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             && running_font.isRightToLeft()
1098             // ERT is an exception, it should be output with no
1099             // decorations at all
1100             && inset->lyxCode() != ERT_CODE) {
1101                 if (runparams.use_polyglossia) {
1102                         os << "\\LRE{";
1103                 } else if (running_font.language()->lang() == "farsi"
1104                            || running_font.language()->lang() == "arabic_arabi")
1105                         os << "\\textLR{" << termcmd;
1106                 else
1107                         os << "\\L{";
1108                 close = true;
1109         }
1110
1111         if (open_font && fontswitch_inset) {
1112                 bool lang_closed = false;
1113                 // Close language if needed
1114                 if (closeLanguage && !lang_switched_at_inset) {
1115                         // We need prev_font here as language changes directly at inset
1116                         // will only be started inside the inset.
1117                         Font const prev_font = (i > 0) ?
1118                                                 owner_->getFont(bparams, i - 1, outerfont)
1119                                               : running_font;
1120                         Font tmpfont(basefont);
1121                         tmpfont.setLanguage(prev_font.language());
1122                         bool needPar = false;
1123                         unsigned int count = tmpfont.latexWriteEndChanges(os, bparams, runparams,
1124                                                                           basefont, basefont,
1125                                                                           needPar, closeLanguage);
1126                         column += count;
1127                         lang_closed = count > 0;
1128                 }
1129                 // Update the running_font, making sure, however,
1130                 // to leave the language as it was.
1131                 // FIXME: probably a better way to keep track of the old
1132                 // language, than copying the entire font?
1133                 Font const copy_font(running_font);
1134                 basefont = owner_->getLayoutFont(bparams, outerfont);
1135                 running_font = basefont;
1136                 if (!closeLanguage)
1137                         running_font.setLanguage(copy_font.language());
1138                 OutputParams rp = runparams;
1139                 rp.encoding = basefont.language()->encoding();
1140                 // For these, we use switches, so they should be taken as
1141                 // base inside the inset.
1142                 basefont.fontInfo().setSize(copy_font.fontInfo().size());
1143                 basefont.fontInfo().setFamily(copy_font.fontInfo().family());
1144                 basefont.fontInfo().setSeries(copy_font.fontInfo().series());
1145                 // Now re-do font changes in a way needed here
1146                 // (using switches with multi-par insets)
1147                 InsetText const * textinset = inset->asInsetText();
1148                 bool const cprotect = textinset
1149                         ? textinset->hasCProtectContent(runparams.moving_arg)
1150                           && !textinset->text().isMainText()
1151                           && inset->lyxCode() != BRANCH_CODE
1152                         : false;
1153                 unsigned int count2 = basefont.latexWriteStartChanges(os, bparams,
1154                                                       rp, running_font,
1155                                                       basefont, true,
1156                                                       cprotect);
1157                 open_font = true;
1158                 column += count2;
1159                 if (count2 == 0 && (lang_closed || lang_switched_at_inset))
1160                         // All fonts closed
1161                         open_font = false;
1162                 if (closeLanguage)
1163                         runparams.local_font = &basefont;
1164         }
1165
1166         if (fontswitch_inset && !closeLanguage) {
1167                 // The directionality has been switched at inset.
1168                 // Force markup inside.
1169                 runparams.local_font = &basefont;
1170         }
1171
1172         size_t const previous_row_count = os.texrow().rows();
1173
1174         try {
1175                 runparams.lastid = id_;
1176                 runparams.lastpos = i;
1177                 inset->latex(os, runparams);
1178         } catch (EncodingException & e) {
1179                 // add location information and throw again.
1180                 e.par_id = id_;
1181                 e.pos = i;
1182                 throw;
1183         }
1184
1185         if (close)
1186                 os << '}';
1187
1188         if (os.texrow().rows() > previous_row_count) {
1189                 os.texrow().start(owner_->id(), i + 1);
1190                 column = 0;
1191         } else {
1192                 column += (unsigned int)(os.os().tellp() - len);
1193         }
1194
1195         if (owner_->isDeleted(i))
1196                 --runparams.inDeletedInset;
1197 }
1198
1199
1200 void Paragraph::Private::latexSpecialChar(otexstream & os,
1201                                           BufferParams const & bparams,
1202                                           OutputParams const & runparams,
1203                                           Font const & running_font,
1204                                           string & alien_script,
1205                                           Layout const & style,
1206                                           pos_type & i,
1207                                           pos_type end_pos,
1208                                           unsigned int & column) const
1209 {
1210         char_type const c = owner_->getUChar(bparams, runparams, i);
1211
1212         if (style.pass_thru || runparams.pass_thru || (runparams.for_searchAdv != OutputParams::NoSearch)
1213             || contains(style.pass_thru_chars, c)
1214             || contains(runparams.pass_thru_chars, c)) {
1215                 if (runparams.for_searchAdv != OutputParams::NoSearch) {
1216                         if (c == '\\')
1217                                 os << "\\\\";
1218                         else if (c == '{')
1219                                 os << "\\braceleft ";
1220                         else if (c == '}')
1221                                 os << "\\braceright ";
1222                         else if (c != '\0')
1223                                 os.put(c);
1224                 }
1225                 else if (c != '\0') {
1226                         Encoding const * const enc = runparams.encoding;
1227                         if (enc && !enc->encodable(c))
1228                                 throw EncodingException(c);
1229                         os.put(c);
1230                 }
1231                 return;
1232         }
1233
1234         // TIPA uses its own T3 encoding
1235         if (runparams.inIPA && latexSpecialT3(c, os, i, column))
1236                 return;
1237         // If T1 font encoding is used, use the special
1238         // characters it provides.
1239         // NOTE: Some languages reset the font encoding internally to a
1240         //       non-standard font encoding. If we are using such a language,
1241         //       we do not output special T1 chars.
1242         if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1243             && !runparams.isFullUnicode() && runparams.main_fontenc == "T1"
1244             && latexSpecialT1(c, os, i, column))
1245                 return;
1246         // NOTE: "fontspec" (non-TeX fonts) sets the font encoding to "TU" (untill 2017 "EU1" or "EU2")
1247         else if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1248                  && runparams.isFullUnicode() && latexSpecialTU(c, os, i, column))
1249                      return;
1250
1251         // Otherwise, we use what LaTeX provides us.
1252         switch (c) {
1253         case '\\':
1254                 os << "\\textbackslash" << termcmd;
1255                 column += 15;
1256                 break;
1257         case '<':
1258                 os << "\\textless" << termcmd;
1259                 column += 10;
1260                 break;
1261         case '>':
1262                 os << "\\textgreater" << termcmd;
1263                 column += 13;
1264                 break;
1265         case '|':
1266                 os << "\\textbar" << termcmd;
1267                 column += 9;
1268                 break;
1269         case '-':
1270                 os << '-';
1271                 if (i + 1 < static_cast<pos_type>(text_.size()) &&
1272                     (end_pos == -1 || i + 1 < end_pos) &&
1273                     text_[i+1] == '-') {
1274                         // Prevent "--" becoming an en dash and "---" an em dash.
1275                         // (Within \ttfamily, "---" is merged to en dash + hyphen.)
1276                         os << "{}";
1277                         column += 2;
1278                 }
1279                 break;
1280         case '\"':
1281                 os << "\\textquotedbl" << termcmd;
1282                 column += 14;
1283                 break;
1284
1285         case '$': case '&':
1286         case '%': case '#': case '{':
1287         case '}': case '_':
1288                 os << '\\';
1289                 os.put(c);
1290                 column += 1;
1291                 break;
1292
1293         case '~':
1294                 os << "\\textasciitilde" << termcmd;
1295                 column += 16;
1296                 break;
1297
1298         case '^':
1299                 os << "\\textasciicircum" << termcmd;
1300                 column += 17;
1301                 break;
1302
1303         case '*':
1304         case '[':
1305         case ']':
1306                 // avoid being mistaken for optional arguments
1307                 os << '{';
1308                 os.put(c);
1309                 os << '}';
1310                 column += 2;
1311                 break;
1312
1313         case ' ':
1314                 // Blanks are printed before font switching.
1315                 // Sure? I am not! (try nice-latex)
1316                 // I am sure it's correct. LyX might be smarter
1317                 // in the future, but for now, nothing wrong is
1318                 // written. (Asger)
1319                 break;
1320
1321         case 0x2013:
1322         case 0x2014:
1323                 // XeTeX's dash behaviour is determined via a global setting
1324                 if (bparams.use_dash_ligatures
1325                     && owner_->getFontSettings(bparams, i).fontInfo().family() != TYPEWRITER_FAMILY
1326                     && !runparams.inIPA
1327                         // TODO #10961: && not in inset Flex Code
1328                         // TODO #10961: && not in layout LyXCode
1329                     && (!bparams.useNonTeXFonts || runparams.flavor != Flavor::XeTeX)) {
1330                         if (c == 0x2013) {
1331                                 // en-dash
1332                                 os << "--";
1333                                 column +=2;
1334                         } else {
1335                                 // em-dash
1336                                 os << "---";
1337                                 column +=3;
1338                         }
1339                         break;
1340                 }
1341                 // fall through
1342         default:
1343                 if (c == '\0')
1344                         return;
1345
1346                 Encoding const & encoding = *(runparams.encoding);
1347                 char_type next = '\0';
1348                 if (i + 1 < int(text_.size())) {
1349                         next = text_[i + 1];
1350                         if (Encodings::isCombiningChar(next)) {
1351                                 column += latexSurrogatePair(bparams, os, c, next, runparams) - 1;
1352                                 ++i;
1353                                 break;
1354                         }
1355                 }
1356                 pair<docstring, bool> latex = encoding.latexChar(c);
1357                 docstring nextlatex;
1358                 bool nexttipas = false;
1359                 string nexttipashortcut;
1360                 if (next != '\0' && next != META_INSET && !encoding.encodable(next)) {
1361                         nextlatex = encoding.latexChar(next).first;
1362                         if (runparams.inIPA) {
1363                                 nexttipashortcut = Encodings::TIPAShortcut(next);
1364                                 nexttipas = !nexttipashortcut.empty();
1365                         }
1366                 }
1367                 bool tipas = false;
1368                 if (runparams.inIPA) {
1369                         string const tipashortcut = Encodings::TIPAShortcut(c);
1370                         if (!tipashortcut.empty()) {
1371                                 latex.first = from_ascii(tipashortcut);
1372                                 latex.second = false;
1373                                 tipas = true;
1374                         }
1375                 }
1376                 // eventually close "script wrapper" command (see `Paragraph::latex`)
1377                 if (!alien_script.empty()
1378                         && alien_script != Encodings::isKnownScriptChar(next)) {
1379                         column += latex.first.length();
1380                         alien_script.clear();
1381                         os << latex.first << "}";
1382                         break;
1383                 }
1384                 if (latex.second
1385                          && ((!prefixIs(nextlatex, '\\')
1386                                && !prefixIs(nextlatex, '{')
1387                                && !prefixIs(nextlatex, '}'))
1388                              || (nexttipas
1389                                  && !prefixIs(from_ascii(nexttipashortcut), '\\')))
1390                          && !tipas) {
1391                         // Prevent eating of a following space or command corruption by
1392                         // following characters
1393                         if (next == ' ' || next == '\0') {
1394                                 column += latex.first.length() + 1;
1395                                 os << latex.first << "{}";
1396                         } else {
1397                                 column += latex.first.length();
1398                                 os << latex.first << " ";
1399                         }
1400                 } else {
1401                         column += latex.first.length() - 1;
1402                         os << latex.first;
1403                 }
1404                 break;
1405         }
1406 }
1407
1408
1409 bool Paragraph::Private::latexSpecialT1(char_type const c, otexstream & os,
1410         pos_type i, unsigned int & column) const
1411 {
1412         switch (c) {
1413         case '>':
1414         case '<':
1415                 os.put(c);
1416                 // In T1 encoding, these characters exist
1417                 // but we should avoid ligatures
1418                 if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
1419                         return true;
1420                 os << "\\textcompwordmark" << termcmd;
1421                 column += 19;
1422                 return true;
1423         case '|':
1424                 os.put(c);
1425                 return true;
1426         case '\"':
1427                 // soul.sty breaks with \char`\"
1428                 os << "\\textquotedbl" << termcmd;
1429                 column += 14;
1430                 return true;
1431         default:
1432                 return false;
1433         }
1434 }
1435
1436
1437 bool Paragraph::Private::latexSpecialTU(char_type const c, otexstream & os,
1438         pos_type i, unsigned int & column) const
1439 {
1440         // TU encoding is currently on par with T1.
1441         return latexSpecialT1(c, os, i, column);
1442 }
1443
1444
1445 bool Paragraph::Private::latexSpecialT3(char_type const c, otexstream & os,
1446         pos_type /*i*/, unsigned int & column) const
1447 {
1448         switch (c) {
1449         case '*':
1450         case '[':
1451         case ']':
1452         case '\"':
1453                 os.put(c);
1454                 return true;
1455         case '|':
1456                 os << "\\textvertline" << termcmd;
1457                 column += 14;
1458                 return true;
1459         default:
1460                 return false;
1461         }
1462 }
1463
1464
1465 void Paragraph::Private::validate(LaTeXFeatures & features) const
1466 {
1467         if (layout_->inpreamble && inset_owner_) {
1468                 // FIXME: Using a string stream here circumvents the encoding
1469                 // switching machinery of odocstream. Therefore the
1470                 // output is wrong if this paragraph contains content
1471                 // that needs to switch encoding.
1472                 Buffer const & buf = inset_owner_->buffer();
1473                 otexstringstream os;
1474                 os << layout_->preamble();
1475                 size_t const length = os.length();
1476                 TeXOnePar(buf, *inset_owner_->getText(int(buf.getParFromID(owner_->id()).idx())),
1477                           buf.getParFromID(owner_->id()).pit(), os,
1478                           features.runparams(), string(), 0, -1, true);
1479                 if (os.length() > length)
1480                         features.addPreambleSnippet(os.release(), true);
1481         }
1482
1483         if (features.runparams().flavor == Flavor::Html
1484             && layout_->htmltitle()) {
1485                 features.setHTMLTitle(owner_->asString(AS_STR_INSETS | AS_STR_SKIPDELETE));
1486         }
1487
1488         // check the params.
1489         if (!params_.spacing().isDefault())
1490                 features.require("setspace");
1491
1492         // then the layouts
1493         features.useLayout(layout_->name());
1494
1495         // then the fonts
1496         fontlist_.validate(features);
1497
1498         // then the indentation
1499         if (!params_.leftIndent().zero())
1500                 features.require("ParagraphLeftIndent");
1501
1502         // then the insets
1503         InsetList::const_iterator icit = insetlist_.begin();
1504         InsetList::const_iterator iend = insetlist_.end();
1505         for (; icit != iend; ++icit) {
1506                 if (icit->inset) {
1507                         features.inDeletedInset(owner_->isDeleted(icit->pos));
1508                         if (icit->inset->lyxCode() == FOOT_CODE) {
1509                                 // FIXME: an item inset would make things much easier.
1510                                 if ((layout_->latextype == LATEX_LIST_ENVIRONMENT
1511                                      || (layout_->latextype == LATEX_ITEM_ENVIRONMENT
1512                                          && layout_->margintype == MARGIN_FIRST_DYNAMIC))
1513                                     && (icit->pos < begin_of_body_
1514                                         || (icit->pos == begin_of_body_
1515                                             && (icit->pos == 0 || text_[icit->pos - 1] != ' '))))
1516                                         features.saveNoteEnv("description");
1517                         }
1518                         icit->inset->validate(features);
1519                         features.inDeletedInset(false);
1520                         if (layout_->needprotect &&
1521                             icit->inset->lyxCode() == FOOT_CODE)
1522                                 features.require("NeedLyXFootnoteCode");
1523                 }
1524         }
1525
1526         // then the contents
1527         BufferParams const bp = features.runparams().is_child
1528                 ? features.buffer().masterParams() : features.buffer().params();
1529         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1530                 char_type c = text_[i];
1531                 CharInfo const & ci = Encodings::unicodeCharInfo(c);
1532                 if (c == 0x0022) {
1533                         if (features.runparams().isFullUnicode() && bp.useNonTeXFonts)
1534                                 features.require("textquotedblp");
1535                         else if (features.runparams().main_fontenc != "T1"
1536                                  || ((&owner_->getFontSettings(bp, i))->language()->internalFontEncoding()))
1537                                 features.require("textquotedbl");
1538                 } else if (ci.textfeature() && contains(ci.textpreamble(), '=')) {
1539                         // features that depend on the font or input encoding
1540                         string feats = ci.textpreamble();
1541                         string fontenc = (&owner_->getFontSettings(bp, i))->language()->fontenc(bp);
1542                         if (fontenc.empty())
1543                                 fontenc = features.runparams().main_fontenc;
1544                         while (!feats.empty()) {
1545                                 string feat;
1546                                 feats = split(feats, feat, ',');
1547                                 if (contains(feat, "!=")) {
1548                                         // a feature that is required except for the spcified
1549                                         // font or input encodings
1550                                         string realfeature;
1551                                         string const contexts = ltrim(split(feat, realfeature, '!'), "=");
1552                                         // multiple encodings are separated by semicolon
1553                                         vector<string> context = getVectorFromString(contexts, ";");
1554                                         // require feature if the context matches neither current font
1555                                         // nor input encoding
1556                                         if (std::find(context.begin(), context.end(), fontenc) == context.end()
1557                                             && std::find(context.begin(), context.end(),
1558                                                          features.runparams().encoding->name()) == context.end())
1559                                                 features.require(realfeature);
1560                                 } else if (contains(feat, '=')) {
1561                                         // a feature that is required only for the spcified
1562                                         // font or input encodings
1563                                         string realfeature;
1564                                         string const contexts = split(feat, realfeature, '=');
1565                                         // multiple encodings are separated by semicolon
1566                                         vector<string> context = getVectorFromString(contexts, ";");
1567                                         // require feature if the context matches either current font
1568                                         // or input encoding
1569                                         if (std::find(context.begin(), context.end(), fontenc) != context.end()
1570                                             || std::find(context.begin(), context.end(),
1571                                                          features.runparams().encoding->name()) != context.end())
1572                                                 features.require(realfeature);
1573                                 }
1574                         }
1575                 } else if (!bp.use_dash_ligatures
1576                            && (c == 0x2013 || c == 0x2014)
1577                            && bp.useNonTeXFonts
1578                            && features.runparams().flavor == Flavor::XeTeX)
1579                         // XeTeX's dash behaviour is determined via a global setting
1580                         features.require("xetexdashbreakstate");
1581                 BufferEncodings::validate(c, features);
1582         }
1583 }
1584
1585 /////////////////////////////////////////////////////////////////////
1586 //
1587 // Paragraph
1588 //
1589 /////////////////////////////////////////////////////////////////////
1590
1591 namespace {
1592         Layout const emptyParagraphLayout;
1593 }
1594
1595 Paragraph::Paragraph()
1596         : d(new Paragraph::Private(this, emptyParagraphLayout))
1597 {
1598         itemdepth = 0;
1599         d->params_.clear();
1600 }
1601
1602
1603 Paragraph::Paragraph(Paragraph const & par)
1604         : itemdepth(par.itemdepth),
1605         d(new Paragraph::Private(*par.d, this))
1606 {
1607         registerWords();
1608 }
1609
1610
1611 Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
1612         : itemdepth(par.itemdepth),
1613         d(new Paragraph::Private(*par.d, this, beg, end))
1614 {
1615         registerWords();
1616 }
1617
1618
1619 Paragraph & Paragraph::operator=(Paragraph const & par)
1620 {
1621         // needed as we will destroy the private part before copying it
1622         if (&par != this) {
1623                 itemdepth = par.itemdepth;
1624
1625                 deregisterWords();
1626                 delete d;
1627                 d = new Private(*par.d, this);
1628                 registerWords();
1629         }
1630         return *this;
1631 }
1632
1633
1634 Paragraph::~Paragraph()
1635 {
1636         deregisterWords();
1637         delete d;
1638 }
1639
1640
1641 namespace {
1642
1643 // this shall be called just before every "os << ..." action.
