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 Kimberly 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 "Buffer.h"
  24 #include "BufferParams.h"
  25 #include "BufferEncodings.h"
  26 #include "Changes.h"
  27 #include "Counters.h"
  28 #include "InsetList.h"
  29 #include "Language.h"
  30 #include "LaTeXFeatures.h"
  31 #include "Layout.h"
  32 #include "Font.h"
  33 #include "FontList.h"
  34 #include "LyXRC.h"
  35 #include "OutputParams.h"
  36 #include "output_latex.h"
  37 #include "output_xhtml.h"
  38 #include "output_docbook.h"
  39 #include "ParagraphParameters.h"
  40 #include "Session.h"
  41 #include "SpellChecker.h"
  42 #include "texstream.h"
  43 #include "TexRow.h"
  44 #include "Text.h"
  45 #include "TextClass.h"
  46 #include "WordLangTuple.h"
  47 #include "WordList.h"
  48
  49 #include "frontends/alert.h"
  50
  51 #include "insets/InsetArgument.h"
  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 and msvc < 2015 do not support C++11 thread_local
  75 #if defined(__APPLE__)
  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 /////////////////////////////////////////////////////////////////////
  86 //
  87 // SpellResultRange
  88 //
  89 /////////////////////////////////////////////////////////////////////
  90
  91 class SpellResultRange {
  92 public:
  93         SpellResultRange(FontSpan range, SpellChecker::Result result)
  94         : range_(range), result_(result)
  95         {}
  96         ///
  97         FontSpan const & range() const { return range_; }
  98         ///
  99         void range(FontSpan const & r) { range_ = r; }
 100         ///
 101         SpellChecker::Result result() const { return result_; }
 102         ///
 103         void result(SpellChecker::Result r) { result_ = r; }
 104         ///
 105         bool contains(pos_type pos) const { return range_.contains(pos); }
 106         ///
 107         bool covered(FontSpan const & r) const
 108         {
 109                 // 1. first of new range inside current range or
 110                 // 2. last of new range inside current range or
 111                 // 3. first of current range inside new range or
 112                 // 4. last of current range inside new range
 113                 //FIXME: is this the same as !range_.intersect(r).empty() ?
 114                 return range_.contains(r.first) || range_.contains(r.last) ||
 115                         r.contains(range_.first) || r.contains(range_.last);
 116         }
 117         ///
 118         void shift(pos_type pos, int offset)
 119         {
 120                 if (range_.first > pos) {
 121                         range_.first += offset;
 122                         range_.last += offset;
 123                 } else if (range_.last >= pos) {
 124                         range_.last += offset;
 125                 }
 126         }
 127 private:
 128         FontSpan range_ ;
 129         SpellChecker::Result result_ ;
 130 };
 131
 132
 133 /////////////////////////////////////////////////////////////////////
 134 //
 135 // SpellCheckerState
 136 //
 137 /////////////////////////////////////////////////////////////////////
 138
 139 class SpellCheckerState {
 140 public:
 141         SpellCheckerState()
 142         {
 143                 needs_refresh_ = true;
 144                 current_change_number_ = 0;
 145         }
 146
 147         void setRange(FontSpan const & fp, SpellChecker::Result state)
 148         {
 149                 Ranges result;
 150                 RangesIterator et = ranges_.end();
 151                 RangesIterator it = ranges_.begin();
 152                 for (; it != et; ++it) {
 153                         if (!it->covered(fp))
 154                                 result.push_back(SpellResultRange(it->range(), it->result()));
 155                         else if (state == SpellChecker::WORD_OK) {
 156                                 // trim or split the current misspelled range
 157                                 // store misspelled ranges only
 158                                 FontSpan range = it->range();
 159                                 if (fp.first > range.first) {
 160                                         // misspelled area in front of WORD_OK
 161                                         range.last = fp.first - 1;
 162                                         result.push_back(SpellResultRange(range, it->result()));
 163                                         range = it->range();
 164                                 }
 165                                 if (fp.last < range.last) {
 166                                         // misspelled area after WORD_OK range
 167                                         range.first = fp.last + 1;
 168                                         result.push_back(SpellResultRange(range, it->result()));
 169                                 }
 170                         }
 171                 }
 172                 ranges_ = result;
 173                 if (state != SpellChecker::WORD_OK)
 174                         ranges_.push_back(SpellResultRange(fp, state));
 175         }
 176
 177         void increasePosAfterPos(pos_type pos)
 178         {
 179                 correctRangesAfterPos(pos, 1);
 180                 needsRefresh(pos);
 181         }
 182
 183         void decreasePosAfterPos(pos_type pos)
 184         {
 185                 correctRangesAfterPos(pos, -1);
 186                 needsRefresh(pos);
 187         }
 188
 189         void refreshLast(pos_type pos)
 190         {
 191                 if (pos < refresh_.last)
 192                         refresh_.last = pos;
 193         }
 194
 195         SpellChecker::Result getState(pos_type pos) const
 196         {
 197                 SpellChecker::Result result = SpellChecker::WORD_OK;
 198                 RangesIterator et = ranges_.end();
 199                 RangesIterator it = ranges_.begin();
 200                 for (; it != et; ++it) {
 201                         if(it->contains(pos)) {
 202                                 return it->result();
 203                         }
 204                 }
 205                 return result;
 206         }
 207
 208         FontSpan const & getRange(pos_type pos) const
 209         {
 210                 /// empty span to indicate mismatch
 211                 static FontSpan empty_;
 212                 RangesIterator et = ranges_.end();
 213                 RangesIterator it = ranges_.begin();
 214                 for (; it != et; ++it) {
 215                         if(it->contains(pos)) {
 216                                 return it->range();
 217                         }
 218                 }
 219                 return empty_;
 220         }
 221
 222         bool needsRefresh() const
 223         {
 224                 return needs_refresh_;
 225         }
 226
 227         SpellChecker::ChangeNumber currentChangeNumber() const
 228         {
 229                 return current_change_number_;
 230         }
 231
 232         void refreshRange(pos_type & first, pos_type & last) const
 233         {
 234                 first = refresh_.first;
 235                 last = refresh_.last;
 236         }
 237
 238         void needsRefresh(pos_type pos)
 239         {
 240                 if (needs_refresh_ && pos != -1) {
 241                         if (pos < refresh_.first)
 242                                 refresh_.first = pos;
 243                         if (pos > refresh_.last)
 244                                 refresh_.last = pos;
 245                 } else if (pos != -1) {
 246                         // init request check for neighbour positions too
 247                         refresh_.first = pos > 0 ? pos - 1 : 0;
 248                         // no need for special end of paragraph check
 249                         refresh_.last = pos + 1;
 250                 }
 251                 needs_refresh_ = pos != -1;
 252         }
 253
 254         void needsCompleteRefresh(SpellChecker::ChangeNumber change_number)
 255         {
 256                 needs_refresh_ = true;
 257                 refresh_.first = 0;
 258                 refresh_.last = -1;
 259                 current_change_number_ = change_number;
 260         }
 261 private:
 262         typedef vector<SpellResultRange> Ranges;
 263         typedef Ranges::const_iterator RangesIterator;
 264         Ranges ranges_;
 265         /// the area of the paragraph with pending spell check
 266         FontSpan refresh_;
 267         /// spell state cache version number
 268         SpellChecker::ChangeNumber current_change_number_;
 269         bool needs_refresh_;
 270
 271
 272         void correctRangesAfterPos(pos_type pos, int offset)
 273         {
 274                 RangesIterator et = ranges_.end();
 275                 Ranges::iterator it = ranges_.begin();
 276                 for (; it != et; ++it) {
 277                         it->shift(pos, offset);
 278                 }
 279         }
 280
 281 };
 282
 283 /////////////////////////////////////////////////////////////////////
 284 //
 285 // Paragraph::Private
 286 //
 287 /////////////////////////////////////////////////////////////////////
 288
 289 class Paragraph::Private
 290 {
 291         // Enforce our own "copy" constructor
 292         Private(Private const &) = delete;
 293         Private & operator=(Private const &) = delete;
 294         // Unique ID generator
 295         static int make_id();
 296 public:
 297         ///
 298         Private(Paragraph * owner, Layout const & layout);
 299         /// "Copy constructor"
 300         Private(Private const &, Paragraph * owner);
 301         /// Copy constructor from \p beg  to \p end
 302         Private(Private const &, Paragraph * owner, pos_type beg, pos_type end);
 303
 304         ///
 305         void insertChar(pos_type pos, char_type c, Change const & change);
 306
 307         /// Output the surrogate pair formed by \p c and \p next to \p os.
 308         /// \return the number of characters written.
 309         int latexSurrogatePair(BufferParams const &, otexstream & os,
 310                                char_type c, char_type next,
 311                                OutputParams const &) const;
 312
 313         /// Output a space in appropriate formatting (or a surrogate pair
 314         /// if the next character is a combining character).
 315         /// \return whether a surrogate pair was output.
 316         bool simpleTeXBlanks(BufferParams const &,
 317                              OutputParams const &,
 318                              otexstream &,
 319                              pos_type i,
 320                              unsigned int & column,
 321                              Font const & font,
 322                              Layout const & style) const;
 323
 324         /// This could go to ParagraphParameters if we want to.
 325         int startTeXParParams(BufferParams const &, otexstream &,
 326                               OutputParams const &) const;
 327
 328         /// This could go to ParagraphParameters if we want to.
 329         bool endTeXParParams(BufferParams const &, otexstream &,
 330                              OutputParams const &) const;
 331
 332         ///
 333         void latexInset(BufferParams const &,
 334                                    otexstream &,
 335                                    OutputParams &,
 336                                    Font & running_font,
 337                                    Font & basefont,
 338                                    Font const & outerfont,
 339                                    bool & open_font,
 340                                    Change & running_change,
 341                                    Layout const & style,
 342                                    pos_type & i,
 343                                    unsigned int & column,
 344                                    bool const fontswitch_inset,
 345                                    bool const closeLanguage,
 346                                    bool const lang_switched_at_inset) const;
 347
 348         ///
 349         void latexSpecialChar(
 350                                    otexstream & os,
 351                                    BufferParams const & bparams,
 352                                    OutputParams const & runparams,
 353                                    Font const & running_font,
 354                                    string & alien_script,
 355                                    Layout const & style,
 356                                    pos_type & i,
 357                                    pos_type end_pos,
 358                                    unsigned int & column) const;
 359
 360         ///
 361         bool latexSpecialT1(
 362                 char_type const c,
 363                 otexstream & os,
 364                 pos_type i,
 365                 unsigned int & column) const;
 366         ///
 367         bool latexSpecialTU(
 368                 char_type const c,
 369                 otexstream & os,
 370                 pos_type i,
 371                 unsigned int & column) const;
 372         ///
 373         bool latexSpecialT3(
 374                 char_type const c,
 375                 otexstream & os,
 376                 pos_type i,
 377                 unsigned int & column) const;
 378
 379         ///
 380         void validate(LaTeXFeatures & features) const;
 381
 382         /// Checks if the paragraph contains only text and no inset or font change.
 383         bool onlyText(Buffer const & buf, Font const & outerfont,
 384                       pos_type initial) const;
 385
 386         /// a vector of speller skip positions
 387         typedef vector<FontSpan> SkipPositions;
 388         typedef SkipPositions::const_iterator SkipPositionsIterator;
 389
 390         void appendSkipPosition(SkipPositions & skips, pos_type const pos) const;
 391
 392         Language * getSpellLanguage(pos_type const from) const;
 393
 394         Language * locateSpellRange(pos_type & from, pos_type & to,
 395                                     SkipPositions & skips) const;
 396
 397         bool hasSpellerChange() const
 398         {
 399                 SpellChecker::ChangeNumber speller_change_number = 0;
 400                 if (theSpellChecker())
 401                         speller_change_number = theSpellChecker()->changeNumber();
 402                 return speller_change_number > speller_state_.currentChangeNumber();
 403         }
 404
 405         bool ignoreWord(docstring const & word) const;
 406
 407         void setMisspelled(pos_type from, pos_type to, SpellChecker::Result state)
 408         {
 409                 pos_type textsize = owner_->size();
 410                 // check for sane arguments
 411                 if (to <= from || from >= textsize)
 412                         return;
 413                 FontSpan fp = FontSpan(from, to - 1);
 414                 speller_state_.setRange(fp, state);
 415         }
 416
 417         void requestSpellCheck(pos_type pos)
 418         {
 419                 if (pos == -1)
 420                         speller_state_.needsCompleteRefresh(speller_state_.currentChangeNumber());
 421                 else
 422                         speller_state_.needsRefresh(pos);
 423         }
 424
 425         void readySpellCheck()
 426         {
 427                 speller_state_.needsRefresh(-1);
 428         }
 429
 430         bool needsSpellCheck() const
 431         {
 432                 return speller_state_.needsRefresh();
 433         }
 434
 435         void rangeOfSpellCheck(pos_type & first, pos_type & last) const
 436         {
 437                 speller_state_.refreshRange(first, last);
 438                 if (last == -1) {
 439                         last = owner_->size();
 440                         return;
 441                 }
 442                 pos_type endpos = last;
 443                 owner_->locateWord(first, endpos, WHOLE_WORD, true);
 444                 if (endpos < last) {
 445                         endpos = last;
 446                         owner_->locateWord(last, endpos, WHOLE_WORD, true);
 447                 }
 448                 last = endpos;
 449         }
 450
 451         int countSkips(SkipPositionsIterator & it, SkipPositionsIterator const et,
 452                             int & start) const
 453         {
 454                 int numskips = 0;
 455                 while (it != et && it->first < start) {
 456                         long skip = it->last - it->first + 1;
 457                         start += skip;
 458                         numskips += skip;
 459                         ++it;
 460                 }
 461                 return numskips;
 462         }
 463
 464         void markMisspelledWords(Language const * lang,
 465                                  pos_type const & first, pos_type const & last,
 466                                  SpellChecker::Result result,
 467                                  docstring const & word,
 468                                  SkipPositions const & skips);
 469
 470         InsetCode ownerCode() const
 471         {
 472                 return inset_owner_ ? inset_owner_->lyxCode() : NO_CODE;
 473         }
 474
 475         /// Which Paragraph owns us?
 476         Paragraph * owner_;
 477
 478         /// In which Inset?
 479         Inset const * inset_owner_;
 480
 481         ///
 482         FontList fontlist_;
 483
 484         ///
 485         ParagraphParameters params_;
 486
 487         /// for recording and looking up changes
 488         Changes changes_;
 489
 490         ///
 491         InsetList insetlist_;
 492
 493         /// end of label
 494         pos_type begin_of_body_;
 495
 496         typedef docstring TextContainer;
 497         ///
 498         TextContainer text_;
 499
 500         typedef set<docstring> Words;
 501         typedef map<string, Words> LangWordsMap;
 502         ///
 503         LangWordsMap words_;
 504         ///
 505         Layout const * layout_;
 506         ///
 507         SpellCheckerState speller_state_;
 508         ///
 509         int id_;
 510 };
 511
 512
 513 Paragraph::Private::Private(Paragraph * owner, Layout const & layout)
 514         : owner_(owner), inset_owner_(nullptr), begin_of_body_(0), layout_(&layout), id_(-1)
 515 {
 516         text_.reserve(100);
 517 }
 518
 519
 520 //static
 521 int Paragraph::Private::make_id()
 522 {
 523         // The id is unique per session across buffers because it is used in
 524         // LFUN_PARAGRAPH_GOTO to switch to a different buffer, for instance in the
 525         // outliner.
 526         // (thread-safe)
 527         static int next_id(0);
 528         return next_id++;
 529 }
 530
 531
 532 Paragraph::Private::Private(Private const & p, Paragraph * owner)
 533         : owner_(owner), inset_owner_(p.inset_owner_), fontlist_(p.fontlist_),
 534           params_(p.params_), changes_(p.changes_), insetlist_(p.insetlist_),
 535           begin_of_body_(p.begin_of_body_), text_(p.text_), words_(p.words_),
 536           layout_(p.layout_), id_(make_id())
 537 {
 538         requestSpellCheck(p.text_.size());
 539 }
 540
 541
 542 Paragraph::Private::Private(Private const & p, Paragraph * owner,
 543         pos_type beg, pos_type end)
 544         : owner_(owner), inset_owner_(p.inset_owner_),
 545           params_(p.params_), changes_(p.changes_),
 546           insetlist_(p.insetlist_, beg, end),
 547           begin_of_body_(p.begin_of_body_), words_(p.words_),
 548           layout_(p.layout_), id_(make_id())
 549 {
 550         if (beg >= pos_type(p.text_.size()))
 551                 return;
 552         text_ = p.text_.substr(beg, end - beg);
 553
 554         FontList::const_iterator fcit = fontlist_.begin();
 555         FontList::const_iterator fend = fontlist_.end();
 556         for (; fcit != fend; ++fcit) {
 557                 if (fcit->pos() < beg)
 558                         continue;
 559                 if (fcit->pos() >= end) {
 560                         // Add last entry in the fontlist_.
 561                         fontlist_.set(text_.size() - 1, fcit->font());
 562                         break;
 563                 }
 564                 // Add a new entry in the fontlist_.
 565                 fontlist_.set(fcit->pos() - beg, fcit->font());
 566         }
 567         requestSpellCheck(p.text_.size());
 568 }
 569
 570
 571 /////////////////////////////////////////////////////////////////////
 572 //
 573 // Paragraph
 574 //
 575 /////////////////////////////////////////////////////////////////////
 576
 577 namespace {
 578
 579 /** This helper class should be instantiated at the start of methods
 580  * that can create or merge changes. If as a result the value of
 581  * Paragraph::isChanged is modified, it makes sure that updateBuffer()
 582  * will be run.
 583  */
 584 struct ChangesMonitor {
 585         ///
 586         ChangesMonitor(Paragraph & par)
 587                 : par_(par), was_changed_(par.isChanged()) {}
 588         ///
 589         ~ChangesMonitor()
 590         {
 591                 /* We may need to run updateBuffer to check whether the buffer
 592                  * contains changes (and toggle the changes toolbar). We do it
 593                  * when:
 594                  * 1. the `changedness' of the paragraph has changed,
 595                  * 2. and we are not in the situation where the buffer has changes
 596                  * and new changes are added to the paragraph.
 597                  */
 598                 try {
 599                         if (par_.isChanged() != was_changed_
 600                                 && par_.inInset().isBufferValid()
 601                                 && !(par_.inInset().buffer().areChangesPresent() && par_.isChanged()))
 602                                 par_.inInset().buffer().forceUpdate();
 603                 } catch(support::ExceptionMessage const &) {}
 604         }
 605
 606 private:
 607         ///
 608         Paragraph const & par_;
 609         ///
 610         bool was_changed_;
 611 };
 612
 613 }
 614
 615 void Paragraph::addChangesToToc(DocIterator const & cdit, Buffer const & buf,
 616                                 bool output_active, TocBackend & backend) const
 617 {
 618         d->changes_.addToToc(cdit, buf, output_active, backend);
 619 }
 620
 621
 622 bool Paragraph::isDeleted(pos_type start, pos_type end) const
 623 {
 624         LASSERT(start >= 0 && start <= size(), return false);
 625         LASSERT(end > start && end <= size() + 1, return false);
 626
 627         return d->changes_.isDeleted(start, end);
 628 }
 629
 630
 631 bool Paragraph::isChanged(pos_type start, pos_type end) const
 632 {
 633         LASSERT(start >= 0 && start <= size(), return false);
 634         LASSERT(end > start && end <= size() + 1, return false);
 635
 636         return d->changes_.isChanged(start, end);
 637 }
 638
 639 // FIXME: Ideally the diverse isChanged() methods should account for that!
 640 bool Paragraph::hasChangedInsets(pos_type start, pos_type end) const
 641 {
 642         LASSERT(start >= 0 && start <= size(), return false);
 643         LASSERT(end > start && end <= size() + 1, return false);
 644
 645         for (auto const & icit : d->insetlist_) {
 646                 if (icit.pos < start)
 647                         continue;
 648                 if (icit.pos >= end)
 649                         break;
 650                 if (icit.inset && icit.inset->isChanged())
 651                         return true;
 652         }
 653         return false;
 654 }
 655
 656 bool Paragraph::isChanged() const
 657 {
 658         return d->changes_.isChanged();
 659 }
 660
 661
 662 bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const
 663 {
 664         // keep the logic here in sync with the logic of eraseChars()
 665         if (!trackChanges)
 666                 return true;
 667
 668         Change const & change = d->changes_.lookup(size());
 669         return change.inserted() && change.currentAuthor();
 670 }
 671
 672 Change Paragraph::parEndChange() const
 673 {
 674         return d->changes_.lookup(size());
 675 }
 676
 677
 678 void Paragraph::setChange(Change const & change)
 679 {
 680         // Make sure that Buffer::hasChangesPresent is updated
 681         ChangesMonitor cm(*this);
 682
 683         // beware of the imaginary end-of-par character!
 684         d->changes_.set(change, 0, size() + 1);
 685
 686         /*
 687          * Propagate the change recursively - but not in case of DELETED!
 688          *
 689          * Imagine that your co-author makes changes in an existing inset. He
 690          * sends your document to you and you come to the conclusion that the
 691          * inset should go completely. If you erase it, LyX must not delete all
 692          * text within the inset. Otherwise, the change tracked insertions of
 693          * your co-author get lost and there is no way to restore them later.
