FT_DBAR);
3093
3094                 // wavy line
3095                 curstate = font.fontInfo().uwave();
3096                 if (font_old.uwave() != curstate)
3097                         doFontSwitch(tagsToOpen, tagsToClose, wave_flag, curstate, html::FT_WAVE);
3098
3099                 // bold
3100                 // a little hackish, but allows us to reuse what we have.
3101                 curstate = (font.fontInfo().series() == BOLD_SERIES ? FONT_ON : FONT_OFF);
3102                 if (font_old.series() != font.fontInfo().series())
3103                         doFontSwitch(tagsToOpen, tagsToClose, bold_flag, curstate, html::FT_BOLD);
3104
3105                 // Font shape
3106                 curr_fs = font.fontInfo().shape();
3107                 FontShape old_fs = font_old.shape();
3108                 if (old_fs != curr_fs) {
3109                         if (shap_flag) {
3110                                 switch (old_fs) {
3111                                 case ITALIC_SHAPE:
3112                                         tagsToClose.push_back(html::EndFontTag(html::FT_ITALIC));
3113                                         break;
3114                                 case SLANTED_SHAPE:
3115                                         tagsToClose.push_back(html::EndFontTag(html::FT_SLANTED));
3116                                         break;
3117                                 case SMALLCAPS_SHAPE:
3118                                         tagsToClose.push_back(html::EndFontTag(html::FT_SMALLCAPS));
3119                                         break;
3120                                 case UP_SHAPE:
3121                                 case INHERIT_SHAPE:
3122                                         break;
3123                                 default:
3124                                         // the other tags are for internal use
3125                                         LATTEST(false);
3126                                         break;
3127                                 }
3128                                 shap_flag = false;
3129                         }
3130                         switch (curr_fs) {
3131                         case ITALIC_SHAPE:
3132                                 tagsToOpen.push_back(html::FontTag(html::FT_ITALIC));
3133                                 shap_flag = true;
3134                                 break;
3135                         case SLANTED_SHAPE:
3136                                 tagsToOpen.push_back(html::FontTag(html::FT_SLANTED));
3137                                 shap_flag = true;
3138                                 break;
3139                         case SMALLCAPS_SHAPE:
3140                                 tagsToOpen.push_back(html::FontTag(html::FT_SMALLCAPS));
3141                                 shap_flag = true;
3142                                 break;
3143                         case UP_SHAPE:
3144                         case INHERIT_SHAPE:
3145                                 break;
3146                         default:
3147                                 // the other tags are for internal use
3148                                 LATTEST(false);
3149                                 break;
3150                         }
3151                 }
3152
3153                 // Font family
3154                 curr_fam = font.fontInfo().family();
3155                 FontFamily old_fam = font_old.family();
3156                 if (old_fam != curr_fam) {
3157                         if (faml_flag) {
3158                                 switch (old_fam) {
3159                                 case ROMAN_FAMILY:
3160                                         tagsToClose.push_back(html::EndFontTag(html::FT_ROMAN));
3161                                         break;
3162                                 case SANS_FAMILY:
3163                                         tagsToClose.push_back(html::EndFontTag(html::FT_SANS));
3164                                         break;
3165                                 case TYPEWRITER_FAMILY:
3166                                         tagsToClose.push_back(html::EndFontTag(html::FT_TYPE));
3167                                         break;
3168                                 case INHERIT_FAMILY:
3169                                         break;
3170                                 default:
3171                                         // the other tags are for internal use
3172                                         LATTEST(false);
3173                                         break;
3174                                 }
3175                                 faml_flag = false;
3176                         }
3177                         switch (curr_fam) {
3178                         case ROMAN_FAMILY:
3179                                 // we will treat a "default" font family as roman, since we have
3180                                 // no other idea what to do.
3181                                 if (default_family != "rmdefault" && default_family != "default") {
3182                                         tagsToOpen.push_back(html::FontTag(html::FT_ROMAN));
3183                                         faml_flag = true;
3184                                 }
3185                                 break;
3186                         case SANS_FAMILY:
3187                                 if (default_family != "sfdefault") {
3188                                         tagsToOpen.push_back(html::FontTag(html::FT_SANS));
3189                                         faml_flag = true;
3190                                 }
3191                                 break;
3192                         case TYPEWRITER_FAMILY:
3193                                 if (default_family != "ttdefault") {
3194                                         tagsToOpen.push_back(html::FontTag(html::FT_TYPE));
3195                                         faml_flag = true;
3196                                 }
3197                                 break;
3198                         case INHERIT_FAMILY:
3199                                 break;
3200                         default:
3201                                 // the other tags are for internal use
3202                                 LATTEST(false);
3203                                 break;
3204                         }
3205                 }
3206
3207                 // Font size
3208                 curr_size = font.fontInfo().size();
3209                 FontSize old_size = font_old.size();
3210                 if (old_size != curr_size) {
3211                         if (size_flag) {
3212                                 switch (old_size) {
3213                                 case TINY_SIZE:
3214                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_TINY));
3215                                         break;
3216                                 case SCRIPT_SIZE:
3217                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SCRIPT));
3218                                         break;
3219                                 case FOOTNOTE_SIZE:
3220                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_FOOTNOTE));
3221                                         break;
3222                                 case SMALL_SIZE:
3223                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_SMALL));
3224                                         break;
3225                                 case LARGE_SIZE:
3226                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGE));
3227                                         break;
3228                                 case LARGER_SIZE:
3229                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGER));
3230                                         break;
3231                                 case LARGEST_SIZE:
3232                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_LARGEST));
3233                                         break;
3234                                 case HUGE_SIZE:
3235                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGE));