1644 void flushString(ostream & os, docstring & s)
1645 {
1646         os << to_utf8(s);
1647         s.erase();
1648 }
1649
1650 } // namespace
1651
1652
1653 void Paragraph::write(ostream & os, BufferParams const & bparams,
1654         depth_type & depth) const
1655 {
1656         // The beginning or end of a deeper (i.e. nested) area?
1657         if (depth != d->params_.depth()) {
1658                 if (d->params_.depth() > depth) {
1659                         while (d->params_.depth() > depth) {
1660                                 os << "\n\\begin_deeper";
1661                                 ++depth;
1662                         }
1663                 } else {
1664                         while (d->params_.depth() < depth) {
1665                                 os << "\n\\end_deeper";
1666                                 --depth;
1667                         }
1668                 }
1669         }
1670
1671         // First write the layout
1672         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1673
1674         d->params_.write(os);
1675
1676         Font font1(inherit_font, bparams.language);
1677
1678         Change running_change = Change(Change::UNCHANGED);
1679
1680         // this string is used as a buffer to avoid repetitive calls
1681         // to to_utf8(), which turn out to be expensive (JMarc)
1682         docstring write_buffer;
1683
1684         int column = 0;
1685         for (pos_type i = 0; i <= size(); ++i) {
1686
1687                 Change const & change = lookupChange(i);
1688                 if (change != running_change)
1689                         flushString(os, write_buffer);
1690                 Changes::lyxMarkChange(os, bparams, column, running_change, change);
1691                 running_change = change;
1692
1693                 if (i == size())
1694                         break;
1695
1696                 // Write font changes
1697                 Font font2 = getFontSettings(bparams, i);
1698                 if (font2 != font1) {
1699                         flushString(os, write_buffer);
1700                         font2.lyxWriteChanges(font1, os);
1701                         column = 0;
1702                         font1 = font2;
1703                 }
1704
1705                 char_type const c = d->text_[i];
1706                 switch (c) {
1707                 case META_INSET:
1708                         if (Inset const * inset = getInset(i)) {
1709                                 flushString(os, write_buffer);
1710                                 if (inset->directWrite()) {
1711                                         // international char, let it write
1712                                         // code directly so it's shorter in
1713                                         // the file
1714                                         inset->write(os);
1715                                 } else {
1716                                         if (i)
1717                                                 os << '\n';
1718                                         os << "\\begin_inset ";
1719                                         inset->write(os);
1720                                         os << "\n\\end_inset\n\n";
1721                                         column = 0;
1722                                 }
1723                                 // FIXME This can be removed again once the mystery
1724                                 // crash has been resolved.
1725                                 os << flush;
1726                         }
1727                         break;
1728                 case '\\':
1729                         flushString(os, write_buffer);
1730                         os << "\n\\backslash\n";
1731                         column = 0;
1732                         break;
1733                 case '.':
1734                         flushString(os, write_buffer);
1735                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1736                                 os << ".\n";
1737                                 column = 0;
1738                         } else
1739                                 os << '.';
1740                         break;
1741                 default:
1742                         if ((column > 70 && c == ' ')
1743                             || column > 79) {
1744                                 flushString(os, write_buffer);
1745                                 os << '\n';
1746                                 column = 0;
1747                         }
1748                         // this check is to amend a bug. LyX sometimes
1749                         // inserts '\0' this could cause problems.
1750                         if (c != '\0')
1751                                 write_buffer.push_back(c);
1752                         else
1753                                 LYXERR0("NUL char in structure.");
1754                         ++column;
1755                         break;
1756                 }
1757         }
1758
1759         flushString(os, write_buffer);
1760         os << "\n\\end_layout\n";
1761         // FIXME This can be removed again once the mystery
1762         // crash has been resolved.
1763         os << flush;
1764 }
1765
1766
1767 void Paragraph::validate(LaTeXFeatures & features) const
1768 {
1769         d->validate(features);
1770         bool fragile = features.runparams().moving_arg;
1771         fragile |= layout().needprotect;
1772         if (inInset().getLayout().isNeedProtect())
1773                 fragile = true;
1774         if (needsCProtection(fragile))
1775                 features.require("cprotect");
1776 }
1777
1778
1779 void Paragraph::insert(pos_type pos, docstring const & str,
1780                        Font const & font, Change const & change)
1781 {
1782         for (size_t i = 0, n = str.size(); i != n ; ++i)
1783                 insertChar(pos + i, str[i], font, change);
1784 }
1785
1786
1787 void Paragraph::appendChar(char_type c, Font const & font,
1788                 Change const & change)
1789 {
1790         // Make sure that Buffer::hasChangesPresent is updated
1791         ChangesMonitor cm(*this);
1792
1793         // track change
1794         d->changes_.insert(change, d->text_.size());
1795         // when appending characters, no need to update tables
1796         d->text_.push_back(c);
1797         setFont(d->text_.size() - 1, font);
1798         d->requestSpellCheck(d->text_.size() - 1);
1799 }
1800
1801
1802 void Paragraph::appendString(docstring const & s, Font const & font,
1803                 Change const & change)
1804 {
1805         // Make sure that Buffer::hasChangesPresent is updated
1806         ChangesMonitor cm(*this);
1807
1808         pos_type end = s.size();
1809         size_t oldsize = d->text_.size();
1810         size_t newsize = oldsize + end;
1811         size_t capacity = d->text_.capacity();
1812         if (newsize >= capacity)
1813                 d->text_.reserve(max(capacity + 100, newsize));
1814
1815         // when appending characters, no need to update tables
1816         d->text_.append(s);
1817
1818         // FIXME: Optimize this!
1819         for (size_t i = oldsize; i != newsize; ++i) {
1820                 // track change
1821                 d->changes_.insert(change, i);
1822                 d->requestSpellCheck(i);
1823         }
1824         d->fontlist_.set(oldsize, font);
1825         d->fontlist_.set(newsize - 1, font);
1826 }
1827
1828
1829 void Paragraph::insertChar(pos_type pos, char_type c,
1830                            bool trackChanges)
1831 {
1832         d->insertChar(pos, c, Change(trackChanges ?
1833                            Change::INSERTED : Change::UNCHANGED));
1834 }
1835
1836
1837 void Paragraph::insertChar(pos_type pos, char_type c,
1838                            Font const & font, bool trackChanges)
1839 {
1840         d->insertChar(pos, c, Change(trackChanges ?
1841                            Change::INSERTED : Change::UNCHANGED));
1842         setFont(pos, font);
1843 }
1844
1845
1846 void Paragraph::insertChar(pos_type pos, char_type c,
1847                            Font const & font, Change const & change)
1848 {
1849         d->insertChar(pos, c, change);
1850         setFont(pos, font);
1851 }
1852
1853
1854 void Paragraph::resetFonts(Font const & font)
1855 {
1856         d->fontlist_.clear();
1857         d->fontlist_.set(0, font);
1858         d->fontlist_.set(d->text_.size() - 1, font);
1859 }
1860
1861 // Gets uninstantiated font setting at position.
1862 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1863                                          pos_type pos) const
1864 {
1865         if (pos > size()) {
1866                 LYXERR0("pos: " << pos << " size: " << size());
1867                 LBUFERR(false);
1868         }
1869
1870         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1871         if (cit != d->fontlist_.end())
1872                 return cit->font();
1873
1874         if (pos == size() && !empty())
1875                 return getFontSettings(bparams, pos - 1);
1876
1877         // Optimisation: avoid a full font instantiation if there is no
1878         // language change from previous call.
1879         static Font previous_font;
1880         static Language const * previous_lang = nullptr;
1881         Language const * lang = getParLanguage(bparams);
1882         if (lang != previous_lang) {
1883                 previous_lang = lang;
1884                 previous_font = Font(inherit_font, lang);
1885         }
1886         return previous_font;
1887 }
1888
1889
1890 FontSpan Paragraph::fontSpan(pos_type pos) const
1891 {
1892         LBUFERR(pos <= size());
1893
1894         if (pos == size())
1895                 return FontSpan(pos, pos);
1896
1897         pos_type start = 0;
1898         FontList::const_iterator cit = d->fontlist_.begin();
1899         FontList::const_iterator end = d->fontlist_.end();
1900         for (; cit != end; ++cit) {
1901                 if (cit->pos() >= pos) {
1902                         if (pos >= beginOfBody())
1903                                 return FontSpan(max(start, beginOfBody()),
1904                                                 cit->pos());
1905                         else
1906                                 return FontSpan(start,
1907                                                 min(beginOfBody() - 1,
1908                                                          cit->pos()));
1909                 }
1910                 start = cit->pos() + 1;
1911         }
1912
1913         // This should not happen, but if so, we take no chances.
1914         LYXERR0("Paragraph::fontSpan: position not found in fontinfo table!");
1915         LASSERT(false, return FontSpan(pos, pos));
1916 }
1917
1918
1919 // Gets uninstantiated font setting at position 0
1920 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1921 {
1922         if (!empty() && !d->fontlist_.empty())
1923                 return d->fontlist_.begin()->font();
1924
1925         // Optimisation: avoid a full font instantiation if there is no
1926         // language change from previous call.
1927         static Font previous_font;
1928         static Language const * previous_lang = nullptr;
1929         if (bparams.language != previous_lang) {
1930                 previous_lang = bparams.language;
1931                 previous_font = Font(inherit_font, bparams.language);
1932         }
1933
1934         return previous_font;
1935 }
1936
1937
1938 // Gets the fully instantiated font at a given position in a paragraph
1939 // This is basically the same function as Text::GetFont() in text2.cpp.
1940 // The difference is that this one is used for generating the LaTeX file,
1941 // and thus cosmetic "improvements" are disallowed: This has to deliver
1942 // the true picture of the buffer. (Asger)
1943 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1944                                  Font const & outerfont) const
1945 {
1946         LBUFERR(pos >= 0);
1947
1948         Font font = getFontSettings(bparams, pos);
1949
1950         pos_type const body_pos = beginOfBody();
1951         FontInfo & fi = font.fontInfo();
1952         if (pos < body_pos)
1953                 fi.realize(d->layout_->labelfont);
1954         else
1955                 fi.realize(d->layout_->font);
1956
1957         fi.realize(outerfont.fontInfo());
1958         fi.realize(bparams.getFont().fontInfo());
1959
1960         return font;
1961 }
1962
1963
1964 Font const Paragraph::getLabelFont
1965         (BufferParams const & bparams, Font const & outerfont) const
1966 {
1967         FontInfo tmpfont = d->layout_->labelfont;
1968         tmpfont.realize(outerfont.fontInfo());
1969         tmpfont.realize(bparams.getFont().fontInfo());
1970         return Font(tmpfont, getParLanguage(bparams));
1971 }
1972
1973
1974 Font const Paragraph::getLayoutFont
1975         (BufferParams const & bparams, Font const & outerfont) const
1976 {
1977         FontInfo tmpfont = d->layout_->font;
1978         tmpfont.realize(outerfont.fontInfo());
1979         tmpfont.realize(bparams.getFont().fontInfo());
1980         return Font(tmpfont, getParLanguage(bparams));
1981 }
1982
1983
1984 char_type Paragraph::getUChar(BufferParams const & bparams,
1985                               OutputParams const & rp,
1986                               pos_type pos) const
1987 {
1988         char_type c = d->text_[pos];
1989
1990         // Return unchanged character in LTR languages
1991         // or if we use poylglossia/bidi (XeTeX).
1992         if (rp.useBidiPackage()
1993             || !getFontSettings(bparams, pos).isRightToLeft())
1994                 return c;
1995
1996         // Without polyglossia/bidi, we need to account for some special cases.
1997         // FIXME This needs to be audited!
1998         // Check if:
1999         // * The input is as expected for all delimiters
2000         //   => checked for Hebrew!
2001         // * The output matches the display in the LyX workarea
2002         //   => checked for Hebrew!
2003         // * The special cases below are really necessary
2004         //   => checked for Hebrew!
2005         // * In arabic_arabi, brackets are transformed to Arabic
2006         //   Ornate Parentheses. Is this is really wanted?
2007
2008         string const & lang = getFontSettings(bparams, pos).language()->lang();
2009         char_type uc = c;
2010
2011         // 1. In the following languages, parentheses need to be reversed.
2012         //    Also with polyglodia/luabidi
2013         bool const reverseparens = (lang == "hebrew" || rp.use_polyglossia);
2014
2015         // 2. In the following languages, brackets don't need to be reversed.
2016         bool const reversebrackets = lang != "arabic_arabtex"
2017                         && lang != "arabic_arabi"
2018                         && lang != "farsi";
2019
2020         // Now swap delimiters if needed.