 694          *
 695          * Conclusion: An inset's content should remain untouched if you delete it
 696          */
 697
 698         if (!change.deleted()) {
 699                 for (pos_type pos = 0; pos < size(); ++pos) {
 700                         if (Inset * inset = getInset(pos))
 701                                 inset->setChange(change);
 702                 }
 703         }
 704 }
 705
 706
 707 void Paragraph::setChange(pos_type pos, Change const & change)
 708 {
 709         // Make sure that Buffer::hasChangesPresent is updated
 710         ChangesMonitor cm(*this);
 711
 712         LASSERT(pos >= 0 && pos <= size(), return);
 713         d->changes_.set(change, pos);
 714
 715         // see comment in setChange(Change const &) above
 716         if (!change.deleted() && pos < size())
 717                 if (Inset * inset = getInset(pos))
 718                         inset->setChange(change);
 719 }
 720
 721
 722 Change const & Paragraph::lookupChange(pos_type pos) const
 723 {
 724         LBUFERR(pos >= 0 && pos <= size());
 725         return d->changes_.lookup(pos);
 726 }
 727
 728
 729 void Paragraph::acceptChanges(pos_type start, pos_type end)
 730 {
 731         // Make sure that Buffer::hasChangesPresent is updated
 732         ChangesMonitor cm(*this);
 733
 734         LASSERT(start >= 0 && start <= size(), return);
 735         LASSERT(end > start && end <= size() + 1, return);
 736
 737         for (pos_type pos = start; pos < end; ++pos) {
 738                 switch (lookupChange(pos).type) {
 739                         case Change::UNCHANGED:
 740                                 // accept changes in nested inset
 741                                 if (Inset * inset = getInset(pos))
 742                                         inset->acceptChanges();
 743                                 break;
 744
 745                         case Change::INSERTED:
 746                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 747                                 // also accept changes in nested inset
 748                                 if (Inset * inset = getInset(pos))
 749                                         inset->acceptChanges();
 750                                 break;
 751
 752                         case Change::DELETED:
 753                                 // Suppress access to non-existent
 754                                 // "end-of-paragraph char"
 755                                 if (pos < size()) {
 756                                         eraseChar(pos, false);
 757                                         --end;
 758                                         --pos;
 759                                 }
 760                                 break;
 761                 }
 762
 763         }
 764 }
 765
 766
 767 void Paragraph::rejectChanges(pos_type start, pos_type end)
 768 {
 769         LASSERT(start >= 0 && start <= size(), return);
 770         LASSERT(end > start && end <= size() + 1, return);
 771
 772         // Make sure that Buffer::hasChangesPresent is updated
 773         ChangesMonitor cm(*this);
 774
 775         for (pos_type pos = start; pos < end; ++pos) {
 776                 switch (lookupChange(pos).type) {
 777                         case Change::UNCHANGED:
 778                                 // reject changes in nested inset
 779                                 if (Inset * inset = getInset(pos))
 780                                                 inset->rejectChanges();
 781                                 break;
 782
 783                         case Change::INSERTED:
 784                                 // Suppress access to non-existent
 785                                 // "end-of-paragraph char"
 786                                 if (pos < size()) {
 787                                         eraseChar(pos, false);
 788                                         --end;
 789                                         --pos;
 790                                 }
 791                                 break;
 792
 793                         case Change::DELETED:
 794                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 795
 796                                 // Do NOT reject changes within a deleted inset!
 797                                 // There may be insertions of a co-author inside of it!
 798
 799                                 break;
 800                 }
 801         }
 802 }
 803
 804
 805 void Paragraph::Private::insertChar(pos_type pos, char_type c,
 806                 Change const & change)
 807 {
 808         LASSERT(pos >= 0 && pos <= int(text_.size()), return);
 809
 810         // Make sure that Buffer::hasChangesPresent is updated
 811         ChangesMonitor cm(*owner_);
 812
 813         // track change
 814         changes_.insert(change, pos);
 815
 816         // This is actually very common when parsing buffers (and
 817         // maybe inserting ascii text)
 818         if (pos == pos_type(text_.size())) {
 819                 // when appending characters, no need to update tables
 820                 text_.push_back(c);
 821                 // but we want spell checking
 822                 requestSpellCheck(pos);
 823                 return;
 824         }
 825
 826         text_.insert(text_.begin() + pos, c);
 827
 828         // Update the font table.
 829         fontlist_.increasePosAfterPos(pos);
 830
 831         // Update the insets
 832         insetlist_.increasePosAfterPos(pos);
 833
 834         // Update list of misspelled positions
 835         speller_state_.increasePosAfterPos(pos);
 836
 837         // Update bookmarks
 838         if (inset_owner_ && inset_owner_->isBufferValid())
 839                 theSession().bookmarks().adjustPosAfterPos(inset_owner_->buffer().fileName(),
 840                                                        id_, pos, 1);
 841 }
 842
 843
 844 bool Paragraph::insertInset(pos_type pos, Inset * inset,
 845                                    Font const & font, Change const & change)
 846 {
 847         LASSERT(inset, return false);
 848         LASSERT(pos >= 0 && pos <= size(), return false);
 849
 850         // Make sure that Buffer::hasChangesPresent is updated
 851         ChangesMonitor cm(*this);
 852
 853         // Paragraph::insertInset() can be used in cut/copy/paste operation where
 854         // d->inset_owner_ is not set yet.
 855         if (d->inset_owner_ && !d->inset_owner_->insetAllowed(inset->lyxCode()))
 856                 return false;
 857
 858         d->insertChar(pos, META_INSET, change);
 859         LASSERT(d->text_[pos] == META_INSET, return false);
 860
 861         // Add a new entry in the insetlist_.
 862         d->insetlist_.insert(inset, pos);
 863
 864         // Some insets require run of spell checker
 865         requestSpellCheck(pos);
 866         setFont(pos, font);
 867         return true;
 868 }
 869
 870
 871 bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
 872 {
 873         LASSERT(pos >= 0 && pos <= size(), return false);
 874
 875         // Make sure that Buffer::hasChangesPresent is updated
 876         ChangesMonitor cm(*this);
 877
 878         // keep the logic here in sync with the logic of isMergedOnEndOfParDeletion()
 879
 880         if (trackChanges) {
 881                 Change change = d->changes_.lookup(pos);
 882
 883                 // set the character to DELETED if
 884                 //  a) it was previously unchanged or
 885                 //  b) it was inserted by a co-author
 886
 887                 if (!change.changed() ||
 888                       (change.inserted() && !change.currentAuthor())) {
 889                         setChange(pos, Change(Change::DELETED));
 890                         // request run of spell checker
 891                         requestSpellCheck(pos);
 892                         return false;
 893                 }
 894
 895                 if (change.deleted())
 896                         return false;
 897         }
 898
 899         // Don't physically access the imaginary end-of-paragraph character.
 900         // eraseChar() can only mark it as DELETED. A physical deletion of
 901         // end-of-par must be handled externally.
 902         if (pos == size()) {
 903                 return false;
 904         }
 905
 906         // track change
 907         d->changes_.erase(pos);
 908
 909         // if it is an inset, delete the inset entry
 910         if (d->text_[pos] == META_INSET)
 911                 d->insetlist_.erase(pos);
 912
 913         d->text_.erase(d->text_.begin() + pos);
 914
 915         // Update the fontlist_
 916         d->fontlist_.erase(pos);
 917
 918         // Update the insetlist_
 919         d->insetlist_.decreasePosAfterPos(pos);
 920
 921         // Update list of misspelled positions
 922         d->speller_state_.decreasePosAfterPos(pos);
 923         d->speller_state_.refreshLast(size());
 924
 925         // Update bookmarks
 926         if (d->inset_owner_ && d->inset_owner_->isBufferValid())
 927                 theSession().bookmarks().adjustPosAfterPos(d->inset_owner_->buffer().fileName(),
 928                                                        d->id_, pos, -1);
 929
 930         return true;
 931 }
 932
 933
 934 int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
 935 {
 936         LASSERT(start >= 0 && start <= size(), return 0);
 937         LASSERT(end >= start && end <= size() + 1, return 0);
 938
 939         pos_type i = start;
 940         for (pos_type count = end - start; count; --count) {
 941                 if (!eraseChar(i, trackChanges))
 942                         ++i;
 943         }
 944         return end - i;
 945 }
 946
 947
 948 // Handle combining characters
 949 int Paragraph::Private::latexSurrogatePair(BufferParams const & bparams,
 950                 otexstream & os, char_type c, char_type next,
 951                 OutputParams const & runparams) const
 952 {
 953         // Writing next here may circumvent a possible font change between
 954         // c and next. Since next is only output if it forms a surrogate pair
 955         // with c we can ignore this:
 956         // A font change inside a surrogate pair does not make sense and is
 957         // hopefully impossible to input.
 958         // FIXME: change tracking
 959         // Is this correct WRT change tracking?
 960         Encoding const & encoding = *(runparams.encoding);
 961         docstring latex1 = encoding.latexChar(next).first;
 962         if (runparams.inIPA) {
 963                 string const tipashortcut = Encodings::TIPAShortcut(next);
 964                 if (!tipashortcut.empty()) {
 965                         latex1 = from_ascii(tipashortcut);
 966                 }
 967         }
 968         docstring latex2 = encoding.latexChar(c).first;
 969
 970         if (bparams.useNonTeXFonts || docstring(1, next) == latex1) {
 971                 // Encoding supports the combination:
 972                 // output as is (combining char after base char).
 973                 os << latex2 << latex1;
 974                 return latex1.length() + latex2.length();
 975         }
 976
 977         os << latex1 << "{" << latex2 << "}";
 978         return latex1.length() + latex2.length() + 2;
 979 }
 980
 981
 982 bool Paragraph::Private::simpleTeXBlanks(BufferParams const & bparams,
 983                                        OutputParams const & runparams,
 984                                        otexstream & os,
 985                                        pos_type i,
 986                                        unsigned int & column,
 987                                        Font const & font,
 988                                        Layout const & style) const
 989 {
 990         if (style.pass_thru || runparams.pass_thru)
 991                 return false;
 992
 993         if (i + 1 < int(text_.size())) {
 994                 char_type next = text_[i + 1];
 995                 if (Encodings::isCombiningChar(next)) {
 996                         // This space has an accent, so we must always output it.
 997                         column += latexSurrogatePair(bparams, os, ' ', next, runparams) - 1;
 998                         return true;
 999                 }
1000         }
1001
1002         if (runparams.linelen > 0
1003             && column > runparams.linelen
1004             && i
1005             && text_[i - 1] != ' '
1006             && (i + 1 < int(text_.size()))
1007             // same in FreeSpacing mode
1008             && !owner_->isFreeSpacing()
1009             // In typewriter mode, we want to avoid
1010             // ! . ? : at the end of a line
1011             && !(font.fontInfo().family() == TYPEWRITER_FAMILY
1012                  && (text_[i - 1] == '.'
1013                      || text_[i - 1] == '?'
1014                      || text_[i - 1] == ':'
1015                      || text_[i - 1] == '!'))) {
1016                 os << '\n';
1017                 os.texrow().start(owner_->id(), i + 1);
1018                 column = 0;
1019         } else if (style.free_spacing) {
1020                 os << '~';
1021         } else {
1022                 os << ' ';
1023         }
1024         return false;
1025 }
1026
1027
1028 void Paragraph::Private::latexInset(BufferParams const & bparams,
1029                                     otexstream & os,
1030                                     OutputParams & runparams,
1031                                     Font & running_font,
1032                                     Font & basefont,
1033                                     Font const & outerfont,
1034                                     bool & open_font,
1035                                     Change & running_change,
1036                                     Layout const & style,
1037                                     pos_type & i,
1038                                     unsigned int & column,
1039                                     bool const fontswitch_inset,
1040                                     bool const closeLanguage,
1041                                     bool const lang_switched_at_inset) const
1042 {
1043         Inset * inset = owner_->getInset(i);
1044         LBUFERR(inset);
1045
1046         if (style.pass_thru) {
1047                 odocstringstream ods;
1048                 inset->plaintext(ods, runparams);
1049                 os << ods.str();
1050                 return;
1051         }
1052
1053         // FIXME: move this to InsetNewline::latex
1054         if (inset->lyxCode() == NEWLINE_CODE || inset->lyxCode() == SEPARATOR_CODE) {
1055                 // newlines are handled differently here than
1056                 // the default in simpleTeXSpecialChars().
1057                 if (!style.newline_allowed) {
1058                         os << '\n';
1059                 } else {
1060                         if (open_font) {
1061                                 bool needPar = false;
1062                                 column += running_font.latexWriteEndChanges(
1063                                         os, bparams, runparams,
1064                                         basefont, basefont, needPar);
1065                                 open_font = false;
1066                         }
1067
1068                         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY)
1069                                 os << '~';
1070
1071                         basefont = owner_->getLayoutFont(bparams, outerfont);
1072                         running_font = basefont;
1073
1074                         if (runparams.moving_arg)
1075                                 os << "\\protect ";
1076
1077                 }
1078                 os.texrow().start(owner_->id(), i + 1);
1079                 column = 0;
1080         }
1081
1082         if (owner_->isDeleted(i)) {
1083                 if( ++runparams.inDeletedInset == 1)
1084                         runparams.changeOfDeletedInset = owner_->lookupChange(i);
1085         }
1086
1087         if (inset->canTrackChanges()) {
1088                 column += Changes::latexMarkChange(os, bparams, running_change,
1089                         Change(Change::UNCHANGED), runparams);
1090                 running_change = Change(Change::UNCHANGED);
1091         }
1092
1093         bool close = false;
1094         odocstream::pos_type const len = os.os().tellp();
1095
1096         if (inset->forceLTR(runparams)
1097             // babel with Xe/LuaTeX does not need a switch
1098             // and \L is not defined there.
1099             && (!runparams.isFullUnicode() || !runparams.use_babel)
1100             && running_font.isRightToLeft()
1101             // ERT is an exception, it should be output with no
1102             // decorations at all
1103             && inset->lyxCode() != ERT_CODE) {
1104                 if (runparams.use_polyglossia)
1105                         // (lua)bidi
1106                         os << "\\LRE{";
1107                 else if (running_font.language()->lang() == "farsi"
1108                          || running_font.language()->lang() == "arabic_arabi")
1109                         os << "\\textLR{" << termcmd;
1110                 else
1111                         // babel classic
1112                         os << "\\L{";
1113                 close = true;
1114         }
1115
1116         if (open_font && fontswitch_inset) {
1117                 bool lang_closed = false;
1118                 // Close language if needed
1119                 if (closeLanguage && !lang_switched_at_inset) {
1120                         // We need prev_font here as language changes directly at inset
1121                         // will only be started inside the inset.
1122                         Font const prev_font = (i > 0) ?
1123                                                 owner_->getFont(bparams, i - 1, outerfont)
1124                                               : running_font;
1125                         Font tmpfont(basefont);
1126                         tmpfont.setLanguage(prev_font.language());
1127                         bool needPar = false;
1128                         unsigned int count = tmpfont.latexWriteEndChanges(os, bparams, runparams,
1129                                                                           basefont, basefont,
1130                                                                           needPar, closeLanguage);
1131                         column += count;
1132                         lang_closed = count > 0;
1133                 }
1134                 // Update the running_font, making sure, however,
1135                 // to leave the language as it was.
1136                 // FIXME: probably a better way to keep track of the old
1137                 // language, than copying the entire font?
1138                 Font const copy_font(running_font);
1139                 basefont = owner_->getLayoutFont(bparams, outerfont);
1140                 running_font = basefont;
1141                 if (!closeLanguage)
1142                         running_font.setLanguage(copy_font.language());
1143                 OutputParams rp = runparams;
1144                 rp.encoding = basefont.language()->encoding();
1145                 // For these, we use switches, so they should be taken as
1146                 // base inside the inset.
1147                 basefont.fontInfo().setSize(copy_font.fontInfo().size());
1148                 basefont.fontInfo().setFamily(copy_font.fontInfo().family());
1149                 basefont.fontInfo().setSeries(copy_font.fontInfo().series());
1150                 // Now re-do font changes in a way needed here
1151                 // (using switches with multi-par insets)
1152                 InsetText const * textinset = inset->asInsetText();
1153                 bool const cprotect = textinset
1154                         ? textinset->hasCProtectContent(runparams.moving_arg)
1155                           && !textinset->text().isMainText()
1156                           && inset->lyxCode() != BRANCH_CODE
1157                         : false;
1158                 unsigned int count2 = basefont.latexWriteStartChanges(os, bparams,
1159                                                       rp, running_font,
1160                                                       basefont, true,
1161                                                       cprotect);
1162                 open_font = true;
1163                 column += count2;
1164                 if (count2 == 0 && (lang_closed || lang_switched_at_inset))
1165                         // All fonts closed
1166                         open_font = false;
1167                 if (closeLanguage)
1168                         runparams.local_font = &basefont;
1169         }
1170
1171         if (fontswitch_inset && !closeLanguage) {
1172                 // The directionality has been switched at inset.
1173                 // Force markup inside.
1174                 runparams.local_font = &basefont;
1175         }
1176
1177         size_t const previous_row_count = os.texrow().rows();
1178
1179         try {
1180                 runparams.lastid = id_;
1181                 runparams.lastpos = i;
1182                 inset->latex(os, runparams);
1183         } catch (EncodingException & e) {
1184                 // add location information and throw again.
1185                 e.par_id = id_;
1186                 e.pos = i;
1187                 throw;
1188         }
1189
1190         if (close)
1191                 os << '}';
1192
1193         if (os.texrow().rows() > previous_row_count) {
1194                 os.texrow().start(owner_->id(), i + 1);
1195                 column = 0;
1196         } else {
1197                 column += (unsigned int)(os.os().tellp() - len);
1198         }
1199
1200         if (owner_->isDeleted(i))
1201                 --runparams.inDeletedInset;
1202 }
1203
1204
1205 void Paragraph::Private::latexSpecialChar(otexstream & os,
1206                                           BufferParams const & bparams,
1207                                           OutputParams const & runparams,
1208                                           Font const & running_font,
1209                                           string & alien_script,
1210                                           Layout const & style,
1211                                           pos_type & i,
1212                                           pos_type end_pos,
1213                                           unsigned int & column) const
1214 {
1215         char_type const c = owner_->getUChar(bparams, runparams, i);
1216
1217         if (style.pass_thru || runparams.pass_thru || runparams.find_effective()
1218             || contains(style.pass_thru_chars, c)
1219             || contains(runparams.pass_thru_chars, c)) {
1220                 if (runparams.find_effective()) {
1221                         switch (c) {
1222                         case '\\':
1223                                 os << "\\\\";
1224                                 return;
1225                         case '{':
1226                                 os << "\\braceleft ";
1227                                 return;
1228                         case '}':
1229                                 os << "\\braceright ";
1230                                 return;
1231                         case '$':
1232                                 os << "\\lyxdollar ";
1233                                 return;
1234                         case '~':
1235                                 os << "\\lyxtilde ";
1236                                 return;
1237                         case ' ':
1238                         case '\0':
1239                                 break;
1240                         default:
1241                                 os.put(c);
1242                                 return;
1243                         }
1244                 }
1245                 else if (c != '\0') {
1246                         Encoding const * const enc = runparams.encoding;
1247                         if (enc && !enc->encodable(c))
1248                                 throw EncodingException(c);
1249                         os.put(c);
1250                         return;
1251                 }
1252         }
1253
1254         // TIPA uses its own T3 encoding
1255         if (runparams.inIPA && latexSpecialT3(c, os, i, column))
1256                 return;
1257         // If T1 font encoding is used, use the special
1258         // characters it provides.
1259         // NOTE: Some languages reset the font encoding internally to a
1260         //       non-standard font encoding. If we are using such a language,
1261         //       we do not output special T1 chars.
1262         if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1263             && !runparams.isFullUnicode() && runparams.main_fontenc == "T1"
1264             && latexSpecialT1(c, os, i, column))
1265                 return;
1266         // NOTE: "fontspec" (non-TeX fonts) sets the font encoding to "TU" (untill 2017 "EU1" or "EU2")
1267         else if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1268                  && runparams.isFullUnicode() && latexSpecialTU(c, os, i, column))
1269                      return;
1270
1271         // Otherwise, we use what LaTeX provides us.
1272         switch (c) {
1273         case '\\':
1274                 os << "\\textbackslash" << termcmd;
1275                 column += 15;
1276                 break;
1277         case '<':
1278                 os << "\\textless" << termcmd;
1279                 column += 10;
1280                 break;
1281         case '>':
1282                 os << "\\textgreater" << termcmd;
1283                 column += 13;
1284                 break;
1285         case '|':
1286                 os << "\\textbar" << termcmd;
1287                 column += 9;
1288                 break;
1289         case '-':
1290                 os << '-';
1291                 if (i + 1 < static_cast<pos_type>(text_.size()) &&
1292                     (end_pos == -1 || i + 1 < end_pos) &&
1293                     text_[i+1] == '-') {
1294                         // Prevent "--" becoming an en dash and "---" an em dash.
1295                         // (Within \ttfamily, "---" is merged to en dash + hyphen.)
1296                         os << "{}";
1297                         column += 2;
1298                 }
1299                 break;
1300         case '\"':
1301                 os << "\\textquotedbl" << termcmd;
1302                 column += 14;
1303                 break;
1304
1305         case '$': case '&':
1306         case '%': case '#': case '{':
1307         case '}': case '_':
1308                 os << '\\';
1309                 os.put(c);
1310                 column += 1;
1311                 break;
1312
1313         case '~':
1314                 os << "\\textasciitilde" << termcmd;
1315                 column += 16;
1316                 break;
1317
1318         case '^':
1319                 os << "\\textasciicircum" << termcmd;
1320                 column += 17;
1321                 break;
1322
1323         case '*':
1324         case '[':
1325         case ']':
1326                 // avoid being mistaken for optional arguments
1327                 os << '{';
1328                 os.put(c);
1329                 os << '}';
1330                 column += 2;
1331                 break;
1332
1333         case ' ':
1334                 // Blanks are printed before font switching.