3236                                         break;
3237                                 case HUGER_SIZE:
3238                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_HUGER));
3239                                         break;
3240                                 case INCREASE_SIZE:
3241                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_INCREASE));
3242                                         break;
3243                                 case DECREASE_SIZE:
3244                                         tagsToClose.push_back(html::EndFontTag(html::FT_SIZE_DECREASE));
3245                                         break;
3246                                 case INHERIT_SIZE:
3247                                 case NORMAL_SIZE:
3248                                         break;
3249                                 default:
3250                                         // the other tags are for internal use
3251                                         LATTEST(false);
3252                                         break;
3253                                 }
3254                                 size_flag = false;
3255                         }
3256                         switch (curr_size) {
3257                         case TINY_SIZE:
3258                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_TINY));
3259                                 size_flag = true;
3260                                 break;
3261                         case SCRIPT_SIZE:
3262                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SCRIPT));
3263                                 size_flag = true;
3264                                 break;
3265                         case FOOTNOTE_SIZE:
3266                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_FOOTNOTE));
3267                                 size_flag = true;
3268                                 break;
3269                         case SMALL_SIZE:
3270                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_SMALL));
3271                                 size_flag = true;
3272                                 break;
3273                         case LARGE_SIZE:
3274                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGE));
3275                                 size_flag = true;
3276                                 break;
3277                         case LARGER_SIZE:
3278                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGER));
3279                                 size_flag = true;
3280                                 break;
3281                         case LARGEST_SIZE:
3282                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_LARGEST));
3283                                 size_flag = true;
3284                                 break;
3285                         case HUGE_SIZE:
3286                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGE));
3287                                 size_flag = true;
3288                                 break;
3289                         case HUGER_SIZE:
3290                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_HUGER));
3291                                 size_flag = true;
3292                                 break;
3293                         case INCREASE_SIZE:
3294                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_INCREASE));
3295                                 size_flag = true;
3296                                 break;
3297                         case DECREASE_SIZE:
3298                                 tagsToOpen.push_back(html::FontTag(html::FT_SIZE_DECREASE));
3299                                 size_flag = true;
3300                                 break;
3301                         case NORMAL_SIZE:
3302                         case INHERIT_SIZE:
3303                                 break;
3304                         default:
3305                                 // the other tags are for internal use
3306                                 LATTEST(false);
3307                                 break;
3308                         }
3309                 }
3310
3311                 // FIXME XHTML
3312                 // Other such tags? What about the other text ranges?
3313
3314                 vector<html::EndFontTag>::const_iterator cit = tagsToClose.begin();
3315                 vector<html::EndFontTag>::const_iterator cen = tagsToClose.end();
3316                 for (; cit != cen; ++cit)
3317                         xs << *cit;
3318
3319                 vector<html::FontTag>::const_iterator sit = tagsToOpen.begin();
3320                 vector<html::FontTag>::const_iterator sen = tagsToOpen.end();
3321                 for (; sit != sen; ++sit)
3322                         xs << *sit;
3323
3324                 tagsToClose.clear();
3325                 tagsToOpen.clear();
3326
3327                 Inset const * inset = getInset(i);
3328                 if (inset) {
3329                         if (!runparams.for_toc || inset->isInToc()) {
3330                                 OutputParams np = runparams;
3331                                 np.local_font = &font;
3332                                 // If the paragraph has size 1, then we are in the "special
3333                                 // case" where we do not output the containing paragraph info
3334                                 if (!inset->getLayout().htmlisblock() && size() != 1)
3335                                         np.html_in_par = true;
3336                                 retval += inset->xhtml(xs, np);
3337                         }
3338                 } else {
3339                         char_type c = getUChar(buf.masterBuffer()->params(),
3340                                                runparams, i);
3341                         if (c == ' ' && (style.free_spacing || runparams.free_spacing))
3342                                 xs << XHTMLStream::ESCAPE_NONE << "&nbsp;";
3343                         else
3344                                 xs << c;
3345                 }
3346                 font_old = font.fontInfo();
3347         }
3348
3349         // FIXME XHTML
3350         // I'm worried about what happens if a branch, say, is itself
3351         // wrapped in some font stuff. I think that will not work.
3352         xs.closeFontTags();
3353         if (close_paragraph)
3354                 xs.endDivision();
3355
3356         return retval;
3357 }
3358
3359
3360 bool Paragraph::isHfill(pos_type pos) const
3361 {
3362         Inset const * inset = getInset(pos);
3363         return inset && inset->isHfill();
3364 }
3365
3366
3367 bool Paragraph::isNewline(pos_type pos) const
3368 {
3369         // U+2028 LINE SEPARATOR
3370         // U+2029 PARAGRAPH SEPARATOR
3371         char_type const c = d->text_[pos];
3372         if (c == 0x2028 || c == 0x2029)
3373                 return true;
3374         Inset const * inset = getInset(pos);
3375         return inset && inset->lyxCode() == NEWLINE_CODE;
3376 }
3377
3378
3379 bool Paragraph::isEnvSeparator(pos_type pos) const
3380 {
3381         Inset const * inset = getInset(pos);
3382         return inset && inset->lyxCode() == SEPARATOR_CODE;
3383 }
3384
3385
3386 bool Paragraph::isLineSeparator(pos_type pos) const
3387 {
3388         char_type const c = d->text_[pos];
3389         if (isLineSeparatorChar(c))
3390                 return true;
3391         Inset const * inset = getInset(pos);
3392         return inset && inset->isLineSeparator();
3393 }
3394
3395
3396 bool Paragraph::isWordSeparator(pos_type pos) const
3397 {
3398         if (pos == size())
3399                 return true;
3400         if (Inset const * inset = getInset(pos))
3401                 return !inset->isLetter();
3402         // if we have a hard hyphen (no en- or emdash) or apostrophe
3403         // we pass this to the spell checker
3404         // FIXME: this method is subject to change, visit
3405         // https://bugzilla.mozilla.org/show_bug.cgi?id=355178
3406         // to get an impression how complex this is.