2021         switch (c) {
2022         case '(':
2023                 if (reverseparens)
2024                         uc = ')';
2025                 break;
2026         case ')':
2027                 if (reverseparens)
2028                         uc = '(';
2029                 break;
2030         case '[':
2031                 if (reversebrackets)
2032                         uc = ']';
2033                 break;
2034         case ']':
2035                 if (reversebrackets)
2036                         uc = '[';
2037                 break;
2038         case '{':
2039                 uc = '}';
2040                 break;
2041         case '}':
2042                 uc = '{';
2043                 break;
2044         case '<':
2045                 uc = '>';
2046                 break;
2047         case '>':
2048                 uc = '<';
2049                 break;
2050         }
2051
2052         return uc;
2053 }
2054
2055
2056 void Paragraph::setFont(pos_type pos, Font const & font)
2057 {
2058         LASSERT(pos <= size(), return);
2059
2060         // First, reduce font against layout/label font
2061         // Update: The setCharFont() routine in text2.cpp already
2062         // reduces font, so we don't need to do that here. (Asger)
2063
2064         d->fontlist_.set(pos, font);
2065 }
2066
2067
2068 void Paragraph::makeSameLayout(Paragraph const & par)
2069 {
2070         d->layout_ = par.d->layout_;
2071         d->params_ = par.d->params_;
2072 }
2073
2074
2075 bool Paragraph::stripLeadingSpaces(bool trackChanges)
2076 {
2077         if (isFreeSpacing())
2078                 return false;
2079
2080         int pos = 0;
2081         int count = 0;
2082
2083         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
2084                 if (eraseChar(pos, trackChanges))
2085                         ++count;
2086                 else
2087                         ++pos;
2088         }
2089
2090         return count > 0 || pos > 0;
2091 }
2092
2093
2094 bool Paragraph::hasSameLayout(Paragraph const & par) const
2095 {
2096         return par.d->layout_ == d->layout_
2097                 && d->params_.sameLayout(par.d->params_);
2098 }
2099
2100
2101 depth_type Paragraph::getDepth() const
2102 {
2103         return d->params_.depth();
2104 }
2105
2106
2107 depth_type Paragraph::getMaxDepthAfter() const
2108 {
2109         if (d->layout_->isEnvironment())
2110                 return d->params_.depth() + 1;
2111         else
2112                 return d->params_.depth();
2113 }
2114
2115
2116 LyXAlignment Paragraph::getAlign(BufferParams const & bparams) const
2117 {
2118         if (d->params_.align() == LYX_ALIGN_LAYOUT)
2119                 return getDefaultAlign(bparams);
2120         else
2121                 return d->params_.align();
2122 }
2123
2124
2125 LyXAlignment Paragraph::getDefaultAlign(BufferParams const & bparams) const
2126 {
2127         LyXAlignment res = layout().align;
2128         if (isRTL(bparams)) {
2129                 // Swap sides
2130                 if (res == LYX_ALIGN_LEFT)
2131                         res = LYX_ALIGN_RIGHT;
2132                 else if  (res == LYX_ALIGN_RIGHT)
2133                         res = LYX_ALIGN_LEFT;
2134         }
2135         return res;
2136 }
2137
2138
2139 docstring const & Paragraph::labelString() const
2140 {
2141         return d->params_.labelString();
2142 }
2143
2144
2145 // the next two functions are for the manual labels
2146 docstring const Paragraph::getLabelWidthString() const
2147 {
2148         if (d->layout_->margintype == MARGIN_MANUAL
2149             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
2150                 return d->params_.labelWidthString();
2151         else
2152                 return _("Senseless with this layout!");
2153 }
2154
2155
2156 void Paragraph::setLabelWidthString(docstring const & s)
2157 {
2158         d->params_.labelWidthString(s);
2159 }
2160
2161
2162 docstring Paragraph::expandLabel(Layout const & layout,
2163                 BufferParams const & bparams) const
2164 {
2165         return expandParagraphLabel(layout, bparams, true);
2166 }
2167
2168
2169 docstring Paragraph::expandParagraphLabel(Layout const & layout,
2170                 BufferParams const & bparams, bool process_appendix) const
2171 {
2172         DocumentClass const & tclass = bparams.documentClass();
2173         string const & lang = getParLanguage(bparams)->code();
2174         bool const in_appendix = process_appendix && d->params_.appendix();
2175         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
2176
2177         if (fmt.empty() && !layout.counter.empty())
2178                 return tclass.counters().theCounter(layout.counter, lang);
2179
2180         // handle 'inherited level parts' in 'fmt',
2181         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
2182         size_t const i = fmt.find('@', 0);
2183         if (i != docstring::npos) {
2184                 size_t const j = fmt.find('@', i + 1);
2185                 if (j != docstring::npos) {
2186                         docstring parent(fmt, i + 1, j - i - 1);
2187                         docstring label = from_ascii("??");
2188                         if (tclass.hasLayout(parent))
2189                                 label = expandParagraphLabel(tclass[parent], bparams,
2190                                                       process_appendix);
2191                         fmt = docstring(fmt, 0, i) + label
2192                                 + docstring(fmt, j + 1, docstring::npos);
2193                 }
2194         }
2195
2196         return tclass.counters().counterLabel(fmt, lang);
2197 }
2198
2199
2200 void Paragraph::applyLayout(Layout const & new_layout)
2201 {
2202         d->layout_ = &new_layout;
2203         LyXAlignment const oldAlign = d->params_.align();
2204
2205         if (!(oldAlign & d->layout_->alignpossible)) {
2206                 frontend::Alert::warning(_("Alignment not permitted"),
2207                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
2208                 d->params_.align(LYX_ALIGN_LAYOUT);
2209         }
2210 }
2211
2212
2213 pos_type Paragraph::beginOfBody() const
2214 {
2215         return d->begin_of_body_;
2216 }
2217
2218
2219 void Paragraph::setBeginOfBody()
2220 {
2221         if (d->layout_->labeltype != LABEL_MANUAL) {
2222                 d->begin_of_body_ = 0;
2223                 return;
2224         }
2225
2226         // Unroll the first two cycles of the loop
2227         // and remember the previous character to
2228         // remove unnecessary getChar() calls
2229         pos_type i = 0;
2230         pos_type end = size();
2231         bool prev_char_deleted = false;
2232         if (i < end && (!(isNewline(i) || isEnvSeparator(i)) || isDeleted(i))) {
2233                 ++i;
2234                 if (i < end) {
2235                         char_type previous_char = d->text_[i];
2236                         if (!(isNewline(i) || isEnvSeparator(i))) {
2237                                 ++i;
2238                                 while (i < end && (previous_char != ' ' || prev_char_deleted)) {
2239                                         char_type temp = d->text_[i];
2240                                         prev_char_deleted = isDeleted(i);
2241                                         if (!isDeleted(i) && (isNewline(i) || isEnvSeparator(i)))
2242                                                 break;
2243                                         ++i;
2244                                         previous_char = temp;
2245                                 }
2246                         }
2247                 }
2248         }
2249
2250         d->begin_of_body_ = i;
2251 }
2252
2253
2254 bool Paragraph::allowParagraphCustomization() const
2255 {
2256         return inInset().allowParagraphCustomization();
2257 }
2258
2259
2260 bool Paragraph::usePlainLayout() const
2261 {
2262         return inInset().usePlainLayout();
2263 }
2264
2265
2266 bool Paragraph::isPassThru() const
2267 {
2268         return inInset().isPassThru() || d->layout_->pass_thru;
2269 }
2270
2271
2272 bool Paragraph::parbreakIsNewline() const
2273 {
2274         return inInset().getLayout().parbreakIsNewline() || d->layout_->parbreak_is_newline;
2275 }
2276
2277
2278 bool Paragraph::isPartOfTextSequence() const
2279 {
2280         for (pos_type i = 0; i < size(); ++i) {
2281                 if (!isInset(i) || getInset(i)->isPartOfTextSequence())
2282                         return true;
2283         }
2284         return false;
2285 }
2286
2287 namespace {
2288
2289 // paragraphs inside floats need different alignment tags to avoid
2290 // unwanted space
2291
2292 bool noTrivlistCentering(InsetCode code)
2293 {
2294         return code == FLOAT_CODE
2295                || code == WRAP_CODE
2296                || code == CELL_CODE;
2297 }
2298
2299
2300 string correction(string const & orig)
2301 {
2302         if (orig == "flushleft")
2303                 return "raggedright";
2304         if (orig == "flushright")
2305                 return "raggedleft";
2306         if (orig == "center")
2307                 return "centering";
2308         return orig;
2309 }
2310
2311
2312 bool corrected_env(otexstream & os, string const & suffix, string const & env,
2313         InsetCode code, bool const lastpar, int & col)
2314 {
2315         string macro = suffix + "{";
2316         if (noTrivlistCentering(code)) {
2317                 if (lastpar) {
2318                         // the last paragraph in non-trivlist-aligned
2319                         // context is special (to avoid unwanted whitespace)
2320                         if (suffix == "\\begin") {
2321                                 macro = "\\" + correction(env) + "{}";
2322                                 os << from_ascii(macro);
2323                                 col += macro.size();
2324                                 return true;
2325                         }
2326                         return false;
2327                 }
2328                 macro += correction(env);
2329         } else
2330                 macro += env;
2331         macro += "}";
2332         if (suffix == "\\par\\end") {
2333                 os << breakln;
2334                 col = 0;
2335         }
2336         os << from_ascii(macro);
2337         col += macro.size();
2338         if (suffix == "\\begin") {
2339                 os << breakln;
2340                 col = 0;
2341         }
2342         return true;
2343 }
2344
2345 } // namespace
2346
2347
2348 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
2349                         otexstream & os, OutputParams const & runparams) const
2350 {
2351         int column = 0;
2352
2353         bool canindent =
2354                 (bparams.paragraph_separation == BufferParams::ParagraphIndentSeparation) ?
2355                         (layout_->toggle_indent != ITOGGLE_NEVER) :
2356                         (layout_->toggle_indent == ITOGGLE_ALWAYS);
2357
2358         LyXAlignment const curAlign = params_.align();
2359
2360         // Do not output \\noindent for paragraphs
2361         // 1. that cannot have indentation or are indented always,
2362         // 2. that are not part of the immediate text sequence (e.g., contain only floats),
2363         // 3. that are PassThru,
2364         // 4. or that are centered.
2365         if (canindent && params_.noindent()
2366             && owner_->isPartOfTextSequence()
2367             && !layout_->pass_thru
2368             && curAlign != LYX_ALIGN_CENTER) {
2369                 if (!owner_->empty()
2370                     && (owner_->isInset(0)
2371                         && owner_->getInset(0)->lyxCode() == VSPACE_CODE))
2372                         // If the paragraph starts with a vspace, the \\noindent
2373                         // needs to come after that (as it leaves vmode).
2374                         // If the paragraph consists only of the vspace,
2375                         // \\noindent is not needed at all.
2376                         runparams.need_noindent = owner_->size() > 1;
2377                 else {
2378                         os << "\\noindent" << termcmd;
2379                         column += 10;
2380                 }
2381         }
2382
2383         if (curAlign == layout_->align)
2384                 return column;
2385
2386         switch (curAlign) {
2387         case LYX_ALIGN_NONE:
2388         case LYX_ALIGN_BLOCK:
2389         case LYX_ALIGN_LAYOUT:
2390         case LYX_ALIGN_SPECIAL:
2391         case LYX_ALIGN_DECIMAL:
2392                 break;
2393         case LYX_ALIGN_LEFT:
2394         case LYX_ALIGN_RIGHT:
2395         case LYX_ALIGN_CENTER:
2396                 if (runparams.moving_arg) {
2397                         os << "\\protect";
2398                         column += 8;
2399                 }
2400                 break;
2401         }
2402
2403         string const begin_tag = "\\begin";
2404         InsetCode code = ownerCode();
2405         bool const lastpar = runparams.isLastPar;
2406         // RTL in classic (PDF)LaTeX (without the Bidi package)
2407         // Luabibdi (used by LuaTeX) behaves like classic
2408         bool const rtl_classic = owner_->getParLanguage(bparams)->rightToLeft()
2409                 && !runparams.useBidiPackage();
2410
2411         switch (curAlign) {
2412         case LYX_ALIGN_NONE:
2413         case LYX_ALIGN_BLOCK:
2414         case LYX_ALIGN_LAYOUT:
2415         case LYX_ALIGN_SPECIAL:
2416         case LYX_ALIGN_DECIMAL:
2417                 break;
2418         case LYX_ALIGN_LEFT: {
2419                 if (rtl_classic)
2420                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2421                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2422                 else
2423                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2424                 break;
2425         } case LYX_ALIGN_RIGHT: {
2426                 if (rtl_classic)
2427                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2428                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2429                 else
2430                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2431                 break;
2432         } case LYX_ALIGN_CENTER: {
2433                 corrected_env(os, begin_tag, "center", code, lastpar, column);
2434                 break;
2435         }
2436         }
2437
2438         return column;
2439 }
2440
2441
2442 bool Paragraph::Private::endTeXParParams(BufferParams const & bparams,
2443                         otexstream & os, OutputParams const & runparams) const
2444 {
2445         LyXAlignment const curAlign = params_.align();
2446
2447         if (curAlign == layout_->align)
2448                 return false;
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         case LYX_ALIGN_RIGHT:
2459         case LYX_ALIGN_CENTER:
2460                 if (runparams.moving_arg)
2461                         os << "\\protect";
2462                 break;
2463         }
2464
2465         bool output = false;
2466         int col = 0;
2467         string const end_tag = "\\par\\end";
2468         InsetCode code = ownerCode();
2469         bool const lastpar = runparams.isLastPar;
2470         // RTL in classic (PDF)LaTeX (without the Bidi package)
2471         // Luabibdi (used by LuaTeX) behaves like classic
2472         bool const rtl_classic = owner_->getParLanguage(bparams)->rightToLeft()
2473                 && !runparams.useBidiPackage();
2474
2475         switch (curAlign) {
2476         case LYX_ALIGN_NONE:
2477         case LYX_ALIGN_BLOCK:
2478         case LYX_ALIGN_LAYOUT:
2479         case LYX_ALIGN_SPECIAL:
2480         case LYX_ALIGN_DECIMAL:
2481                 break;
2482         case LYX_ALIGN_LEFT: {
2483                 if (rtl_classic)
2484                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2485                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2486                 else
2487                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2488                 break;
2489         } case LYX_ALIGN_RIGHT: {
2490                 if (rtl_classic)
2491                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2492                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2493                 else
2494                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2495                 break;
2496         } case LYX_ALIGN_CENTER: {
2497                 corrected_env(os, end_tag, "center", code, lastpar, col);
2498                 break;
2499         }
2500         }
2501
2502         return output || lastpar;
2503 }
2504
2505
2506 // This one spits out the text of the paragraph
2507 void Paragraph::latex(BufferParams const & bparams,
2508         Font const & outerfont,
2509         otexstream & os,
2510         OutputParams const & runparams,
2511         int start_pos, int end_pos, bool force) const
2512 {
2513         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
2514
2515         // FIXME This check should not be needed. Perhaps issue an
2516         // error if it triggers.
2517         Layout const & style = inInset().forcePlainLayout() ?
2518                 bparams.documentClass().plainLayout() : *d->layout_;
2519
2520         if (!force && style.inpreamble)
2521                 return;
2522
2523         bool const allowcust = allowParagraphCustomization();
2524
2525         // Current base font for all inherited font changes, without any
2526         // change caused by an individual character, except for the language:
2527         // It is set to the language of the first character.
2528         // As long as we are in the label, this font is the base font of the
2529         // label. Before the first body character it is set to the base font
2530         // of the body.
2531         Font basefont;
2532
2533         // If there is an open font-encoding changing command (script wrapper),
2534         // alien_script is set to its name
2535         string alien_script;
2536         string script;
2537
2538         // Maybe we have to create a optional argument.
2539         pos_type body_pos = beginOfBody();
2540         unsigned int column = 0;
2541
2542         // If we are inside an non inheritFont() inset, the outerfont is the default font
2543         Font const real_outerfont =
2544                 inInset().inheritFont() ? outerfont : Font(bparams.documentClass().defaultfont());
2545
2546         if (body_pos > 0) {
2547                 // the optional argument is kept in curly brackets in
2548                 // case it contains a ']'
2549                 // This is not strictly needed, but if this is changed it
2550                 // would be a file format change, and tex2lyx would need
2551                 // to be adjusted, since it unconditionally removes the
2552                 // braces when it parses \item.
2553                 os << "[{";
2554                 column += 2;
2555                 basefont = getLabelFont(bparams, real_outerfont);
2556         } else {
2557                 basefont = getLayoutFont(bparams, real_outerfont);
2558         }
2559
2560         // Which font is currently active?
2561         Font running_font(basefont);
2562         // Do we have an open font change?
2563         bool open_font = false;
2564
2565         Change runningChange =
2566             runparams.inDeletedInset && !inInset().canTrackChanges()
2567             ? runparams.changeOfDeletedInset : Change(Change::UNCHANGED);
2568
2569         Encoding const * const prev_encoding = runparams.encoding;
2570
2571         os.texrow().start(id(), 0);
2572
2573         // if the paragraph is empty, the loop will not be entered at all
2574         if (empty()) {
2575                 // For InTitle commands, we have already opened a group
2576                 // in output_latex::TeXOnePar.
2577                 if (style.isCommand() && !style.intitle) {
2578                         os << '{';
2579                         ++column;
2580                 }
2581                 if (!style.leftdelim().empty()) {
2582                         os << style.leftdelim();
2583                         column += style.leftdelim().size();
2584                 }
2585                 if (allowcust)
2586                         column += d->startTeXParParams(bparams, os, runparams);
2587         }
2588
2589         // Whether a \par can be issued for insets typeset inline with text.
2590         // Yes if greater than 0. This has to be static.
2591         THREAD_LOCAL_STATIC int parInline = 0;
2592
2593         for (pos_type i = 0; i < size(); ++i) {
2594                 // First char in paragraph or after label?
2595                 if (i == body_pos) {
2596                         if (body_pos > 0) {
2597                                 if (open_font) {
2598                                         bool needPar = false;
2599                                         column += running_font.latexWriteEndChanges(
2600                                                 os, bparams, runparams,
2601                                                 basefont, basefont, needPar);
2602                                         open_font = false;
2603                                 }
2604                                 basefont = getLayoutFont(bparams, real_outerfont);
2605                                 running_font = basefont;
2606
2607                                 column += Changes::latexMarkChange(os, bparams,
2608                                                 runningChange, Change(Change::UNCHANGED),
2609                                                 runparams);
2610                                 runningChange = Change(Change::UNCHANGED);
2611
2612                                 os << (isEnvSeparator(i) ? "}]~" : "}] ");
2613                                 column +=3;
2614                         }
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
2622                         if (!style.leftdelim().empty()) {
2623                                 os << style.leftdelim();
2624                                 column += style.leftdelim().size();
2625                         }
2626
2627                         if (allowcust)
2628                                 column += d->startTeXParParams(bparams, os,
2629                                                             runparams);
2630                 }
2631
2632                 runparams.wasDisplayMath = runparams.inDisplayMath;
2633                 runparams.inDisplayMath = false;
2634                 bool deleted_display_math = false;
2635                 Change const & change = runparams.inDeletedInset
2636                         ? runparams.changeOfDeletedInset : lookupChange(i);
2637
2638                 char_type const c = d->text_[i];
2639
2640                 // Check whether a display math inset follows
2641                 bool output_changes;
2642                 if (runparams.for_searchAdv == OutputParams::NoSearch)
2643                         output_changes = bparams.output_changes;
2644                 else
2645                         output_changes = (runparams.for_searchAdv == OutputParams::SearchWithDeleted);
2646                 if (c == META_INSET
2647                     && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2648                         if (isDeleted(i))
2649                                 runparams.ctObject = getInset(i)->getCtObject(runparams);
2650
2651                         InsetMath const * im = getInset(i)->asInsetMath();
2652                         if (im && im->asHullInset()
2653                             && im->asHullInset()->outerDisplay()) {
2654                                 runparams.inDisplayMath = true;
2655                                 // runparams.inDeletedInset will be set by
2656                                 // latexInset later, but we need this info
2657                                 // before it is called. On the other hand, we
2658                                 // cannot set it here because it is a counter.
2659                                 deleted_display_math = isDeleted(i);
2660                         }
2661                         if (output_changes && deleted_display_math
2662                             && runningChange == change
2663                             && change.type == Change::DELETED
2664                             && !os.afterParbreak()) {
2665                                 // A display math in the same paragraph follows.