1335                 // Sure? I am not! (try nice-latex)
1336                 // I am sure it's correct. LyX might be smarter
1337                 // in the future, but for now, nothing wrong is
1338                 // written. (Asger)
1339                 break;
1340
1341         case 0x2013:
1342         case 0x2014:
1343                 // XeTeX's dash behaviour is determined via a global setting
1344                 if (bparams.use_dash_ligatures
1345                     && owner_->getFontSettings(bparams, i).fontInfo().family() != TYPEWRITER_FAMILY
1346                     && !runparams.inIPA
1347                         // TODO #10961: && not in inset Flex Code
1348                         // TODO #10961: && not in layout LyXCode
1349                     && (!bparams.useNonTeXFonts || runparams.flavor != Flavor::XeTeX)) {
1350                         if (c == 0x2013) {
1351                                 // en-dash
1352                                 os << "--";
1353                                 column +=2;
1354                         } else {
1355                                 // em-dash
1356                                 os << "---";
1357                                 column +=3;
1358                         }
1359                         break;
1360                 }
1361                 // fall through
1362         default:
1363                 if (c == '\0')
1364                         return;
1365
1366                 Encoding const & encoding = *(runparams.encoding);
1367                 char_type next = '\0';
1368                 if (i + 1 < int(text_.size())) {
1369                         next = text_[i + 1];
1370                         if (Encodings::isCombiningChar(next)) {
1371                                 column += latexSurrogatePair(bparams, os, c, next, runparams) - 1;
1372                                 ++i;
1373                                 break;
1374                         }
1375                 }
1376                 pair<docstring, bool> latex = encoding.latexChar(c);
1377                 docstring nextlatex;
1378                 bool nexttipas = false;
1379                 string nexttipashortcut;
1380                 if (next != '\0' && next != META_INSET && !encoding.encodable(next)) {
1381                         nextlatex = encoding.latexChar(next).first;
1382                         if (runparams.inIPA) {
1383                                 nexttipashortcut = Encodings::TIPAShortcut(next);
1384                                 nexttipas = !nexttipashortcut.empty();
1385                         }
1386                 }
1387                 bool tipas = false;
1388                 if (runparams.inIPA) {
1389                         string const tipashortcut = Encodings::TIPAShortcut(c);
1390                         if (!tipashortcut.empty()) {
1391                                 latex.first = from_ascii(tipashortcut);
1392                                 latex.second = false;
1393                                 tipas = true;
1394                         }
1395                 }
1396                 // eventually close "script wrapper" command (see `Paragraph::latex`)
1397                 if (!alien_script.empty()
1398                         && alien_script != Encodings::isKnownScriptChar(next)) {
1399                         column += latex.first.length();
1400                         alien_script.clear();
1401                         os << latex.first << "}";
1402                         break;
1403                 }
1404                 if (latex.second
1405                          && ((!prefixIs(nextlatex, '\\')
1406                                && !prefixIs(nextlatex, '{')
1407                                && !prefixIs(nextlatex, '}'))
1408                              || (nexttipas
1409                                  && !prefixIs(from_ascii(nexttipashortcut), '\\')))
1410                          && !tipas) {
1411                         // Prevent eating of a following space or command corruption by
1412                         // following characters
1413                         if (next == ' ' || next == '\0') {
1414                                 column += latex.first.length() + 1;
1415                                 os << latex.first << "{}";
1416                         } else {
1417                                 column += latex.first.length();
1418                                 os << latex.first << " ";
1419                         }
1420                 } else {
1421                         column += latex.first.length() - 1;
1422                         os << latex.first;
1423                 }
1424                 break;
1425         }
1426 }
1427
1428
1429 bool Paragraph::Private::latexSpecialT1(char_type const c, otexstream & os,
1430         pos_type i, unsigned int & column) const
1431 {
1432         switch (c) {
1433         case '>':
1434         case '<':
1435                 os.put(c);
1436                 // In T1 encoding, these characters exist
1437                 // but we should avoid ligatures
1438                 if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
1439                         return true;
1440                 os << "\\textcompwordmark" << termcmd;
1441                 column += 19;
1442                 return true;
1443         case '|':
1444                 os.put(c);
1445                 return true;
1446         case '\"':
1447                 // soul.sty breaks with \char`\"
1448                 os << "\\textquotedbl" << termcmd;
1449                 column += 14;
1450                 return true;
1451         default:
1452                 return false;
1453         }
1454 }
1455
1456
1457 bool Paragraph::Private::latexSpecialTU(char_type const c, otexstream & os,
1458         pos_type i, unsigned int & column) const
1459 {
1460         // TU encoding is currently on par with T1.
1461         return latexSpecialT1(c, os, i, column);
1462 }
1463
1464
1465 bool Paragraph::Private::latexSpecialT3(char_type const c, otexstream & os,
1466         pos_type /*i*/, unsigned int & column) const
1467 {
1468         switch (c) {
1469         case '*':
1470         case '[':
1471         case ']':
1472         case '\"':
1473                 os.put(c);
1474                 return true;
1475         case '|':
1476                 os << "\\textvertline" << termcmd;
1477                 column += 14;
1478                 return true;
1479         default:
1480                 return false;
1481         }
1482 }
1483
1484
1485 void Paragraph::Private::validate(LaTeXFeatures & features) const
1486 {
1487         if (layout_->inpreamble && inset_owner_) {
1488                 // FIXME: Using a string stream here circumvents the encoding
1489                 // switching machinery of odocstream. Therefore the
1490                 // output is wrong if this paragraph contains content
1491                 // that needs to switch encoding.
1492                 Buffer const & buf = inset_owner_->buffer();
1493                 otexstringstream os;
1494                 os << layout_->preamble();
1495                 size_t const length = os.length();
1496                 TeXOnePar(buf, *inset_owner_->getText(int(buf.getParFromID(owner_->id()).idx())),
1497                           buf.getParFromID(owner_->id()).pit(), os,
1498                           features.runparams(), string(), 0, -1, true);
1499                 if (os.length() > length)
1500                         features.addPreambleSnippet(os.release(), true);
1501         }
1502
1503         if (features.runparams().flavor == Flavor::Html
1504             && layout_->htmltitle()) {
1505                 features.setHTMLTitle(owner_->asString(AS_STR_INSETS | AS_STR_SKIPDELETE));
1506         }
1507
1508         // check the params.
1509         if (!params_.spacing().isDefault())
1510                 features.require("setspace");
1511
1512         // then the layouts
1513         features.useLayout(layout_->name());
1514
1515         // then the fonts
1516         fontlist_.validate(features);
1517
1518         // then the indentation
1519         if (!params_.leftIndent().zero())
1520                 features.require("ParagraphLeftIndent");
1521
1522         // then the insets
1523         InsetList::const_iterator icit = insetlist_.begin();
1524         InsetList::const_iterator iend = insetlist_.end();
1525         for (; icit != iend; ++icit) {
1526                 if (icit->inset) {
1527                         features.inDeletedInset(owner_->isDeleted(icit->pos));
1528                         if (icit->inset->lyxCode() == FOOT_CODE) {
1529                                 // FIXME: an item inset would make things much easier.
1530                                 if ((layout_->latextype == LATEX_LIST_ENVIRONMENT
1531                                      || (layout_->latextype == LATEX_ITEM_ENVIRONMENT
1532                                          && layout_->margintype == MARGIN_FIRST_DYNAMIC))
1533                                     && (icit->pos < begin_of_body_
1534                                         || (icit->pos == begin_of_body_
1535                                             && (icit->pos == 0 || text_[icit->pos - 1] != ' '))))
1536                                         features.saveNoteEnv("description");
1537                         }
1538                         icit->inset->validate(features);
1539                         features.inDeletedInset(false);
1540                         if (layout_->needprotect &&
1541                             icit->inset->lyxCode() == FOOT_CODE)
1542                                 features.require("NeedLyXFootnoteCode");
1543                 }
1544         }
1545
1546         // then the contents
1547         BufferParams const bp = features.runparams().is_child
1548                 ? features.buffer().masterParams() : features.buffer().params();
1549         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1550                 char_type c = text_[i];
1551                 CharInfo const & ci = Encodings::unicodeCharInfo(c);
1552                 if (c == 0x0022) {
1553                         if (features.runparams().isFullUnicode() && bp.useNonTeXFonts)
1554                                 features.require("textquotedblp");
1555                         else if (features.runparams().main_fontenc != "T1"
1556                                  || ((&owner_->getFontSettings(bp, i))->language()->internalFontEncoding()))
1557                                 features.require("textquotedbl");
1558                 } else if (ci.textFeature() && contains(ci.textPreamble(), '=')) {
1559                         // features that depend on the font or input encoding
1560                         string feats = ci.textPreamble();
1561                         string fontenc = (&owner_->getFontSettings(bp, i))->language()->fontenc(bp);
1562                         if (fontenc.empty())
1563                                 fontenc = features.runparams().main_fontenc;
1564                         while (!feats.empty()) {
1565                                 string feat;
1566                                 feats = split(feats, feat, ',');
1567                                 if (contains(feat, "!=")) {
1568                                         // a feature that is required except for the spcified
1569                                         // font or input encodings
1570                                         string realfeature;
1571                                         string const contexts = ltrim(split(feat, realfeature, '!'), "=");
1572                                         // multiple encodings are separated by semicolon
1573                                         vector<string> context = getVectorFromString(contexts, ";");
1574                                         // require feature if the context matches neither current font
1575                                         // nor input encoding
1576                                         if (std::find(context.begin(), context.end(), fontenc) == context.end()
1577                                             && std::find(context.begin(), context.end(),
1578                                                          features.runparams().encoding->name()) == context.end())
1579                                                 features.require(realfeature);
1580                                 } else if (contains(feat, '=')) {
1581                                         // a feature that is required only for the spcified
1582                                         // font or input encodings
1583                                         string realfeature;
1584                                         string const contexts = split(feat, realfeature, '=');
1585                                         // multiple encodings are separated by semicolon
1586                                         vector<string> context = getVectorFromString(contexts, ";");
1587                                         // require feature if the context matches either current font
1588                                         // or input encoding
1589                                         if (std::find(context.begin(), context.end(), fontenc) != context.end()
1590                                             || std::find(context.begin(), context.end(),
1591                                                          features.runparams().encoding->name()) != context.end())
1592                                                 features.require(realfeature);
1593                                 }
1594                         }
1595                 } else if (!bp.use_dash_ligatures
1596                            && (c == 0x2013 || c == 0x2014)
1597                            && bp.useNonTeXFonts
1598                            && features.runparams().flavor == Flavor::XeTeX)
1599                         // XeTeX's dash behaviour is determined via a global setting
1600                         features.require("xetexdashbreakstate");
1601                 BufferEncodings::validate(c, features);
1602         }
1603 }
1604
1605 /////////////////////////////////////////////////////////////////////
1606 //
1607 // Paragraph
1608 //
1609 /////////////////////////////////////////////////////////////////////
1610
1611 namespace {
1612         Layout const emptyParagraphLayout;
1613 }
1614
1615 Paragraph::Paragraph()
1616         : d(new Paragraph::Private(this, emptyParagraphLayout))
1617 {
1618         itemdepth = 0;
1619         d->params_.clear();
1620 }
1621
1622
1623 Paragraph::Paragraph(Paragraph const & par)
1624         : itemdepth(par.itemdepth),
1625         d(new Paragraph::Private(*par.d, this))
1626 {
1627         registerWords();
1628 }
1629
1630
1631 Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
1632         : itemdepth(par.itemdepth),
1633         d(new Paragraph::Private(*par.d, this, beg, end))
1634 {
1635         registerWords();
1636 }
1637
1638
1639 Paragraph & Paragraph::operator=(Paragraph const & par)
1640 {
1641         // needed as we will destroy the private part before copying it
1642         if (&par != this) {
1643                 itemdepth = par.itemdepth;
1644
1645                 deregisterWords();
1646                 delete d;
1647                 d = new Private(*par.d, this);
1648                 registerWords();
1649         }
1650         return *this;
1651 }
1652
1653
1654 Paragraph::~Paragraph()
1655 {
1656         deregisterWords();
1657         delete d;
1658 }
1659
1660
1661 namespace {
1662
1663 // this shall be called just before every "os << ..." action.
1664 void flushString(ostream & os, docstring & s)
1665 {
1666         os << to_utf8(s);
1667         s.erase();
1668 }
1669
1670 } // namespace
1671
1672
1673 void Paragraph::write(ostream & os, BufferParams const & bparams,
1674         depth_type & depth) const
1675 {
1676         // The beginning or end of a deeper (i.e. nested) area?
1677         if (depth != d->params_.depth()) {
1678                 if (d->params_.depth() > depth) {
1679                         while (d->params_.depth() > depth) {
1680                                 os << "\n\\begin_deeper";
1681                                 ++depth;
1682                         }
1683                 } else {
1684                         while (d->params_.depth() < depth) {
1685                                 os << "\n\\end_deeper";
1686                                 --depth;
1687                         }
1688                 }
1689         }
1690
1691         // First write the layout
1692         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1693
1694         d->params_.write(os);
1695
1696         Font font1(inherit_font, bparams.language);
1697
1698         Change running_change = Change(Change::UNCHANGED);
1699
1700         // this string is used as a buffer to avoid repetitive calls
1701         // to to_utf8(), which turn out to be expensive (JMarc)
1702         docstring write_buffer;
1703
1704         int column = 0;
1705         for (pos_type i = 0; i <= size(); ++i) {
1706
1707                 Change const & change = lookupChange(i);
1708                 if (change != running_change)
1709                         flushString(os, write_buffer);
1710                 Changes::lyxMarkChange(os, bparams, column, running_change, change);
1711                 running_change = change;
1712
1713                 if (i == size())
1714                         break;
1715
1716                 // Write font changes
1717                 Font font2 = getFontSettings(bparams, i);
1718                 if (font2 != font1) {
1719                         flushString(os, write_buffer);
1720                         font2.lyxWriteChanges(font1, os);
1721                         column = 0;
1722                         font1 = font2;
1723                 }
1724
1725                 char_type const c = d->text_[i];
1726                 switch (c) {
1727                 case META_INSET:
1728                         if (Inset const * inset = getInset(i)) {
1729                                 flushString(os, write_buffer);
1730                                 if (inset->directWrite()) {
1731                                         // international char, let it write
1732                                         // code directly so it's shorter in
1733                                         // the file
1734                                         inset->write(os);
1735                                 } else {
1736                                         if (i)
1737                                                 os << '\n';
1738                                         os << "\\begin_inset ";
1739                                         inset->write(os);
1740                                         os << "\n\\end_inset\n\n";
1741                                         column = 0;
1742                                 }
1743                                 // FIXME This can be removed again once the mystery
1744                                 // crash has been resolved.
1745                                 os << flush;
1746                         }
1747                         break;
1748                 case '\\':
1749                         flushString(os, write_buffer);
1750                         os << "\n\\backslash\n";
1751                         column = 0;
1752                         break;
1753                 case '.':
1754                 case '!':
1755                 case '?':
1756                 case ':':
1757                 case ';':
1758                 case ',':
1759                 case 0x061F:  // ؟ U+061F  ARABIC QUESTION MARK
1760                 case 0x061B:  // ؛ U+061B  ARABIC SEMICOLON
1761                 case 0x060C:  // ، U+060C  ARABIC COMMA
1762                         flushString(os, write_buffer);
1763                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1764                                 os << to_utf8(docstring(1, c)) << '\n';
1765                                 column = 0;
1766                         } else
1767                                 os << to_utf8(docstring(1, c));
1768                         break;
1769                 case 0x2014:  // — U+2014  EM DASH
1770                 case 0x3002:  // 。 U+3002  IDEOGRAPHIC FULL STOP
1771                 case 0xFF01:  // ！ U+FF01  FULLWIDTH EXCLAMATION MARK
1772                 case 0xFF1F:  // ？ U+FF1F  FULLWIDTH QUESTION MARK
1773                 case 0xFF1A:  // ： U+FF1A  FULLWIDTH COLON
1774                 case 0xFF1B:  // ； U+FF1B  FULLWIDTH SEMICOLON
1775                 case 0xFF0C:  // ， U+FF0C  FULLWIDTH COMMA
1776                         flushString(os, write_buffer);
1777                         os << to_utf8(docstring(1, c)) << '\n';
1778                         column = 0;
1779                         break;
1780                 default:
1781                         if (column > 500) {
1782                                 flushString(os, write_buffer);
1783                                 os << '\n';
1784                                 column = 0;
1785                         }
1786                         // this check is to amend a bug. LyX sometimes
1787                         // inserts '\0' this could cause problems.
1788                         if (c != '\0')
1789                                 write_buffer.push_back(c);
1790                         else
1791                                 LYXERR0("NUL char in structure.");
1792                         ++column;
1793                         break;
1794                 }
1795         }
1796
1797         flushString(os, write_buffer);
1798         os << "\n\\end_layout\n";
1799         // FIXME This can be removed again once the mystery
1800         // crash has been resolved.
1801         os << flush;
1802 }
1803
1804
1805 void Paragraph::validate(LaTeXFeatures & features) const
1806 {
1807         d->validate(features);
1808         bool fragile = features.runparams().moving_arg;
1809         fragile |= layout().needprotect;
1810         if (inInset().getLayout().isNeedProtect())
1811                 fragile = true;
1812         if (needsCProtection(fragile))
1813                 features.require("cprotect");
1814 }
1815
1816
1817 void Paragraph::insert(pos_type pos, docstring const & str,
1818                        Font const & font, Change const & change)
1819 {
1820         for (size_t i = 0, n = str.size(); i != n ; ++i)
1821                 insertChar(pos + i, str[i], font, change);
1822 }
1823
1824
1825 void Paragraph::appendChar(char_type c, Font const & font,
1826                 Change const & change)
1827 {
1828         // Make sure that Buffer::hasChangesPresent is updated
1829         ChangesMonitor cm(*this);
1830
1831         // track change
1832         d->changes_.insert(change, d->text_.size());
1833         // when appending characters, no need to update tables
1834         d->text_.push_back(c);
1835         setFont(d->text_.size() - 1, font);
1836         d->requestSpellCheck(d->text_.size() - 1);
1837 }
1838
1839
1840 void Paragraph::appendString(docstring const & s, Font const & font,
1841                 Change const & change)
1842 {
1843         // Make sure that Buffer::hasChangesPresent is updated
1844         ChangesMonitor cm(*this);
1845
1846         pos_type end = s.size();
1847         size_t oldsize = d->text_.size();
1848         size_t newsize = oldsize + end;
1849         size_t capacity = d->text_.capacity();
1850         if (newsize >= capacity)
1851                 d->text_.reserve(max(capacity + 100, newsize));
1852
1853         // when appending characters, no need to update tables
1854         d->text_.append(s);
1855
1856         // FIXME: Optimize this!
1857         for (size_t i = oldsize; i != newsize; ++i) {
1858                 // track change
1859                 d->changes_.insert(change, i);
1860                 d->requestSpellCheck(i);
1861         }
1862         d->fontlist_.set(oldsize, font);
1863         d->fontlist_.set(newsize - 1, font);
1864 }
1865
1866
1867 void Paragraph::insertChar(pos_type pos, char_type c,
1868                            bool trackChanges)
1869 {
1870         d->insertChar(pos, c, Change(trackChanges ?
1871                            Change::INSERTED : Change::UNCHANGED));
1872 }
1873
1874
1875 void Paragraph::insertChar(pos_type pos, char_type c,
1876                            Font const & font, bool trackChanges)
1877 {
1878         d->insertChar(pos, c, Change(trackChanges ?
1879                            Change::INSERTED : Change::UNCHANGED));
1880         setFont(pos, font);
1881 }
1882
1883
1884 void Paragraph::insertChar(pos_type pos, char_type c,
1885                            Font const & font, Change const & change)
1886 {
1887         d->insertChar(pos, c, change);
1888         setFont(pos, font);
1889 }
1890
1891
1892 void Paragraph::resetFonts(Font const & font)
1893 {
1894         d->fontlist_.clear();
1895         d->fontlist_.set(0, font);
1896         d->fontlist_.set(d->text_.size() - 1, font);
1897 }
1898
1899 // Gets uninstantiated font setting at position.
1900 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1901                                          pos_type pos) const
1902 {
1903         if (pos > size()) {
1904                 LYXERR0("pos: " << pos << " size: " << size());
1905                 LBUFERR(false);
1906         }
1907
1908         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1909         if (cit != d->fontlist_.end())
1910                 return cit->font();
1911
1912         if (pos == size() && !empty())
1913                 return getFontSettings(bparams, pos - 1);
1914
1915         // Optimisation: avoid a full font instantiation if there is no
1916         // language change from previous call.
1917         static Font previous_font;
1918         static Language const * previous_lang = nullptr;
1919         Language const * lang = getParLanguage(bparams);
1920         if (lang != previous_lang) {
1921                 previous_lang = lang;
1922                 previous_font = Font(inherit_font, lang);
1923         }
1924         return previous_font;
1925 }
1926
1927
1928 FontSpan Paragraph::fontSpan(pos_type pos) const
1929 {
1930         LBUFERR(pos <= size());
1931
1932         if (pos == size())
1933                 return FontSpan(pos, pos);
1934
1935         pos_type start = 0;
1936         FontList::const_iterator cit = d->fontlist_.begin();
1937         FontList::const_iterator end = d->fontlist_.end();
1938         for (; cit != end; ++cit) {
1939                 if (cit->pos() >= pos) {
1940                         if (pos >= beginOfBody())
1941                                 return FontSpan(max(start, beginOfBody()),
1942                                                 cit->pos());
1943                         else
1944                                 return FontSpan(start,
1945                                                 min(beginOfBody() - 1,
1946                                                          cit->pos()));
1947                 }
1948                 start = cit->pos() + 1;
1949         }
1950
1951         // This should not happen, but if so, we take no chances.
1952         LYXERR0("Paragraph::fontSpan: position not found in fontinfo table!");
1953         LASSERT(false, /**/);
1954         return FontSpan(pos, pos);
1955 }
1956
1957
1958 // Gets uninstantiated font setting at position 0
1959 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1960 {
1961         if (!empty() && !d->fontlist_.empty())
1962                 return d->fontlist_.begin()->font();
1963
1964         // Optimisation: avoid a full font instantiation if there is no
1965         // language change from previous call.
1966         static Font previous_font;
1967         static Language const * previous_lang = nullptr;
1968         if (bparams.language != previous_lang) {
1969                 previous_lang = bparams.language;
1970                 previous_font = Font(inherit_font, bparams.language);
1971         }
1972
1973         return previous_font;
1974 }
1975
1976
1977 // Gets the fully instantiated font at a given position in a paragraph
1978 // This is basically the same function as Text::GetFont() in text2.cpp.
1979 // The difference is that this one is used for generating the LaTeX file,
1980 // and thus cosmetic "improvements" are disallowed: This has to deliver
1981 // the true picture of the buffer. (Asger)
1982 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1983                                  Font const & outerfont) const
1984 {
1985         LBUFERR(pos >= 0);
1986
1987         Font font = getFontSettings(bparams, pos);
1988
1989         pos_type const body_pos = beginOfBody();
1990         FontInfo & fi = font.fontInfo();
1991         if (pos < body_pos)
1992                 fi.realize(d->layout_->labelfont);
1993         else
1994                 fi.realize(d->layout_->font);
1995
1996         fi.realize(outerfont.fontInfo());
1997         fi.realize(bparams.getFont().fontInfo());
1998
1999         return font;
2000 }
2001
2002
2003 Font const Paragraph::getLabelFont
2004         (BufferParams const & bparams, Font const & outerfont) const
2005 {
2006         FontInfo tmpfont = d->layout_->labelfont;
2007         tmpfont.realize(outerfont.fontInfo());
2008         tmpfont.realize(bparams.getFont().fontInfo());
2009         return Font(tmpfont, getParLanguage(bparams));
2010 }
2011
2012
2013 Font const Paragraph::getLayoutFont
2014         (BufferParams const & bparams, Font const & outerfont) const
2015 {
2016         FontInfo tmpfont = d->layout_->font;
2017         tmpfont.realize(outerfont.fontInfo());
2018         tmpfont.realize(bparams.getFont().fontInfo());
2019         return Font(tmpfont, getParLanguage(bparams));
2020 }
2021
2022
2023 char_type Paragraph::getUChar(BufferParams const & bparams,
2024                               OutputParams const & rp,
2025                               pos_type pos) const
2026 {
2027         char_type c = d->text_[pos];
2028
2029         // Return unchanged character in LTR languages
2030         // or if we use poylglossia/bidi (XeTeX).
