src/Paragraph.cpp

   1 /**
   2  * \file Paragraph.cpp
   3  * This file is part of LyX, the document processor.
   4  * Licence details can be found in the file COPYING.
   5  *
   6  * \author Asger Alstrup
   7  * \author Lars Gullik Bjønnes
   8  * \author Richard Heck (XHTML output)
   9  * \author Jean-Marc Lasgouttes
  10  * \author Angus Leeming
  11  * \author John Levon
  12  * \author André Pönitz
  13  * \author Dekel Tsur
  14  * \author Jürgen Vigna
  15  *
  16  * Full author contact details are available in file CREDITS.
  17  */
  18
  19 #include <config.h>
  20
  21 #include "Paragraph.h"
  22
  23 #include "LayoutFile.h"
  24 #include "Buffer.h"
  25 #include "BufferParams.h"
  26 #include "Changes.h"
  27 #include "Counters.h"
  28 #include "BufferEncodings.h"
  29 #include "InsetList.h"
  30 #include "Language.h"
  31 #include "LaTeXFeatures.h"
  32 #include "Layout.h"
  33 #include "Font.h"
  34 #include "FontList.h"
  35 #include "LyXRC.h"
  36 #include "OutputParams.h"
  37 #include "output_latex.h"
  38 #include "output_xhtml.h"
  39 #include "output_docbook.h"
  40 #include "ParagraphParameters.h"
  41 #include "SpellChecker.h"
  42 #include "xml.h"
  43 #include "texstream.h"
  44 #include "TextClass.h"
  45 #include "TexRow.h"
  46 #include "Text.h"
  47 #include "WordLangTuple.h"
  48 #include "WordList.h"
  49
  50 #include "frontends/alert.h"
  51
  52 #include "insets/InsetBibitem.h"
  53 #include "insets/InsetLabel.h"
  54 #include "insets/InsetSpecialChar.h"
  55 #include "insets/InsetText.h"
  56
  57 #include "mathed/InsetMathHull.h"
  58
  59 #include "support/debug.h"
  60 #include "support/docstring_list.h"
  61 #include "support/ExceptionMessage.h"
  62 #include "support/gettext.h"
  63 #include "support/lassert.h"
  64 #include "support/Length.h"
  65 #include "support/lstrings.h"
  66 #include "support/textutils.h"
  67 #include "output_docbook.h"
  68
  69 #include <atomic>
  70 #include <sstream>
  71 #include <vector>
  72
  73 using namespace std;
  74 using namespace lyx::support;
  75
  76 // OSX clang, gcc < 4.8.0, and msvc < 2015 do not support C++11 thread_local
  77 #if defined(__APPLE__) || (defined(__GNUC__) && __GNUC__ == 4 && __GNUC_MINOR__ < 8)
  78 #define THREAD_LOCAL_STATIC static __thread
  79 #elif defined(_MSC_VER) && (_MSC_VER < 1900)
  80 #define THREAD_LOCAL_STATIC static __declspec(thread)
  81 #else
  82 #define THREAD_LOCAL_STATIC thread_local static
  83 #endif
  84
  85 namespace lyx {
  86
  87 /////////////////////////////////////////////////////////////////////
  88 //
  89 // SpellResultRange
  90 //
  91 /////////////////////////////////////////////////////////////////////
  92
  93 class SpellResultRange {
  94 public:
  95         SpellResultRange(FontSpan range, SpellChecker::Result result)
  96         : range_(range), result_(result)
  97         {}
  98         ///
  99         FontSpan const & range() const { return range_; }
 100         ///
 101         void range(FontSpan const & r) { range_ = r; }
 102         ///
 103         SpellChecker::Result result() const { return result_; }
 104         ///
 105         void result(SpellChecker::Result r) { result_ = r; }
 106         ///
 107         bool contains(pos_type pos) const { return range_.contains(pos); }
 108         ///
 109         bool covered(FontSpan const & r) const
 110         {
 111                 // 1. first of new range inside current range or
 112                 // 2. last of new range inside current range or
 113                 // 3. first of current range inside new range or
 114                 // 4. last of current range inside new range
 115                 //FIXME: is this the same as !range_.intersect(r).empty() ?