3407         if (isHardHyphenOrApostrophe(pos))
3408                 return false;
3409         char_type const c = d->text_[pos];
3410         // We want to pass the escape chars to the spellchecker
3411         docstring const escape_chars = from_utf8(lyxrc.spellchecker_esc_chars);
3412         return !isLetterChar(c) && !isDigitASCII(c) && !contains(escape_chars, c);
3413 }
3414
3415
3416 bool Paragraph::isHardHyphenOrApostrophe(pos_type pos) const
3417 {
3418         pos_type const psize = size();
3419         if (pos >= psize)
3420                 return false;
3421         char_type const c = d->text_[pos];
3422         if (c != '-' && c != '\'')
3423                 return false;
3424         int nextpos = pos + 1;
3425         int prevpos = pos > 0 ? pos - 1 : 0;
3426         if ((nextpos == psize || isSpace(nextpos))
3427                 && (pos == 0 || isSpace(prevpos)))
3428                 return false;
3429         return true;
3430 }
3431
3432
3433 bool Paragraph::needsCProtection(bool const fragile) const
3434 {
3435         // first check the layout of the paragraph, but only in insets
3436         InsetText const * textinset = inInset().asInsetText();
3437         bool const maintext = textinset
3438                 ? textinset->text().isMainText()
3439                 : false;
3440
3441         if (!maintext && layout().needcprotect) {
3442                 // Environments need cprotection regardless the content
3443                 if (layout().latextype == LATEX_ENVIRONMENT)
3444                         return true;
3445
3446                 // Commands need cprotection if they contain specific chars
3447                 int const nchars_escape = 9;
3448                 static char_type const chars_escape[nchars_escape] = {
3449                         '&', '_', '$', '%', '#', '^', '{', '}', '\\'};
3450
3451                 docstring const pars = asString();
3452                 for (int k = 0; k < nchars_escape; k++) {
3453                         if (contains(pars, chars_escape[k]))
3454                                 return true;
3455                 }
3456         }
3457
3458         // now check whether we have insets that need cprotection
3459         pos_type size = pos_type(d->text_.size());
3460         for (pos_type i = 0; i < size; ++i) {
3461                 if (!isInset(i))
3462                         continue;
3463                 Inset const * ins = getInset(i);
3464                 if (ins->needsCProtection(maintext, fragile))
3465                         return true;
3466                 if (ins->getLayout().latextype() == InsetLayout::ENVIRONMENT)
3467                         // Environments need cprotection regardless the content
3468                         return true;
3469                 // Now check math environments
3470                 InsetMath const * im = getInset(i)->asInsetMath();
3471                 if (!im || im->cell(0).empty())
3472                         continue;
3473                 switch(im->cell(0)[0]->lyxCode()) {
3474                 case MATH_AMSARRAY_CODE:
3475                 case MATH_SUBSTACK_CODE:
3476                 case MATH_ENV_CODE:
3477                 case MATH_XYMATRIX_CODE:
3478                         // these need cprotection
3479                         return true;
3480                 default:
3481                         break;
3482                 }
3483         }
3484
3485         return false;
3486 }
3487
3488
3489 FontSpan const & Paragraph::getSpellRange(pos_type pos) const
3490 {
3491         return d->speller_state_.getRange(pos);
3492 }
3493
3494
3495 bool Paragraph::isChar(pos_type pos) const
3496 {
3497         if (Inset const * inset = getInset(pos))
3498                 return inset->isChar();
3499         char_type const c = d->text_[pos];
3500         return !isLetterChar(c) && !isDigitASCII(c) && !lyx::isSpace(c);
3501 }
3502
3503
3504 bool Paragraph::isSpace(pos_type pos) const
3505 {
3506         if (Inset const * inset = getInset(pos))
3507                 return inset->isSpace();
3508         char_type const c = d->text_[pos];
3509         return lyx::isSpace(c);
3510 }
3511
3512
3513 Language const *
3514 Paragraph::getParLanguage(BufferParams const & bparams) const
3515 {
3516         if (!empty())
3517                 return getFirstFontSettings(bparams).language();
3518         // FIXME: we should check the prev par as well (Lgb)
3519         return bparams.language;
3520 }
3521
3522
3523 bool Paragraph::isRTL(BufferParams const & bparams) const
3524 {
3525         return getParLanguage(bparams)->rightToLeft()
3526                 && !inInset().getLayout().forceLTR();
3527 }
3528
3529
3530 void Paragraph::changeLanguage(BufferParams const & bparams,
3531                                Language const * from, Language const * to)
3532 {
3533         // change language including dummy font change at the end
3534         for (pos_type i = 0; i <= size(); ++i) {
3535                 Font font = getFontSettings(bparams, i);
3536                 if (font.language() == from) {
3537                         font.setLanguage(to);
3538                         setFont(i, font);
3539                         d->requestSpellCheck(i);
3540                 }
3541         }
3542 }
3543
3544
3545 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
3546 {
3547         Language const * doc_language = bparams.language;
3548         for (auto const & f : d->fontlist_)
3549                 if (f.font().language() != ignore_language &&
3550                     f.font().language() != latex_language &&
3551                     f.font().language() != doc_language)
3552                         return true;
3553         return false;
3554 }
3555
3556
3557 void Paragraph::getLanguages(std::set<Language const *> & langs) const