2031         if (rp.useBidiPackage()
2032             || !getFontSettings(bparams, pos).isRightToLeft())
2033                 return c;
2034
2035         // Without polyglossia/bidi, we need to account for some special cases.
2036         // FIXME This needs to be audited!
2037         // Check if:
2038         // * The input is as expected for all delimiters
2039         //   => checked for Hebrew!
2040         // * The output matches the display in the LyX workarea
2041         //   => checked for Hebrew!
2042         // * The special cases below are really necessary
2043         //   => checked for Hebrew!
2044         // * In arabic_arabi, brackets are transformed to Arabic
2045         //   Ornate Parentheses. Is this is really wanted?
2046
2047         string const & lang = getFontSettings(bparams, pos).language()->lang();
2048         char_type uc = c;
2049
2050         // 1. In the following languages, parentheses need to be reversed.
2051         //    Also with polyglodia/luabidi
2052         bool const reverseparens = (lang == "hebrew" || rp.use_polyglossia);
2053
2054         // 2. In the following languages, brackets don't need to be reversed.
2055         bool const reversebrackets = lang != "arabic_arabtex"
2056                         && lang != "arabic_arabi"
2057                         && lang != "farsi";
2058
2059         // Now swap delimiters if needed.
2060         switch (c) {
2061         case '(':
2062                 if (reverseparens)
2063                         uc = ')';
2064                 break;
2065         case ')':
2066                 if (reverseparens)
2067                         uc = '(';
2068                 break;
2069         case '[':
2070                 if (reversebrackets)
2071                         uc = ']';
2072                 break;
2073         case ']':
2074                 if (reversebrackets)
2075                         uc = '[';
2076                 break;
2077         case '{':
2078                 uc = '}';
2079                 break;
2080         case '}':
2081                 uc = '{';
2082                 break;
2083         case '<':
2084                 uc = '>';
2085                 break;
2086         case '>':
2087                 uc = '<';
2088                 break;
2089         }
2090
2091         return uc;
2092 }
2093
2094
2095 void Paragraph::setFont(pos_type pos, Font const & font)
2096 {
2097         LASSERT(pos <= size(), return);
2098
2099         // First, reduce font against layout/label font
2100         // Update: The setCharFont() routine in text2.cpp already
2101         // reduces font, so we don't need to do that here. (Asger)
2102
2103         d->fontlist_.set(pos, font);
2104 }
2105
2106
2107 void Paragraph::makeSameLayout(Paragraph const & par)
2108 {
2109         d->layout_ = par.d->layout_;
2110         d->params_ = par.d->params_;
2111 }
2112
2113
2114 bool Paragraph::stripLeadingSpaces(bool trackChanges)
2115 {
2116         if (isFreeSpacing())
2117                 return false;
2118
2119         int pos = 0;
2120         int count = 0;
2121
2122         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
2123                 if (eraseChar(pos, trackChanges))
2124                         ++count;
2125                 else
2126                         ++pos;
2127         }
2128
2129         return count > 0 || pos > 0;
2130 }
2131
2132
2133 bool Paragraph::hasSameLayout(Paragraph const & par) const
2134 {
2135         return par.d->layout_ == d->layout_
2136                 && d->params_.sameLayout(par.d->params_);
2137 }
2138
2139
2140 depth_type Paragraph::getDepth() const
2141 {
2142         return d->params_.depth();
2143 }
2144
2145
2146 depth_type Paragraph::getMaxDepthAfter() const
2147 {
2148         if (d->layout_->isEnvironment())
2149                 return d->params_.depth() + 1;
2150         else
2151                 return d->params_.depth();
2152 }
2153
2154
2155 LyXAlignment Paragraph::getAlign(BufferParams const & bparams) const
2156 {
2157         if (d->params_.align() == LYX_ALIGN_LAYOUT)
2158                 return getDefaultAlign(bparams);
2159         else
2160                 return d->params_.align();
2161 }
2162
2163
2164 LyXAlignment Paragraph::getDefaultAlign(BufferParams const & bparams) const
2165 {
2166         LyXAlignment res = layout().align;
2167         if (isRTL(bparams)) {
2168                 // Swap sides
2169                 if (res == LYX_ALIGN_LEFT)
2170                         res = LYX_ALIGN_RIGHT;
2171                 else if  (res == LYX_ALIGN_RIGHT)
2172                         res = LYX_ALIGN_LEFT;
2173         }
2174         return res;
2175 }
2176
2177
2178 docstring const & Paragraph::labelString() const
2179 {
2180         return d->params_.labelString();
2181 }
2182
2183
2184 // the next two functions are for the manual labels
2185 docstring const Paragraph::getLabelWidthString() const
2186 {
2187         if (d->layout_->margintype == MARGIN_MANUAL
2188             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
2189                 return d->params_.labelWidthString();
2190         else
2191                 return _("Senseless with this layout!");
2192 }
2193
2194
2195 void Paragraph::setLabelWidthString(docstring const & s)
2196 {
2197         d->params_.labelWidthString(s);
2198 }
2199
2200
2201 docstring Paragraph::expandLabel(Layout const & layout,
2202                 BufferParams const & bparams) const
2203 {
2204         return expandParagraphLabel(layout, bparams, true);
2205 }
2206
2207
2208 docstring Paragraph::expandParagraphLabel(Layout const & layout,
2209                 BufferParams const & bparams, bool process_appendix) const
2210 {
2211         DocumentClass const & tclass = bparams.documentClass();
2212         string const & lang = getParLanguage(bparams)->code();
2213         bool const in_appendix = process_appendix && d->params_.appendix();
2214         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
2215
2216         if (fmt.empty() && !layout.counter.empty())
2217                 return tclass.counters().theCounter(layout.counter, lang);
2218
2219         // handle 'inherited level parts' in 'fmt',
2220         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
2221         size_t const i = fmt.find('@', 0);
2222         if (i != docstring::npos) {
2223                 size_t const j = fmt.find('@', i + 1);
2224                 if (j != docstring::npos) {
2225                         docstring parent(fmt, i + 1, j - i - 1);
2226                         docstring label = from_ascii("??");
2227                         if (tclass.hasLayout(parent))
2228                                 label = expandParagraphLabel(tclass[parent], bparams,
2229                                                       process_appendix);
2230                         fmt = docstring(fmt, 0, i) + label
2231                                 + docstring(fmt, j + 1, docstring::npos);
2232                 }
2233         }
2234
2235         return tclass.counters().counterLabel(fmt, lang);
2236 }
2237
2238
2239 void Paragraph::applyLayout(Layout const & new_layout)
2240 {
2241         d->layout_ = &new_layout;
2242         LyXAlignment const oldAlign = d->params_.align();
2243
2244         if (!(oldAlign & d->layout_->alignpossible)) {
2245                 frontend::Alert::warning(_("Alignment not permitted"),
2246                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
2247                 d->params_.align(LYX_ALIGN_LAYOUT);
2248         }
2249 }
2250
2251
2252 pos_type Paragraph::beginOfBody() const
2253 {
2254         return d->begin_of_body_;
2255 }
2256
2257
2258 void Paragraph::setBeginOfBody()
2259 {
2260         if (d->layout_->labeltype != LABEL_MANUAL) {
2261                 d->begin_of_body_ = 0;
2262                 return;
2263         }
2264
2265         // Unroll the first two cycles of the loop
2266         // and remember the previous character to
2267         // remove unnecessary getChar() calls
2268         pos_type i = 0;
2269         pos_type end = size();
2270         bool prev_char_deleted = false;
2271         if (i < end && (!(isNewline(i) || isEnvSeparator(i)) || isDeleted(i))) {
2272                 ++i;
2273                 if (i < end) {
2274                         char_type previous_char = d->text_[i];
2275                         if (!(isNewline(i) || isEnvSeparator(i))) {
2276                                 ++i;
2277                                 while (i < end && (previous_char != ' ' || prev_char_deleted)) {
2278                                         char_type temp = d->text_[i];
2279                                         prev_char_deleted = isDeleted(i);
2280                                         if (!isDeleted(i) && (isNewline(i) || isEnvSeparator(i)))
2281                                                 break;
2282                                         ++i;
2283                                         previous_char = temp;
2284                                 }
2285                         }
2286                 }
2287         }
2288
2289         d->begin_of_body_ = i;
2290 }
2291
2292
2293 bool Paragraph::allowParagraphCustomization() const
2294 {
2295         return inInset().allowParagraphCustomization();
2296 }
2297
2298
2299 bool Paragraph::usePlainLayout() const
2300 {
2301         return inInset().usePlainLayout();
2302 }
2303
2304
2305 bool Paragraph::isPassThru() const
2306 {
2307         return inInset().isPassThru() || d->layout_->pass_thru;
2308 }
2309
2310
2311 bool Paragraph::parbreakIsNewline() const
2312 {
2313         return inInset().getLayout().parbreakIsNewline() || d->layout_->parbreak_is_newline;
2314 }
2315
2316
2317 bool Paragraph::isPartOfTextSequence() const
2318 {
2319         for (pos_type i = 0; i < size(); ++i) {
2320                 if (!isInset(i) || getInset(i)->isPartOfTextSequence())
2321                         return true;
2322         }
2323         return false;
2324 }
2325
2326 namespace {
2327
2328 // paragraphs inside floats need different alignment tags to avoid
2329 // unwanted space
2330
2331 bool noTrivlistCentering(InsetCode code)
2332 {
2333         return code == FLOAT_CODE
2334                || code == WRAP_CODE
2335                || code == CELL_CODE;
2336 }
2337
2338
2339 string correction(string const & orig)
2340 {
2341         if (orig == "flushleft")
2342                 return "raggedright";
2343         if (orig == "flushright")
2344                 return "raggedleft";
2345         if (orig == "center")
2346                 return "centering";
2347         return orig;
2348 }
2349
2350
2351 bool corrected_env(otexstream & os, string const & suffix, string const & env,
2352         InsetCode code, bool const lastpar, int & col)
2353 {
2354         string macro = suffix + "{";
2355         if (noTrivlistCentering(code)) {
2356                 if (lastpar) {
2357                         // the last paragraph in non-trivlist-aligned
2358                         // context is special (to avoid unwanted whitespace)
2359                         if (suffix == "\\begin") {
2360                                 macro = "\\" + correction(env) + "{}";
2361                                 os << from_ascii(macro);
2362                                 col += macro.size();
2363                                 return true;
2364                         }
2365                         return false;
2366                 }
2367                 macro += correction(env);
2368         } else
2369                 macro += env;
2370         macro += "}";
2371         if (suffix == "\\par\\end") {
2372                 os << breakln;
2373                 col = 0;
2374         }
2375         os << from_ascii(macro);
2376         col += macro.size();
2377         if (suffix == "\\begin") {
2378                 os << breakln;
2379                 col = 0;
2380         }
2381         return true;
2382 }
2383
2384 } // namespace
2385
2386
2387 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
2388                         otexstream & os, OutputParams const & runparams) const
2389 {
2390         int column = 0;
2391
2392         bool canindent =
2393                 (bparams.paragraph_separation == BufferParams::ParagraphIndentSeparation) ?
2394                         (layout_->toggle_indent != ITOGGLE_NEVER) :
2395                         (layout_->toggle_indent == ITOGGLE_ALWAYS);
2396
2397         LyXAlignment const curAlign = params_.align();
2398
2399         // Do not output \\noindent for paragraphs
2400         // 1. that cannot have indentation or are indented always,
2401         // 2. that are not part of the immediate text sequence (e.g., contain only floats),
2402         // 3. that are PassThru,
2403         // 4. or that are centered.
2404         if (canindent && params_.noindent()
2405             && owner_->isPartOfTextSequence()
2406             && !layout_->pass_thru
2407             && curAlign != LYX_ALIGN_CENTER) {
2408                 if (!owner_->empty()
2409                     && owner_->getInset(0)
2410                     && owner_->getInset(0)->lyxCode() == VSPACE_CODE)
2411                         // If the paragraph starts with a vspace, the \\noindent
2412                         // needs to come after that (as it leaves vmode).
2413                         // If the paragraph consists only of the vspace,
2414                         // \\noindent is not needed at all.
2415                         runparams.need_noindent = owner_->size() > 1;
2416                 else {
2417                         os << "\\noindent" << termcmd;
2418                         column += 10;
2419                 }
2420         }
2421
2422         if (curAlign == layout_->align)
2423                 return column;
2424
2425         switch (curAlign) {
2426         case LYX_ALIGN_NONE:
2427         case LYX_ALIGN_BLOCK:
2428         case LYX_ALIGN_LAYOUT:
2429         case LYX_ALIGN_SPECIAL:
2430         case LYX_ALIGN_DECIMAL:
2431                 break;
2432         case LYX_ALIGN_LEFT:
2433         case LYX_ALIGN_RIGHT:
2434         case LYX_ALIGN_CENTER:
2435                 if (runparams.moving_arg) {
2436                         os << "\\protect";
2437                         column += 8;
2438                 }
2439                 break;
2440         }
2441
2442         string const begin_tag = "\\begin";
2443         InsetCode code = ownerCode();
2444         bool const lastpar = runparams.isLastPar;
2445         // RTL in classic (PDF)LaTeX (without the Bidi package)
2446         // Luabibdi (used by LuaTeX) behaves like classic
2447         bool const rtl_classic = owner_->getParLanguage(bparams)->rightToLeft()
2448                 && !runparams.useBidiPackage();
2449
2450         switch (curAlign) {
2451         case LYX_ALIGN_NONE:
2452         case LYX_ALIGN_BLOCK:
2453         case LYX_ALIGN_LAYOUT:
2454         case LYX_ALIGN_SPECIAL:
2455         case LYX_ALIGN_DECIMAL:
2456                 break;
2457         case LYX_ALIGN_LEFT: {
2458                 if (rtl_classic)
2459                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2460                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2461                 else
2462                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2463                 break;
2464         } case LYX_ALIGN_RIGHT: {
2465                 if (rtl_classic)
2466                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2467                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2468                 else
2469                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2470                 break;
2471         } case LYX_ALIGN_CENTER: {
2472                 corrected_env(os, begin_tag, "center", code, lastpar, column);
2473                 break;
2474         }
2475         }
2476
2477         return column;
2478 }
2479
2480
2481 bool Paragraph::Private::endTeXParParams(BufferParams const & bparams,
2482                         otexstream & os, OutputParams const & runparams) const
2483 {
2484         LyXAlignment const curAlign = params_.align();
2485
2486         if (curAlign == layout_->align)
2487                 return false;
2488
2489         switch (curAlign) {
2490         case LYX_ALIGN_NONE:
2491         case LYX_ALIGN_BLOCK:
2492         case LYX_ALIGN_LAYOUT:
2493         case LYX_ALIGN_SPECIAL:
2494         case LYX_ALIGN_DECIMAL:
2495                 break;
2496         case LYX_ALIGN_LEFT:
2497         case LYX_ALIGN_RIGHT:
2498         case LYX_ALIGN_CENTER:
2499                 if (runparams.moving_arg)
2500                         os << "\\protect";
2501                 break;
2502         }
2503
2504         bool output = false;
2505         int col = 0;
2506         string const end_tag = "\\par\\end";
2507         InsetCode code = ownerCode();
2508         bool const lastpar = runparams.isLastPar;
2509         // RTL in classic (PDF)LaTeX (without the Bidi package)
2510         // Luabibdi (used by LuaTeX) behaves like classic
2511         bool const rtl_classic = owner_->getParLanguage(bparams)->rightToLeft()
2512                 && !runparams.useBidiPackage();
2513
2514         switch (curAlign) {
2515         case LYX_ALIGN_NONE:
2516         case LYX_ALIGN_BLOCK:
2517         case LYX_ALIGN_LAYOUT:
2518         case LYX_ALIGN_SPECIAL:
2519         case LYX_ALIGN_DECIMAL:
2520                 break;
2521         case LYX_ALIGN_LEFT: {
2522                 if (rtl_classic)
2523                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2524                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2525                 else
2526                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2527                 break;
2528         } case LYX_ALIGN_RIGHT: {
2529                 if (rtl_classic)
2530                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2531                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2532                 else
2533                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2534                 break;
2535         } case LYX_ALIGN_CENTER: {
2536                 corrected_env(os, end_tag, "center", code, lastpar, col);
2537                 break;
2538         }
2539         }
2540
2541         return output || lastpar;
2542 }
2543
2544
2545 // This one spits out the text of the paragraph
2546 void Paragraph::latex(BufferParams const & bparams,
2547         Font const & outerfont,
2548         otexstream & os,
2549         OutputParams const & runparams,
2550         int start_pos, int end_pos, bool force) const
2551 {
2552         LYXERR(Debug::OUTFILE, "Paragraph::latex...     " << this);
2553
2554         // FIXME This check should not be needed. Perhaps issue an
2555         // error if it triggers.
2556         Layout const & style = inInset().forcePlainLayout() ?
2557                 bparams.documentClass().plainLayout() : *d->layout_;
2558
2559         if (!force && style.inpreamble)
2560                 return;
2561
2562         bool const allowcust = allowParagraphCustomization();
2563
2564         // Current base font for all inherited font changes, without any
2565         // change caused by an individual character, except for the language:
2566         // It is set to the language of the first character.
2567         // As long as we are in the label, this font is the base font of the
2568         // label. Before the first body character it is set to the base font
2569         // of the body.
2570         Font basefont;
2571
2572         // If there is an open font-encoding changing command (script wrapper),
2573         // alien_script is set to its name
2574         string alien_script;
2575         string script;
2576
2577         // Maybe we have to create a optional argument.
2578         pos_type body_pos = beginOfBody();
2579         unsigned int column = 0;
2580
2581         // If we are inside an non inheritFont() inset,
2582         // the outerfont is the buffer's main font
2583         Font const real_outerfont =
2584                 inInset().inheritFont() ? outerfont : Font(bparams.getFont());
2585
2586         if (body_pos > 0) {
2587                 // the optional argument is kept in curly brackets in
2588                 // case it contains a ']'
2589                 // This is not strictly needed, but if this is changed it
2590                 // would be a file format change, and tex2lyx would need
2591                 // to be adjusted, since it unconditionally removes the
2592                 // braces when it parses \item.
2593                 os << "[{";
2594                 column += 2;
2595                 basefont = getLabelFont(bparams, real_outerfont);
2596         } else {
2597                 basefont = getLayoutFont(bparams, real_outerfont);
2598         }
2599
2600         // Which font is currently active?
2601         Font running_font(basefont);
2602         // Do we have an open font change?
2603         bool open_font = false;
2604
2605         Change runningChange =
2606             runparams.inDeletedInset && !inInset().canTrackChanges()
2607             ? runparams.changeOfDeletedInset : Change(Change::UNCHANGED);
2608
2609         Encoding const * const prev_encoding = runparams.encoding;
2610
2611         os.texrow().start(id(), 0);
2612
2613         // if the paragraph is empty, the loop will not be entered at all
2614         if (empty()) {
2615                 // For InTitle commands, we have already opened a group
2616                 // in output_latex::TeXOnePar.
2617                 if (style.isCommand() && !style.intitle) {
2618                         os << '{';
2619                         ++column;
2620                 }
2621                 if (!style.leftdelim().empty()) {
2622                         os << style.leftdelim();
2623                         column += style.leftdelim().size();
2624                 }
2625                 if (allowcust)
2626                         column += d->startTeXParParams(bparams, os, runparams);
2627         }
2628
2629         // Whether a \par can be issued for insets typeset inline with text.
2630         // Yes if greater than 0. This has to be static.
2631         THREAD_LOCAL_STATIC int parInline = 0;
2632
2633         for (pos_type i = 0; i < size(); ++i) {
2634                 // First char in paragraph or after label?
2635                 if (i == body_pos) {
2636                         if (body_pos > 0) {
2637                                 if (open_font) {
2638                                         bool needPar = false;
2639                                         column += running_font.latexWriteEndChanges(
2640                                                 os, bparams, runparams,
2641                                                 basefont, basefont, needPar);
2642                                         open_font = false;
2643                                 }
2644                                 basefont = getLayoutFont(bparams, real_outerfont);
2645                                 running_font = basefont;
2646
2647                                 column += Changes::latexMarkChange(os, bparams,
2648                                                 runningChange, Change(Change::UNCHANGED),
2649                                                 runparams);
2650                                 runningChange = Change(Change::UNCHANGED);
2651
2652                                 os << ((isEnvSeparator(i) && !runparams.find_effective()) ? "}]~" : "}] ");
2653                                 column +=3;
2654                         }
2655                         // For InTitle commands, we have already opened a group
2656                         // in output_latex::TeXOnePar.
2657                         if (style.isCommand() && !style.intitle) {
2658                                 os << '{';
2659                                 ++column;
2660                         }
2661
2662                         if (!style.leftdelim().empty()) {
2663                                 os << style.leftdelim();
2664                                 column += style.leftdelim().size();
2665                         }
2666
2667                         if (allowcust)
2668                                 column += d->startTeXParParams(bparams, os,
2669                                                             runparams);
2670                 }
2671
2672                 runparams.wasDisplayMath = runparams.inDisplayMath;
2673                 runparams.inDisplayMath = false;
2674                 bool deleted_display_math = false;
2675                 Change const & change = runparams.inDeletedInset
2676                         ? runparams.changeOfDeletedInset : lookupChange(i);
2677
2678                 char_type const c = d->text_[i];
2679
2680                 // Check whether a display math inset follows
2681                 bool output_changes;
2682                 if (!runparams.find_effective())
2683                         output_changes = bparams.output_changes;
2684                 else
2685                         output_changes = runparams.find_with_deleted();
2686                 if (c == META_INSET
2687                     && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2688                         if (isDeleted(i))
2689                                 runparams.ctObject = getInset(i)->getCtObject(runparams);
2690
2691                         InsetMath const * im = getInset(i)->asInsetMath();
2692                         if (im && im->asHullInset()
2693                             && im->asHullInset()->outerDisplay()) {
2694                                 runparams.inDisplayMath = true;
2695                                 // runparams.inDeletedInset will be set by
2696                                 // latexInset later, but we need this info
2697                                 // before it is called. On the other hand, we
2698                                 // cannot set it here because it is a counter.