 116                 return range_.contains(r.first) || range_.contains(r.last) ||
 117                         r.contains(range_.first) || r.contains(range_.last);
 118         }
 119         ///
 120         void shift(pos_type pos, int offset)
 121         {
 122                 if (range_.first > pos) {
 123                         range_.first += offset;
 124                         range_.last += offset;
 125                 } else if (range_.last >= pos) {
 126                         range_.last += offset;
 127                 }
 128         }
 129 private:
 130         FontSpan range_ ;
 131         SpellChecker::Result result_ ;
 132 };
 133
 134
 135 /////////////////////////////////////////////////////////////////////
 136 //
 137 // SpellCheckerState
 138 //
 139 /////////////////////////////////////////////////////////////////////
 140
 141 class SpellCheckerState {
 142 public:
 143         SpellCheckerState()
 144         {
 145                 needs_refresh_ = true;
 146                 current_change_number_ = 0;
 147         }
 148
 149         void setRange(FontSpan const & fp, SpellChecker::Result state)
 150         {
 151                 Ranges result;
 152                 RangesIterator et = ranges_.end();
 153                 RangesIterator it = ranges_.begin();
 154                 for (; it != et; ++it) {
 155                         if (!it->covered(fp))
 156                                 result.push_back(SpellResultRange(it->range(), it->result()));
 157                         else if (state == SpellChecker::WORD_OK) {
 158                                 // trim or split the current misspelled range
 159                                 // store misspelled ranges only
 160                                 FontSpan range = it->range();
 161                                 if (fp.first > range.first) {
 162                                         // misspelled area in front of WORD_OK
 163                                         range.last = fp.first - 1;
 164                                         result.push_back(SpellResultRange(range, it->result()));
 165                                         range = it->range();
 166                                 }
 167                                 if (fp.last < range.last) {
 168                                         // misspelled area after WORD_OK range
 169                                         range.first = fp.last + 1;
 170                                         result.push_back(SpellResultRange(range, it->result()));
 171                                 }
 172                         }
 173                 }
 174                 ranges_ = result;
 175                 if (state != SpellChecker::WORD_OK)
 176                         ranges_.push_back(SpellResultRange(fp, state));
 177         }
 178
 179         void increasePosAfterPos(pos_type pos)
 180         {
 181                 correctRangesAfterPos(pos, 1);
 182                 needsRefresh(pos);
 183         }
 184
 185         void decreasePosAfterPos(pos_type pos)
 186         {
 187                 correctRangesAfterPos(pos, -1);
 188                 needsRefresh(pos);
 189         }
 190
 191         void refreshLast(pos_type pos)
 192         {
 193                 if (pos < refresh_.last)
 194                         refresh_.last = pos;
 195         }
 196
 197         SpellChecker::Result getState(pos_type pos) const
 198         {
 199                 SpellChecker::Result result = SpellChecker::WORD_OK;
 200                 RangesIterator et = ranges_.end();
 201                 RangesIterator it = ranges_.begin();
 202                 for (; it != et; ++it) {
 203                         if(it->contains(pos)) {
 204                                 return it->result();
 205                         }
 206                 }
 207                 return result;
 208         }
 209
 210         FontSpan const & getRange(pos_type pos) const
 211         {
 212                 /// empty span to indicate mismatch
 213                 static FontSpan empty_;
 214                 RangesIterator et = ranges_.end();
 215                 RangesIterator it = ranges_.begin();
 216                 for (; it != et; ++it) {
 217                         if(it->contains(pos)) {
 218                                 return it->range();
 219                         }
 220                 }
 221                 return empty_;
 222         }
 223
 224         bool needsRefresh() const
 225         {
 226                 return needs_refresh_;
 227         }
 228
 229         SpellChecker::ChangeNumber currentChangeNumber() const
 230         {
 231                 return current_change_number_;
 232         }
 233
 234         void refreshRange(pos_type & first, pos_type & last) const
 235         {
 236                 first = refresh_.first;
 237                 last = refresh_.last;
 238         }
 239
 240         void needsRefresh(pos_type pos)
 241         {
 242                 if (needs_refresh_ && pos != -1) {
 243                         if (pos < refresh_.first)
 244                                 refresh_.first = pos;
 245                         if (pos > refresh_.last)
 246                                 refresh_.last = pos;
 247                 } else if (pos != -1) {
 248                         // init request check for neighbour positions too
 249                         refresh_.first = pos > 0 ? pos - 1 : 0;
 250                         // no need for special end of paragraph check