3558 {
3559         for (auto const & f : d->fontlist_) {
3560                 Language const * lang = f.font().language();
3561                 if (lang != ignore_language &&
3562                     lang != latex_language)
3563                         langs.insert(lang);
3564         }
3565 }
3566
3567
3568 docstring Paragraph::asString(int options) const
3569 {
3570         return asString(0, size(), options);
3571 }
3572
3573
3574 docstring Paragraph::asString(pos_type beg, pos_type end, int options, const OutputParams *runparams) const
3575 {
3576         odocstringstream os;
3577
3578         if (beg == 0
3579             && options & AS_STR_LABEL
3580             && !d->params_.labelString().empty())
3581                 os << d->params_.labelString() << ' ';
3582
3583         for (pos_type i = beg; i < end; ++i) {
3584                 if ((options & AS_STR_SKIPDELETE) && isDeleted(i))
3585                         continue;
3586                 char_type const c = d->text_[i];
3587                 if (isPrintable(c) || c == '\t'
3588                     || (c == '\n' && (options & AS_STR_NEWLINES)))
3589                         os.put(c);
3590                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
3591                         if (c == META_INSET && (options & AS_STR_PLAINTEXT)) {
3592                                 LASSERT(runparams != 0, return docstring());
3593                                 getInset(i)->plaintext(os, *runparams);
3594                         } else {
3595                                 getInset(i)->toString(os);
3596                         }
3597                 }
3598         }
3599
3600         return os.str();
3601 }
3602
3603
3604 void Paragraph::forOutliner(docstring & os, size_t const maxlen,
3605                             bool const shorten, bool const label) const
3606 {
3607         size_t tmplen = shorten ? maxlen + 1 : maxlen;
3608         if (label && !labelString().empty())
3609                 os += labelString() + ' ';
3610         if (!layout().isTocCaption())
3611                 return;
3612         for (pos_type i = 0; i < size() && os.length() < tmplen; ++i) {
3613                 if (isDeleted(i))
3614                         continue;
3615                 char_type const c = d->text_[i];
3616                 if (isPrintable(c))
3617                         os += c;
3618                 else if (c == META_INSET)
3619                         getInset(i)->forOutliner(os, tmplen, false);
3620         }
3621         if (shorten)
3622                 Text::shortenForOutliner(os, maxlen);
3623 }
3624
3625
3626 void Paragraph::setInsetOwner(Inset const * inset)
3627 {
3628         d->inset_owner_ = inset;
3629 }
3630
3631
3632 int Paragraph::id() const
3633 {
3634         return d->id_;
3635 }
3636
3637
3638 void Paragraph::setId(int id)
3639 {
3640         d->id_ = id;
3641 }
3642
3643
3644 Layout const & Paragraph::layout() const
3645 {
3646         return *d->layout_;
3647 }
3648
3649
3650 void Paragraph::setLayout(Layout const & layout)
3651 {
3652         d->layout_ = &layout;
3653 }
3654
3655
3656 void Paragraph::setDefaultLayout(DocumentClass const & tc)
3657 {
3658         setLayout(tc.defaultLayout());
3659 }
3660
3661
3662 void Paragraph::setPlainLayout(DocumentClass const & tc)
3663 {
3664         setLayout(tc.plainLayout());
3665 }
3666
3667
3668 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
3669 {
3670         if (usePlainLayout())
3671                 setPlainLayout(tclass);
3672         else
3673                 setDefaultLayout(tclass);
3674 }
3675
3676
3677 Inset const & Paragraph::inInset() const
3678 {
3679         LBUFERR(d->inset_owner_);
3680         return *d->inset_owner_;
3681 }
3682
3683
3684 ParagraphParameters & Paragraph::params()
3685 {
3686         return d->params_;
3687 }
3688
3689
3690 ParagraphParameters const & Paragraph::params() const
3691 {
3692         return d->params_;
3693 }
3694
3695
3696 bool Paragraph::isFreeSpacing() const
3697 {
3698         if (d->layout_->free_spacing)
3699                 return true;
3700         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
3701 }
3702
3703
3704 bool Paragraph::allowEmpty() const
3705 {
3706         if (d->layout_->keepempty)
3707                 return true;
3708         return d->inset_owner_ && d->inset_owner_->allowEmpty();
3709 }
3710
3711
3712 bool Paragraph::brokenBiblio() const
3713 {
3714         // There is a problem if there is no bibitem at position 0 in
3715         // paragraphs that need one, if there is another bibitem in the
3716         // paragraph or if this paragraph is not supposed to have
3717         // a bibitem inset at all.
3718         return ((d->layout_->labeltype == LABEL_BIBLIO
3719                 && (d->insetlist_.find(BIBITEM_CODE) != 0
3720                     || d->insetlist_.find(BIBITEM_CODE, 1) > 0))
3721                 || (d->layout_->labeltype != LABEL_BIBLIO
3722                     && d->insetlist_.find(BIBITEM_CODE) != -1));
3723 }
3724
3725
3726 int Paragraph::fixBiblio(Buffer const & buffer)
3727 {
3728         // FIXME: when there was already an inset at 0, the return value is 1,
3729         // which does not tell whether another inset has been remove; the
3730         // cursor cannot be correctly updated.
3731
3732         bool const track_changes = buffer.params().track_changes;
3733         int bibitem_pos = d->insetlist_.find(BIBITEM_CODE);
3734
3735         // The case where paragraph is not BIBLIO
3736         if (d->layout_->labeltype != LABEL_BIBLIO) {
3737                 if (bibitem_pos == -1)
3738                         // No InsetBibitem => OK
3739                         return 0;
3740                 // There is an InsetBibitem: remove it!