2699                                 deleted_display_math = isDeleted(i);
2700                         }
2701                         if (output_changes && deleted_display_math
2702                             && runningChange == change
2703                             && change.type == Change::DELETED
2704                             && !os.afterParbreak()) {
2705                                 // A display math in the same paragraph follows.
2706                                 // We have to close and then reopen \lyxdeleted,
2707                                 // otherwise the math will be shifted up.
2708                                 OutputParams rp = runparams;
2709                                 if (open_font) {
2710                                         bool needPar = false;
2711                                         column += running_font.latexWriteEndChanges(
2712                                                 os, bparams, rp, basefont,
2713                                                 basefont, needPar);
2714                                         open_font = false;
2715                                 }
2716                                 basefont = (body_pos > i) ? getLabelFont(bparams, real_outerfont)
2717                                                           : getLayoutFont(bparams, real_outerfont);
2718                                 running_font = basefont;
2719                                 column += Changes::latexMarkChange(os, bparams,
2720                                         Change(Change::INSERTED), change, rp);
2721                         }
2722                 }
2723
2724                 if (output_changes && runningChange != change) {
2725                         if (!alien_script.empty()) {
2726                                 column += 1;
2727                                 os << "}";
2728                                 alien_script.clear();
2729                         }
2730                         if (open_font) {
2731                                 bool needPar = false;
2732                                 column += running_font.latexWriteEndChanges(
2733                                                 os, bparams, runparams,
2734                                                 basefont, basefont, needPar);
2735                                 open_font = false;
2736                         }
2737                         basefont = (body_pos > i) ? getLabelFont(bparams, real_outerfont)
2738                                                   : getLayoutFont(bparams, real_outerfont);
2739                         running_font = basefont;
2740                         column += Changes::latexMarkChange(os, bparams, runningChange,
2741                                                            change, runparams);
2742                         runningChange = change;
2743                 }
2744
2745                 // do not output text which is marked deleted
2746                 // if change tracking output is disabled
2747                 if (!output_changes && change.deleted()) {
2748                         continue;
2749                 }
2750
2751                 ++column;
2752
2753                 // Fully instantiated font
2754                 Font current_font = getFont(bparams, i, outerfont);
2755                 // Previous font
2756                 Font const prev_font = (i > 0) ?
2757                                         getFont(bparams, i - 1, outerfont)
2758                                       : current_font;
2759
2760                 Font const last_font = running_font;
2761                 bool const in_ct_deletion = (output_changes
2762                                              && runningChange == change
2763                                              && change.type == Change::DELETED
2764                                              && !os.afterParbreak());
2765                 // Insets where font switches are used (rather than font commands)
2766                 bool const fontswitch_inset =
2767                                 c == META_INSET
2768                                 && getInset(i)
2769                                 && getInset(i)->allowMultiPar()
2770                                 && getInset(i)->lyxCode() != ERT_CODE
2771                                 && (getInset(i)->producesOutput()
2772                                     // FIXME Something more general?
2773                                     // Comments do not "produce output" but are still
2774                                     // part of the TeX source and require font switches
2775                                     // to be closed (otherwise LaTeX fails).
2776                                     || getInset(i)->layoutName() == "Note:Comment");
2777
2778                 bool closeLanguage = false;
2779                 bool lang_switched_at_inset = false;
2780                 if (fontswitch_inset) {
2781                         // Some insets cannot be inside a font change command.
2782                         // However, even such insets *can* be placed in \L or \R
2783                         // or their equivalents (for RTL language switches),
2784                         // so we don't close the language in those cases
2785                         // (= differing isRightToLeft()).
2786                         // ArabTeX, though, doesn't seem to handle this special behavior.
2787                         closeLanguage = basefont.isRightToLeft() == current_font.isRightToLeft()
2788                                         || basefont.language()->lang() == "arabic_arabtex"
2789                                         || current_font.language()->lang() == "arabic_arabtex";
2790                         // We need to check prev_font as language changes directly at inset
2791                         // will only be started inside the inset.
2792                         lang_switched_at_inset = prev_font.language() != current_font.language();
2793                 }
2794
2795                 // Do we need to close the previous font?
2796                 bool langClosed = false;
2797                 if (open_font &&
2798                     ((current_font != running_font
2799                       || current_font.language() != running_font.language())
2800                      || (fontswitch_inset
2801                          && (current_font == prev_font))))
2802                 {
2803                         // ensure there is no open script-wrapper
2804                         if (!alien_script.empty()) {
2805                                 column += 1;
2806                                 os << "}";
2807                                 alien_script.clear();
2808                         }
2809                         if (in_ct_deletion) {
2810                                 // We have to close and then reopen \lyxdeleted,
2811                                 // as strikeout needs to be on lowest level.
2812                                 os << '}';
2813                                 column += 1;
2814                         }
2815                         if (closeLanguage) {
2816                                 // Force language closing
2817                                 current_font.setLanguage(basefont.language());
2818                                 langClosed = true;
2819                         }
2820                         Font const nextfont = (i == body_pos-1) ? basefont : current_font;
2821                         bool needPar = false;
2822                         column += running_font.latexWriteEndChanges(
2823                                     os, bparams, runparams, basefont,
2824                                     nextfont, needPar);
2825                         if (in_ct_deletion) {
2826                                 // We have to close and then reopen \lyxdeleted,
2827                                 // as strikeout needs to be on lowest level.
2828                                 OutputParams rp = runparams;
2829                                 column += Changes::latexMarkChange(os, bparams,
2830                                         Change(Change::UNCHANGED), Change(Change::DELETED), rp);
2831                         }
2832                         // Has the language been closed in the latexWriteEndChanges() call above?
2833                         langClosed |= running_font.language() != basefont.language()
2834                                         && running_font.language() != nextfont.language()
2835                                         && (running_font.language()->encoding()->package() != Encoding::CJK);
2836                         running_font = basefont;
2837                         open_font &= !langClosed;
2838                 }
2839
2840                 // if necessary, close language environment before opening CJK
2841                 string const running_lang = running_font.language()->babel();
2842                 string const lang_end_command = lyxrc.language_command_end;
2843                 if (!lang_end_command.empty() && !bparams.useNonTeXFonts
2844                         && !running_lang.empty()
2845                         && running_lang == openLanguageName()
2846                         && current_font.language()->encoding()->package() == Encoding::CJK) {
2847                         string end_tag = subst(lang_end_command, "$$lang", running_lang);
2848                         os << from_ascii(end_tag);
2849                         column += end_tag.length();
2850                         if (!languageStackEmpty())
2851                                 popLanguageName();
2852                 }
2853
2854                 // Switch file encoding if necessary (and allowed)
2855                 if ((!fontswitch_inset || closeLanguage)
2856                     && !runparams.pass_thru && !style.pass_thru &&
2857                     runparams.encoding->package() != Encoding::none &&
2858                     current_font.language()->encoding()->package() != Encoding::none) {
2859                         pair<bool, int> const enc_switch =
2860                                 switchEncoding(os.os(), bparams, runparams,
2861                                         *(current_font.language()->encoding()));
2862                         if (enc_switch.first) {
2863                                 column += enc_switch.second;
2864                                 runparams.encoding = current_font.language()->encoding();
2865                         }
2866                 }
2867
2868                 // A display math inset inside an ulem command will be output
2869                 // as a box of width \linewidth, so we have to either disable
2870                 // indentation if the inset starts a paragraph, or start a new
2871                 // line to accommodate such box. This has to be done before
2872                 // writing any font changing commands.
2873                 if (runparams.inDisplayMath && !deleted_display_math
2874                     && runparams.inulemcmd) {
2875                         if (os.afterParbreak())
2876                                 os << "\\noindent";
2877                         else
2878                                 os << "\\\\\n";
2879                 }
2880
2881                 // Do we need to change font?
2882                 if ((current_font != running_font ||
2883                      current_font.language() != running_font.language())
2884                     && i != body_pos - 1)
2885                 {
2886                         if (!fontswitch_inset) {
2887                                 if (in_ct_deletion) {
2888                                         // We have to close and then reopen \lyxdeleted,
2889                                         // as strikeout needs to be on lowest level.
2890                                         OutputParams rp = runparams;
2891                                         bool needPar = false;
2892                                         column += running_font.latexWriteEndChanges(
2893                                                 os, bparams, rp, basefont,
2894                                                 basefont, needPar);
2895                                         os << '}';
2896                                         column += 1;
2897                                 }
2898                                 otexstringstream ots;
2899                                 InsetText const * textinset = inInset().asInsetText();
2900                                 bool const cprotect = textinset
2901                                         ? textinset->hasCProtectContent(runparams.moving_arg)
2902                                           && !textinset->text().isMainText()
2903                                           && inInset().lyxCode() != BRANCH_CODE
2904                                         : false;
2905                                 column += current_font.latexWriteStartChanges(ots, bparams,
2906                                                                               runparams, basefont, last_font, false,
2907                                                                               cprotect);
2908                                 // Check again for display math in ulem commands as a
2909                                 // font change may also occur just before a math inset.
2910                                 if (runparams.inDisplayMath && !deleted_display_math
2911                                     && runparams.inulemcmd) {
2912                                         if (os.afterParbreak())
2913                                                 os << "\\noindent";
2914                                         else
2915                                                 os << "\\\\\n";
2916                                 }
2917                                 running_font = current_font;
2918                                 open_font = true;
2919                                 docstring fontchange = ots.str();
2920                                 os << fontchange;
2921                                 // check whether the fontchange ends with a \\textcolor
2922                                 // modifier and the text starts with a space. If so we
2923                                 // need to add } in order to prevent \\textcolor from gobbling
2924                                 // the space (bug 4473).
2925                                 docstring const last_modifier = rsplit(fontchange, '\\');
2926                                 if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2927                                         os << from_ascii("{}");
2928                                 else if (ots.terminateCommand())
2929                                         os << termcmd;
2930                                 if (in_ct_deletion) {
2931                                         // We have to close and then reopen \lyxdeleted,
2932                                         // as strikeout needs to be on lowest level.
2933                                         OutputParams rp = runparams;
2934                                         column += Changes::latexMarkChange(os, bparams,
2935                                                 Change(Change::UNCHANGED), change, rp);
2936                                 }
2937                         } else {// if fontswitch_inset
2938                                 if (current_font != running_font || !langClosed)
2939                                         // font is still open in fontswitch_insets if we have
2940                                         // a non-lang font difference or if the language
2941                                         // is the only difference but has not been forcedly
2942                                         // closed meanwhile
2943                                         open_font = true;
2944                                 running_font = current_font;
2945                         }
2946                 }
2947
2948                 // FIXME: think about end_pos implementation...
2949                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2950                         // FIXME: integrate this case in latexSpecialChar
2951                         // Do not print the separation of the optional argument
2952                         // if style.pass_thru is false. This works because
2953                         // latexSpecialChar ignores spaces if
2954                         // style.pass_thru is false.
2955                         if (i != body_pos - 1) {
2956                                 if (d->simpleTeXBlanks(bparams, runparams, os,
2957                                                 i, column, current_font, style)) {
2958                                         // A surrogate pair was output. We
2959                                         // must not call latexSpecialChar
2960                                         // in this iteration, since it would output
2961                                         // the combining character again.
2962                                         ++i;
2963                                         continue;
2964                                 }
2965                         }
2966                 }
2967
2968                 OutputParams rp = runparams;
2969                 rp.free_spacing = style.free_spacing;
2970                 rp.local_font = &current_font;
2971                 rp.intitle = style.intitle;
2972
2973                 // Two major modes:  LaTeX or plain
2974                 // Handle here those cases common to both modes
2975                 // and then split to handle the two modes separately.
2976                 if (c == META_INSET) {
2977                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2978                                 // Greyedout notes and, in general, all insets
2979                                 // with InsetLayout::isDisplay() == false,
2980                                 // are typeset inline with the text. So, we
2981                                 // can add a \par to the last paragraph of
2982                                 // such insets only if nothing else follows.
2983                                 bool incremented = false;
2984                                 Inset const * inset = getInset(i);
2985                                 InsetText const * textinset = inset
2986                                                         ? inset->asInsetText()
2987                                                         : nullptr;
2988                                 if (i + 1 == size() && textinset
2989                                     && !inset->getLayout().isDisplay()) {
2990                                         ParagraphList const & pars =
2991                                                 textinset->text().paragraphs();
2992                                         pit_type const pit = pars.size() - 1;
2993                                         Font const lastfont =
2994                                                 pit < 0 || pars[pit].empty()
2995                                                 ? pars[pit].getLayoutFont(
2996                                                                 bparams,
2997                                                                 real_outerfont)
2998                                                 : pars[pit].getFont(bparams,
2999                                                         pars[pit].size() - 1,
3000                                                         real_outerfont);
3001                                         if (lastfont.fontInfo().size() !=
3002                                             basefont.fontInfo().size()) {
3003                                                 ++parInline;
3004                                                 incremented = true;
3005                                         }
3006                                 }
3007                                 // We need to restore parts of this after insets with
3008                                 // allowMultiPar() true
3009                                 Font const save_basefont = basefont;
3010                                 d->latexInset(bparams, os, rp, running_font,
3011                                                 basefont, real_outerfont, open_font,
3012                                                 runningChange, style, i, column, fontswitch_inset,
3013                                                 closeLanguage, (lang_switched_at_inset || langClosed));
3014                                 if (fontswitch_inset) {
3015                                         if (open_font) {
3016                                                 bool needPar = false;
3017                                                 column += running_font.latexWriteEndChanges(
3018                                                         os, bparams, runparams,
3019                                                         basefont, basefont, needPar);
3020                                                 open_font = false;
3021                                         }
3022                                         basefont.fontInfo().setSize(save_basefont.fontInfo().size());
3023                                         basefont.fontInfo().setFamily(save_basefont.fontInfo().family());
3024                                         basefont.fontInfo().setSeries(save_basefont.fontInfo().series());
3025                                 }
3026                                 if (incremented)
3027                                         --parInline;
3028
3029                                 if (runparams.ctObject == CtObject::DisplayObject
3030                                     || runparams.ctObject == CtObject::UDisplayObject) {
3031                                         // Close \lyx*deleted and force its
3032                                         // reopening (if needed)
3033                                         os << '}';
3034                                         column++;
3035                                         runningChange = Change(Change::UNCHANGED);
3036                                         runparams.ctObject = CtObject::Normal;
3037                                 }
3038                         }
3039                 } else if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
3040                         if (!bparams.useNonTeXFonts)
3041                           script = Encodings::isKnownScriptChar(c);
3042                         if (script != alien_script) {
3043                                 if (!alien_script.empty()) {
3044                                         os << "}";
3045                                         alien_script.clear();
3046                                 }
3047                                 string fontenc = running_font.language()->fontenc(bparams);
3048                                 if (!script.empty()
3049                                         && !Encodings::fontencSupportsScript(fontenc, script)) {
3050                                         column += script.length() + 2;
3051                                         os << "\\" << script << "{";
3052                                         alien_script = script;
3053                                 }
3054                         }
3055                         try {
3056                                 d->latexSpecialChar(os, bparams, rp, running_font,
3057                                                                         alien_script, style, i, end_pos, column);
3058                         } catch (EncodingException & e) {
3059                                 if (runparams.dryrun) {
3060                                         os << "<" << _("LyX Warning: ")
3061                                            << _("uncodable character") << " '";
3062                                         os.put(c);
3063                                         os << "'>";
3064                                 } else {
3065                                         // add location information and throw again.
3066                                         e.par_id = id();
3067                                         e.pos = i;
3068                                         throw;
3069                                 }
3070                         }
3071                 }
3072
3073                 // Set the encoding to that returned from latexSpecialChar (see
3074                 // comment for encoding member in OutputParams.h)
3075                 runparams.encoding = rp.encoding;
3076
3077                 // Also carry on the info on a closed ulem command for insets
3078                 // such as Note that do not produce any output, so that no
3079                 // command is ever executed but its opening was recorded.
3080                 runparams.inulemcmd = rp.inulemcmd;
3081
3082                 // These need to be passed upstream as well
3083                 runparams.need_maketitle = rp.need_maketitle;
3084                 runparams.have_maketitle = rp.have_maketitle;
3085
3086                 // And finally, pass the post_macros upstream
3087                 runparams.post_macro = rp.post_macro;
3088         }
3089
3090         // Close wrapper for alien script
3091         if (!alien_script.empty()) {
3092                 os << "}";
3093                 alien_script.clear();
3094         }
3095
3096         Font const font = empty()
3097                 ? getLayoutFont(bparams, real_outerfont)
3098                 : getFont(bparams, size() - 1, real_outerfont);
3099
3100         InsetText const * textinset = inInset().asInsetText();
3101
3102         bool const maintext = textinset
3103                 ? textinset->text().isMainText()
3104                 : false;
3105
3106         size_t const numpars = textinset
3107                 ? textinset->text().paragraphs().size()
3108                 : 0;
3109
3110         bool needPar = false;
3111
3112         if (style.resfont.size() != font.fontInfo().size()
3113             && (!runparams.isLastPar || maintext
3114                 || (numpars > 1 && d->ownerCode() != CELL_CODE
3115                     && (inInset().getLayout().isDisplay()
3116                         || parInline)))
3117             && !style.isCommand()) {
3118                 needPar = true;
3119         }
3120
3121         // If we have an open font definition, we have to close it
3122         if (open_font) {
3123                 // Make sure that \\par is done with the font of the last
3124                 // character if this has another size as the default.
3125                 // This is necessary because LaTeX (and LyX on the screen)
3126                 // calculates the space between the baselines according
3127                 // to this font. (Matthias)
3128                 //
3129                 // We must not change the font for the last paragraph
3130                 // of non-multipar insets, tabular cells or commands,
3131                 // since this produces unwanted whitespace.
3132 #ifdef FIXED_LANGUAGE_END_DETECTION
3133                 if (next_) {
3134                         running_font.latexWriteEndChanges(os, bparams,
3135                                         runparams, basefont,
3136                                         next_->getFont(bparams, 0, outerfont),
3137                                                        needPar);
3138                 } else {
3139                         running_font.latexWriteEndChanges(os, bparams,
3140                                         runparams, basefont, basefont, needPar);
3141                 }
3142 #else
3143 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
3144 //FIXME: there as we start another \selectlanguage with the next paragraph if
3145 //FIXME: we are in need of this. This should be fixed sometime (Jug)
3146                 running_font.latexWriteEndChanges(os, bparams, runparams,
3147                                 basefont, basefont, needPar);
3148 #endif
3149         }
3150         if (needPar) {
3151                 // The \par could not be inserted at the same nesting
3152                 // level of the font size change, so do it now.
3153                 os << "{\\" << font.latexSize() << "\\par}";
3154         }
3155
3156         if (!runparams.inDeletedInset || inInset().canTrackChanges())
3157                 column += Changes::latexMarkChange(os, bparams, runningChange,
3158                                         Change(Change::UNCHANGED), runparams);
3159
3160         // Needed if there is an optional argument but no contents.
3161         if (body_pos > 0 && body_pos == size()) {
3162                 os << "}]~";
3163         }
3164
3165         if (!style.rightdelim().empty()) {
3166                 os << style.rightdelim();
3167                 column += style.rightdelim().size();
3168         }
3169
3170         if (allowcust && d->endTeXParParams(bparams, os, runparams)
3171             && runparams.encoding != prev_encoding) {
3172                 runparams.encoding = prev_encoding;
3173                 os << setEncoding(prev_encoding->iconvName());
3174         }
3175
3176         LYXERR(Debug::OUTFILE, "Paragraph::latex... done " << this);
3177 }
3178
3179
3180 bool Paragraph::emptyTag() const
3181 {
3182         for (pos_type i = 0; i < size(); ++i) {
3183                 if (Inset const * inset = getInset(i)) {
3184                         InsetCode lyx_code = inset->lyxCode();
3185                         // FIXME testing like that is wrong. What is
3186                         // the intent?