2666                                 // We have to close and then reopen \lyxdeleted,
2667                                 // otherwise the math will be shifted up.
2668                                 OutputParams rp = runparams;
2669                                 if (open_font) {
2670                                         bool needPar = false;
2671                                         column += running_font.latexWriteEndChanges(
2672                                                 os, bparams, rp, basefont,
2673                                                 basefont, needPar);
2674                                         open_font = false;
2675                                 }
2676                                 basefont = (body_pos > i) ? getLabelFont(bparams, real_outerfont)
2677                                                           : getLayoutFont(bparams, real_outerfont);
2678                                 running_font = basefont;
2679                                 column += Changes::latexMarkChange(os, bparams,
2680                                         Change(Change::INSERTED), change, rp);
2681                         }
2682                 }
2683
2684                 if (output_changes && runningChange != change) {
2685                         if (!alien_script.empty()) {
2686                                 column += 1;
2687                                 os << "}";
2688                                 alien_script.clear();
2689                         }
2690                         if (open_font) {
2691                                 bool needPar = false;
2692                                 column += running_font.latexWriteEndChanges(
2693                                                 os, bparams, runparams,
2694                                                 basefont, basefont, needPar);
2695                                 open_font = false;
2696                         }
2697                         basefont = (body_pos > i) ? getLabelFont(bparams, real_outerfont)
2698                                                   : getLayoutFont(bparams, real_outerfont);
2699                         running_font = basefont;
2700                         column += Changes::latexMarkChange(os, bparams, runningChange,
2701                                                            change, runparams);
2702                         runningChange = change;
2703                 }
2704
2705                 // do not output text which is marked deleted
2706                 // if change tracking output is disabled
2707                 if (!output_changes && change.deleted()) {
2708                         continue;
2709                 }
2710
2711                 ++column;
2712
2713                 // Fully instantiated font
2714                 Font current_font = getFont(bparams, i, outerfont);
2715                 // Previous font
2716                 Font const prev_font = (i > 0) ?
2717                                         getFont(bparams, i - 1, outerfont)
2718                                       : current_font;
2719
2720                 Font const last_font = running_font;
2721                 bool const in_ct_deletion = (output_changes
2722                                              && runningChange == change
2723                                              && change.type == Change::DELETED
2724                                              && !os.afterParbreak());
2725                 // Insets where font switches are used (rather than font commands)
2726                 bool const fontswitch_inset =
2727                                 c == META_INSET
2728                                 && getInset(i)
2729                                 && getInset(i)->allowMultiPar()
2730                                 && getInset(i)->lyxCode() != ERT_CODE
2731                                 && (getInset(i)->producesOutput()
2732                                     // FIXME Something more general?
2733                                     // Comments do not "produce output" but are still
2734                                     // part of the TeX source and require font switches
2735                                     // to be closed (otherwise LaTeX fails).
2736                                     || getInset(i)->layoutName() == "Note:Comment");
2737
2738                 bool closeLanguage = false;
2739                 bool lang_switched_at_inset = false;
2740                 if (fontswitch_inset) {
2741                         // Some insets cannot be inside a font change command.
2742                         // However, even such insets *can* be placed in \L or \R
2743                         // or their equivalents (for RTL language switches),
2744                         // so we don't close the language in those cases
2745                         // (= differing isRightToLeft()).
2746                         // ArabTeX, though, doesn't seem to handle this special behavior.
2747                         closeLanguage = basefont.isRightToLeft() == current_font.isRightToLeft()
2748                                         || basefont.language()->lang() == "arabic_arabtex"
2749                                         || current_font.language()->lang() == "arabic_arabtex";
2750                         // We need to check prev_font as language changes directly at inset
2751                         // will only be started inside the inset.
2752                         lang_switched_at_inset = prev_font.language() != current_font.language();
2753                 }
2754
2755                 // Do we need to close the previous font?
2756                 bool langClosed = false;
2757                 if (open_font &&
2758                     ((current_font != running_font
2759                       || current_font.language() != running_font.language())
2760                      || (fontswitch_inset
2761                          && (current_font == prev_font))))
2762                 {
2763                         // ensure there is no open script-wrapper
2764                         if (!alien_script.empty()) {
2765                                 column += 1;
2766                                 os << "}";
2767                                 alien_script.clear();
2768                         }
2769                         if (in_ct_deletion) {
2770                                 // We have to close and then reopen \lyxdeleted,
2771                                 // as strikeout needs to be on lowest level.
2772                                 os << '}';
2773                                 column += 1;
2774                         }
2775                         if (closeLanguage)
2776                                 // Force language closing
2777                                 current_font.setLanguage(basefont.language());
2778                         Font const nextfont = (i == body_pos-1) ? basefont : current_font;
2779                         bool needPar = false;
2780                         column += running_font.latexWriteEndChanges(
2781                                     os, bparams, runparams, basefont,
2782                                     nextfont, needPar);
2783                         if (in_ct_deletion) {
2784                                 // We have to close and then reopen \lyxdeleted,
2785                                 // as strikeout needs to be on lowest level.
2786                                 OutputParams rp = runparams;
2787                                 column += Changes::latexMarkChange(os, bparams,
2788                                         Change(Change::UNCHANGED), Change(Change::DELETED), rp);
2789                         }
2790                         // Has the language been closed in the latexWriteEndChanges() call above?
2791                         langClosed = running_font.language() != basefont.language()
2792                                         && running_font.language() != nextfont.language()
2793                                         && (running_font.language()->encoding()->package() != Encoding::CJK);
2794                         running_font = basefont;
2795                         open_font &= !langClosed;
2796                 }
2797
2798                 // if necessary, close language environment before opening CJK
2799                 string const running_lang = running_font.language()->babel();
2800                 string const lang_end_command = lyxrc.language_command_end;
2801                 if (!lang_end_command.empty() && !bparams.useNonTeXFonts
2802                         && !running_lang.empty()
2803                         && running_lang == openLanguageName()
2804                         && current_font.language()->encoding()->package() == Encoding::CJK) {
2805                         string end_tag = subst(lang_end_command, "$$lang", running_lang);
2806                         os << from_ascii(end_tag);
2807                         column += end_tag.length();
2808                         if (!languageStackEmpty())
2809                                 popLanguageName();
2810                 }
2811
2812                 // Switch file encoding if necessary (and allowed)
2813                 if ((!fontswitch_inset || closeLanguage)
2814                     && !runparams.pass_thru && !style.pass_thru &&
2815                     runparams.encoding->package() != Encoding::none &&
2816                     current_font.language()->encoding()->package() != Encoding::none) {
2817                         pair<bool, int> const enc_switch =
2818                                 switchEncoding(os.os(), bparams, runparams,
2819                                         *(current_font.language()->encoding()));
2820                         if (enc_switch.first) {
2821                                 column += enc_switch.second;
2822                                 runparams.encoding = current_font.language()->encoding();
2823                         }
2824                 }
2825
2826                 // A display math inset inside an ulem command will be output
2827                 // as a box of width \linewidth, so we have to either disable
2828                 // indentation if the inset starts a paragraph, or start a new
2829                 // line to accommodate such box. This has to be done before
2830                 // writing any font changing commands.
2831                 if (runparams.inDisplayMath && !deleted_display_math
2832                     && runparams.inulemcmd) {
2833                         if (os.afterParbreak())
2834                                 os << "\\noindent";
2835                         else
2836                                 os << "\\\\\n";
2837                 }
2838
2839                 // Do we need to change font?
2840                 if ((current_font != running_font ||
2841                      current_font.language() != running_font.language())
2842                     && i != body_pos - 1)
2843                 {
2844                         if (!fontswitch_inset) {
2845                                 if (in_ct_deletion) {
2846                                         // We have to close and then reopen \lyxdeleted,
2847                                         // as strikeout needs to be on lowest level.
2848                                         OutputParams rp = runparams;
2849                                         bool needPar = false;
2850                                         column += running_font.latexWriteEndChanges(
2851                                                 os, bparams, rp, basefont,
2852                                                 basefont, needPar);
2853                                         os << '}';
2854                                         column += 1;
2855                                 }
2856                                 otexstringstream ots;
2857                                 InsetText const * textinset = inInset().asInsetText();
2858                                 bool const cprotect = textinset
2859                                         ? textinset->hasCProtectContent(runparams.moving_arg)
2860                                           && !textinset->text().isMainText()
2861                                           && inInset().lyxCode() != BRANCH_CODE
2862                                         : false;
2863                                 column += current_font.latexWriteStartChanges(ots, bparams,
2864                                                                               runparams, basefont, last_font, false,
2865                                                                               cprotect);
2866                                 // Check again for display math in ulem commands as a
2867                                 // font change may also occur just before a math inset.
2868                                 if (runparams.inDisplayMath && !deleted_display_math
2869                                     && runparams.inulemcmd) {
2870                                         if (os.afterParbreak())
2871                                                 os << "\\noindent";
2872                                         else
2873                                                 os << "\\\\\n";
2874                                 }
2875                                 running_font = current_font;
2876                                 open_font = true;
2877                                 docstring fontchange = ots.str();
2878                                 os << fontchange;
2879                                 // check whether the fontchange ends with a \\textcolor
2880                                 // modifier and the text starts with a space. If so we
2881                                 // need to add } in order to prevent \\textcolor from gobbling
2882                                 // the space (bug 4473).
2883                                 docstring const last_modifier = rsplit(fontchange, '\\');
2884                                 if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2885                                         os << from_ascii("{}");
2886                                 else if (ots.terminateCommand())
2887                                         os << termcmd;
2888                                 if (in_ct_deletion) {
2889                                         // We have to close and then reopen \lyxdeleted,
2890                                         // as strikeout needs to be on lowest level.
2891                                         OutputParams rp = runparams;
2892                                         column += Changes::latexMarkChange(os, bparams,
2893                                                 Change(Change::UNCHANGED), change, rp);
2894                                 }
2895                         } else {
2896                                 running_font = current_font;
2897                                 open_font &= !langClosed;
2898                         }
2899                 }
2900
2901                 // FIXME: think about end_pos implementation...
2902                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2903                         // FIXME: integrate this case in latexSpecialChar
2904                         // Do not print the separation of the optional argument
2905                         // if style.pass_thru is false. This works because
2906                         // latexSpecialChar ignores spaces if
2907                         // style.pass_thru is false.
2908                         if (i != body_pos - 1) {
2909                                 if (d->simpleTeXBlanks(bparams, runparams, os,
2910                                                 i, column, current_font, style)) {
2911                                         // A surrogate pair was output. We
2912                                         // must not call latexSpecialChar
2913                                         // in this iteration, since it would output
2914                                         // the combining character again.
2915                                         ++i;
2916                                         continue;
2917                                 }
2918                         }
2919                 }
2920
2921                 OutputParams rp = runparams;
2922                 rp.free_spacing = style.free_spacing;
2923                 rp.local_font = &current_font;
2924                 rp.intitle = style.intitle;
2925
2926                 // Two major modes:  LaTeX or plain
2927                 // Handle here those cases common to both modes
2928                 // and then split to handle the two modes separately.
2929                 if (c == META_INSET) {
2930                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2931                                 // Greyedout notes and, in general, all insets
2932                                 // with InsetLayout::isDisplay() == false,
2933                                 // are typeset inline with the text. So, we
2934                                 // can add a \par to the last paragraph of
2935                                 // such insets only if nothing else follows.
2936                                 bool incremented = false;
2937                                 Inset const * inset = getInset(i);
2938                                 InsetText const * textinset = inset
2939                                                         ? inset->asInsetText()
2940                                                         : nullptr;
2941                                 if (i + 1 == size() && textinset
2942                                     && !inset->getLayout().isDisplay()) {
2943                                         ParagraphList const & pars =
2944                                                 textinset->text().paragraphs();
2945                                         pit_type const pit = pars.size() - 1;
2946                                         Font const lastfont =
2947                                                 pit < 0 || pars[pit].empty()
2948                                                 ? pars[pit].getLayoutFont(
2949                                                                 bparams,
2950                                                                 real_outerfont)
2951                                                 : pars[pit].getFont(bparams,
2952                                                         pars[pit].size() - 1,
2953                                                         real_outerfont);
2954                                         if (lastfont.fontInfo().size() !=
2955                                             basefont.fontInfo().size()) {
2956                                                 ++parInline;
2957                                                 incremented = true;
2958                                         }
2959                                 }
2960                                 // We need to restore parts of this after insets with
2961                                 // allowMultiPar() true
2962                                 Font const save_basefont = basefont;
2963                                 d->latexInset(bparams, os, rp, running_font,
2964                                                 basefont, real_outerfont, open_font,
2965                                                 runningChange, style, i, column, fontswitch_inset,
2966                                                 closeLanguage, lang_switched_at_inset);
2967                                 if (fontswitch_inset) {
2968                                         if (open_font) {
2969                                                 bool needPar = false;
2970                                                 column += running_font.latexWriteEndChanges(
2971                                                         os, bparams, runparams,
2972                                                         basefont, basefont, needPar);
2973                                                 open_font = false;
2974                                         }
2975                                         basefont.fontInfo().setSize(save_basefont.fontInfo().size());
2976                                         basefont.fontInfo().setFamily(save_basefont.fontInfo().family());
2977                                         basefont.fontInfo().setSeries(save_basefont.fontInfo().series());
2978                                 }
2979                                 if (incremented)
2980                                         --parInline;
2981
2982                                 if (runparams.ctObject == CtObject::DisplayObject
2983                                     || runparams.ctObject == CtObject::UDisplayObject) {
2984                                         // Close \lyx*deleted and force its
2985                                         // reopening (if needed)
2986                                         os << '}';
2987                                         column++;
2988                                         runningChange = Change(Change::UNCHANGED);
2989                                         runparams.ctObject = CtObject::Normal;
2990                                 }
2991                         }
2992                 } else if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2993                         if (!bparams.useNonTeXFonts)
2994                           script = Encodings::isKnownScriptChar(c);
2995                         if (script != alien_script) {
2996                                 if (!alien_script.empty()) {
2997                                         os << "}";
2998                                         alien_script.clear();
2999                                 }
3000                                 string fontenc = running_font.language()->fontenc(bparams);
3001                                 if (!script.empty()
3002                                         && !Encodings::fontencSupportsScript(fontenc, script)) {
3003                                         column += script.length() + 2;
3004                                         os << "\\" << script << "{";
3005                                         alien_script = script;
3006                                 }
3007                         }
3008                         try {
3009                                 d->latexSpecialChar(os, bparams, rp, running_font,
3010                                                                         alien_script, style, i, end_pos, column);
3011                         } catch (EncodingException & e) {
3012                                 if (runparams.dryrun) {
3013                                         os << "<" << _("LyX Warning: ")
3014                                            << _("uncodable character") << " '";
3015                                         os.put(c);
3016                                         os << "'>";
3017                                 } else {
3018                                         // add location information and throw again.
3019                                         e.par_id = id();
3020                                         e.pos = i;
3021                                         throw;
3022                                 }
3023                         }
3024                 }
3025
3026                 // Set the encoding to that returned from latexSpecialChar (see
3027                 // comment for encoding member in OutputParams.h)
3028                 runparams.encoding = rp.encoding;
3029
3030                 // Also carry on the info on a closed ulem command for insets
3031                 // such as Note that do not produce any output, so that no
3032                 // command is ever executed but its opening was recorded.
3033                 runparams.inulemcmd = rp.inulemcmd;
3034
3035                 // These need to be passed upstream as well
3036                 runparams.need_maketitle = rp.need_maketitle;
3037                 runparams.have_maketitle = rp.have_maketitle;
3038
3039                 // And finally, pass the post_macros upstream
3040                 runparams.post_macro = rp.post_macro;
3041         }
3042
3043         // Close wrapper for alien script
3044         if (!alien_script.empty()) {
3045                 os << "}";
3046                 alien_script.clear();
3047         }
3048
3049         Font const font = empty()
3050                 ? getLayoutFont(bparams, real_outerfont)
3051                 : getFont(bparams, size() - 1, real_outerfont);
3052
3053         InsetText const * textinset = inInset().asInsetText();
3054
3055         bool const maintext = textinset
3056                 ? textinset->text().isMainText()
3057                 : false;
3058
3059         size_t const numpars = textinset
3060                 ? textinset->text().paragraphs().size()
3061                 : 0;
3062
3063         bool needPar = false;
3064
3065         if (style.resfont.size() != font.fontInfo().size()
3066             && (!runparams.isLastPar || maintext
3067                 || (numpars > 1 && d->ownerCode() != CELL_CODE
3068                     && (inInset().getLayout().isDisplay()
3069                         || parInline)))
3070             && !style.isCommand()) {
3071                 needPar = true;
3072         }
3073
3074         // If we have an open font definition, we have to close it
3075         if (open_font) {
3076                 // Make sure that \\par is done with the font of the last
3077                 // character if this has another size as the default.
3078                 // This is necessary because LaTeX (and LyX on the screen)
3079                 // calculates the space between the baselines according
3080                 // to this font. (Matthias)
3081                 //
3082                 // We must not change the font for the last paragraph
3083                 // of non-multipar insets, tabular cells or commands,
3084                 // since this produces unwanted whitespace.
3085 #ifdef FIXED_LANGUAGE_END_DETECTION
3086                 if (next_) {
3087                         running_font.latexWriteEndChanges(os, bparams,
3088                                         runparams, basefont,
3089                                         next_->getFont(bparams, 0, outerfont),
3090                                                        needPar);
3091                 } else {
3092                         running_font.latexWriteEndChanges(os, bparams,
3093                                         runparams, basefont, basefont, needPar);
3094                 }
3095 #else
3096 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
3097 //FIXME: there as we start another \selectlanguage with the next paragraph if
3098 //FIXME: we are in need of this. This should be fixed sometime (Jug)
3099                 running_font.latexWriteEndChanges(os, bparams, runparams,
3100                                 basefont, basefont, needPar);
3101 #endif
3102         }
3103         if (needPar) {
3104                 // The \par could not be inserted at the same nesting
3105                 // level of the font size change, so do it now.