3741                 d->insetlist_.release(bibitem_pos);
3742                 eraseChar(bibitem_pos, track_changes);
3743                 return (bibitem_pos == 0) ? -1 : -bibitem_pos;
3744         }
3745
3746         bool const hasbibitem0 = bibitem_pos == 0;
3747         if (hasbibitem0) {
3748                 bibitem_pos = d->insetlist_.find(BIBITEM_CODE, 1);
3749                 // There was an InsetBibitem at pos 0,
3750                 // and no other one => OK
3751                 if (bibitem_pos == -1)
3752                         return 0;
3753                 // there is a bibitem at the 0 position, but since
3754                 // there is a second one, we copy the second on the
3755                 // first. We're assuming there are at most two of
3756                 // these, which there should be.
3757                 // FIXME: why does it make sense to do that rather
3758                 // than keep the first? (JMarc)
3759                 Inset * inset = releaseInset(bibitem_pos);
3760                 d->insetlist_.begin()->inset = inset;
3761                 return -bibitem_pos;
3762         }
3763
3764         // We need to create an inset at the beginning
3765         Inset * inset = nullptr;
3766         if (bibitem_pos > 0) {
3767                 // there was one somewhere in the paragraph, let's move it
3768                 inset = d->insetlist_.release(bibitem_pos);
3769                 eraseChar(bibitem_pos, track_changes);
3770         } else
3771                 // make a fresh one
3772                 inset = new InsetBibitem(const_cast<Buffer *>(&buffer),
3773                                          InsetCommandParams(BIBITEM_CODE));
3774
3775         Font font(inherit_font, buffer.params().language);
3776         insertInset(0, inset, font, Change(track_changes ? Change::INSERTED
3777                                                    : Change::UNCHANGED));
3778
3779         // This is needed to get the counters right
3780         buffer.updateBuffer();
3781         return 1;
3782 }
3783
3784
3785 void Paragraph::checkAuthors(AuthorList const & authorList)
3786 {
3787         d->changes_.checkAuthors(authorList);
3788 }
3789
3790
3791 bool Paragraph::isChanged(pos_type pos) const
3792 {
3793         return lookupChange(pos).changed();
3794 }
3795
3796
3797 bool Paragraph::isInserted(pos_type pos) const
3798 {
3799         return lookupChange(pos).inserted();
3800 }
3801
3802
3803 bool Paragraph::isDeleted(pos_type pos) const
3804 {
3805         return lookupChange(pos).deleted();
3806 }
3807
3808
3809 InsetList const & Paragraph::insetList() const
3810 {
3811         return d->insetlist_;
3812 }
3813
3814
3815 void Paragraph::setInsetBuffers(Buffer & b)
3816 {
3817         d->insetlist_.setBuffer(b);
3818 }
3819
3820
3821 void Paragraph::resetBuffer()
3822 {
3823         d->insetlist_.resetBuffer();
3824 }
3825
3826
3827 Inset * Paragraph::releaseInset(pos_type pos)
3828 {
3829         Inset * inset = d->insetlist_.release(pos);
3830         /// does not honour change tracking!
3831         eraseChar(pos, false);
3832         return inset;
3833 }
3834
3835
3836 Inset * Paragraph::getInset(pos_type pos)
3837 {
3838         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3839                  ? d->insetlist_.get(pos) : 0;
3840 }
3841
3842
3843 Inset const * Paragraph::getInset(pos_type pos) const
3844 {
3845         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
3846                  ? d->insetlist_.get(pos) : 0;
3847 }
3848
3849
3850 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
3851                 pos_type & right, TextCase action)
3852 {
3853         // process sequences of modified characters; in change
3854         // tracking mode, this approach results in much better
3855         // usability than changing case on a char-by-char basis
3856         // We also need to track the current font, since font
3857         // changes within sequences can occur.
3858         vector<pair<char_type, Font> > changes;
3859
3860         bool const trackChanges = bparams.track_changes;
3861
3862         bool capitalize = true;
3863
3864         for (; pos < right; ++pos) {
3865                 char_type oldChar = d->text_[pos];
3866                 char_type newChar = oldChar;
3867
3868                 // ignore insets and don't play with deleted text!