 251                         refresh_.last = pos + 1;
 252                 }
 253                 needs_refresh_ = pos != -1;
 254         }
 255
 256         void needsCompleteRefresh(SpellChecker::ChangeNumber change_number)
 257         {
 258                 needs_refresh_ = true;
 259                 refresh_.first = 0;
 260                 refresh_.last = -1;
 261                 current_change_number_ = change_number;
 262         }
 263 private:
 264         typedef vector<SpellResultRange> Ranges;
 265         typedef Ranges::const_iterator RangesIterator;
 266         Ranges ranges_;
 267         /// the area of the paragraph with pending spell check
 268         FontSpan refresh_;
 269         /// spell state cache version number
 270         SpellChecker::ChangeNumber current_change_number_;
 271         bool needs_refresh_;
 272
 273
 274         void correctRangesAfterPos(pos_type pos, int offset)
 275         {
 276                 RangesIterator et = ranges_.end();
 277                 Ranges::iterator it = ranges_.begin();
 278                 for (; it != et; ++it) {
 279                         it->shift(pos, offset);
 280                 }
 281         }
 282
 283 };
 284
 285 /////////////////////////////////////////////////////////////////////
 286 //
 287 // Paragraph::Private
 288 //
 289 /////////////////////////////////////////////////////////////////////
 290
 291 class Paragraph::Private
 292 {
 293         // Enforce our own "copy" constructor
 294         Private(Private const &) = delete;
 295         Private & operator=(Private const &) = delete;
 296         // Unique ID generator
 297         static int make_id();
 298 public:
 299         ///
 300         Private(Paragraph * owner, Layout const & layout);
 301         /// "Copy constructor"
 302         Private(Private const &, Paragraph * owner);
 303         /// Copy constructor from \p beg  to \p end
 304         Private(Private const &, Paragraph * owner, pos_type beg, pos_type end);
 305
 306         ///
 307         void insertChar(pos_type pos, char_type c, Change const & change);
 308
 309         /// Output the surrogate pair formed by \p c and \p next to \p os.
 310         /// \return the number of characters written.
 311         int latexSurrogatePair(BufferParams const &, otexstream & os,
 312                                char_type c, char_type next,
 313                                OutputParams const &);
 314
 315         /// Output a space in appropriate formatting (or a surrogate pair
 316         /// if the next character is a combining character).
 317         /// \return whether a surrogate pair was output.
 318         bool simpleTeXBlanks(BufferParams const &,
 319                              OutputParams const &,
 320                              otexstream &,
 321                              pos_type i,
 322                              unsigned int & column,
 323                              Font const & font,
 324                              Layout const & style);
 325
 326         /// This could go to ParagraphParameters if we want to.
 327         int startTeXParParams(BufferParams const &, otexstream &,
 328                               OutputParams const &) const;
 329
 330         /// This could go to ParagraphParameters if we want to.
 331         bool endTeXParParams(BufferParams const &, otexstream &,
 332                              OutputParams const &) const;
 333
 334         ///
 335         void latexInset(BufferParams const &,
 336                                    otexstream &,
 337                                    OutputParams &,
 338                                    Font & running_font,
 339                                    Font & basefont,
 340                                    Font const & outerfont,
 341                                    bool & open_font,
 342                                    Change & running_change,
 343                                    Layout const & style,
 344                                    pos_type & i,
 345                                    unsigned int & column);
 346
 347         ///
 348         void latexSpecialChar(
 349                                    otexstream & os,
 350                                    BufferParams const & bparams,
 351                                    OutputParams const & runparams,
 352                                    Font const & running_font,
 353                                    string & alien_script,
 354                                    Layout const & style,
 355                                    pos_type & i,
 356                                    pos_type end_pos,
 357                                    unsigned int & column);
 358
 359         ///
 360         bool latexSpecialT1(
 361                 char_type const c,
 362                 otexstream & os,
 363                 pos_type i,
 364                 unsigned int & column);
 365         ///
 366         bool latexSpecialTU(
 367                 char_type const c,
 368                 otexstream & os,
 369                 pos_type i,
 370                 unsigned int & column);
 371         ///
 372         bool latexSpecialT3(
 373                 char_type const c,
 374                 otexstream & os,
 375                 pos_type i,
 376                 unsigned int & column);
 377
 378         ///
 379         void validate(LaTeXFeatures & features) const;
 380
 381         /// Checks if the paragraph contains only text and no inset or font change.