3106                 os << "{\\" << font.latexSize() << "\\par}";
3107         }
3108
3109         if (!runparams.inDeletedInset || inInset().canTrackChanges())
3110                 column += Changes::latexMarkChange(os, bparams, runningChange,
3111                                         Change(Change::UNCHANGED), runparams);
3112
3113         // Needed if there is an optional argument but no contents.
3114         if (body_pos > 0 && body_pos == size()) {
3115                 os << "}]~";
3116         }
3117
3118         if (!style.rightdelim().empty()) {
3119                 os << style.rightdelim();
3120                 column += style.rightdelim().size();
3121         }
3122
3123         if (allowcust && d->endTeXParParams(bparams, os, runparams)
3124             && runparams.encoding != prev_encoding) {
3125                 runparams.encoding = prev_encoding;
3126                 os << setEncoding(prev_encoding->iconvName());
3127         }
3128
3129         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
3130 }
3131
3132
3133 bool Paragraph::emptyTag() const
3134 {
3135         for (pos_type i = 0; i < size(); ++i) {
3136                 if (Inset const * inset = getInset(i)) {
3137                         InsetCode lyx_code = inset->lyxCode();
3138                         // FIXME testing like that is wrong. What is
3139                         // the intent?
3140                         if (lyx_code != TOC_CODE &&
3141                             lyx_code != INCLUDE_CODE &&
3142                             lyx_code != GRAPHICS_CODE &&
3143                             lyx_code != ERT_CODE &&
3144                             lyx_code != LISTINGS_CODE &&
3145                             lyx_code != FLOAT_CODE &&
3146                             lyx_code != TABULAR_CODE) {
3147                                 return false;
3148                         }
3149                 } else {
3150                         char_type c = d->text_[i];
3151                         if (c != ' ' && c != '\t')
3152                                 return false;
3153                 }
3154         }
3155         return true;
3156 }
3157
3158
3159 string Paragraph::getID(Buffer const &, OutputParams const &)
3160         const
3161 {
3162         for (pos_type i = 0; i < size(); ++i) {
3163                 if (Inset const * inset = getInset(i)) {
3164                         InsetCode lyx_code = inset->lyxCode();
3165                         if (lyx_code == LABEL_CODE) {
3166                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
3167                                 docstring const & id = il->getParam("name");
3168                                 return "id='" + to_utf8(xml::cleanID(id)) + "'";
3169                         }
3170                 }
3171         }
3172         return string();
3173 }
3174
3175
3176 pos_type Paragraph::firstWordDocBook(XMLStream & xs, OutputParams const & runparams) const
3177 {
3178         pos_type i;
3179         for (i = 0; i < size(); ++i) {
3180                 if (Inset const * inset = getInset(i)) {
3181                         inset->docbook(xs, runparams);
3182                 } else {
3183                         char_type c = d->text_[i];
3184                         if (c == ' ')
3185                                 break;
3186                         xs << c;
3187                 }
3188         }
3189         return i;
3190 }
3191
3192
3193 pos_type Paragraph::firstWordLyXHTML(XMLStream & xs, OutputParams const & runparams)
3194         const
3195 {
3196         pos_type i;
3197         for (i = 0; i < size(); ++i) {
3198                 if (Inset const * inset = getInset(i)) {
3199                         inset->xhtml(xs, runparams);
3200                 } else {
3201                         char_type c = d->text_[i];
3202                         if (c == ' ')
3203                                 break;
3204                         xs << c;
3205                 }
3206         }
3207         return i;
3208 }
3209
3210
3211 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
3212 {
3213         Font font_old;
3214         pos_type size = text_.size();
3215         for (pos_type i = initial; i < size; ++i) {
3216                 Font font = owner_->getFont(buf.params(), i, outerfont);
3217                 if (text_[i] == META_INSET)
3218                         return false;
3219                 if (i != initial && font != font_old)
3220                         return false;
3221                 font_old = font;
3222         }
3223
3224         return true;
3225 }
3226
3227
3228 namespace {
3229
3230 void doFontSwitchDocBook(vector<xml::FontTag> & tagsToOpen,
3231                   vector<xml::EndFontTag> & tagsToClose,
3232                   bool & flag, FontState curstate, xml::FontTypes type)
3233 {
3234         if (curstate == FONT_ON) {
3235                 tagsToOpen.push_back(docbookStartFontTag(type));
3236                 flag = true;
3237         } else if (flag) {
3238                 tagsToClose.push_back(docbookEndFontTag(type));
3239                 flag = false;
3240         }
3241 }
3242
3243 class OptionalFontType {
3244 public:
3245         xml::FontTypes ft;
3246         bool has_value;
3247
3248         OptionalFontType(): ft(xml::FT_EMPH), has_value(false) {} // A possible value at random for ft.
3249         OptionalFontType(xml::FontTypes ft): ft(ft), has_value(true) {}
3250 };
3251
3252 OptionalFontType fontShapeToXml(FontShape fs)
3253 {
3254         switch (fs) {
3255         case ITALIC_SHAPE:
3256                 return {xml::FT_ITALIC};
3257         case SLANTED_SHAPE:
3258                 return {xml::FT_SLANTED};
3259         case SMALLCAPS_SHAPE:
3260                 return {xml::FT_SMALLCAPS};
3261         case UP_SHAPE:
3262         case INHERIT_SHAPE:
3263                 return {};
3264         default:
3265                 // the other tags are for internal use
3266                 LATTEST(false);
3267                 return {};
3268         }
3269 }
3270
3271 OptionalFontType fontFamilyToXml(FontFamily fm)
3272 {
3273         switch (fm) {
3274         case ROMAN_FAMILY:
3275                 return {xml::FT_ROMAN};
3276         case SANS_FAMILY:
3277                 return {xml::FT_SANS};
3278         case TYPEWRITER_FAMILY:
3279                 return {xml::FT_TYPE};
3280         case INHERIT_FAMILY:
3281                 return {};
3282         default:
3283                 // the other tags are for internal use
3284                 LATTEST(false);
3285                 return {};
3286         }
3287 }
3288
3289 OptionalFontType fontSizeToXml(FontSize fs)
3290 {
3291         switch (fs) {
3292         case TINY_SIZE:
3293                 return {xml::FT_SIZE_TINY};
3294         case SCRIPT_SIZE:
3295                 return {xml::FT_SIZE_SCRIPT};
3296         case FOOTNOTE_SIZE:
3297                 return {xml::FT_SIZE_FOOTNOTE};
3298         case SMALL_SIZE:
3299                 return {xml::FT_SIZE_SMALL};
3300         case LARGE_SIZE:
3301                 return {xml::FT_SIZE_LARGE};
3302         case LARGER_SIZE:
3303                 return {xml::FT_SIZE_LARGER};
3304         case LARGEST_SIZE:
3305                 return {xml::FT_SIZE_LARGEST};
3306         case HUGE_SIZE:
3307                 return {xml::FT_SIZE_HUGE};
3308         case HUGER_SIZE:
3309                 return {xml::FT_SIZE_HUGER};
3310         case INCREASE_SIZE:
3311                 return {xml::FT_SIZE_INCREASE};
3312         case DECREASE_SIZE:
3313                 return {xml::FT_SIZE_DECREASE};
3314         case INHERIT_SIZE:
3315         case NORMAL_SIZE:
3316                 return {};
3317         default:
3318                 // the other tags are for internal use
3319                 LATTEST(false);
3320                 return {};
3321         }
3322 }
3323
3324 struct DocBookFontState
3325 {
3326         FontShape  curr_fs   = INHERIT_SHAPE;
3327         FontFamily curr_fam  = INHERIT_FAMILY;
3328         FontSize   curr_size = INHERIT_SIZE;
3329
3330         // track whether we have opened these tags
3331         bool emph_flag = false;
3332         bool bold_flag = false;
3333         bool noun_flag = false;
3334         bool ubar_flag = false;
3335         bool dbar_flag = false;
3336         bool sout_flag = false;
3337         bool xout_flag = false;
3338         bool wave_flag = false;
3339         // shape tags
3340         bool shap_flag = false;
3341         // family tags
3342         bool faml_flag = false;
3343         // size tags
3344         bool size_flag = false;
3345 };
3346
3347 std::tuple<vector<xml::FontTag>, vector<xml::EndFontTag>> computeDocBookFontSwitch(FontInfo const & font_old,
3348                                                                                            Font const & font,
3349                                                                                            std::string const & default_family,
3350                                                                                            DocBookFontState & fs)
3351 {
3352         vector<xml::FontTag> tagsToOpen;
3353         vector<xml::EndFontTag> tagsToClose;
3354
3355         // emphasis
3356         FontState curstate = font.fontInfo().emph();
3357         if (font_old.emph() != curstate)
3358                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.emph_flag, curstate, xml::FT_EMPH);
3359
3360         // noun
3361         curstate = font.fontInfo().noun();
3362         if (font_old.noun() != curstate)
3363                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.noun_flag, curstate, xml::FT_NOUN);
3364
3365         // underbar
3366         curstate = font.fontInfo().underbar();
3367         if (font_old.underbar() != curstate)
3368                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.ubar_flag, curstate, xml::FT_UBAR);
3369
3370         // strikeout
3371         curstate = font.fontInfo().strikeout();
3372         if (font_old.strikeout() != curstate)
3373                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.sout_flag, curstate, xml::FT_SOUT);
3374
3375         // xout
3376         curstate = font.fontInfo().xout();
3377         if (font_old.xout() != curstate)
3378                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.xout_flag, curstate, xml::FT_XOUT);
3379
3380         // double underbar
3381         curstate = font.fontInfo().uuline();
3382         if (font_old.uuline() != curstate)
3383                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.dbar_flag, curstate, xml::FT_DBAR);
3384
3385         // wavy line
3386         curstate = font.fontInfo().uwave();
3387         if (font_old.uwave() != curstate)
3388                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.wave_flag, curstate, xml::FT_WAVE);
3389
3390         // bold
3391         // a little hackish, but allows us to reuse what we have.
3392         curstate = (font.fontInfo().series() == BOLD_SERIES ? FONT_ON : FONT_OFF);
3393         if (font_old.series() != font.fontInfo().series())
3394                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.bold_flag, curstate, xml::FT_BOLD);
3395
3396         // Font shape
3397         fs.curr_fs = font.fontInfo().shape();
3398         FontShape old_fs = font_old.shape();
3399         if (old_fs != fs.curr_fs) {
3400                 if (fs.shap_flag) {
3401                         OptionalFontType tag = fontShapeToXml(old_fs);
3402                         if (tag.has_value)
3403                                 tagsToClose.push_back(docbookEndFontTag(tag.ft));
3404                         fs.shap_flag = false;
3405                 }
3406
3407                 OptionalFontType tag = fontShapeToXml(fs.curr_fs);
3408                 if (tag.has_value)
3409                         tagsToOpen.push_back(docbookStartFontTag(tag.ft));
3410         }
3411
3412         // Font family
3413         fs.curr_fam = font.fontInfo().family();
3414         FontFamily old_fam = font_old.family();
3415         if (old_fam != fs.curr_fam) {
3416                 if (fs.faml_flag) {
3417                         OptionalFontType tag = fontFamilyToXml(old_fam);
3418                         if (tag.has_value)
3419                                 tagsToClose.push_back(docbookEndFontTag(tag.ft));
3420                         fs.faml_flag = false;
3421                 }
3422                 switch (fs.curr_fam) {
3423                         case ROMAN_FAMILY:
3424                                 // we will treat a "default" font family as roman, since we have
3425                                 // no other idea what to do.
3426                                 if (default_family != "rmdefault" && default_family != "default") {
3427                                         tagsToOpen.push_back(docbookStartFontTag(xml::FT_ROMAN));
3428                                         fs.faml_flag = true;
3429                                 }
3430                                 break;
3431                         case SANS_FAMILY:
3432                                 if (default_family != "sfdefault") {
3433                                         tagsToOpen.push_back(docbookStartFontTag(xml::FT_SANS));
3434                                         fs.faml_flag = true;
3435                                 }
3436                                 break;
3437                         case TYPEWRITER_FAMILY:
3438                                 if (default_family != "ttdefault") {
3439                                         tagsToOpen.push_back(docbookStartFontTag(xml::FT_TYPE));
3440                                         fs.faml_flag = true;
3441                                 }
3442                                 break;
3443                         case INHERIT_FAMILY:
3444                                 break;
3445                         default:
3446                                 // the other tags are for internal use
3447                                 LATTEST(false);
3448                                 break;
3449                 }
3450         }
3451
3452         // Font size
3453         fs.curr_size = font.fontInfo().size();
3454         FontSize old_size = font_old.size();
3455         if (old_size != fs.curr_size) {
3456                 if (fs.size_flag) {
3457                         OptionalFontType tag = fontSizeToXml(old_size);
3458                         if (tag.has_value)
3459                                 tagsToClose.push_back(docbookEndFontTag(tag.ft));
3460                         fs.size_flag = false;
3461                 }
3462
3463                 OptionalFontType tag = fontSizeToXml(fs.curr_size);
3464                 if (tag.has_value) {
3465                         tagsToOpen.push_back(docbookStartFontTag(tag.ft));
3466                         fs.size_flag = true;
3467                 }
3468         }
3469
3470         return std::tuple<vector<xml::FontTag>, vector<xml::EndFontTag>>(tagsToOpen, tagsToClose);
3471 }
3472
3473 } // anonymous namespace
3474
3475
3476 std::tuple<std::vector<docstring>, std::vector<docstring>, std::vector<docstring>>
3477     Paragraph::simpleDocBookOnePar(Buffer const & buf,
3478                                                       OutputParams const & runparams,
3479                                                       Font const & outerfont,
3480                                                       pos_type initial,
3481                                                       bool is_last_par,
3482                                                       bool ignore_fonts) const
3483 {
3484         std::vector<docstring> prependedParagraphs;
3485         std::vector<docstring> generatedParagraphs;
3486         std::vector<docstring> appendedParagraphs;
3487         odocstringstream os;
3488
3489         // If there is an argument that must be output before the main tag, do it before handling the rest of the paragraph.
3490         // Also tag all arguments that shouldn't go in the main content right now, so that they are never generated at the
3491         // wrong place.
3492         OutputParams rp = runparams;
3493     for (pos_type i = initial; i < size(); ++i) {
3494         if (getInset(i) && getInset(i)->lyxCode() == ARG_CODE) {
3495             const InsetArgument * arg = getInset(i)->asInsetArgument();
3496             if (arg->docbookargumentbeforemaintag()) {
3497                 auto xs_local = XMLStream(os);
3498                 arg->docbook(xs_local, rp);
3499
3500                 prependedParagraphs.push_back(os.str());
3501                 os.str(from_ascii(""));
3502
3503                 rp.docbook_prepended_arguments.insert(arg);
3504             } else if (arg->docbookargumentaftermaintag()) {
3505                 rp.docbook_appended_arguments.insert(arg);
3506             }
3507         }
3508     }
3509
3510     // State variables for the main loop.
3511     auto xs = new XMLStream(os); // XMLStream has no copy constructor: to create a new object, the only solution
3512     // is to hold a pointer to the XMLStream (xs = XMLStream(os) is not allowed once the first object is built).
3513     std::vector<char_type> delayedChars; // When a font tag ends with a space, output it after the closing font tag.
3514     // This requires to store delayed characters at some point.
3515
3516     DocBookFontState fs; // Track whether we have opened font tags
3517     DocBookFontState old_fs = fs;
3518
3519     Layout const & style = *d->layout_;
3520     FontInfo font_old = style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
3521     string const default_family = buf.masterBuffer()->params().fonts_default_family;
3522
3523     vector<xml::FontTag> tagsToOpen;
3524     vector<xml::EndFontTag> tagsToClose;
3525
3526         // Parsing main loop.
3527         for (pos_type i = initial; i < size(); ++i) {
3528             bool ignore_fonts_i = ignore_fonts
3529                               || style.docbooknofontinside()
3530                               || (getInset(i) && getInset(i)->getLayout().docbooknofontinside());
3531
3532                 // Don't show deleted material in the output.
3533                 if (isDeleted(i))
3534                         continue;
3535
3536                 // If this is an InsetNewline, generate a new paragraph. Also reset the fonts, so that tags are closed in
3537                 // this paragraph.
3538                 if (getInset(i) && getInset(i)->lyxCode() == NEWLINE_CODE) {
3539                         if (!ignore_fonts_i)
3540                                 xs->closeFontTags();
3541
3542                         // Output one paragraph (i.e. one string entry in generatedParagraphs).
3543                         generatedParagraphs.push_back(os.str());
3544
3545                         // Create a new XMLStream for the new paragraph, completely independent from the previous one. This implies
3546                         // that the string stream must be reset.
3547                         os.str(from_ascii(""));
3548                         delete xs;
3549                         xs = new XMLStream(os);
3550
3551                         // Restore the fonts for the new paragraph, so that the right tags are opened for the new entry.
3552                         if (!ignore_fonts_i) {
3553                                 font_old = outerfont.fontInfo();
3554                                 fs = old_fs;
3555                         }
3556                 }
3557
3558                 // Determine which tags should be opened or closed regarding fonts.