3187                         if (lyx_code != TOC_CODE &&
3188                             lyx_code != INCLUDE_CODE &&
3189                             lyx_code != GRAPHICS_CODE &&
3190                             lyx_code != ERT_CODE &&
3191                             lyx_code != LISTINGS_CODE &&
3192                             lyx_code != FLOAT_CODE &&
3193                             lyx_code != TABULAR_CODE) {
3194                                 return false;
3195                         }
3196                 } else {
3197                         char_type c = d->text_[i];
3198                         if (c != ' ' && c != '\t')
3199                                 return false;
3200                 }
3201         }
3202         return true;
3203 }
3204
3205
3206 string Paragraph::getID(Buffer const &, OutputParams const &)
3207         const
3208 {
3209         for (pos_type i = 0; i < size(); ++i) {
3210                 if (Inset const * inset = getInset(i)) {
3211                         InsetCode lyx_code = inset->lyxCode();
3212                         if (lyx_code == LABEL_CODE) {
3213                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
3214                                 docstring const & id = il->getParam("name");
3215                                 return "id='" + to_utf8(xml::cleanID(id)) + "'";
3216                         }
3217                 }
3218         }
3219         return string();
3220 }
3221
3222
3223 pos_type Paragraph::firstWordDocBook(XMLStream & xs, OutputParams const & runparams) const
3224 {
3225         pos_type i;
3226         for (i = 0; i < size(); ++i) {
3227                 if (Inset const * inset = getInset(i)) {
3228                         inset->docbook(xs, runparams);
3229                 } else {
3230                         char_type c = d->text_[i];
3231                         if (c == ' ')
3232                                 break;
3233                         xs << c;
3234                 }
3235         }
3236         return i;
3237 }
3238
3239
3240 pos_type Paragraph::firstWordLyXHTML(XMLStream & xs, OutputParams const & runparams)
3241         const
3242 {
3243         pos_type i;
3244         for (i = 0; i < size(); ++i) {
3245                 if (Inset const * inset = getInset(i)) {
3246                         inset->xhtml(xs, runparams);
3247                 } else {
3248                         char_type c = d->text_[i];
3249                         if (c == ' ')
3250                                 break;
3251                         xs << c;
3252                 }
3253         }
3254         return i;
3255 }
3256
3257
3258 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
3259 {
3260         Font font_old;
3261         pos_type size = text_.size();
3262         for (pos_type i = initial; i < size; ++i) {
3263                 Font font = owner_->getFont(buf.params(), i, outerfont);
3264                 if (text_[i] == META_INSET)
3265                         return false;
3266                 if (i != initial && font != font_old)
3267                         return false;
3268                 font_old = font;
3269         }
3270
3271         return true;
3272 }
3273
3274
3275 namespace {
3276
3277 void doFontSwitchDocBook(vector<xml::FontTag> & tagsToOpen,
3278                   vector<xml::EndFontTag> & tagsToClose,
3279                   bool & flag, FontState curstate, xml::FontTypes type)
3280 {
3281         if (curstate == FONT_ON) {
3282                 tagsToOpen.push_back(docbookStartFontTag(type));
3283                 flag = true;
3284         } else if (flag) {
3285                 tagsToClose.push_back(docbookEndFontTag(type));
3286                 flag = false;
3287         }
3288 }
3289
3290 class OptionalFontType {
3291 public:
3292         xml::FontTypes ft;
3293         bool has_value;
3294
3295         OptionalFontType(): ft(xml::FT_EMPH), has_value(false) {} // A possible value at random for ft.
3296         OptionalFontType(xml::FontTypes ft): ft(ft), has_value(true) {}
3297 };
3298
3299 OptionalFontType fontShapeToXml(FontShape fs)
3300 {
3301         switch (fs) {
3302         case ITALIC_SHAPE:
3303                 return {xml::FT_ITALIC};
3304         case SLANTED_SHAPE:
3305                 return {xml::FT_SLANTED};
3306         case SMALLCAPS_SHAPE:
3307                 return {xml::FT_SMALLCAPS};
3308         case UP_SHAPE:
3309         case INHERIT_SHAPE:
3310                 return {};
3311         default:
3312                 // the other tags are for internal use
3313                 LATTEST(false);
3314                 return {};
3315         }
3316 }
3317
3318 OptionalFontType fontFamilyToXml(FontFamily fm)
3319 {
3320         switch (fm) {
3321         case ROMAN_FAMILY:
3322                 return {xml::FT_ROMAN};
3323         case SANS_FAMILY:
3324                 return {xml::FT_SANS};
3325         case TYPEWRITER_FAMILY:
3326                 return {xml::FT_TYPE};
3327         case INHERIT_FAMILY:
3328                 return {};
3329         default:
3330                 // the other tags are for internal use
3331                 LATTEST(false);
3332                 return {};
3333         }
3334 }
3335
3336 OptionalFontType fontSizeToXml(FontSize fs)
3337 {
3338         switch (fs) {
3339         case TINY_SIZE:
3340                 return {xml::FT_SIZE_TINY};
3341         case SCRIPT_SIZE:
3342                 return {xml::FT_SIZE_SCRIPT};
3343         case FOOTNOTE_SIZE:
3344                 return {xml::FT_SIZE_FOOTNOTE};
3345         case SMALL_SIZE:
3346                 return {xml::FT_SIZE_SMALL};
3347         case LARGE_SIZE:
3348                 return {xml::FT_SIZE_LARGE};
3349         case LARGER_SIZE:
3350                 return {xml::FT_SIZE_LARGER};
3351         case LARGEST_SIZE:
3352                 return {xml::FT_SIZE_LARGEST};
3353         case HUGE_SIZE:
3354                 return {xml::FT_SIZE_HUGE};
3355         case HUGER_SIZE:
3356                 return {xml::FT_SIZE_HUGER};
3357         case INCREASE_SIZE:
3358                 return {xml::FT_SIZE_INCREASE};
3359         case DECREASE_SIZE:
3360                 return {xml::FT_SIZE_DECREASE};
3361         case INHERIT_SIZE:
3362         case NORMAL_SIZE:
3363                 return {};
3364         default:
3365                 // the other tags are for internal use
3366                 LATTEST(false);
3367                 return {};
3368         }
3369 }
3370
3371 struct DocBookFontState
3372 {
3373         FontShape  curr_fs   = INHERIT_SHAPE;
3374         FontFamily curr_fam  = INHERIT_FAMILY;
3375         FontSize   curr_size = INHERIT_SIZE;
3376
3377         // track whether we have opened these tags
3378         bool emph_flag = false;
3379         bool bold_flag = false;
3380         bool noun_flag = false;
3381         bool ubar_flag = false;
3382         bool dbar_flag = false;
3383         bool sout_flag = false;
3384         bool xout_flag = false;
3385         bool wave_flag = false;
3386         // shape tags
3387         bool shap_flag = false;
3388         // family tags
3389         bool faml_flag = false;
3390         // size tags
3391         bool size_flag = false;
3392 };
3393
3394 std::tuple<vector<xml::FontTag>, vector<xml::EndFontTag>> computeDocBookFontSwitch(FontInfo const & font_old,
3395                                                                                            Font const & font,
3396                                                                                            std::string const & default_family,
3397                                                                                            DocBookFontState & fs)
3398 {
3399         vector<xml::FontTag> tagsToOpen;
3400         vector<xml::EndFontTag> tagsToClose;
3401
3402         // emphasis
3403         FontState curstate = font.fontInfo().emph();
3404         if (font_old.emph() != curstate)
3405                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.emph_flag, curstate, xml::FT_EMPH);
3406
3407         // noun
3408         curstate = font.fontInfo().noun();
3409         if (font_old.noun() != curstate)
3410                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.noun_flag, curstate, xml::FT_NOUN);
3411
3412         // underbar
3413         curstate = font.fontInfo().underbar();
3414         if (font_old.underbar() != curstate)
3415                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.ubar_flag, curstate, xml::FT_UBAR);
3416
3417         // strikeout
3418         curstate = font.fontInfo().strikeout();
3419         if (font_old.strikeout() != curstate)
3420                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.sout_flag, curstate, xml::FT_SOUT);
3421
3422         // xout
3423         curstate = font.fontInfo().xout();
3424         if (font_old.xout() != curstate)
3425                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.xout_flag, curstate, xml::FT_XOUT);
3426
3427         // double underbar
3428         curstate = font.fontInfo().uuline();
3429         if (font_old.uuline() != curstate)
3430                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.dbar_flag, curstate, xml::FT_DBAR);
3431
3432         // wavy line
3433         curstate = font.fontInfo().uwave();
3434         if (font_old.uwave() != curstate)
3435                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.wave_flag, curstate, xml::FT_WAVE);
3436
3437         // bold
3438         // a little hackish, but allows us to reuse what we have.
3439         curstate = (font.fontInfo().series() == BOLD_SERIES ? FONT_ON : FONT_OFF);
3440         if (font_old.series() != font.fontInfo().series())
3441                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.bold_flag, curstate, xml::FT_BOLD);
3442
3443         // Font shape
3444         fs.curr_fs = font.fontInfo().shape();
3445         FontShape old_fs = font_old.shape();
3446         if (old_fs != fs.curr_fs) {
3447                 if (fs.shap_flag) {
3448                         OptionalFontType tag = fontShapeToXml(old_fs);
3449                         if (tag.has_value)
3450                                 tagsToClose.push_back(docbookEndFontTag(tag.ft));
3451                         fs.shap_flag = false;
3452                 }
3453
3454                 OptionalFontType tag = fontShapeToXml(fs.curr_fs);
3455                 if (tag.has_value)
3456                         tagsToOpen.push_back(docbookStartFontTag(tag.ft));
3457         }
3458
3459         // Font family
3460         fs.curr_fam = font.fontInfo().family();
3461         FontFamily old_fam = font_old.family();
3462         if (old_fam != fs.curr_fam) {
3463                 if (fs.faml_flag) {
3464                         OptionalFontType tag = fontFamilyToXml(old_fam);
3465                         if (tag.has_value)
3466                                 tagsToClose.push_back(docbookEndFontTag(tag.ft));
3467                         fs.faml_flag = false;
3468                 }
3469                 switch (fs.curr_fam) {
3470                         case ROMAN_FAMILY:
3471                                 // we will treat a "default" font family as roman, since we have
3472                                 // no other idea what to do.
3473                                 if (default_family != "rmdefault" && default_family != "default") {
3474                                         tagsToOpen.push_back(docbookStartFontTag(xml::FT_ROMAN));
3475                                         fs.faml_flag = true;
3476                                 }
3477                                 break;
3478                         case SANS_FAMILY:
3479                                 if (default_family != "sfdefault") {
3480                                         tagsToOpen.push_back(docbookStartFontTag(xml::FT_SANS));
3481                                         fs.faml_flag = true;
3482                                 }
3483                                 break;
3484                         case TYPEWRITER_FAMILY:
3485                                 if (default_family != "ttdefault") {
3486                                         tagsToOpen.push_back(docbookStartFontTag(xml::FT_TYPE));
3487                                         fs.faml_flag = true;
3488                                 }
3489                                 break;
3490                         case INHERIT_FAMILY:
3491                                 break;
3492                         default:
3493                                 // the other tags are for internal use
3494                                 LATTEST(false);
3495                                 break;
3496                 }
3497         }
3498
3499         // Font size
3500         fs.curr_size = font.fontInfo().size();
3501         FontSize old_size = font_old.size();
3502         if (old_size != fs.curr_size) {
3503                 if (fs.size_flag) {
3504                         OptionalFontType tag = fontSizeToXml(old_size);
3505                         if (tag.has_value)
3506                                 tagsToClose.push_back(docbookEndFontTag(tag.ft));
3507                         fs.size_flag = false;
3508                 }
3509
3510                 OptionalFontType tag = fontSizeToXml(fs.curr_size);
3511                 if (tag.has_value) {
3512                         tagsToOpen.push_back(docbookStartFontTag(tag.ft));
3513                         fs.size_flag = true;
3514                 }
3515         }
3516
3517         return std::tuple<vector<xml::FontTag>, vector<xml::EndFontTag>>(tagsToOpen, tagsToClose);
3518 }
3519
3520 } // anonymous namespace
3521
3522
3523 std::tuple<std::vector<docstring>, std::vector<docstring>, std::vector<docstring>>
3524     Paragraph::simpleDocBookOnePar(Buffer const & buf,
3525                                                       OutputParams const & runparams,
3526                                                       Font const & outerfont,
3527                                                       pos_type initial,
3528                                                       bool is_last_par,
3529                                                       bool ignore_fonts) const
3530 {
3531         // Return values: segregation of the content of this paragraph.
3532         std::vector<docstring> prependedParagraphs; // Anything that must be output before the main tag of this paragraph.
3533         std::vector<docstring> generatedParagraphs; // The main content of the paragraph.
3534         std::vector<docstring> appendedParagraphs;  // Anything that must be output after the main tag of this paragraph.
3535
3536         // Internal string stream to store the output before being added to one of the previous lists.
3537         odocstringstream os;
3538
3539         // If there is an argument that must be output before the main tag, do it before handling the rest of the paragraph.
3540         // Also tag all arguments that shouldn't go in the main content right now, so that they are never generated at the
3541         // wrong place.
3542         OutputParams rp = runparams;
3543     for (pos_type i = initial; i < size(); ++i) {
3544         if (getInset(i) && getInset(i)->lyxCode() == ARG_CODE) {
3545             const InsetArgument * arg = getInset(i)->asInsetArgument();
3546             if (arg->docbookargumentbeforemaintag()) {
3547                 auto xs_local = XMLStream(os);
3548                 arg->docbook(xs_local, rp);
3549
3550                 prependedParagraphs.push_back(os.str());
3551                 os.str(from_ascii(""));
3552
3553                 rp.docbook_prepended_arguments.insert(arg);
3554             } else if (arg->docbookargumentaftermaintag()) {
3555                 rp.docbook_appended_arguments.insert(arg);
3556             }
3557         }
3558     }
3559
3560     // State variables for the main loop.
3561     auto xs = new XMLStream(os); // XMLStream has no copy constructor: to create a new object, the only solution
3562     // is to hold a pointer to the XMLStream (xs = XMLStream(os) is not allowed once the first object is built).
3563     std::vector<docstring> delayedChars; // When a font tag ends with a space, output it after the closing font tag.
3564     // This requires to store delayed characters at some point.
3565
3566         // Track whether we have opened font tags
3567     DocBookFontState fs;
3568     DocBookFontState old_fs = fs;
3569
3570     Layout const & style = *d->layout_;
3571
3572         // Conversion of the font opening/closing into DocBook tags.
3573     vector<xml::FontTag> tagsToOpen;
3574     vector<xml::EndFontTag> tagsToClose;
3575
3576         // Parsing main loop.
3577         for (pos_type i = initial; i < size(); ++i) {
3578             bool ignore_fonts_i = ignore_fonts
3579                               || style.docbooknofontinside()
3580                               || (getInset(i) && getInset(i)->getLayout().docbooknofontinside());
3581
3582                 // Don't show deleted material in the output.
3583                 if (isDeleted(i))
3584                         continue;
3585
3586                 // If this is an InsetNewline, generate a new paragraph. Also reset the fonts, so that tags are closed in
3587                 // this paragraph.
3588                 if (getInset(i) && getInset(i)->lyxCode() == NEWLINE_CODE) {
3589                         if (!ignore_fonts_i)
3590                                 xs->closeFontTags();
3591
3592                         // Output one paragraph (i.e. one string entry in generatedParagraphs).
3593                         generatedParagraphs.push_back(os.str());
3594
3595                         // Create a new XMLStream for the new paragraph, completely independent of the previous one. This implies
3596                         // that the string stream must be reset.
3597                         os.str(from_ascii(""));
3598                         delete xs;
3599                         xs = new XMLStream(os);
3600
3601                         // Restore the fonts for the new paragraph, so that the right tags are opened for the new entry.
3602                         if (!ignore_fonts_i) {
3603                                 fs = old_fs;
3604                         }
3605                 }
3606
3607                 // Determine which tags should be opened or closed regarding fonts.
3608                 FontInfo const font_old = (i == 0 ?
3609                                 (style.labeltype == LABEL_MANUAL ? style.labelfont : style.font) :
3610                                 getFont(buf.masterBuffer()->params(), i - 1, outerfont).fontInfo());
3611                 Font const font = getFont(buf.masterBuffer()->params(), i, outerfont);
3612         tie(tagsToOpen, tagsToClose) = computeDocBookFontSwitch(
3613                                 font_old, font, buf.masterBuffer()->params().fonts_default_family, fs);
3614
3615                 if (!ignore_fonts_i) {
3616             vector<xml::EndFontTag>::const_iterator cit = tagsToClose.begin();
3617             vector<xml::EndFontTag>::const_iterator cen = tagsToClose.end();
3618             for (; cit != cen; ++cit)
3619                 *xs << *cit;
3620         }
3621
3622         // Deal with the delayed characters *after* closing font tags.
3623         if (!delayedChars.empty()) {
3624             for (const docstring& c: delayedChars)
3625                 *xs << XMLStream::ESCAPE_NONE << c;
3626             delayedChars.clear();
3627         }
3628
3629         if (!ignore_fonts_i) {
3630                         vector<xml::FontTag>::const_iterator sit = tagsToOpen.begin();
3631                         vector<xml::FontTag>::const_iterator sen = tagsToOpen.end();
3632                         for (; sit != sen; ++sit)
3633                                 *xs << *sit;
3634
3635                         tagsToClose.clear();
3636                         tagsToOpen.clear();
3637                 }
3638
3639         // Finally, write the next character or inset.
3640                 if (Inset const * inset = getInset(i)) {
3641                     bool inset_is_argument_elsewhere = getInset(i)->asInsetArgument() &&
3642                             rp.docbook_appended_arguments.find(inset->asInsetArgument()) != rp.docbook_appended_arguments.end() &&
3643                             rp.docbook_prepended_arguments.find(inset->asInsetArgument()) != rp.docbook_prepended_arguments.end();
3644
3645                         if ((!rp.for_toc || inset->isInToc()) && !inset_is_argument_elsewhere) {
3646                             // Arguments may need to be output
3647                                 OutputParams np = rp;
3648                                 np.local_font = &font;
3649
3650                                 // TODO: special case will bite here.
3651                                 np.docbook_in_par = true;
3652                                 inset->docbook(*xs, np);
3653                         }
3654                 } else {
3655                         char_type c = getUChar(buf.masterBuffer()->params(), rp, i);
3656                         if (lyx::isSpace(c) && !ignore_fonts) { // Delay spaces *after* the font-tag closure for cleaner output.
3657                                 if (c == ' ' && (style.free_spacing || rp.free_spacing)) {
3658                                         delayedChars.push_back(from_ascii("&#160;"));
3659                                 } else {
3660                                         delayedChars.emplace_back(1, c);
3661                                 }
3662                         } else { // No need to delay the character.
3663                                 if (c == '\'' && !ignore_fonts)
3664                                         *xs << XMLStream::ESCAPE_NONE << "&#8217;";
3665                                 else
3666                                         *xs << c;
3667                         }
3668                 }
3669         }
3670
3671         // FIXME, this code is just imported from XHTML
3672         // I'm worried about what happens if a branch, say, is itself
3673         // wrapped in some font stuff. I think that will not work.
3674         if (!ignore_fonts)
3675                 xs->closeFontTags();
3676
3677         // Deal with the delayed characters *after* closing font tags.
3678         if (!delayedChars.empty()) {
3679                 for (const docstring &c: delayedChars)
3680                         *xs << XMLStream::ESCAPE_NONE << c;
3681                 delayedChars.clear();
3682         }
3683
3684         // In listings, new lines (i.e. \n characters in the output) are very important. Avoid generating one for the
3685         // last line to get a clean output.
3686         if (rp.docbook_in_listing && !is_last_par)
3687                 *xs << xml::CR();
3688
3689         // Finalise the last (and most likely only) paragraph.
3690         generatedParagraphs.push_back(os.str());
3691     os.str(from_ascii(""));
3692     delete xs;
3693
3694     // If there is an argument that must be output after the main tag, do it after handling the rest of the paragraph.
3695     for (pos_type i = initial; i < size(); ++i) {
3696         if (getInset(i) && getInset(i)->lyxCode() == ARG_CODE) {
3697             const InsetArgument * arg = getInset(i)->asInsetArgument();
3698             if (arg->docbookargumentaftermaintag()) {
3699                 // Don't use rp, as this argument would not generate anything.
3700                 auto xs_local = XMLStream(os);
3701                 arg->docbook(xs_local, runparams);
3702
3703                 appendedParagraphs.push_back(os.str());
3704                 os.str(from_ascii(""));
3705             }
3706         }
3707     }
3708
3709         return std::make_tuple(prependedParagraphs, generatedParagraphs, appendedParagraphs);
3710 }
3711
3712
3713 namespace {
3714
3715 void doFontSwitchXHTML(vector<xml::FontTag> & tagsToOpen,
3716                   vector<xml::EndFontTag> & tagsToClose,
3717                   bool & flag, FontState curstate, xml::FontTypes type)
3718 {
3719         if (curstate == FONT_ON) {
3720                 tagsToOpen.push_back(xhtmlStartFontTag(type));
3721                 flag = true;
3722         } else if (flag) {
3723                 tagsToClose.push_back(xhtmlEndFontTag(type));
3724                 flag = false;
3725         }
3726 }
3727
3728 } // anonymous namespace
3729
3730
3731 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
3732                                     XMLStream & xs,
3733                                     OutputParams const & runparams,
3734                                     Font const & outerfont,
3735                                     bool start_paragraph, bool close_paragraph,
3736                                     pos_type initial) const
3737 {
3738         docstring retval;
3739
3740         // track whether we have opened these tags
3741         bool emph_flag = false;
3742         bool bold_flag = false;
3743         bool noun_flag = false;
3744         bool ubar_flag = false;
3745         bool dbar_flag = false;
3746         bool sout_flag = false;
3747         bool xout_flag = false;
3748         bool wave_flag = false;
3749         // shape tags
3750         bool shap_flag = false;
3751         // family tags
3752         bool faml_flag = false;
3753         // size tags
3754         bool size_flag = false;
3755
3756         Layout const & style = *d->layout_;
3757
3758         if (start_paragraph)
3759                 xs.startDivision(allowEmpty());
3760
3761         FontInfo font_old =
3762                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
3763
3764         FontShape  curr_fs   = INHERIT_SHAPE;
3765         FontFamily curr_fam  = INHERIT_FAMILY;
3766         FontSize   curr_size = INHERIT_SIZE;
3767
3768         string const default_family =
3769                 buf.masterBuffer()->params().fonts_default_family;
3770
3771         vector<xml::FontTag> tagsToOpen;
3772         vector<xml::EndFontTag> tagsToClose;
3773
3774         // parsing main loop
3775         for (pos_type i = initial; i < size(); ++i) {
3776                 // let's not show deleted material in the output
3777                 if (isDeleted(i))
3778                         continue;
3779
3780                 Font const font = getFont(buf.masterBuffer()->params(), i, outerfont);
3781
3782                 // emphasis
3783                 FontState curstate = font.fontInfo().emph();
3784                 if (font_old.emph() != curstate)
3785                         doFontSwitchXHTML(tagsToOpen, tagsToClose, emph_flag, curstate, xml::FT_EMPH);
3786
3787                 // noun
3788                 curstate = font.fontInfo().noun();
3789                 if (font_old.noun() != curstate)
3790                         doFontSwitchXHTML(tagsToOpen, tagsToClose, noun_flag, curstate, xml::FT_NOUN);
3791
3792                 // underbar
3793                 curstate = font.fontInfo().underbar();
3794                 if (font_old.underbar() != curstate)
3795                         doFontSwitchXHTML(tagsToOpen, tagsToClose, ubar_flag, curstate, xml::FT_UBAR);
3796
3797                 // strikeout
3798                 curstate = font.fontInfo().strikeout();
3799                 if (font_old.strikeout() != curstate)
3800                         doFontSwitchXHTML(tagsToOpen, tagsToClose, sout_flag, curstate, xml::FT_SOUT);
3801
3802                 // xout
3803                 curstate = font.fontInfo().xout();
3804                 if (font_old.xout() != curstate)
3805                         doFontSwitchXHTML(tagsToOpen, tagsToClose, xout_flag, curstate, xml::FT_XOUT);
3806
3807                 // double underbar
3808                 curstate = font.fontInfo().uuline();
3809                 if (font_old.uuline() != curstate)
3810                         doFontSwitchXHTML(tagsToOpen, tagsToClose, dbar_flag, curstate, xml::FT_DBAR);
3811
3812                 // wavy line
3813                 curstate = font.fontInfo().uwave();
3814                 if (font_old.uwave() != curstate)
3815                         doFontSwitchXHTML(tagsToOpen, tagsToClose, wave_flag, curstate, xml::FT_WAVE);
3816
3817                 // bold
3818                 // a little hackish, but allows us to reuse what we have.