 382         bool onlyText(Buffer const & buf, Font const & outerfont,
 383                       pos_type initial) const;
 384
 385         /// a vector of speller skip positions
 386         typedef vector<FontSpan> SkipPositions;
 387         typedef SkipPositions::const_iterator SkipPositionsIterator;
 388
 389         void appendSkipPosition(SkipPositions & skips, pos_type const pos) const;
 390
 391         Language * getSpellLanguage(pos_type const from) const;
 392
 393         Language * locateSpellRange(pos_type & from, pos_type & to,
 394                                     SkipPositions & skips) const;
 395
 396         bool hasSpellerChange() const
 397         {
 398                 SpellChecker::ChangeNumber speller_change_number = 0;
 399                 if (theSpellChecker())
 400                         speller_change_number = theSpellChecker()->changeNumber();
 401                 return speller_change_number > speller_state_.currentChangeNumber();
 402         }
 403
 404         bool ignoreWord(docstring const & word) const ;
 405
 406         void setMisspelled(pos_type from, pos_type to, SpellChecker::Result state)
 407         {
 408                 pos_type textsize = owner_->size();
 409                 // check for sane arguments
 410                 if (to <= from || from >= textsize)
 411                         return;
 412                 FontSpan fp = FontSpan(from, to - 1);
 413                 speller_state_.setRange(fp, state);
 414         }
 415
 416         void requestSpellCheck(pos_type pos)
 417         {
 418                 if (pos == -1)
 419                         speller_state_.needsCompleteRefresh(speller_state_.currentChangeNumber());
 420                 else
 421                         speller_state_.needsRefresh(pos);
 422         }
 423
 424         void readySpellCheck()
 425         {
 426                 speller_state_.needsRefresh(-1);
 427         }
 428
 429         bool needsSpellCheck() const
 430         {
 431                 return speller_state_.needsRefresh();
 432         }
 433
 434         void rangeOfSpellCheck(pos_type & first, pos_type & last) const
 435         {
 436                 speller_state_.refreshRange(first, last);
 437                 if (last == -1) {
 438                         last = owner_->size();
 439                         return;
 440                 }
 441                 pos_type endpos = last;
 442                 owner_->locateWord(first, endpos, WHOLE_WORD, true);
 443                 if (endpos < last) {
 444                         endpos = last;
 445                         owner_->locateWord(last, endpos, WHOLE_WORD, true);
 446                 }
 447                 last = endpos;
 448         }
 449
 450         int countSkips(SkipPositionsIterator & it, SkipPositionsIterator const et,
 451                             int & start) const
 452         {
 453                 int numskips = 0;
 454                 while (it != et && it->first < start) {
 455                         long skip = it->last - it->first + 1;
 456                         start += skip;
 457                         numskips += skip;
 458                         ++it;
 459                 }
 460                 return numskips;
 461         }
 462
 463         void markMisspelledWords(pos_type const & first, pos_type const & last,
 464                                                          SpellChecker::Result result,
 465                                                          docstring const & word,
 466                                                          SkipPositions const & skips);
 467
 468         InsetCode ownerCode() const
 469         {
 470                 return inset_owner_ ? inset_owner_->lyxCode() : NO_CODE;
 471         }
 472
 473         /// Which Paragraph owns us?
 474         Paragraph * owner_;
 475
 476         /// In which Inset?
 477         Inset const * inset_owner_;
 478
 479         ///
 480         FontList fontlist_;
 481
 482         ///
 483         ParagraphParameters params_;
 484
 485         /// for recording and looking up changes
 486         Changes changes_;
 487
 488         ///
 489         InsetList insetlist_;
 490
 491         /// end of label
 492         pos_type begin_of_body_;
 493
 494         typedef docstring TextContainer;
 495         ///
 496         TextContainer text_;
 497
 498         typedef set<docstring> Words;
 499         typedef map<string, Words> LangWordsMap;
 500         ///
 501         LangWordsMap words_;
 502         ///
 503         Layout const * layout_;
 504         ///
 505         SpellCheckerState speller_state_;
 506         ///
 507         int id_;
 508 };
 509
 510
 511 Paragraph::Private::Private(Paragraph * owner, Layout const & layout)
 512         : owner_(owner), inset_owner_(nullptr), begin_of_body_(0), layout_(&layout), id_(-1)
 513 {
 514         text_.reserve(100);
 515 }
 516
 517
 518 //static
 519 int Paragraph::Private::make_id()
 520 {
 521         // The id is unique per session across buffers because it is used in
 522         // LFUN_PARAGRAPH_GOTO to switch to a different buffer, for instance in the
 523         // outliner.