3559                 Font const font = getFont(buf.masterBuffer()->params(), i, outerfont);
3560         tie(tagsToOpen, tagsToClose) = computeDocBookFontSwitch(font_old, font, default_family, fs);
3561
3562                 if (!ignore_fonts_i) {
3563             vector<xml::EndFontTag>::const_iterator cit = tagsToClose.begin();
3564             vector<xml::EndFontTag>::const_iterator cen = tagsToClose.end();
3565             for (; cit != cen; ++cit)
3566                 *xs << *cit;
3567         }
3568
3569         // Deal with the delayed characters *after* closing font tags.
3570         if (!delayedChars.empty()) {
3571             for (char_type c: delayedChars)
3572                 *xs << c;
3573             delayedChars.clear();
3574         }
3575
3576         if (!ignore_fonts_i) {
3577                         vector<xml::FontTag>::const_iterator sit = tagsToOpen.begin();
3578                         vector<xml::FontTag>::const_iterator sen = tagsToOpen.end();
3579                         for (; sit != sen; ++sit)
3580                                 *xs << *sit;
3581
3582                         tagsToClose.clear();
3583                         tagsToOpen.clear();
3584                 }
3585
3586         // Finally, write the next character or inset.
3587                 if (Inset const * inset = getInset(i)) {
3588                     bool inset_is_argument_elsewhere = getInset(i)->asInsetArgument() &&
3589                             rp.docbook_appended_arguments.find(inset->asInsetArgument()) != rp.docbook_appended_arguments.end() &&
3590                             rp.docbook_prepended_arguments.find(inset->asInsetArgument()) != rp.docbook_prepended_arguments.end();
3591
3592                         if ((!rp.for_toc || inset->isInToc()) && !inset_is_argument_elsewhere) {
3593                             // Arguments may need to be output
3594                                 OutputParams np = rp;
3595                                 np.local_font = &font;
3596
3597                                 // TODO: special case will bite here.
3598                                 np.docbook_in_par = true;
3599                                 inset->docbook(*xs, np);
3600                         }
3601                 } else {
3602                         char_type c = getUChar(buf.masterBuffer()->params(), rp, i);
3603                         if (lyx::isSpace(c) && !ignore_fonts)
3604                                 delayedChars.push_back(c);
3605                         else
3606                                 *xs << c;
3607                 }
3608                 font_old = font.fontInfo();
3609         }
3610
3611         // FIXME, this code is just imported from XHTML
3612         // I'm worried about what happens if a branch, say, is itself
3613         // wrapped in some font stuff. I think that will not work.
3614         if (!ignore_fonts)
3615                 xs->closeFontTags();
3616
3617         // Deal with the delayed characters *after* closing font tags.
3618         if (!delayedChars.empty())
3619                 for (char_type c: delayedChars)
3620                         *xs << c;
3621
3622         // In listings, new lines (i.e. \n characters in the output) are very important. Avoid generating one for the
3623         // last line to get a clean output.
3624         if (rp.docbook_in_listing && !is_last_par)
3625                 *xs << xml::CR();
3626
3627         // Finalise the last (and most likely only) paragraph.
3628         generatedParagraphs.push_back(os.str());
3629     os.str(from_ascii(""));
3630     delete xs;
3631
3632     // If there is an argument that must be output after the main tag, do it after handling the rest of the paragraph.
3633     for (pos_type i = initial; i < size(); ++i) {
3634         if (getInset(i) && getInset(i)->lyxCode() == ARG_CODE) {
3635             const InsetArgument * arg = getInset(i)->asInsetArgument();
3636             if (arg->docbookargumentaftermaintag()) {
3637                 // Don't use rp, as this argument would not generate anything.
3638                 auto xs_local = XMLStream(os);
3639                 arg->docbook(xs_local, runparams);
3640
3641                 appendedParagraphs.push_back(os.str());
3642                 os.str(from_ascii(""));
3643             }
3644         }
3645     }
3646
3647         return std::make_tuple(prependedParagraphs, generatedParagraphs, appendedParagraphs);
3648 }
3649
3650
3651 namespace {
3652
3653 void doFontSwitchXHTML(vector<xml::FontTag> & tagsToOpen,
3654                   vector<xml::EndFontTag> & tagsToClose,
3655                   bool & flag, FontState curstate, xml::FontTypes type)
3656 {
3657         if (curstate == FONT_ON) {
3658                 tagsToOpen.push_back(xhtmlStartFontTag(type));
3659                 flag = true;
3660         } else if (flag) {
3661                 tagsToClose.push_back(xhtmlEndFontTag(type));
3662                 flag = false;
3663         }
3664 }
3665
3666 } // anonymous namespace
3667
3668
3669 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
3670                                     XMLStream & xs,
3671                                     OutputParams const & runparams,
3672                                     Font const & outerfont,
3673                                     bool start_paragraph, bool close_paragraph,
3674                                     pos_type initial) const
3675 {
3676         docstring retval;
3677
3678         // track whether we have opened these tags
3679         bool emph_flag = false;
3680         bool bold_flag = false;
3681         bool noun_flag = false;
3682         bool ubar_flag = false;
3683         bool dbar_flag = false;
3684         bool sout_flag = false;
3685         bool xout_flag = false;
3686         bool wave_flag = false;
3687         // shape tags
3688         bool shap_flag = false;
3689         // family tags
3690         bool faml_flag = false;
3691         // size tags
3692         bool size_flag = false;
3693
3694         Layout const & style = *d->layout_;
3695
3696         if (start_paragraph)
3697                 xs.startDivision(allowEmpty());
3698
3699         FontInfo font_old =
3700                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
3701
3702         FontShape  curr_fs   = INHERIT_SHAPE;
3703         FontFamily curr_fam  = INHERIT_FAMILY;
3704         FontSize   curr_size = INHERIT_SIZE;
3705
3706         string const default_family =
3707                 buf.masterBuffer()->params().fonts_default_family;
3708
3709         vector<xml::FontTag> tagsToOpen;
3710         vector<xml::EndFontTag> tagsToClose;
3711
3712         // parsing main loop
3713         for (pos_type i = initial; i < size(); ++i) {
3714                 // let's not show deleted material in the output
3715                 if (isDeleted(i))
3716                         continue;
3717
3718                 Font const font = getFont(buf.masterBuffer()->params(), i, outerfont);
3719
3720                 // emphasis
3721                 FontState curstate = font.fontInfo().emph();
3722                 if (font_old.emph() != curstate)
3723                         doFontSwitchXHTML(tagsToOpen, tagsToClose, emph_flag, curstate, xml::FT_EMPH);
3724
3725                 // noun
3726                 curstate = font.fontInfo().noun();
3727                 if (font_old.noun() != curstate)
3728                         doFontSwitchXHTML(tagsToOpen, tagsToClose, noun_flag, curstate, xml::FT_NOUN);
3729
3730                 // underbar
3731                 curstate = font.fontInfo().underbar();
3732                 if (font_old.underbar() != curstate)
3733                         doFontSwitchXHTML(tagsToOpen, tagsToClose, ubar_flag, curstate, xml::FT_UBAR);
3734
3735                 // strikeout
3736                 curstate = font.fontInfo().strikeout();
3737                 if (font_old.strikeout() != curstate)
3738                         doFontSwitchXHTML(tagsToOpen, tagsToClose, sout_flag, curstate, xml::FT_SOUT);
3739
3740                 // xout
3741                 curstate = font.fontInfo().xout();
3742                 if (font_old.xout() != curstate)
3743                         doFontSwitchXHTML(tagsToOpen, tagsToClose, xout_flag, curstate, xml::FT_XOUT);
3744
3745                 // double underbar
3746                 curstate = font.fontInfo().uuline();
3747                 if (font_old.uuline() != curstate)
3748                         doFontSwitchXHTML(tagsToOpen, tagsToClose, dbar_flag, curstate, xml::FT_DBAR);
3749
3750                 // wavy line
3751                 curstate = font.fontInfo().uwave();
3752                 if (font_old.uwave() != curstate)
3753                         doFontSwitchXHTML(tagsToOpen, tagsToClose, wave_flag, curstate, xml::FT_WAVE);
3754
3755                 // bold
3756                 // a little hackish, but allows us to reuse what we have.
3757                 curstate = (font.fontInfo().series() == BOLD_SERIES ? FONT_ON : FONT_OFF);
3758                 if (font_old.series() != font.fontInfo().series())
3759                         doFontSwitchXHTML(tagsToOpen, tagsToClose, bold_flag, curstate, xml::FT_BOLD);
3760
3761                 // Font shape
3762                 curr_fs = font.fontInfo().shape();
3763                 FontShape old_fs = font_old.shape();
3764                 if (old_fs != curr_fs) {
3765                         if (shap_flag) {
3766                                 switch (old_fs) {
3767                                 case ITALIC_SHAPE:
3768                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_ITALIC));
3769                                         break;
3770                                 case SLANTED_SHAPE:
3771                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SLANTED));
3772                                         break;
3773                                 case SMALLCAPS_SHAPE:
3774                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SMALLCAPS));
3775                                         break;
3776                                 case UP_SHAPE:
3777                                 case INHERIT_SHAPE:
3778                                         break;
3779                                 default:
3780                                         // the other tags are for internal use
3781                                         LATTEST(false);
3782                                         break;
3783                                 }
3784                                 shap_flag = false;
3785                         }
3786                         switch (curr_fs) {
3787                         case ITALIC_SHAPE:
3788                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_ITALIC));
3789                                 shap_flag = true;
3790                                 break;
3791                         case SLANTED_SHAPE:
3792                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SLANTED));
3793                                 shap_flag = true;
3794                                 break;
3795                         case SMALLCAPS_SHAPE:
3796                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SMALLCAPS));
3797                                 shap_flag = true;
3798                                 break;
3799                         case UP_SHAPE:
3800                         case INHERIT_SHAPE:
3801                                 break;
3802                         default:
3803                                 // the other tags are for internal use
3804                                 LATTEST(false);
3805                                 break;
3806                         }
3807                 }
3808
3809                 // Font family
3810                 curr_fam = font.fontInfo().family();
3811                 FontFamily old_fam = font_old.family();
3812                 if (old_fam != curr_fam) {
3813                         if (faml_flag) {
3814                                 switch (old_fam) {
3815                                 case ROMAN_FAMILY:
3816                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_ROMAN));
3817                                         break;
3818                                 case SANS_FAMILY:
3819                                     tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SANS));
3820                                     break;
3821                                 case TYPEWRITER_FAMILY:
3822                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_TYPE));
3823                                         break;
3824                                 case INHERIT_FAMILY:
3825                                         break;
3826                                 default:
3827                                         // the other tags are for internal use
3828                                         LATTEST(false);
3829                                         break;
3830                                 }
3831                                 faml_flag = false;
3832                         }
3833                         switch (curr_fam) {
3834                         case ROMAN_FAMILY:
3835                                 // we will treat a "default" font family as roman, since we have
3836                                 // no other idea what to do.
3837                                 if (default_family != "rmdefault" && default_family != "default") {
3838                                         tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_ROMAN));
3839                                         faml_flag = true;
3840                                 }
3841                                 break;
3842                         case SANS_FAMILY:
3843                                 if (default_family != "sfdefault") {
3844                                         tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SANS));
3845                                         faml_flag = true;
3846                                 }
3847                                 break;
3848                         case TYPEWRITER_FAMILY:
3849                                 if (default_family != "ttdefault") {
3850                                         tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_TYPE));
3851                                         faml_flag = true;
3852                                 }
3853                                 break;
3854                         case INHERIT_FAMILY:
3855                                 break;
3856                         default:
3857                                 // the other tags are for internal use
3858                                 LATTEST(false);
3859                                 break;
3860                         }
3861                 }
3862
3863                 // Font size
3864                 curr_size = font.fontInfo().size();
3865                 FontSize old_size = font_old.size();
3866                 if (old_size != curr_size) {
3867                         if (size_flag) {
3868                                 switch (old_size) {
3869                                 case TINY_SIZE:
3870                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_TINY));
3871                                         break;
3872                                 case SCRIPT_SIZE:
3873                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_SCRIPT));
3874                                         break;
3875                                 case FOOTNOTE_SIZE:
3876                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_FOOTNOTE));
3877                                         break;
3878                                 case SMALL_SIZE:
3879                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_SMALL));
3880                                         break;
3881                                 case LARGE_SIZE:
3882                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_LARGE));
3883                                         break;
3884                                 case LARGER_SIZE:
3885                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_LARGER));
3886                                         break;
3887                                 case LARGEST_SIZE:
3888                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_LARGEST));
3889                                         break;
3890                                 case HUGE_SIZE:
3891                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_HUGE));
3892                                         break;
3893                                 case HUGER_SIZE:
3894                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_HUGER));
3895                                         break;
3896                                 case INCREASE_SIZE:
3897                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_INCREASE));
3898                                         break;
3899                                 case DECREASE_SIZE:
3900                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_DECREASE));
3901                                         break;
3902                                 case INHERIT_SIZE:
3903                                 case NORMAL_SIZE:
3904                                         break;
3905                                 default:
3906                                         // the other tags are for internal use
3907                                         LATTEST(false);
3908                                         break;
3909                                 }
3910                                 size_flag = false;
3911                         }
3912                         switch (curr_size) {
3913                         case TINY_SIZE:
3914                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_TINY));
3915                                 size_flag = true;
3916                                 break;
3917                         case SCRIPT_SIZE:
3918                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_SCRIPT));
3919                                 size_flag = true;
3920                                 break;
3921                         case FOOTNOTE_SIZE:
3922                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_FOOTNOTE));
3923                                 size_flag = true;
3924                                 break;
3925                         case SMALL_SIZE:
3926                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_SMALL));
3927                                 size_flag = true;
3928                                 break;
3929                         case LARGE_SIZE:
3930                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_LARGE));
3931                                 size_flag = true;
3932                                 break;
3933                         case LARGER_SIZE:
3934                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_LARGER));
3935                                 size_flag = true;
3936                                 break;
3937                         case LARGEST_SIZE:
3938                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_LARGEST));
3939                                 size_flag = true;
3940                                 break;
3941                         case HUGE_SIZE:
3942                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_HUGE));
3943                                 size_flag = true;
3944                                 break;
3945                         case HUGER_SIZE:
3946                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_HUGER));
3947                                 size_flag = true;
3948                                 break;
3949                         case INCREASE_SIZE:
3950                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_INCREASE));
3951                                 size_flag = true;
3952                                 break;
3953                         case DECREASE_SIZE:
3954                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_DECREASE));
3955                                 size_flag = true;
3956                                 break;
3957                         case INHERIT_SIZE:
3958                         case NORMAL_SIZE:
3959                                 break;
3960                         default:
3961                                 // the other tags are for internal use
3962                                 LATTEST(false);
3963                                 break;
3964                         }
3965                 }
3966
3967                 // FIXME XHTML
3968                 // Other such tags? What about the other text ranges?
3969
3970                 vector<xml::EndFontTag>::const_iterator cit = tagsToClose.begin();
3971                 vector<xml::EndFontTag>::const_iterator cen = tagsToClose.end();
3972                 for (; cit != cen; ++cit)
3973                         xs << *cit;
3974
3975                 vector<xml::FontTag>::const_iterator sit = tagsToOpen.begin();
3976                 vector<xml::FontTag>::const_iterator sen = tagsToOpen.end();
3977                 for (; sit != sen; ++sit)
3978                         xs << *sit;
3979
3980                 tagsToClose.clear();
3981                 tagsToOpen.clear();
3982
3983                 Inset const * inset = getInset(i);
3984                 if (inset) {
3985                         if (!runparams.for_toc || inset->isInToc()) {
3986                                 OutputParams np = runparams;
3987                                 np.local_font = &font;
3988                                 // If the paragraph has size 1, then we are in the "special
3989                                 // case" where we do not output the containing paragraph info
3990                                 if (!inset->getLayout().htmlisblock() && size() != 1)
3991                                         np.html_in_par = true;
3992                                 retval += inset->xhtml(xs, np);
3993                         }
3994                 } else {
3995                         char_type c = getUChar(buf.masterBuffer()->params(),
3996                                                runparams, i);
3997                         if (c == ' ' && (style.free_spacing || runparams.free_spacing))
3998                                 xs << XMLStream::ESCAPE_NONE << "&nbsp;";
3999                         else
4000                                 xs << c;
4001                 }
4002                 font_old = font.fontInfo();
4003         }
4004
4005         // FIXME XHTML
4006         // I'm worried about what happens if a branch, say, is itself
4007         // wrapped in some font stuff. I think that will not work.
4008         xs.closeFontTags();
4009         if (close_paragraph)
4010                 xs.endDivision();
4011
4012         return retval;
4013 }
4014
4015
4016 bool Paragraph::isHfill(pos_type pos) const
4017 {
4018         Inset const * inset = getInset(pos);
4019         return inset && inset->isHfill();
4020 }
4021
4022
4023 bool Paragraph::isNewline(pos_type pos) const
4024 {
4025         // U+2028 LINE SEPARATOR
4026         // U+2029 PARAGRAPH SEPARATOR
4027         char_type const c = d->text_[pos];
4028         if (c == 0x2028 || c == 0x2029)
4029                 return true;
4030         Inset const * inset = getInset(pos);
4031         return inset && inset->lyxCode() == NEWLINE_CODE;
4032 }
4033
4034
4035 bool Paragraph::isEnvSeparator(pos_type pos) const
4036 {
4037         Inset const * inset = getInset(pos);
4038         return inset && inset->lyxCode() == SEPARATOR_CODE;
4039 }
4040
4041
4042 bool Paragraph::isLineSeparator(pos_type pos) const
4043 {
4044         char_type const c = d->text_[pos];
4045         if (isLineSeparatorChar(c))
4046                 return true;
4047         Inset const * inset = getInset(pos);
4048         return inset && inset->isLineSeparator();
4049 }
4050
4051
4052 bool Paragraph::isWordSeparator(pos_type pos, bool const ignore_deleted) const
4053 {
4054         if (pos == size())
4055                 return true;
4056         if (ignore_deleted && isDeleted(pos))
4057                 return false;
4058         if (Inset const * inset = getInset(pos))
4059                 return !inset->isLetter();
4060         // if we have a hard hyphen (no en- or emdash) or apostrophe
4061         // we pass this to the spell checker
4062         // FIXME: this method is subject to change, visit
4063         // https://bugzilla.mozilla.org/show_bug.cgi?id=355178
4064         // to get an impression how complex this is.