3819                 curstate = (font.fontInfo().series() == BOLD_SERIES ? FONT_ON : FONT_OFF);
3820                 if (font_old.series() != font.fontInfo().series())
3821                         doFontSwitchXHTML(tagsToOpen, tagsToClose, bold_flag, curstate, xml::FT_BOLD);
3822
3823                 // Font shape
3824                 curr_fs = font.fontInfo().shape();
3825                 FontShape old_fs = font_old.shape();
3826                 if (old_fs != curr_fs) {
3827                         if (shap_flag) {
3828                                 switch (old_fs) {
3829                                 case ITALIC_SHAPE:
3830                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_ITALIC));
3831                                         break;
3832                                 case SLANTED_SHAPE:
3833                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SLANTED));
3834                                         break;
3835                                 case SMALLCAPS_SHAPE:
3836                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SMALLCAPS));
3837                                         break;
3838                                 case UP_SHAPE:
3839                                 case INHERIT_SHAPE:
3840                                         break;
3841                                 default:
3842                                         // the other tags are for internal use
3843                                         LATTEST(false);
3844                                         break;
3845                                 }
3846                                 shap_flag = false;
3847                         }
3848                         switch (curr_fs) {
3849                         case ITALIC_SHAPE:
3850                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_ITALIC));
3851                                 shap_flag = true;
3852                                 break;
3853                         case SLANTED_SHAPE:
3854                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SLANTED));
3855                                 shap_flag = true;
3856                                 break;
3857                         case SMALLCAPS_SHAPE:
3858                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SMALLCAPS));
3859                                 shap_flag = true;
3860                                 break;
3861                         case UP_SHAPE:
3862                         case INHERIT_SHAPE:
3863                                 break;
3864                         default:
3865                                 // the other tags are for internal use
3866                                 LATTEST(false);
3867                                 break;
3868                         }
3869                 }
3870
3871                 // Font family
3872                 curr_fam = font.fontInfo().family();
3873                 FontFamily old_fam = font_old.family();
3874                 if (old_fam != curr_fam) {
3875                         if (faml_flag) {
3876                                 switch (old_fam) {
3877                                 case ROMAN_FAMILY:
3878                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_ROMAN));
3879                                         break;
3880                                 case SANS_FAMILY:
3881                                     tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SANS));
3882                                     break;
3883                                 case TYPEWRITER_FAMILY:
3884                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_TYPE));
3885                                         break;
3886                                 case INHERIT_FAMILY:
3887                                         break;
3888                                 default:
3889                                         // the other tags are for internal use
3890                                         LATTEST(false);
3891                                         break;
3892                                 }
3893                                 faml_flag = false;
3894                         }
3895                         switch (curr_fam) {
3896                         case ROMAN_FAMILY:
3897                                 // we will treat a "default" font family as roman, since we have
3898                                 // no other idea what to do.
3899                                 if (default_family != "rmdefault" && default_family != "default") {
3900                                         tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_ROMAN));
3901                                         faml_flag = true;
3902                                 }
3903                                 break;
3904                         case SANS_FAMILY:
3905                                 if (default_family != "sfdefault") {
3906                                         tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SANS));
3907                                         faml_flag = true;
3908                                 }
3909                                 break;
3910                         case TYPEWRITER_FAMILY:
3911                                 if (default_family != "ttdefault") {
3912                                         tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_TYPE));
3913                                         faml_flag = true;
3914                                 }
3915                                 break;
3916                         case INHERIT_FAMILY:
3917                                 break;
3918                         default:
3919                                 // the other tags are for internal use
3920                                 LATTEST(false);
3921                                 break;
3922                         }
3923                 }
3924
3925                 // Font size
3926                 curr_size = font.fontInfo().size();
3927                 FontSize old_size = font_old.size();
3928                 if (old_size != curr_size) {
3929                         if (size_flag) {
3930                                 switch (old_size) {
3931                                 case TINY_SIZE:
3932                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_TINY));
3933                                         break;
3934                                 case SCRIPT_SIZE:
3935                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_SCRIPT));
3936                                         break;
3937                                 case FOOTNOTE_SIZE:
3938                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_FOOTNOTE));
3939                                         break;
3940                                 case SMALL_SIZE:
3941                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_SMALL));
3942                                         break;
3943                                 case LARGE_SIZE:
3944                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_LARGE));
3945                                         break;
3946                                 case LARGER_SIZE:
3947                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_LARGER));
3948                                         break;
3949                                 case LARGEST_SIZE:
3950                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_LARGEST));
3951                                         break;
3952                                 case HUGE_SIZE:
3953                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_HUGE));
3954                                         break;
3955                                 case HUGER_SIZE:
3956                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_HUGER));
3957                                         break;
3958                                 case INCREASE_SIZE:
3959                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_INCREASE));
3960                                         break;
3961                                 case DECREASE_SIZE:
3962                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_DECREASE));
3963                                         break;
3964                                 case INHERIT_SIZE:
3965                                 case NORMAL_SIZE:
3966                                         break;
3967                                 default:
3968                                         // the other tags are for internal use
3969                                         LATTEST(false);
3970                                         break;
3971                                 }
3972                                 size_flag = false;
3973                         }
3974                         switch (curr_size) {
3975                         case TINY_SIZE:
3976                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_TINY));
3977                                 size_flag = true;
3978                                 break;
3979                         case SCRIPT_SIZE:
3980                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_SCRIPT));
3981                                 size_flag = true;
3982                                 break;
3983                         case FOOTNOTE_SIZE:
3984                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_FOOTNOTE));
3985                                 size_flag = true;
3986                                 break;
3987                         case SMALL_SIZE:
3988                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_SMALL));
3989                                 size_flag = true;
3990                                 break;
3991                         case LARGE_SIZE:
3992                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_LARGE));
3993                                 size_flag = true;
3994                                 break;
3995                         case LARGER_SIZE:
3996                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_LARGER));
3997                                 size_flag = true;
3998                                 break;
3999                         case LARGEST_SIZE:
4000                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_LARGEST));
4001                                 size_flag = true;
4002                                 break;
4003                         case HUGE_SIZE:
4004                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_HUGE));
4005                                 size_flag = true;
4006                                 break;
4007                         case HUGER_SIZE:
4008                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_HUGER));
4009                                 size_flag = true;
4010                                 break;
4011                         case INCREASE_SIZE:
4012                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_INCREASE));
4013                                 size_flag = true;
4014                                 break;
4015                         case DECREASE_SIZE:
4016                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_DECREASE));
4017                                 size_flag = true;
4018                                 break;
4019                         case INHERIT_SIZE:
4020                         case NORMAL_SIZE:
4021                                 break;
4022                         default:
4023                                 // the other tags are for internal use
4024                                 LATTEST(false);
4025                                 break;
4026                         }
4027                 }
4028
4029                 // FIXME XHTML
4030                 // Other such tags? What about the other text ranges?
4031
4032                 for (auto const & t : tagsToClose)
4033                         xs << t;
4034                 for (auto const & t : tagsToOpen)
4035                         xs << t;
4036                 tagsToClose.clear();
4037                 tagsToOpen.clear();
4038
4039                 Inset const * inset = getInset(i);
4040                 if (inset) {
4041                         if (!runparams.for_toc || inset->isInToc()) {
4042                                 OutputParams np = runparams;
4043                                 np.local_font = &font;
4044                                 // If the paragraph has size 1, then we are in the "special
4045                                 // case" where we do not output the containing paragraph info
4046                                 if (!inset->getLayout().htmlisblock() && size() != 1)
4047                                         np.html_in_par = true;
4048                                 retval += inset->xhtml(xs, np);
4049                         }
4050                 } else {
4051                         char_type c = getUChar(buf.masterBuffer()->params(),
4052                                                runparams, i);
4053                         if (c == ' ' && (style.free_spacing || runparams.free_spacing))
4054                                 xs << XMLStream::ESCAPE_NONE << "&#160;";
4055                         else if (c == '\'')
4056                                 xs << XMLStream::ESCAPE_NONE << "&#8217;";
4057                         else
4058                                 xs << c;
4059                 }
4060                 font_old = font.fontInfo();
4061         }
4062
4063         // FIXME XHTML
4064         // I'm worried about what happens if a branch, say, is itself
4065         // wrapped in some font stuff. I think that will not work.
4066         xs.closeFontTags();
4067         if (close_paragraph)
4068                 xs.endDivision();
4069
4070         return retval;
4071 }
4072
4073
4074 bool Paragraph::isHfill(pos_type pos) const
4075 {
4076         Inset const * inset = getInset(pos);
4077         return inset && inset->isHfill();
4078 }
4079
4080
4081 bool Paragraph::isNewline(pos_type pos) const
4082 {
4083         // U+2028 LINE SEPARATOR
4084         // U+2029 PARAGRAPH SEPARATOR
4085         char_type const c = d->text_[pos];
4086         if (c == 0x2028 || c == 0x2029)
4087                 return true;
4088         Inset const * inset = getInset(pos);
4089         return inset && inset->lyxCode() == NEWLINE_CODE;
4090 }
4091
4092
4093 bool Paragraph::isEnvSeparator(pos_type pos) const
4094 {
4095         Inset const * inset = getInset(pos);
4096         return inset && inset->lyxCode() == SEPARATOR_CODE;
4097 }
4098
4099
4100 bool Paragraph::isLineSeparator(pos_type pos) const
4101 {
4102         char_type const c = d->text_[pos];
4103         if (isLineSeparatorChar(c))
4104                 return true;
4105         Inset const * inset = getInset(pos);
4106         return inset && inset->isLineSeparator();
4107 }
4108
4109
4110 bool Paragraph::isWordSeparator(pos_type pos, bool const ignore_deleted) const
4111 {
4112         if (pos == size())
4113                 return true;
4114         if (ignore_deleted && isDeleted(pos))
4115                 return false;
4116         if (Inset const * inset = getInset(pos))
4117                 return !inset->isLetter();
4118         // if we have a hard hyphen (no en- or emdash) or apostrophe
4119         // we pass this to the spell checker
4120         // FIXME: this method is subject to change, visit
4121         // https://bugzilla.mozilla.org/show_bug.cgi?id=355178
4122         // to get an impression how complex this is.
4123         if (isHardHyphenOrApostrophe(pos))
4124                 return false;
4125         char_type const c = d->text_[pos];
4126         // We want to pass the escape chars to the spellchecker
4127         docstring const escape_chars = from_utf8(lyxrc.spellchecker_esc_chars);
4128         return !isLetterChar(c) && !isDigitASCII(c) && !contains(escape_chars, c);
4129 }
4130
4131
4132 bool Paragraph::isHardHyphenOrApostrophe(pos_type pos) const
4133 {
4134         pos_type const psize = size();
4135         if (pos >= psize)
4136                 return false;
4137         char_type const c = d->text_[pos];
4138         if (c != '-' && c != '\'')
4139                 return false;
4140         pos_type nextpos = pos + 1;
4141         pos_type prevpos = pos > 0 ? pos - 1 : 0;
4142         if ((nextpos == psize || isSpace(nextpos))
4143                 && (pos == 0 || isSpace(prevpos)))
4144                 return false;
4145         return true;
4146 }
4147
4148
4149 bool Paragraph::needsCProtection(bool const fragile) const
4150 {
4151         // first check the layout of the paragraph, but only in insets
4152         InsetText const * textinset = inInset().asInsetText();
4153         bool const maintext = textinset
4154                 ? textinset->text().isMainText()
4155                 : false;
4156
4157         if (!maintext && layout().needcprotect) {
4158                 // Environments need cprotection regardless the content
4159                 if (layout().latextype == LATEX_ENVIRONMENT)
4160                         return true;
4161
4162                 // Commands need cprotection if they contain specific chars
4163                 int const nchars_escape = 9;
4164                 static char_type const chars_escape[nchars_escape] = {
4165                         '&', '_', '$', '%', '#', '^', '{', '}', '\\'};
4166
4167                 docstring const pars = asString();
4168                 for (int k = 0; k < nchars_escape; k++) {
4169                         if (contains(pars, chars_escape[k]))
4170                                 return true;
4171                 }
4172         }
4173
4174         // now check whether we have insets that need cprotection
4175         for (auto const & icit : d->insetlist_) {
4176                 Inset const * ins = icit.inset;
4177                 if (!ins)
4178                         continue;
4179                 if (ins->needsCProtection(maintext, fragile))
4180                         return true;
4181         }
4182
4183         return false;
4184 }
4185
4186
4187 FontSpan const & Paragraph::getSpellRange(pos_type pos) const
4188 {
4189         return d->speller_state_.getRange(pos);
4190 }
4191
4192
4193 bool Paragraph::isChar(pos_type pos) const
4194 {
4195         if (Inset const * inset = getInset(pos))
4196                 return inset->isChar();
4197         char_type const c = d->text_[pos];
4198         return !isLetterChar(c) && !isDigitASCII(c) && !lyx::isSpace(c);
4199 }
4200
4201
4202 bool Paragraph::isSpace(pos_type pos) const
4203 {
4204         if (Inset const * inset = getInset(pos))
4205                 return inset->isSpace();
4206         char_type const c = d->text_[pos];
4207         return lyx::isSpace(c);
4208 }
4209
4210
4211 Language const *
4212 Paragraph::getParLanguage(BufferParams const & bparams) const
4213 {
4214         if (!empty())
4215                 return getFirstFontSettings(bparams).language();
4216         // FIXME: we should check the prev par as well (Lgb)
4217         return bparams.language;
4218 }
4219
4220
4221 bool Paragraph::isRTL(BufferParams const & bparams) const
4222 {
4223         return getParLanguage(bparams)->rightToLeft()
4224                 && !inInset().getLayout().forceLTR();
4225 }
4226
4227
4228 void Paragraph::changeLanguage(BufferParams const & bparams,
4229                                Language const * from, Language const * to)
4230 {
4231         // change language including dummy font change at the end
4232         for (pos_type i = 0; i <= size(); ++i) {
4233                 Font font = getFontSettings(bparams, i);
4234                 if (font.language() == from) {
4235                         font.setLanguage(to);
4236                         setFont(i, font);
4237                         d->requestSpellCheck(i);
4238                 }
4239         }
4240 }
4241
4242
4243 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
4244 {
4245         Language const * doc_language = bparams.language;
4246         for (auto const & f : d->fontlist_)
4247                 if (f.font().language() != ignore_language &&
4248                     f.font().language() != latex_language &&
4249                     f.font().language() != doc_language)
4250                         return true;
4251         return false;
4252 }
4253
4254
4255 void Paragraph::getLanguages(std::set<Language const *> & langs) const
4256 {
4257         for (auto const & f : d->fontlist_) {
4258                 Language const * lang = f.font().language();
4259                 if (lang != ignore_language &&
4260                     lang != latex_language)
4261                         langs.insert(lang);
4262         }
4263 }
4264
4265
4266 docstring Paragraph::asString(int options) const
4267 {
4268         return asString(0, size(), options);
4269 }
4270
4271
4272 docstring Paragraph::asString(pos_type beg, pos_type end, int options, const OutputParams *runparams) const
4273 {
4274         odocstringstream os;
4275
4276         if (beg == 0
4277             && options & AS_STR_LABEL
4278             && !d->params_.labelString().empty())
4279                 os << d->params_.labelString() << ' ';
4280
4281         for (pos_type i = beg; i < end; ++i) {
4282                 if ((options & AS_STR_SKIPDELETE) && isDeleted(i))
4283                         continue;
4284                 char_type const c = d->text_[i];
4285                 if (isPrintable(c) || c == '\t'
4286                     || (c == '\n' && (options & AS_STR_NEWLINES)))
4287                         os.put(c);
4288                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
4289                         if (c == META_INSET && (options & AS_STR_PLAINTEXT)) {
4290                                 LASSERT(runparams != nullptr, return docstring());
4291                                 if (runparams->find_effective())
4292                                         getInset(i)->toString(os);
4293                                 else
4294                                         getInset(i)->plaintext(os, *runparams);
4295                         } else if (c == META_INSET && (options & AS_STR_MATHED)
4296                                    && getInset(i)->lyxCode() == REF_CODE) {
4297                                 Buffer const & buf = getInset(i)->buffer();
4298                                 OutputParams rp(&buf.params().encoding());
4299                                 Font const font(inherit_font, buf.params().language);
4300                                 rp.local_font = &font;
4301                                 otexstream ots(os);
4302                                 getInset(i)->latex(ots, rp);
4303                         } else {
4304                                 getInset(i)->toString(os);
4305                         }
4306                 }
4307         }
4308
4309         return os.str();
4310 }
4311
4312
4313 void Paragraph::forOutliner(docstring & os, size_t const maxlen,
4314                             bool const shorten, bool const label) const
4315 {
4316         size_t tmplen = shorten ? maxlen + 1 : maxlen;
4317         if (label && !labelString().empty())
4318                 os += labelString() + ' ';
4319         if (!layout().isTocCaption())
4320                 return;
4321         for (pos_type i = 0; i < size() && os.length() < tmplen; ++i) {
4322                 if (isDeleted(i))
4323                         continue;
4324                 char_type const c = d->text_[i];
4325                 if (isPrintable(c))
4326                         os += c;
4327                 else if (c == META_INSET)
4328                         getInset(i)->forOutliner(os, tmplen, false);
4329         }
4330         if (shorten)
4331                 Text::shortenForOutliner(os, maxlen);
4332 }
4333
4334
4335 void Paragraph::setInsetOwner(Inset const * inset)
4336 {
4337         d->inset_owner_ = inset;
4338 }
4339
4340
4341 int Paragraph::id() const
4342 {
4343         return d->id_;
4344 }
4345
4346
4347 void Paragraph::setId(int id)
4348 {
4349         d->id_ = id;
4350 }
4351
4352
4353 Layout const & Paragraph::layout() const
4354 {
4355         return *d->layout_;
4356 }
4357
4358
4359 void Paragraph::setLayout(Layout const & layout)
4360 {
4361         d->layout_ = &layout;
4362 }
4363
4364
4365 void Paragraph::setDefaultLayout(DocumentClass const & tc)
4366 {
4367         setLayout(tc.defaultLayout());
4368 }
4369
4370
4371 void Paragraph::setPlainLayout(DocumentClass const & tc)
4372 {
4373         setLayout(tc.plainLayout());
4374 }
4375
4376
4377 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tc)
4378 {
4379         if (usePlainLayout())
4380                 setPlainLayout(tc);
4381         else
4382                 setDefaultLayout(tc);
4383 }
4384
4385
4386 Inset const & Paragraph::inInset() const
4387 {
4388         LBUFERR(d->inset_owner_);
4389         return *d->inset_owner_;
4390 }
4391
4392
4393 ParagraphParameters & Paragraph::params()
4394 {
4395         return d->params_;
4396 }
4397
4398
4399 ParagraphParameters const & Paragraph::params() const
4400 {
4401         return d->params_;
4402 }
4403
4404
4405 bool Paragraph::isFreeSpacing() const
4406 {
4407         if (d->layout_->free_spacing)
4408                 return true;
4409         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
4410 }
4411
4412
4413 bool Paragraph::allowEmpty() const
4414 {
4415         if (d->layout_->keepempty)
4416                 return true;
4417         return d->inset_owner_ && d->inset_owner_->allowEmpty();
4418 }
4419
4420
4421 bool Paragraph::brokenBiblio() const
4422 {
4423         // There is a problem if there is no bibitem at position 0 in
4424         // paragraphs that need one, if there is another bibitem in the
4425         // paragraph or if this paragraph is not supposed to have
4426         // a bibitem inset at all.
4427         return ((d->layout_->labeltype == LABEL_BIBLIO
4428                 && (d->insetlist_.find(BIBITEM_CODE) != 0
4429                     || d->insetlist_.find(BIBITEM_CODE, 1) > 0))
4430                 || (d->layout_->labeltype != LABEL_BIBLIO
4431                     && d->insetlist_.find(BIBITEM_CODE) != -1));
4432 }
4433
4434
4435 int Paragraph::fixBiblio(Buffer const & buffer)
4436 {
4437         // FIXME: when there was already an inset at 0, the return value is 1,
4438         // which does not tell whether another inset has been removed; the
4439         // cursor cannot be correctly updated.
4440
4441         bool const track_changes = buffer.params().track_changes;
4442         int bibitem_pos = d->insetlist_.find(BIBITEM_CODE);
4443
4444         // The case where paragraph is not BIBLIO
4445         if (d->layout_->labeltype != LABEL_BIBLIO) {
4446                 if (bibitem_pos == -1)
4447                         // No InsetBibitem => OK
4448                         return 0;
4449                 // There is an InsetBibitem: remove it!
4450                 d->insetlist_.release(bibitem_pos);
4451                 eraseChar(bibitem_pos, track_changes);
4452                 return (bibitem_pos == 0) ? -1 : -bibitem_pos;
4453         }
4454
4455         bool const hasbibitem0 = bibitem_pos == 0;
4456         if (hasbibitem0) {
4457                 bibitem_pos = d->insetlist_.find(BIBITEM_CODE, 1);
4458                 // There was an InsetBibitem at pos 0,
4459                 // and no other one => OK
4460                 if (bibitem_pos == -1)
4461                         return 0;
4462                 // there is a bibitem at the 0 position, but since
4463                 // there is a second one, we copy the second on the
4464                 // first. We're assuming there are at most two of
4465                 // these, which there should be.
4466                 // FIXME: why does it make sense to do that rather
4467                 // than keep the first? (JMarc)
4468                 Inset * inset = releaseInset(bibitem_pos);
4469                 d->insetlist_.begin()->inset = inset;
4470                 // This needs to be done to update the counter (#8499)
4471                 buffer.updateBuffer();
4472                 return -bibitem_pos;
4473         }
4474
4475         // We need to create an inset at the beginning
4476         Inset * inset = nullptr;
4477         if (bibitem_pos > 0) {
4478                 // there was one somewhere in the paragraph, let's move it
4479                 inset = d->insetlist_.release(bibitem_pos);
4480                 eraseChar(bibitem_pos, track_changes);
4481         } else
4482                 // make a fresh one
4483                 inset = new InsetBibitem(const_cast<Buffer *>(&buffer),
4484                                          InsetCommandParams(BIBITEM_CODE));
4485
4486         Font font(inherit_font, buffer.params().language);
4487         insertInset(0, inset, font, Change(track_changes ? Change::INSERTED
4488                                                    : Change::UNCHANGED));
4489
4490         // This is needed to get the counters right
4491         buffer.updateBuffer();
4492         return 1;
4493 }
4494
4495
4496 void Paragraph::checkAuthors(AuthorList const & authorList)
4497 {
4498         d->changes_.checkAuthors(authorList);
4499 }
4500
4501
4502 bool Paragraph::isChanged(pos_type pos) const
4503 {
4504         return lookupChange(pos).changed();
4505 }
4506
4507
4508 bool Paragraph::isInserted(pos_type pos) const
4509 {
4510         return lookupChange(pos).inserted();
4511 }
4512
4513
4514 bool Paragraph::isDeleted(pos_type pos) const
4515 {
4516         return lookupChange(pos).deleted();
4517 }
4518
4519
4520 InsetList const & Paragraph::insetList() const
4521 {
4522         return d->insetlist_;
4523 }
4524
4525
4526 void Paragraph::setInsetBuffers(Buffer & b)
4527 {
4528         d->insetlist_.setBuffer(b);
4529 }
4530
4531
4532 void Paragraph::resetBuffer()
4533 {
4534         d->insetlist_.resetBuffer();
4535 }
4536
4537
4538 Inset * Paragraph::releaseInset(pos_type pos)
4539 {
4540         Inset * inset = d->insetlist_.release(pos);
4541         /// does not honour change tracking!