3869                 if (oldChar != META_INSET && !isDeleted(pos)) {
3870                         switch (action) {
3871                                 case text_lowercase:
3872                                         newChar = lowercase(oldChar);
3873                                         break;
3874                                 case text_capitalization:
3875                                         if (capitalize) {
3876                                                 newChar = uppercase(oldChar);
3877                                                 capitalize = false;
3878                                         }
3879                                         break;
3880                                 case text_uppercase:
3881                                         newChar = uppercase(oldChar);
3882                                         break;
3883                         }
3884                 }
3885
3886                 if (isWordSeparator(pos) || isDeleted(pos)) {
3887                         // permit capitalization again
3888                         capitalize = true;
3889                 }
3890
3891                 if (oldChar != newChar) {
3892                         changes.push_back(make_pair(newChar, getFontSettings(bparams, pos)));
3893                         if (pos != right - 1)
3894                                 continue;
3895                         // step behind the changing area
3896                         pos++;
3897                 }
3898
3899                 int erasePos = pos - changes.size();
3900                 for (size_t i = 0; i < changes.size(); i++) {
3901                         insertChar(pos, changes[i].first,
3902                                    changes[i].second,
3903                                    trackChanges);
3904                         if (!eraseChar(erasePos, trackChanges)) {
3905                                 ++erasePos;
3906                                 ++pos; // advance
3907                                 ++right; // expand selection
3908                         }
3909                 }
3910                 changes.clear();
3911         }
3912 }
3913
3914
3915 int Paragraph::find(docstring const & str, bool cs, bool mw,
3916                 pos_type start_pos, bool del) const
3917 {
3918         pos_type pos = start_pos;
3919         int const strsize = str.length();
3920         int i = 0;
3921         pos_type const parsize = d->text_.size();
3922         for (i = 0; i < strsize && pos < parsize; ++i, ++pos) {
3923                 // Ignore "invisible" letters such as ligature breaks
3924                 // and hyphenation chars while searching
3925                 while (pos < parsize - 1 && isInset(pos)) {
3926                         odocstringstream os;
3927                         getInset(pos)->toString(os);
3928                         if (!getInset(pos)->isLetter() || !os.str().empty())
3929                                 break;
3930                         pos++;
3931                 }
3932                 if (cs && str[i] != d->text_[pos])
3933                         break;
3934                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos]))
3935                         break;
3936                 if (!del && isDeleted(pos))
3937                         break;
3938         }
3939
3940         if (i != strsize)
3941                 return 0;
3942
3943         // if necessary, check whether string matches word
3944         if (mw) {
3945                 if (start_pos > 0 && !isWordSeparator(start_pos - 1))
3946                         return 0;
3947                 if (pos < parsize
3948                         && !isWordSeparator(pos))
3949                         return 0;
3950         }
3951
3952         return pos - start_pos;
3953 }
3954
3955
3956 char_type Paragraph::getChar(pos_type pos) const
3957 {
3958         return d->text_[pos];
3959 }
3960
3961
3962 pos_type Paragraph::size() const
3963 {
3964         return d->text_.size();
3965 }
3966
3967
3968 bool Paragraph::empty() const
3969 {
3970         return d->text_.empty();
3971 }
3972
3973
3974 bool Paragraph::isInset(pos_type pos) const
3975 {
3976         return d->text_[pos] == META_INSET;
3977 }
3978
3979
3980 bool Paragraph::isSeparator(pos_type pos) const
3981 {
3982         //FIXME: Are we sure this can be the only separator?
3983         return d->text_[pos] == ' ';
3984 }
3985
3986
3987 void Paragraph::deregisterWords()
3988 {
3989         Private::LangWordsMap::const_iterator itl = d->words_.begin();
3990         Private::LangWordsMap::const_iterator ite = d->words_.end();
3991         for (; itl != ite; ++itl) {
3992                 WordList & wl = theWordList(itl->first);
3993                 Private::Words::const_iterator it = (itl->second).begin();
3994                 Private::Words::const_iterator et = (itl->second).end();
3995                 for (; it != et; ++it)
3996                         wl.remove(*it);
3997         }
3998         d->words_.clear();
3999 }
4000
4001
4002 void Paragraph::locateWord(pos_type & from, pos_type & to,
4003         word_location const loc) const
4004 {
4005         switch (loc) {
4006         case WHOLE_WORD_STRICT:
4007                 if (from == 0 || from == size()
4008                     || isWordSeparator(from)
4009                     || isWordSeparator(from - 1)) {
4010                         to = from;
4011                         return;
4012                 }
4013                 // fall through
4014
4015         case WHOLE_WORD:
4016                 // If we are already at the beginning of a word, do nothing
4017                 if (!from || isWordSeparator(from - 1))
4018                         break;
4019                 // fall through
4020
4021         case PREVIOUS_WORD:
4022                 // always move the cursor to the beginning of previous word
4023                 while (from && !isWordSeparator(from - 1))
4024                         --from;
4025                 break;
4026         case NEXT_WORD:
4027                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
4028                 break;
4029         case PARTIAL_WORD:
4030                 // no need to move the 'from' cursor
4031                 break;
4032         }
4033         to = from;
4034         while (to < size() && !isWordSeparator(to))
4035                 ++to;
4036 }
4037
4038
4039 void Paragraph::collectWords()
4040 {
4041         for (pos_type pos = 0; pos < size(); ++pos) {
4042                 if (isWordSeparator(pos))
4043                         continue;
4044                 pos_type from = pos;
4045                 locateWord(from, pos, WHOLE_WORD);
4046                 // Work around MSVC warning: The statement
4047                 // if (pos < from + lyxrc.completion_minlength)
4048                 // triggers a signed vs. unsigned warning.
4049                 // I don't know why this happens, it could be a MSVC bug, or
4050                 // related to LLP64 (windows) vs. LP64 (unix) programming
4051                 // model, or the C++ standard might be ambigous in the section
4052                 // defining the "usual arithmetic conversions". However, using
4053                 // a temporary variable is safe and works on all compilers.