4065         if (isHardHyphenOrApostrophe(pos))
4066                 return false;
4067         char_type const c = d->text_[pos];
4068         // We want to pass the escape chars to the spellchecker
4069         docstring const escape_chars = from_utf8(lyxrc.spellchecker_esc_chars);
4070         return !isLetterChar(c) && !isDigitASCII(c) && !contains(escape_chars, c);
4071 }
4072
4073
4074 bool Paragraph::isHardHyphenOrApostrophe(pos_type pos) const
4075 {
4076         pos_type const psize = size();
4077         if (pos >= psize)
4078                 return false;
4079         char_type const c = d->text_[pos];
4080         if (c != '-' && c != '\'')
4081                 return false;
4082         pos_type nextpos = pos + 1;
4083         pos_type prevpos = pos > 0 ? pos - 1 : 0;
4084         if ((nextpos == psize || isSpace(nextpos))
4085                 && (pos == 0 || isSpace(prevpos)))
4086                 return false;
4087         return true;
4088 }
4089
4090
4091 bool Paragraph::needsCProtection(bool const fragile) const
4092 {
4093         // first check the layout of the paragraph, but only in insets
4094         InsetText const * textinset = inInset().asInsetText();
4095         bool const maintext = textinset
4096                 ? textinset->text().isMainText()
4097                 : false;
4098
4099         if (!maintext && layout().needcprotect) {
4100                 // Environments need cprotection regardless the content
4101                 if (layout().latextype == LATEX_ENVIRONMENT)
4102                         return true;
4103
4104                 // Commands need cprotection if they contain specific chars
4105                 int const nchars_escape = 9;
4106                 static char_type const chars_escape[nchars_escape] = {
4107                         '&', '_', '$', '%', '#', '^', '{', '}', '\\'};
4108
4109                 docstring const pars = asString();
4110                 for (int k = 0; k < nchars_escape; k++) {
4111                         if (contains(pars, chars_escape[k]))
4112                                 return true;
4113                 }
4114         }
4115
4116         // now check whether we have insets that need cprotection
4117         for (auto const & icit : d->insetlist_) {
4118                 Inset const * ins = icit.inset;
4119                 if (!ins)
4120                         continue;
4121                 if (ins->needsCProtection(maintext, fragile))
4122                         return true;
4123         }
4124
4125         return false;
4126 }
4127
4128
4129 FontSpan const & Paragraph::getSpellRange(pos_type pos) const
4130 {
4131         return d->speller_state_.getRange(pos);
4132 }
4133
4134
4135 bool Paragraph::isChar(pos_type pos) const
4136 {
4137         if (Inset const * inset = getInset(pos))
4138                 return inset->isChar();
4139         char_type const c = d->text_[pos];
4140         return !isLetterChar(c) && !isDigitASCII(c) && !lyx::isSpace(c);
4141 }
4142
4143
4144 bool Paragraph::isSpace(pos_type pos) const
4145 {
4146         if (Inset const * inset = getInset(pos))
4147                 return inset->isSpace();
4148         char_type const c = d->text_[pos];
4149         return lyx::isSpace(c);
4150 }
4151
4152
4153 Language const *
4154 Paragraph::getParLanguage(BufferParams const & bparams) const
4155 {
4156         if (!empty())
4157                 return getFirstFontSettings(bparams).language();
4158         // FIXME: we should check the prev par as well (Lgb)
4159         return bparams.language;
4160 }
4161
4162
4163 bool Paragraph::isRTL(BufferParams const & bparams) const
4164 {
4165         return getParLanguage(bparams)->rightToLeft()
4166                 && !inInset().getLayout().forceLTR();
4167 }
4168
4169
4170 void Paragraph::changeLanguage(BufferParams const & bparams,
4171                                Language const * from, Language const * to)
4172 {
4173         // change language including dummy font change at the end
4174         for (pos_type i = 0; i <= size(); ++i) {
4175                 Font font = getFontSettings(bparams, i);
4176                 if (font.language() == from) {
4177                         font.setLanguage(to);
4178                         setFont(i, font);
4179                         d->requestSpellCheck(i);
4180                 }
4181         }
4182 }
4183
4184
4185 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
4186 {
4187         Language const * doc_language = bparams.language;
4188         for (auto const & f : d->fontlist_)
4189                 if (f.font().language() != ignore_language &&
4190                     f.font().language() != latex_language &&
4191                     f.font().language() != doc_language)
4192                         return true;
4193         return false;
4194 }
4195
4196
4197 void Paragraph::getLanguages(std::set<Language const *> & langs) const
4198 {
4199         for (auto const & f : d->fontlist_) {
4200                 Language const * lang = f.font().language();
4201                 if (lang != ignore_language &&
4202                     lang != latex_language)
4203                         langs.insert(lang);
4204         }
4205 }
4206
4207
4208 docstring Paragraph::asString(int options) const
4209 {
4210         return asString(0, size(), options);
4211 }
4212
4213
4214 docstring Paragraph::asString(pos_type beg, pos_type end, int options, const OutputParams *runparams) const
4215 {
4216         odocstringstream os;
4217
4218         if (beg == 0
4219             && options & AS_STR_LABEL
4220             && !d->params_.labelString().empty())
4221                 os << d->params_.labelString() << ' ';
4222
4223         for (pos_type i = beg; i < end; ++i) {
4224                 if ((options & AS_STR_SKIPDELETE) && isDeleted(i))
4225                         continue;
4226                 char_type const c = d->text_[i];
4227                 if (isPrintable(c) || c == '\t'
4228                     || (c == '\n' && (options & AS_STR_NEWLINES)))
4229                         os.put(c);
4230                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
4231                         if (c == META_INSET && (options & AS_STR_PLAINTEXT)) {
4232                                 LASSERT(runparams != nullptr, return docstring());
4233                                 getInset(i)->plaintext(os, *runparams);
4234                         } else if (c == META_INSET && (options & AS_STR_MATHED)
4235                                    && getInset(i)->lyxCode() == REF_CODE) {
4236                                 Buffer const & buf = getInset(i)->buffer();
4237                                 OutputParams rp(&buf.params().encoding());
4238                                 Font const font(inherit_font, buf.params().language);
4239                                 rp.local_font = &font;
4240                                 otexstream ots(os);
4241                                 getInset(i)->latex(ots, rp);
4242                         } else {
4243                                 getInset(i)->toString(os);
4244                         }
4245                 }
4246         }
4247
4248         return os.str();
4249 }
4250
4251
4252 void Paragraph::forOutliner(docstring & os, size_t const maxlen,
4253                             bool const shorten, bool const label) const
4254 {
4255         size_t tmplen = shorten ? maxlen + 1 : maxlen;
4256         if (label && !labelString().empty())
4257                 os += labelString() + ' ';
4258         if (!layout().isTocCaption())
4259                 return;
4260         for (pos_type i = 0; i < size() && os.length() < tmplen; ++i) {
4261                 if (isDeleted(i))
4262                         continue;
4263                 char_type const c = d->text_[i];
4264                 if (isPrintable(c))
4265                         os += c;
4266                 else if (c == META_INSET)
4267                         getInset(i)->forOutliner(os, tmplen, false);
4268         }
4269         if (shorten)
4270                 Text::shortenForOutliner(os, maxlen);
4271 }
4272
4273
4274 void Paragraph::setInsetOwner(Inset const * inset)
4275 {
4276         d->inset_owner_ = inset;
4277 }
4278
4279
4280 int Paragraph::id() const
4281 {
4282         return d->id_;
4283 }
4284
4285
4286 void Paragraph::setId(int id)
4287 {
4288         d->id_ = id;
4289 }
4290
4291
4292 Layout const & Paragraph::layout() const
4293 {
4294         return *d->layout_;
4295 }
4296
4297
4298 void Paragraph::setLayout(Layout const & layout)
4299 {
4300         d->layout_ = &layout;
4301 }
4302
4303
4304 void Paragraph::setDefaultLayout(DocumentClass const & tc)
4305 {
4306         setLayout(tc.defaultLayout());
4307 }
4308
4309
4310 void Paragraph::setPlainLayout(DocumentClass const & tc)
4311 {
4312         setLayout(tc.plainLayout());
4313 }
4314
4315
4316 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tc)
4317 {
4318         if (usePlainLayout())
4319                 setPlainLayout(tc);
4320         else
4321                 setDefaultLayout(tc);
4322 }
4323
4324
4325 Inset const & Paragraph::inInset() const
4326 {
4327         LBUFERR(d->inset_owner_);
4328         return *d->inset_owner_;
4329 }
4330
4331
4332 ParagraphParameters & Paragraph::params()
4333 {
4334         return d->params_;
4335 }
4336
4337
4338 ParagraphParameters const & Paragraph::params() const
4339 {
4340         return d->params_;
4341 }
4342
4343
4344 bool Paragraph::isFreeSpacing() const
4345 {
4346         if (d->layout_->free_spacing)
4347                 return true;
4348         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
4349 }
4350
4351
4352 bool Paragraph::allowEmpty() const
4353 {
4354         if (d->layout_->keepempty)
4355                 return true;
4356         return d->inset_owner_ && d->inset_owner_->allowEmpty();
4357 }
4358
4359
4360 bool Paragraph::brokenBiblio() const
4361 {
4362         // There is a problem if there is no bibitem at position 0 in
4363         // paragraphs that need one, if there is another bibitem in the
4364         // paragraph or if this paragraph is not supposed to have
4365         // a bibitem inset at all.
4366         return ((d->layout_->labeltype == LABEL_BIBLIO
4367                 && (d->insetlist_.find(BIBITEM_CODE) != 0
4368                     || d->insetlist_.find(BIBITEM_CODE, 1) > 0))
4369                 || (d->layout_->labeltype != LABEL_BIBLIO
4370                     && d->insetlist_.find(BIBITEM_CODE) != -1));
4371 }
4372
4373
4374 int Paragraph::fixBiblio(Buffer const & buffer)
4375 {
4376         // FIXME: when there was already an inset at 0, the return value is 1,
4377         // which does not tell whether another inset has been removed; the
4378         // cursor cannot be correctly updated.
4379
4380         bool const track_changes = buffer.params().track_changes;
4381         int bibitem_pos = d->insetlist_.find(BIBITEM_CODE);
4382
4383         // The case where paragraph is not BIBLIO
4384         if (d->layout_->labeltype != LABEL_BIBLIO) {
4385                 if (bibitem_pos == -1)
4386                         // No InsetBibitem => OK
4387                         return 0;
4388                 // There is an InsetBibitem: remove it!
4389                 d->insetlist_.release(bibitem_pos);
4390                 eraseChar(bibitem_pos, track_changes);
4391                 return (bibitem_pos == 0) ? -1 : -bibitem_pos;
4392         }
4393
4394         bool const hasbibitem0 = bibitem_pos == 0;
4395         if (hasbibitem0) {
4396                 bibitem_pos = d->insetlist_.find(BIBITEM_CODE, 1);
4397                 // There was an InsetBibitem at pos 0,
4398                 // and no other one => OK
4399                 if (bibitem_pos == -1)
4400                         return 0;
4401                 // there is a bibitem at the 0 position, but since
4402                 // there is a second one, we copy the second on the
4403                 // first. We're assuming there are at most two of
4404                 // these, which there should be.
4405                 // FIXME: why does it make sense to do that rather
4406                 // than keep the first? (JMarc)
4407                 Inset * inset = releaseInset(bibitem_pos);
4408                 d->insetlist_.begin()->inset = inset;
4409                 // This needs to be done to update the counter (#8499)
4410                 buffer.updateBuffer();
4411                 return -bibitem_pos;
4412         }
4413
4414         // We need to create an inset at the beginning
4415         Inset * inset = nullptr;
4416         if (bibitem_pos > 0) {
4417                 // there was one somewhere in the paragraph, let's move it
4418                 inset = d->insetlist_.release(bibitem_pos);
4419                 eraseChar(bibitem_pos, track_changes);
4420         } else
4421                 // make a fresh one
4422                 inset = new InsetBibitem(const_cast<Buffer *>(&buffer),
4423                                          InsetCommandParams(BIBITEM_CODE));
4424
4425         Font font(inherit_font, buffer.params().language);
4426         insertInset(0, inset, font, Change(track_changes ? Change::INSERTED
4427                                                    : Change::UNCHANGED));
4428
4429         // This is needed to get the counters right
4430         buffer.updateBuffer();
4431         return 1;
4432 }
4433
4434
4435 void Paragraph::checkAuthors(AuthorList const & authorList)
4436 {
4437         d->changes_.checkAuthors(authorList);
4438 }
4439
4440
4441 bool Paragraph::isChanged(pos_type pos) const
4442 {
4443         return lookupChange(pos).changed();
4444 }
4445
4446
4447 bool Paragraph::isInserted(pos_type pos) const
4448 {
4449         return lookupChange(pos).inserted();
4450 }
4451
4452
4453 bool Paragraph::isDeleted(pos_type pos) const
4454 {
4455         return lookupChange(pos).deleted();
4456 }
4457
4458
4459 InsetList const & Paragraph::insetList() const
4460 {
4461         return d->insetlist_;
4462 }
4463
4464
4465 void Paragraph::setInsetBuffers(Buffer & b)
4466 {
4467         d->insetlist_.setBuffer(b);
4468 }
4469
4470
4471 void Paragraph::resetBuffer()
4472 {
4473         d->insetlist_.resetBuffer();
4474 }
4475
4476
4477 Inset * Paragraph::releaseInset(pos_type pos)
4478 {
4479         Inset * inset = d->insetlist_.release(pos);
4480         /// does not honour change tracking!
4481         eraseChar(pos, false);
4482         return inset;
4483 }
4484
4485
4486 Inset * Paragraph::getInset(pos_type pos)
4487 {
4488         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
4489                  ? d->insetlist_.get(pos) : nullptr;
4490 }
4491
4492
4493 Inset const * Paragraph::getInset(pos_type pos) const
4494 {
4495         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
4496                  ? d->insetlist_.get(pos) : nullptr;
4497 }
4498
4499
4500 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
4501                 pos_type & right, TextCase action)
4502 {
4503         // process sequences of modified characters; in change
4504         // tracking mode, this approach results in much better
4505         // usability than changing case on a char-by-char basis
4506         // We also need to track the current font, since font
4507         // changes within sequences can occur.
4508         vector<pair<char_type, Font> > changes;
4509
4510         bool const trackChanges = bparams.track_changes;
4511
4512         bool capitalize = true;
4513
4514         for (; pos < right; ++pos) {
4515                 char_type oldChar = d->text_[pos];
4516                 char_type newChar = oldChar;
4517
4518                 // ignore insets and don't play with deleted text!
4519                 if (oldChar != META_INSET && !isDeleted(pos)) {
4520                         switch (action) {
4521                                 case text_lowercase:
4522                                         newChar = lowercase(oldChar);
4523                                         break;
4524                                 case text_capitalization:
4525                                         if (capitalize) {
4526                                                 newChar = uppercase(oldChar);
4527                                                 capitalize = false;
4528                                         }
4529                                         break;
4530                                 case text_uppercase:
4531                                         newChar = uppercase(oldChar);
4532                                         break;
4533                         }
4534                 }
4535
4536                 if (isWordSeparator(pos) || isDeleted(pos)) {
4537                         // permit capitalization again
4538                         capitalize = true;
4539                 }
4540
4541                 if (oldChar != newChar) {
4542                         changes.push_back(make_pair(newChar, getFontSettings(bparams, pos)));
4543                         if (pos != right - 1)
4544                                 continue;
4545                         // step behind the changing area
4546                         pos++;
4547                 }
4548
4549                 int erasePos = pos - changes.size();
4550                 for (auto const & change : changes) {
4551                         insertChar(pos, change.first, change.second, trackChanges);
4552                         if (!eraseChar(erasePos, trackChanges)) {
4553                                 ++erasePos;
4554                                 ++pos; // advance
4555                                 ++right; // expand selection
4556                         }
4557                 }
4558                 changes.clear();
4559         }
4560 }
4561
4562
4563 int Paragraph::find(docstring const & str, bool cs, bool mw,
4564                 pos_type start_pos, bool del) const
4565 {
4566         pos_type pos = start_pos;
4567         int const strsize = str.length();
4568         int i = 0;
4569         pos_type const parsize = d->text_.size();
4570         for (i = 0; i < strsize && pos < parsize; ++i, ++pos) {
4571                 // ignore deleted matter
4572                 if (!del && isDeleted(pos)) {
4573                         if (pos == parsize - 1)
4574                                 break;
4575                         pos++;
4576                         --i;
4577                         continue;
4578                 }
4579                 // Ignore "invisible" letters such as ligature breaks
4580                 // and hyphenation chars while searching
4581                 bool nonmatch = false;
4582                 while (pos < parsize && isInset(pos)) {
4583                         Inset const * inset = getInset(pos);
4584                         if (!inset->isLetter() && !inset->isChar())
4585                                 break;
4586                         odocstringstream os;
4587                         inset->toString(os);
4588                         docstring const insetstring = os.str();
4589                         if (!insetstring.empty()) {
4590                                 int const insetstringsize = insetstring.length();
4591                                 for (int j = 0; j < insetstringsize && pos < parsize; ++i, ++j) {
4592                                         if ((cs && str[i] != insetstring[j])
4593                                             || (!cs && uppercase(str[i]) != uppercase(insetstring[j]))) {
4594                                                 nonmatch = true;
4595                                                 break;
4596                                         }
4597                                 }
4598                         }
4599                         pos++;
4600                 }
4601                 if (nonmatch || i == strsize)
4602                         break;
4603                 if (cs && str[i] != d->text_[pos])
4604                         break;
4605                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos]))
4606                         break;
4607         }
4608
4609         if (i != strsize)
4610                 return 0;
4611
4612         // if necessary, check whether string matches word
4613         if (mw) {
4614                 if (start_pos > 0 && !isWordSeparator(start_pos - 1))
4615                         return 0;
4616                 if (pos < parsize
4617                         && !isWordSeparator(pos))
4618                         return 0;
4619         }
4620
4621         return pos - start_pos;
4622 }
4623
4624
4625 char_type Paragraph::getChar(pos_type pos) const
4626 {
4627         return d->text_[pos];
4628 }
4629
4630
4631 pos_type Paragraph::size() const
4632 {
4633         return d->text_.size();
4634 }
4635
4636
4637 bool Paragraph::empty() const
4638 {
4639         return d->text_.empty();
4640 }
4641
4642
4643 bool Paragraph::isInset(pos_type pos) const
4644 {
4645         return d->text_[pos] == META_INSET;
4646 }
4647
4648
4649 bool Paragraph::isSeparator(pos_type pos) const
4650 {
4651         //FIXME: Are we sure this can be the only separator?