4542         eraseChar(pos, false);
4543         return inset;
4544 }
4545
4546
4547 Inset * Paragraph::getInset(pos_type pos)
4548 {
4549         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
4550                  ? d->insetlist_.get(pos) : nullptr;
4551 }
4552
4553
4554 Inset const * Paragraph::getInset(pos_type pos) const
4555 {
4556         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
4557                  ? d->insetlist_.get(pos) : nullptr;
4558 }
4559
4560
4561 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
4562                 pos_type & right, TextCase action)
4563 {
4564         // process sequences of modified characters; in change
4565         // tracking mode, this approach results in much better
4566         // usability than changing case on a char-by-char basis
4567         // We also need to track the current font, since font
4568         // changes within sequences can occur.
4569         vector<pair<char_type, Font> > changes;
4570
4571         bool const trackChanges = bparams.track_changes;
4572
4573         bool capitalize = true;
4574
4575         for (; pos < right; ++pos) {
4576                 char_type oldChar = d->text_[pos];
4577                 char_type newChar = oldChar;
4578
4579                 // ignore insets and don't play with deleted text!
4580                 if (oldChar != META_INSET && !isDeleted(pos)) {
4581                         switch (action) {
4582                                 case text_lowercase:
4583                                         newChar = lowercase(oldChar);
4584                                         break;
4585                                 case text_capitalization:
4586                                         if (capitalize) {
4587                                                 newChar = uppercase(oldChar);
4588                                                 capitalize = false;
4589                                         }
4590                                         break;
4591                                 case text_uppercase:
4592                                         newChar = uppercase(oldChar);
4593                                         break;
4594                         }
4595                 }
4596
4597                 if (isWordSeparator(pos) || isDeleted(pos)) {
4598                         // permit capitalization again
4599                         capitalize = true;
4600                 }
4601
4602                 if (oldChar != newChar) {
4603                         changes.push_back(make_pair(newChar, getFontSettings(bparams, pos)));
4604                         if (pos != right - 1)
4605                                 continue;
4606                         // step behind the changing area
4607                         pos++;
4608                 }
4609
4610                 int erasePos = pos - changes.size();
4611                 for (auto const & change : changes) {
4612                         insertChar(pos, change.first, change.second, trackChanges);
4613                         if (!eraseChar(erasePos, trackChanges)) {
4614                                 ++erasePos;
4615                                 ++pos; // advance
4616                                 ++right; // expand selection
4617                         }
4618                 }
4619                 changes.clear();
4620         }
4621 }
4622
4623 int Paragraph::find(docstring const & str, bool cs, bool mw,
4624                 pos_type start_pos, bool del) const
4625 {
4626         pos_type pos = start_pos;
4627         int const strsize = str.length();
4628         int i = 0;
4629         pos_type const parsize = d->text_.size();
4630         for (i = 0; i < strsize && pos < parsize; ++i, ++pos) {
4631                 // ignore deleted matter
4632                 if (!del && isDeleted(pos)) {
4633                         if (pos == parsize - 1)
4634                                 break;
4635                         pos++;
4636                         --i;
4637                         continue;
4638                 }
4639                 // Ignore "invisible" letters such as ligature breaks
4640                 // and hyphenation chars while searching
4641                 bool nonmatch = false;
4642                 while (pos < parsize && isInset(pos)) {
4643                         Inset const * inset = getInset(pos);
4644                         if (!inset->isLetter() && !inset->isChar())
4645                                 break;
4646                         odocstringstream os;
4647                         if (inset->lyxCode() == lyx::QUOTE_CODE) {
4648                                 OutputParams op(0);
4649                                 op.find_set_feature(OutputParams::SearchQuick);
4650                                 inset->plaintext(os, op);
4651                         }
4652                         else {
4653                                 inset->toString(os);
4654                         }
4655                         docstring const insetstring = os.str();
4656                         if (!insetstring.empty()) {
4657                                 int const insetstringsize = insetstring.length();
4658                                 for (int j = 0; j < insetstringsize && pos < parsize; ++i, ++j) {
4659                                         if ((cs && str[i] != insetstring[j])
4660                                             || (!cs && uppercase(str[i]) != uppercase(insetstring[j]))) {
4661                                                 nonmatch = true;
4662                                                 break;
4663                                         }
4664                                 }
4665                         }
4666                         pos++;
4667                 }
4668                 if (nonmatch || i == strsize)
4669                         break;
4670                 char_type dp = d->text_[pos];
4671                 if (cs && str[i] != dp)
4672                         break;
4673                 if (!cs && uppercase(str[i]) != uppercase(dp))
4674                         break;
4675         }
4676
4677         if (i != strsize)
4678                 return 0;
4679
4680         // if necessary, check whether string matches word
4681         if (mw) {
4682                 if (start_pos > 0 && !isWordSeparator(start_pos - 1))
4683                         return 0;
4684                 if (pos < parsize
4685                         && !isWordSeparator(pos))
4686                         return 0;
4687         }
4688
4689         return pos - start_pos;
4690 }
4691
4692
4693 char_type Paragraph::getChar(pos_type pos) const
4694 {
4695         return d->text_[pos];
4696 }
4697
4698
4699 pos_type Paragraph::size() const
4700 {
4701         return d->text_.size();
4702 }
4703
4704
4705 bool Paragraph::empty() const
4706 {
4707         return d->text_.empty();
4708 }
4709
4710
4711 bool Paragraph::isInset(pos_type pos) const
4712 {
4713         return d->text_[pos] == META_INSET;
4714 }
4715
4716
4717 bool Paragraph::isSeparator(pos_type pos) const
4718 {
4719         //FIXME: Are we sure this can be the only separator?
4720         return d->text_[pos] == ' ';
4721 }
4722
4723
4724 void Paragraph::deregisterWords()
4725 {
4726         Private::LangWordsMap::const_iterator itl = d->words_.begin();
4727         Private::LangWordsMap::const_iterator ite = d->words_.end();
4728         for (; itl != ite; ++itl) {
4729                 WordList & wl = theWordList(itl->first);
4730                 Private::Words::const_iterator it = (itl->second).begin();
4731                 Private::Words::const_iterator et = (itl->second).end();
4732                 for (; it != et; ++it)
4733                         wl.remove(*it);
4734         }
4735         d->words_.clear();
4736 }
4737
4738
4739 void Paragraph::locateWord(pos_type & from, pos_type & to,
4740         word_location const loc, bool const ignore_deleted) const
4741 {
4742         switch (loc) {
4743         case WHOLE_WORD_STRICT:
4744                 if (from == 0 || from == size()
4745                     || isWordSeparator(from, ignore_deleted)
4746                     || isWordSeparator(from - 1, ignore_deleted)) {
4747                         to = from;
4748                         return;
4749                 }
4750                 // fall through
4751
4752         case WHOLE_WORD:
4753                 // If we are already at the beginning of a word, do nothing
4754                 if (!from || isWordSeparator(from - 1, ignore_deleted))
4755                         break;
4756                 // fall through
4757
4758         case PREVIOUS_WORD:
4759                 // always move the cursor to the beginning of previous word
4760                 while (from && !isWordSeparator(from - 1, ignore_deleted))
4761                         --from;
4762                 break;
4763         case NEXT_WORD:
4764                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
4765                 break;
4766         case PARTIAL_WORD:
4767                 // no need to move the 'from' cursor
4768                 break;
4769         }
4770         to = from;
4771         while (to < size() && !isWordSeparator(to, ignore_deleted))
4772                 ++to;
4773 }
4774
4775
4776 void Paragraph::collectWords()
4777 {
4778         for (pos_type pos = 0; pos < size(); ++pos) {
4779                 if (isWordSeparator(pos))
4780                         continue;
4781                 pos_type from = pos;
4782                 locateWord(from, pos, WHOLE_WORD);
4783                 // Work around MSVC warning: The statement
4784                 // if (pos < from + lyxrc.completion_minlength)
4785                 // triggers a signed vs. unsigned warning.
4786                 // I don't know why this happens, it could be a MSVC bug, or
4787                 // related to LLP64 (windows) vs. LP64 (unix) programming
4788                 // model, or the C++ standard might be ambigous in the section
4789                 // defining the "usual arithmetic conversions". However, using
4790                 // a temporary variable is safe and works on all compilers.
4791                 pos_type const endpos = from + lyxrc.completion_minlength;
4792                 if (pos < endpos)
4793                         continue;
4794                 FontList::const_iterator cit = d->fontlist_.fontIterator(from);
4795                 if (cit == d->fontlist_.end())
4796                         return;
4797                 Language const * lang = cit->font().language();
4798                 docstring const word = asString(from, pos, AS_STR_NONE);
4799                 d->words_[lang->lang()].insert(word);
4800         }
4801 }
4802
4803
4804 void Paragraph::registerWords()
4805 {
4806         Private::LangWordsMap::const_iterator itl = d->words_.begin();
4807         Private::LangWordsMap::const_iterator ite = d->words_.end();
4808         for (; itl != ite; ++itl) {
4809                 WordList & wl = theWordList(itl->first);
4810                 Private::Words::const_iterator it = (itl->second).begin();
4811                 Private::Words::const_iterator et = (itl->second).end();
4812                 for (; it != et; ++it)
4813                         wl.insert(*it);
4814         }
4815 }
4816
4817
4818 void Paragraph::updateWords()
4819 {
4820         deregisterWords();
4821         collectWords();
4822         registerWords();
4823 }
4824
4825
4826 void Paragraph::Private::appendSkipPosition(SkipPositions & skips, pos_type const pos) const
4827 {
4828         SkipPositionsIterator begin = skips.begin();
4829         SkipPositions::iterator end = skips.end();
4830         if (pos > 0 && begin < end) {
4831                 --end;
4832                 if (end->last == pos - 1) {
4833                         end->last = pos;
4834                         return;
4835                 }
4836         }
4837         skips.insert(end, FontSpan(pos, pos));
4838 }
4839
4840
4841 Language * Paragraph::Private::locateSpellRange(
4842         pos_type & from, pos_type & to,
4843         SkipPositions & skips) const
4844 {
4845         // skip leading white space
4846         while (from < to && owner_->isWordSeparator(from))
4847                 ++from;
4848         // don't check empty range
4849         if (from >= to)
4850                 return nullptr;
4851         // get current language
4852         Language * lang = getSpellLanguage(from);
4853         pos_type last = from;
4854         bool samelang = true;
4855         bool sameinset = true;
4856         while (last < to && samelang && sameinset) {
4857                 // hop to end of word
4858                 while (last < to && !owner_->isWordSeparator(last)) {
4859                         Inset const * inset = owner_->getInset(last);
4860                         if (inset && dynamic_cast<const InsetSpecialChar *>(inset)) {
4861                                 // check for "invisible" letters such as ligature breaks
4862                                 odocstringstream os;
4863                                 inset->toString(os);
4864                                 if (os.str().length() != 0) {
4865                                         // avoid spell check of visible special char insets
4866                                         // stop the loop in front of the special char inset
4867                                         sameinset = false;
4868                                         break;
4869                                 }
4870                         } else if (inset) {
4871                                 appendSkipPosition(skips, last);
4872                         } else if (owner_->isDeleted(last)) {
4873                                 appendSkipPosition(skips, last);
4874                         }
4875                         ++last;
4876                 }
4877                 // hop to next word while checking for insets
4878                 while (sameinset && last < to && owner_->isWordSeparator(last)) {
4879                         if (Inset const * inset = owner_->getInset(last))
4880                                 sameinset = inset->isChar() && inset->isLetter();
4881                         if (sameinset && owner_->isDeleted(last)) {
4882                                 appendSkipPosition(skips, last);
4883                         }
4884                         if (sameinset)
4885                                 last++;
4886                 }
4887                 if (sameinset && last < to) {
4888                         // now check for language change
4889                         samelang = lang == getSpellLanguage(last);
4890                 }
4891         }
4892         // if language change detected backstep is needed
4893         if (!samelang)
4894                 --last;
4895         to = last;
4896         return lang;
4897 }
4898
4899
4900 Language * Paragraph::Private::getSpellLanguage(pos_type const from) const
4901 {
4902         Language * lang =
4903                 const_cast<Language *>(owner_->getFontSettings(
4904                         inset_owner_->buffer().params(), from).language());
4905         if (lang == inset_owner_->buffer().params().language
4906                 && !lyxrc.spellchecker_alt_lang.empty()) {
4907                 string lang_code;
4908                 string const lang_variety =
4909                         split(lyxrc.spellchecker_alt_lang, lang_code, '-');
4910                 lang->setCode(lang_code);
4911                 lang->setVariety(lang_variety);
4912         }
4913         return lang;
4914 }
4915
4916
4917 void Paragraph::requestSpellCheck(pos_type pos)
4918 {
4919         d->requestSpellCheck(pos);
4920         if (pos == -1) {
4921                 // Also request spellcheck within (text) insets
4922                 for (auto const & insets : insetList()) {
4923                         if (!insets.inset->asInsetText())
4924                                 continue;
4925                         ParagraphList & inset_pars =
4926                                 insets.inset->asInsetText()->paragraphs();
4927                         ParagraphList::iterator pit = inset_pars.begin();
4928                         ParagraphList::iterator pend = inset_pars.end();
4929                         for (; pit != pend; ++pit)
4930                                 pit->requestSpellCheck();
4931                 }
4932         }
4933 }
4934
4935
4936 bool Paragraph::needsSpellCheck() const
4937 {
4938         SpellChecker::ChangeNumber speller_change_number = 0;
4939         if (theSpellChecker())
4940                 speller_change_number = theSpellChecker()->changeNumber();
4941         if (speller_change_number > d->speller_state_.currentChangeNumber()) {
4942                 d->speller_state_.needsCompleteRefresh(speller_change_number);
4943         }
4944         return d->needsSpellCheck();
4945 }
4946
4947
4948 bool Paragraph::Private::ignoreWord(docstring const & word) const
4949 {
4950         // Ignore words with digits
4951         // FIXME: make this customizable
4952         // (note that some checkers ignore words with digits by default)
4953         docstring::const_iterator cit = word.begin();
4954         docstring::const_iterator const end = word.end();
4955         for (; cit != end; ++cit) {
4956                 if (isNumber((*cit)))
4957                         return true;
4958         }
4959         return false;
4960 }
4961
4962
4963 SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
4964         WordLangTuple & wl, docstring_list & suggestions,
4965         bool do_suggestion, bool check_learned) const
4966 {
4967         SpellChecker::Result result = SpellChecker::WORD_OK;
4968         SpellChecker * speller = theSpellChecker();
4969         if (!speller)
4970                 return result;
4971
4972         if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
4973                 return result;
4974
4975         locateWord(from, to, WHOLE_WORD, true);
4976         if (from == to || from >= size())
4977                 return result;
4978
4979         docstring word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
4980         Language * lang = d->getSpellLanguage(from);
4981
4982         BufferParams const & bparams = d->inset_owner_->buffer().params();
4983
4984         if (getFontSettings(bparams, from).fontInfo().nospellcheck() == FONT_ON)
4985                 return result;
4986
4987         wl = WordLangTuple(word, lang);
4988
4989         if (word.empty())
4990                 return result;
4991
4992         if (needsSpellCheck() || check_learned) {
4993                 pos_type end = to;
4994                 if (!d->ignoreWord(word)) {
4995                         bool const trailing_dot = to < size() && d->text_[to] == '.';
4996                         result = speller->check(wl, bparams.spellignore());
4997                         if (SpellChecker::misspelled(result) && trailing_dot) {
4998                                 wl = WordLangTuple(word.append(from_ascii(".")), lang);
4999                                 result = speller->check(wl, bparams.spellignore());
5000                                 if (!SpellChecker::misspelled(result)) {
5001                                         LYXERR(Debug::GUI, "misspelled word is correct with dot: \"" <<
5002                                            word << "\" [" <<
5003                                            from << ".." << to << "]");
5004                                 } else {
5005                                         // spell check with dot appended failed too
5006                                         // restore original word/lang value
5007                                         word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
5008                                         wl = WordLangTuple(word, lang);
5009                                 }
5010                         }
5011                 }
5012                 if (!SpellChecker::misspelled(result)) {
5013                         // area up to the begin of the next word is not misspelled
5014                         while (end < size() && isWordSeparator(end))
5015                                 ++end;
5016                 }
5017                 d->setMisspelled(from, end, result);
5018         } else {
5019                 result = d->speller_state_.getState(from);
5020         }
5021
5022         if (do_suggestion)
5023                 suggestions.clear();
5024
5025         if (SpellChecker::misspelled(result)) {
5026                 LYXERR(Debug::GUI, "misspelled word: \"" <<
5027                            word << "\" [" <<
5028                            from << ".." << to << "]");
5029                 if (do_suggestion)
5030                         speller->suggest(wl, suggestions);
5031         }
5032         return result;
5033 }
5034
5035
5036 void Paragraph::anonymize()
5037 {
5038         // This is a very crude anonymization for now
5039         for (char_type & c : d->text_)
5040                 if (isLetterChar(c) || isNumber(c))
5041                         c = 'a';
5042 }
5043
5044
5045 void Paragraph::Private::markMisspelledWords(
5046         Language const * lang,
5047         pos_type const & first, pos_type const & last,
5048         SpellChecker::Result result,
5049         docstring const & word,
5050         SkipPositions const & skips)
5051 {
5052         if (!SpellChecker::misspelled(result)) {
5053                 setMisspelled(first, last, SpellChecker::WORD_OK);
5054                 return;
5055         }
5056         int snext = first;
5057         SpellChecker * speller = theSpellChecker();
5058         // locate and enumerate the error positions
5059         int nerrors = speller->numMisspelledWords();
5060         int numskipped = 0;
5061         SkipPositionsIterator it = skips.begin();
5062         SkipPositionsIterator et = skips.end();
5063         for (int index = 0; index < nerrors; ++index) {
5064                 int wstart;
5065                 int wlen = 0;
5066                 speller->misspelledWord(index, wstart, wlen);
5067                 /// should not happen if speller supports range checks
5068                 if (!wlen)
5069                         continue;
5070                 WordLangTuple const candidate(word.substr(wstart, wlen), lang);
5071                 wstart += first + numskipped;
5072                 if (snext < wstart) {
5073                         /// mark the range of correct spelling
5074                         numskipped += countSkips(it, et, wstart);
5075                         setMisspelled(snext,
5076                                 wstart - 1, SpellChecker::WORD_OK);
5077                 }
5078                 snext = wstart + wlen;
5079                 // Check whether the candidate is in the document's local dict
5080                 SpellChecker::Result actresult = result;
5081                 if (inset_owner_->buffer().params().spellignored(candidate))
5082                         actresult = SpellChecker::DOCUMENT_LEARNED_WORD;
5083                 numskipped += countSkips(it, et, snext);
5084                 /// mark the range of misspelling
5085                 setMisspelled(wstart, snext, actresult);
5086                 if (actresult == SpellChecker::DOCUMENT_LEARNED_WORD)
5087                         LYXERR(Debug::GUI, "local dictionary word: \"" <<
5088                                    candidate.word() << "\" [" <<
5089                                    wstart << ".." << (snext-1) << "]");
5090                 else
5091                         LYXERR(Debug::GUI, "misspelled word: \"" <<
5092                                    candidate.word() << "\" [" <<
5093                                    wstart << ".." << (snext-1) << "]");
5094                 ++snext;
5095         }
5096         if (snext <= last) {
5097                 /// mark the range of correct spelling at end
5098                 setMisspelled(snext, last, SpellChecker::WORD_OK);
5099         }
5100 }
5101
5102
5103 void Paragraph::spellCheck() const
5104 {
5105         SpellChecker * speller = theSpellChecker();
5106         if (!speller || empty() ||!needsSpellCheck())
5107                 return;
5108         pos_type start;
5109         pos_type endpos;
5110         d->rangeOfSpellCheck(start, endpos);
5111         if (speller->canCheckParagraph()) {
5112                 // loop until we leave the range
5113                 for (pos_type first = start; first < endpos; ) {
5114                         pos_type last = endpos;
5115                         Private::SkipPositions skips;
5116                         Language * lang = d->locateSpellRange(first, last, skips);
5117                         if (first >= endpos)
5118                                 break;
5119                         // start the spell checker on the unit of meaning
5120                         docstring word = asString(first, last, AS_STR_INSETS + AS_STR_SKIPDELETE);
5121                         WordLangTuple wl = WordLangTuple(word, lang);
5122                         BufferParams const & bparams = d->inset_owner_->buffer().params();
5123                         SpellChecker::Result result = !word.empty() ?
5124                                 speller->check(wl, bparams.spellignore()) : SpellChecker::WORD_OK;
5125                         d->markMisspelledWords(lang, first, last, result, word, skips);
5126                         first = ++last;
5127                 }
5128         } else {
5129                 static docstring_list suggestions;
5130                 pos_type to = endpos;
5131                 while (start < endpos) {
5132                         WordLangTuple wl;
5133                         spellCheck(start, to, wl, suggestions, false);
5134                         start = to + 1;
5135                 }
5136         }
5137         d->readySpellCheck();
5138 }
5139
5140
5141 bool Paragraph::isMisspelled(pos_type pos, bool check_boundary) const
5142 {
5143         bool result = SpellChecker::misspelled(d->speller_state_.getState(pos));
5144         if (result || pos <= 0 || pos > size())
5145                 return result;
5146         if (check_boundary && (pos == size() || isWordSeparator(pos)))
5147                 result = SpellChecker::misspelled(d->speller_state_.getState(pos - 1));
5148         return result;
5149 }
5150
5151
5152 string Paragraph::magicLabel() const
5153 {
5154         stringstream ss;
5155         ss << "magicparlabel-" << id();
5156         return ss.str();
5157 }
5158
5159
5160 } // namespace lyx