4054                 pos_type const endpos = from + lyxrc.completion_minlength;
4055                 if (pos < endpos)
4056                         continue;
4057                 FontList::const_iterator cit = d->fontlist_.fontIterator(from);
4058                 if (cit == d->fontlist_.end())
4059                         return;
4060                 Language const * lang = cit->font().language();
4061                 docstring const word = asString(from, pos, AS_STR_NONE);
4062                 d->words_[lang->lang()].insert(word);
4063         }
4064 }
4065
4066
4067 void Paragraph::registerWords()
4068 {
4069         Private::LangWordsMap::const_iterator itl = d->words_.begin();
4070         Private::LangWordsMap::const_iterator ite = d->words_.end();
4071         for (; itl != ite; ++itl) {
4072                 WordList & wl = theWordList(itl->first);
4073                 Private::Words::const_iterator it = (itl->second).begin();
4074                 Private::Words::const_iterator et = (itl->second).end();
4075                 for (; it != et; ++it)
4076                         wl.insert(*it);
4077         }
4078 }
4079
4080
4081 void Paragraph::updateWords()
4082 {
4083         deregisterWords();
4084         collectWords();
4085         registerWords();
4086 }
4087
4088
4089 void Paragraph::Private::appendSkipPosition(SkipPositions & skips, pos_type const pos) const
4090 {
4091         SkipPositionsIterator begin = skips.begin();
4092         SkipPositions::iterator end = skips.end();
4093         if (pos > 0 && begin < end) {
4094                 --end;
4095                 if (end->last == pos - 1) {
4096                         end->last = pos;
4097                         return;
4098                 }
4099         }
4100         skips.insert(end, FontSpan(pos, pos));
4101 }
4102
4103
4104 Language * Paragraph::Private::locateSpellRange(
4105         pos_type & from, pos_type & to,
4106         SkipPositions & skips) const
4107 {
4108         // skip leading white space
4109         while (from < to && owner_->isWordSeparator(from))
4110                 ++from;
4111         // don't check empty range
4112         if (from >= to)
4113                 return 0;
4114         // get current language
4115         Language * lang = getSpellLanguage(from);
4116         pos_type last = from;
4117         bool samelang = true;
4118         bool sameinset = true;
4119         while (last < to && samelang && sameinset) {
4120                 // hop to end of word
4121                 while (last < to && !owner_->isWordSeparator(last)) {
4122                         if (owner_->getInset(last)) {
4123                                 appendSkipPosition(skips, last);
4124                         } else if (owner_->isDeleted(last)) {
4125                                 appendSkipPosition(skips, last);
4126                         }
4127                         ++last;
4128                 }
4129                 // hop to next word while checking for insets
4130                 while (sameinset && last < to && owner_->isWordSeparator(last)) {
4131                         if (Inset const * inset = owner_->getInset(last))
4132                                 sameinset = inset->isChar() && inset->isLetter();
4133                         if (sameinset && owner_->isDeleted(last)) {
4134                                 appendSkipPosition(skips, last);
4135                         }
4136                         if (sameinset)
4137                                 last++;
4138                 }
4139                 if (sameinset && last < to) {
4140                         // now check for language change
4141                         samelang = lang == getSpellLanguage(last);
4142                 }
4143         }
4144         // if language change detected backstep is needed
4145         if (!samelang)
4146                 --last;
4147         to = last;
4148         return lang;
4149 }
4150
4151
4152 Language * Paragraph::Private::getSpellLanguage(pos_type const from) const
4153 {
4154         Language * lang =
4155                 const_cast<Language *>(owner_->getFontSettings(
4156                         inset_owner_->buffer().params(), from).language());
4157         if (lang == inset_owner_->buffer().params().language
4158                 && !lyxrc.spellchecker_alt_lang.empty()) {
4159                 string lang_code;
4160                 string const lang_variety =
4161                         split(lyxrc.spellchecker_alt_lang, lang_code, '-');
4162                 lang->setCode(lang_code);
4163                 lang->setVariety(lang_variety);
4164         }
4165         return lang;
4166 }
4167
4168
4169 void Paragraph::requestSpellCheck(pos_type pos)
4170 {
4171         d->requestSpellCheck(pos);
4172 }
4173
4174
4175 bool Paragraph::needsSpellCheck() const
4176 {
4177         SpellChecker::ChangeNumber speller_change_number = 0;
4178         if (theSpellChecker())
4179                 speller_change_number = theSpellChecker()->changeNumber();
4180         if (speller_change_number > d->speller_state_.currentChangeNumber()) {
4181                 d->speller_state_.needsCompleteRefresh(speller_change_number);
4182         }
4183         return d->needsSpellCheck();
4184 }
4185
4186
4187 bool Paragraph::Private::ignoreWord(docstring const & word) const
4188 {
4189         // Ignore words with digits
4190         // FIXME: make this customizable
4191         // (note that some checkers ignore words with digits by default)
4192         docstring::const_iterator cit = word.begin();
4193         docstring::const_iterator const end = word.end();
4194         for (; cit != end; ++cit) {
4195                 if (isNumber((*cit)))
4196                         return true;
4197         }
4198         return false;
4199 }
4200
4201
4202 SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
4203         WordLangTuple & wl, docstring_list & suggestions,
4204         bool do_suggestion, bool check_learned) const
4205 {
4206         SpellChecker::Result result = SpellChecker::WORD_OK;
4207         SpellChecker * speller = theSpellChecker();
4208         if (!speller)
4209                 return result;
4210
4211         if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
4212                 return result;
4213
4214         locateWord(from, to, WHOLE_WORD);
4215         if (from == to || from >= size())
4216                 return result;
4217
4218         docstring word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
4219         Language * lang = d->getSpellLanguage(from);
4220
4221         if (getFontSettings(d->inset_owner_->buffer().params(), from).fontInfo().nospellcheck() == FONT_ON)
4222                 return result;
4223
4224         wl = WordLangTuple(word, lang);
4225
4226         if (word.empty())
4227                 return result;
4228
4229         if (needsSpellCheck() || check_learned) {
4230                 pos_type end = to;
4231                 if (!d->ignoreWord(word)) {