4652         return d->text_[pos] == ' ';
4653 }
4654
4655
4656 void Paragraph::deregisterWords()
4657 {
4658         Private::LangWordsMap::const_iterator itl = d->words_.begin();
4659         Private::LangWordsMap::const_iterator ite = d->words_.end();
4660         for (; itl != ite; ++itl) {
4661                 WordList & wl = theWordList(itl->first);
4662                 Private::Words::const_iterator it = (itl->second).begin();
4663                 Private::Words::const_iterator et = (itl->second).end();
4664                 for (; it != et; ++it)
4665                         wl.remove(*it);
4666         }
4667         d->words_.clear();
4668 }
4669
4670
4671 void Paragraph::locateWord(pos_type & from, pos_type & to,
4672         word_location const loc, bool const ignore_deleted) const
4673 {
4674         switch (loc) {
4675         case WHOLE_WORD_STRICT:
4676                 if (from == 0 || from == size()
4677                     || isWordSeparator(from, ignore_deleted)
4678                     || isWordSeparator(from - 1, ignore_deleted)) {
4679                         to = from;
4680                         return;
4681                 }
4682                 // fall through
4683
4684         case WHOLE_WORD:
4685                 // If we are already at the beginning of a word, do nothing
4686                 if (!from || isWordSeparator(from - 1, ignore_deleted))
4687                         break;
4688                 // fall through
4689
4690         case PREVIOUS_WORD:
4691                 // always move the cursor to the beginning of previous word
4692                 while (from && !isWordSeparator(from - 1, ignore_deleted))
4693                         --from;
4694                 break;
4695         case NEXT_WORD:
4696                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
4697                 break;
4698         case PARTIAL_WORD:
4699                 // no need to move the 'from' cursor
4700                 break;
4701         }
4702         to = from;
4703         while (to < size() && !isWordSeparator(to, ignore_deleted))
4704                 ++to;
4705 }
4706
4707
4708 void Paragraph::collectWords()
4709 {
4710         for (pos_type pos = 0; pos < size(); ++pos) {
4711                 if (isWordSeparator(pos))
4712                         continue;
4713                 pos_type from = pos;
4714                 locateWord(from, pos, WHOLE_WORD);
4715                 // Work around MSVC warning: The statement
4716                 // if (pos < from + lyxrc.completion_minlength)
4717                 // triggers a signed vs. unsigned warning.
4718                 // I don't know why this happens, it could be a MSVC bug, or
4719                 // related to LLP64 (windows) vs. LP64 (unix) programming
4720                 // model, or the C++ standard might be ambigous in the section
4721                 // defining the "usual arithmetic conversions". However, using
4722                 // a temporary variable is safe and works on all compilers.
4723                 pos_type const endpos = from + lyxrc.completion_minlength;
4724                 if (pos < endpos)
4725                         continue;
4726                 FontList::const_iterator cit = d->fontlist_.fontIterator(from);
4727                 if (cit == d->fontlist_.end())
4728                         return;
4729                 Language const * lang = cit->font().language();
4730                 docstring const word = asString(from, pos, AS_STR_NONE);
4731                 d->words_[lang->lang()].insert(word);
4732         }
4733 }
4734
4735
4736 void Paragraph::registerWords()
4737 {
4738         Private::LangWordsMap::const_iterator itl = d->words_.begin();
4739         Private::LangWordsMap::const_iterator ite = d->words_.end();
4740         for (; itl != ite; ++itl) {
4741                 WordList & wl = theWordList(itl->first);
4742                 Private::Words::const_iterator it = (itl->second).begin();
4743                 Private::Words::const_iterator et = (itl->second).end();
4744                 for (; it != et; ++it)
4745                         wl.insert(*it);
4746         }
4747 }
4748
4749
4750 void Paragraph::updateWords()
4751 {
4752         deregisterWords();
4753         collectWords();
4754         registerWords();
4755 }
4756
4757
4758 void Paragraph::Private::appendSkipPosition(SkipPositions & skips, pos_type const pos) const
4759 {
4760         SkipPositionsIterator begin = skips.begin();
4761         SkipPositions::iterator end = skips.end();
4762         if (pos > 0 && begin < end) {
4763                 --end;
4764                 if (end->last == pos - 1) {
4765                         end->last = pos;
4766                         return;
4767                 }
4768         }
4769         skips.insert(end, FontSpan(pos, pos));
4770 }
4771
4772
4773 Language * Paragraph::Private::locateSpellRange(
4774         pos_type & from, pos_type & to,
4775         SkipPositions & skips) const
4776 {
4777         // skip leading white space
4778         while (from < to && owner_->isWordSeparator(from))
4779                 ++from;
4780         // don't check empty range
4781         if (from >= to)
4782                 return nullptr;
4783         // get current language
4784         Language * lang = getSpellLanguage(from);
4785         pos_type last = from;
4786         bool samelang = true;
4787         bool sameinset = true;
4788         while (last < to && samelang && sameinset) {
4789                 // hop to end of word
4790                 while (last < to && !owner_->isWordSeparator(last)) {
4791                         Inset const * inset = owner_->getInset(last);
4792                         if (inset && dynamic_cast<const InsetSpecialChar *>(inset)) {
4793                                 // check for "invisible" letters such as ligature breaks
4794                                 odocstringstream os;
4795                                 inset->toString(os);
4796                                 if (os.str().length() != 0) {
4797                                         // avoid spell check of visible special char insets
4798                                         // stop the loop in front of the special char inset
4799                                         sameinset = false;
4800                                         break;
4801                                 }
4802                         } else if (inset) {
4803                                 appendSkipPosition(skips, last);
4804                         } else if (owner_->isDeleted(last)) {
4805                                 appendSkipPosition(skips, last);
4806                         }
4807                         ++last;
4808                 }
4809                 // hop to next word while checking for insets
4810                 while (sameinset && last < to && owner_->isWordSeparator(last)) {
4811                         if (Inset const * inset = owner_->getInset(last))
4812                                 sameinset = inset->isChar() && inset->isLetter();
4813                         if (sameinset && owner_->isDeleted(last)) {
4814                                 appendSkipPosition(skips, last);
4815                         }
4816                         if (sameinset)
4817                                 last++;
4818                 }
4819                 if (sameinset && last < to) {
4820                         // now check for language change
4821                         samelang = lang == getSpellLanguage(last);
4822                 }
4823         }
4824         // if language change detected backstep is needed
4825         if (!samelang)
4826                 --last;
4827         to = last;
4828         return lang;
4829 }
4830
4831
4832 Language * Paragraph::Private::getSpellLanguage(pos_type const from) const
4833 {
4834         Language * lang =
4835                 const_cast<Language *>(owner_->getFontSettings(
4836                         inset_owner_->buffer().params(), from).language());
4837         if (lang == inset_owner_->buffer().params().language
4838                 && !lyxrc.spellchecker_alt_lang.empty()) {
4839                 string lang_code;
4840                 string const lang_variety =
4841                         split(lyxrc.spellchecker_alt_lang, lang_code, '-');
4842                 lang->setCode(lang_code);
4843                 lang->setVariety(lang_variety);
4844         }
4845         return lang;
4846 }
4847
4848
4849 void Paragraph::requestSpellCheck(pos_type pos)
4850 {
4851         d->requestSpellCheck(pos);
4852         if (pos == -1) {
4853                 // Also request spellcheck within (text) insets
4854                 for (auto const & insets : insetList()) {
4855                         if (!insets.inset->asInsetText())
4856                                 continue;
4857                         ParagraphList & inset_pars =
4858                                 insets.inset->asInsetText()->paragraphs();
4859                         ParagraphList::iterator pit = inset_pars.begin();
4860                         ParagraphList::iterator pend = inset_pars.end();
4861                         for (; pit != pend; ++pit)
4862                                 pit->requestSpellCheck();
4863                 }
4864         }
4865 }
4866
4867
4868 bool Paragraph::needsSpellCheck() const
4869 {
4870         SpellChecker::ChangeNumber speller_change_number = 0;
4871         if (theSpellChecker())
4872                 speller_change_number = theSpellChecker()->changeNumber();
4873         if (speller_change_number > d->speller_state_.currentChangeNumber()) {
4874                 d->speller_state_.needsCompleteRefresh(speller_change_number);
4875         }
4876         return d->needsSpellCheck();
4877 }
4878
4879
4880 bool Paragraph::Private::ignoreWord(docstring const & word) const
4881 {
4882         // Ignore words with digits
4883         // FIXME: make this customizable
4884         // (note that some checkers ignore words with digits by default)
4885         docstring::const_iterator cit = word.begin();
4886         docstring::const_iterator const end = word.end();
4887         for (; cit != end; ++cit) {
4888                 if (isNumber((*cit)))
4889                         return true;
4890         }
4891         return false;
4892 }
4893
4894
4895 SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
4896         WordLangTuple & wl, docstring_list & suggestions,
4897         bool do_suggestion, bool check_learned) const
4898 {
4899         SpellChecker::Result result = SpellChecker::WORD_OK;
4900         SpellChecker * speller = theSpellChecker();
4901         if (!speller)
4902                 return result;
4903
4904         if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
4905                 return result;
4906
4907         locateWord(from, to, WHOLE_WORD, true);
4908         if (from == to || from >= size())
4909                 return result;
4910
4911         docstring word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
4912         Language * lang = d->getSpellLanguage(from);
4913
4914         BufferParams const & bparams = d->inset_owner_->buffer().params();
4915
4916         if (getFontSettings(bparams, from).fontInfo().nospellcheck() == FONT_ON)
4917                 return result;
4918
4919         wl = WordLangTuple(word, lang);
4920
4921         if (word.empty())
4922                 return result;
4923
4924         if (needsSpellCheck() || check_learned) {
4925                 pos_type end = to;
4926                 if (!d->ignoreWord(word)) {
4927                         bool const trailing_dot = to < size() && d->text_[to] == '.';
4928                         result = speller->check(wl, bparams.spellignore());
4929                         if (SpellChecker::misspelled(result) && trailing_dot) {
4930                                 wl = WordLangTuple(word.append(from_ascii(".")), lang);
4931                                 result = speller->check(wl, bparams.spellignore());
4932                                 if (!SpellChecker::misspelled(result)) {
4933                                         LYXERR(Debug::GUI, "misspelled word is correct with dot: \"" <<
4934                                            word << "\" [" <<
4935                                            from << ".." << to << "]");
4936                                 } else {
4937                                         // spell check with dot appended failed too
4938                                         // restore original word/lang value
4939                                         word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
4940                                         wl = WordLangTuple(word, lang);
4941                                 }
4942                         }
4943                 }
4944                 if (!SpellChecker::misspelled(result)) {
4945                         // area up to the begin of the next word is not misspelled
4946                         while (end < size() && isWordSeparator(end))
4947                                 ++end;
4948                 }
4949                 d->setMisspelled(from, end, result);
4950         } else {
4951                 result = d->speller_state_.getState(from);
4952         }
4953
4954         if (do_suggestion)
4955                 suggestions.clear();
4956
4957         if (SpellChecker::misspelled(result)) {
4958                 LYXERR(Debug::GUI, "misspelled word: \"" <<
4959                            word << "\" [" <<
4960                            from << ".." << to << "]");
4961                 if (do_suggestion)
4962                         speller->suggest(wl, suggestions);
4963         }
4964         return result;
4965 }
4966
4967
4968 void Paragraph::anonymize()
4969 {
4970         // This is a very crude anonymization for now
4971         for (char_type & c : d->text_)
4972                 if (isLetterChar(c) || isNumber(c))
4973                         c = 'a';
4974 }
4975
4976
4977 void Paragraph::Private::markMisspelledWords(
4978         Language const * lang,
4979         pos_type const & first, pos_type const & last,
4980         SpellChecker::Result result,
4981         docstring const & word,
4982         SkipPositions const & skips)
4983 {
4984         if (!SpellChecker::misspelled(result)) {
4985                 setMisspelled(first, last, SpellChecker::WORD_OK);
4986                 return;
4987         }
4988         int snext = first;
4989         SpellChecker * speller = theSpellChecker();
4990         // locate and enumerate the error positions
4991         int nerrors = speller->numMisspelledWords();
4992         int numskipped = 0;
4993         SkipPositionsIterator it = skips.begin();
4994         SkipPositionsIterator et = skips.end();
4995         for (int index = 0; index < nerrors; ++index) {
4996                 int wstart;
4997                 int wlen = 0;
4998                 speller->misspelledWord(index, wstart, wlen);
4999                 /// should not happen if speller supports range checks
5000                 if (!wlen)
5001                         continue;
5002                 WordLangTuple const candidate(word.substr(wstart, wlen), lang);
5003                 wstart += first + numskipped;
5004                 if (snext < wstart) {
5005                         /// mark the range of correct spelling
5006                         numskipped += countSkips(it, et, wstart);
5007                         setMisspelled(snext,
5008                                 wstart - 1, SpellChecker::WORD_OK);
5009                 }
5010                 snext = wstart + wlen;
5011                 // Check whether the candidate is in the document's local dict
5012                 SpellChecker::Result actresult = result;
5013                 if (inset_owner_->buffer().params().spellignored(candidate))
5014                         actresult = SpellChecker::DOCUMENT_LEARNED_WORD;
5015                 numskipped += countSkips(it, et, snext);
5016                 /// mark the range of misspelling
5017                 setMisspelled(wstart, snext, actresult);
5018                 if (actresult == SpellChecker::DOCUMENT_LEARNED_WORD)
5019                         LYXERR(Debug::GUI, "local dictionary word: \"" <<
5020                                    candidate.word() << "\" [" <<
5021                                    wstart << ".." << (snext-1) << "]");
5022                 else
5023                         LYXERR(Debug::GUI, "misspelled word: \"" <<
5024                                    candidate.word() << "\" [" <<
5025                                    wstart << ".." << (snext-1) << "]");
5026                 ++snext;
5027         }
5028         if (snext <= last) {
5029                 /// mark the range of correct spelling at end
5030                 setMisspelled(snext, last, SpellChecker::WORD_OK);
5031         }
5032 }
5033
5034
5035 void Paragraph::spellCheck() const
5036 {
5037         SpellChecker * speller = theSpellChecker();
5038         if (!speller || empty() ||!needsSpellCheck())
5039                 return;
5040         pos_type start;
5041         pos_type endpos;
5042         d->rangeOfSpellCheck(start, endpos);
5043         if (speller->canCheckParagraph()) {
5044                 // loop until we leave the range
5045                 for (pos_type first = start; first < endpos; ) {
5046                         pos_type last = endpos;
5047                         Private::SkipPositions skips;
5048                         Language * lang = d->locateSpellRange(first, last, skips);
5049                         if (first >= endpos)
5050                                 break;
5051                         // start the spell checker on the unit of meaning
5052                         docstring word = asString(first, last, AS_STR_INSETS + AS_STR_SKIPDELETE);
5053                         WordLangTuple wl = WordLangTuple(word, lang);
5054                         BufferParams const & bparams = d->inset_owner_->buffer().params();
5055                         SpellChecker::Result result = !word.empty() ?
5056                                 speller->check(wl, bparams.spellignore()) : SpellChecker::WORD_OK;
5057                         d->markMisspelledWords(lang, first, last, result, word, skips);
5058                         first = ++last;
5059                 }
5060         } else {
5061                 static docstring_list suggestions;
5062                 pos_type to = endpos;
5063                 while (start < endpos) {
5064                         WordLangTuple wl;
5065                         spellCheck(start, to, wl, suggestions, false);
5066                         start = to + 1;
5067                 }
5068         }
5069         d->readySpellCheck();
5070 }
5071
5072
5073 bool Paragraph::isMisspelled(pos_type pos, bool check_boundary) const
5074 {
5075         bool result = SpellChecker::misspelled(d->speller_state_.getState(pos));
5076         if (result || pos <= 0 || pos > size())
5077                 return result;
5078         if (check_boundary && (pos == size() || isWordSeparator(pos)))
5079                 result = SpellChecker::misspelled(d->speller_state_.getState(pos - 1));
5080         return result;
5081 }
5082
5083
5084 string Paragraph::magicLabel() const
5085 {
5086         stringstream ss;
5087         ss << "magicparlabel-" << id();
5088         return ss.str();
5089 }
5090
5091
5092 } // namespace lyx