4232                         bool const trailing_dot = to < size() && d->text_[to] == '.';
4233                         result = speller->check(wl);
4234                         if (SpellChecker::misspelled(result) && trailing_dot) {
4235                                 wl = WordLangTuple(word.append(from_ascii(".")), lang);
4236                                 result = speller->check(wl);
4237                                 if (!SpellChecker::misspelled(result)) {
4238                                         LYXERR(Debug::GUI, "misspelled word is correct with dot: \"" <<
4239                                            word << "\" [" <<
4240                                            from << ".." << to << "]");
4241                                 } else {
4242                                         // spell check with dot appended failed too
4243                                         // restore original word/lang value
4244                                         word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
4245                                         wl = WordLangTuple(word, lang);
4246                                 }
4247                         }
4248                 }
4249                 if (!SpellChecker::misspelled(result)) {
4250                         // area up to the begin of the next word is not misspelled
4251                         while (end < size() && isWordSeparator(end))
4252                                 ++end;
4253                 }
4254                 d->setMisspelled(from, end, result);
4255         } else {
4256                 result = d->speller_state_.getState(from);
4257         }
4258
4259         if (do_suggestion)
4260                 suggestions.clear();
4261
4262         if (SpellChecker::misspelled(result)) {
4263                 LYXERR(Debug::GUI, "misspelled word: \"" <<
4264                            word << "\" [" <<
4265                            from << ".." << to << "]");
4266                 if (do_suggestion)
4267                         speller->suggest(wl, suggestions);
4268         }
4269         return result;
4270 }
4271
4272
4273 void Paragraph::anonymize()
4274 {
4275         // This is a very crude anonymization for now
4276         for (char_type & c : d->text_)
4277                 if (isLetterChar(c) || isNumber(c))
4278                         c = 'a';
4279 }
4280
4281
4282 void Paragraph::Private::markMisspelledWords(
4283         pos_type const & first, pos_type const & last,
4284         SpellChecker::Result result,
4285         docstring const & word,
4286         SkipPositions const & skips)
4287 {
4288         if (!SpellChecker::misspelled(result)) {
4289                 setMisspelled(first, last, SpellChecker::WORD_OK);
4290                 return;
4291         }
4292         int snext = first;
4293         SpellChecker * speller = theSpellChecker();
4294         // locate and enumerate the error positions
4295         int nerrors = speller->numMisspelledWords();
4296         int numskipped = 0;
4297         SkipPositionsIterator it = skips.begin();
4298         SkipPositionsIterator et = skips.end();
4299         for (int index = 0; index < nerrors; ++index) {
4300                 int wstart;
4301                 int wlen = 0;
4302                 speller->misspelledWord(index, wstart, wlen);
4303                 /// should not happen if speller supports range checks
4304                 if (!wlen) continue;
4305                 docstring const misspelled = word.substr(wstart, wlen);
4306                 wstart += first + numskipped;
4307                 if (snext < wstart) {
4308                         /// mark the range of correct spelling
4309                         numskipped += countSkips(it, et, wstart);
4310                         setMisspelled(snext,
4311                                 wstart - 1, SpellChecker::WORD_OK);
4312                 }
4313                 snext = wstart + wlen;
4314                 numskipped += countSkips(it, et, snext);
4315                 /// mark the range of misspelling
4316                 setMisspelled(wstart, snext, result);
4317                 LYXERR(Debug::GUI, "misspelled word: \"" <<
4318                            misspelled << "\" [" <<
4319                            wstart << ".." << (snext-1) << "]");
4320                 ++snext;
4321         }
4322         if (snext <= last) {
4323                 /// mark the range of correct spelling at end
4324                 setMisspelled(snext, last, SpellChecker::WORD_OK);
4325         }
4326 }
4327
4328
4329 void Paragraph::spellCheck() const
4330 {
4331         SpellChecker * speller = theSpellChecker();
4332         if (!speller || empty() ||!needsSpellCheck())
4333                 return;
4334         pos_type start;
4335         pos_type endpos;
4336         d->rangeOfSpellCheck(start, endpos);
4337         if (speller->canCheckParagraph()) {
4338                 // loop until we leave the range
4339                 for (pos_type first = start; first < endpos; ) {
4340                         pos_type last = endpos;
4341                         Private::SkipPositions skips;
4342                         Language * lang = d->locateSpellRange(first, last, skips);
4343                         if (first >= endpos)
4344                                 break;
4345                         // start the spell checker on the unit of meaning
4346                         docstring word = asString(first, last, AS_STR_INSETS + AS_STR_SKIPDELETE);
4347                         WordLangTuple wl = WordLangTuple(word, lang);
4348                         SpellChecker::Result result = word.size() ?
4349                                 speller->check(wl) : SpellChecker::WORD_OK;
4350                         d->markMisspelledWords(first, last, result, word, skips);
4351                         first = ++last;
4352                 }
4353         } else {
4354                 static docstring_list suggestions;
4355                 pos_type to = endpos;
4356                 while (start < endpos) {
4357                         WordLangTuple wl;
4358                         spellCheck(start, to, wl, suggestions, false);
4359                         start = to + 1;
4360                 }
4361         }
4362         d->readySpellCheck();
4363 }
4364
4365
4366 bool Paragraph::isMisspelled(pos_type pos, bool check_boundary) const
4367 {
4368         bool result = SpellChecker::misspelled(d->speller_state_.getState(pos));
4369         if (result || pos <= 0 || pos > size())
4370                 return result;
4371         if (check_boundary && (pos == size() || isWordSeparator(pos)))
4372                 result = SpellChecker::misspelled(d->speller_state_.getState(pos - 1));
4373         return result;
4374 }
4375
4376
4377 string Paragraph::magicLabel() const
4378 {
4379         stringstream ss;
4380         ss << "magicparlabel-" << id();
4381         return ss.str();
4382 }
4383
4384
4385 } // namespace lyx