 524         // (thread-safe)
 525         static int next_id(0);
 526         return next_id++;
 527 }
 528
 529
 530 Paragraph::Private::Private(Private const & p, Paragraph * owner)
 531         : owner_(owner), inset_owner_(p.inset_owner_), fontlist_(p.fontlist_),
 532           params_(p.params_), changes_(p.changes_), insetlist_(p.insetlist_),
 533           begin_of_body_(p.begin_of_body_), text_(p.text_), words_(p.words_),
 534           layout_(p.layout_), id_(make_id())
 535 {
 536         requestSpellCheck(p.text_.size());
 537 }
 538
 539
 540 Paragraph::Private::Private(Private const & p, Paragraph * owner,
 541         pos_type beg, pos_type end)
 542         : owner_(owner), inset_owner_(p.inset_owner_),
 543           params_(p.params_), changes_(p.changes_),
 544           insetlist_(p.insetlist_, beg, end),
 545           begin_of_body_(p.begin_of_body_), words_(p.words_),
 546           layout_(p.layout_), id_(make_id())
 547 {
 548         if (beg >= pos_type(p.text_.size()))
 549                 return;
 550         text_ = p.text_.substr(beg, end - beg);
 551
 552         FontList::const_iterator fcit = fontlist_.begin();
 553         FontList::const_iterator fend = fontlist_.end();
 554         for (; fcit != fend; ++fcit) {
 555                 if (fcit->pos() < beg)
 556                         continue;
 557                 if (fcit->pos() >= end) {
 558                         // Add last entry in the fontlist_.
 559                         fontlist_.set(text_.size() - 1, fcit->font());
 560                         break;
 561                 }
 562                 // Add a new entry in the fontlist_.
 563                 fontlist_.set(fcit->pos() - beg, fcit->font());
 564         }
 565         requestSpellCheck(p.text_.size());
 566 }
 567
 568
 569 void Paragraph::addChangesToToc(DocIterator const & cdit, Buffer const & buf,
 570                                 bool output_active, TocBackend & backend) const
 571 {
 572         d->changes_.addToToc(cdit, buf, output_active, backend);
 573 }
 574
 575
 576 bool Paragraph::isDeleted(pos_type start, pos_type end) const
 577 {
 578         LASSERT(start >= 0 && start <= size(), return false);
 579         LASSERT(end > start && end <= size() + 1, return false);
 580
 581         return d->changes_.isDeleted(start, end);
 582 }
 583
 584
 585 bool Paragraph::isChanged(pos_type start, pos_type end) const
 586 {
 587         LASSERT(start >= 0 && start <= size(), return false);
 588         LASSERT(end > start && end <= size() + 1, return false);
 589
 590         return d->changes_.isChanged(start, end);
 591 }
 592
 593 // FIXME: Ideally the diverse isChanged() methods should account for that!
 594 bool Paragraph::hasChangedInsets(pos_type start, pos_type end) const
 595 {
 596         LASSERT(start >= 0 && start <= size(), return false);
 597         LASSERT(end > start && end <= size() + 1, return false);
 598
 599         for (auto const & icit : d->insetlist_) {
 600                 if (icit.pos < start)
 601                         continue;
 602                 if (icit.pos >= end)
 603                         break;
 604                 if (icit.inset && icit.inset->isChanged())
 605                         return true;
 606         }
 607         return false;
 608 }
 609
 610 bool Paragraph::isChanged() const
 611 {
 612         return d->changes_.isChanged();
 613 }
 614
 615
 616 bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const
 617 {
 618         // keep the logic here in sync with the logic of eraseChars()
 619         if (!trackChanges)
 620                 return true;
 621
 622         Change const & change = d->changes_.lookup(size());
 623         return change.inserted() && change.currentAuthor();
 624 }
 625
 626 Change Paragraph::parEndChange() const
 627 {
 628         return d->changes_.lookup(size());
 629 }
 630
 631
 632 void Paragraph::setChange(Change const & change)
 633 {
 634         // beware of the imaginary end-of-par character!
 635         d->changes_.set(change, 0, size() + 1);
 636
 637         /*
 638          * Propagate the change recursively - but not in case of DELETED!
 639          *
 640          * Imagine that your co-author makes changes in an existing inset. He
 641          * sends your document to you and you come to the conclusion that the
 642          * inset should go completely. If you erase it, LyX must not delete all
 643          * text within the inset. Otherwise, the change tracked insertions of
 644          * your co-author get lost and there is no way to restore them later.
 645          *
 646          * Conclusion: An inset's content should remain untouched if you delete it
 647          */
 648
 649         if (!change.deleted()) {
 650                 for (pos_type pos = 0; pos < size(); ++pos) {
 651                         if (Inset * inset = getInset(pos))
 652                                 inset->setChange(change);
 653                 }
 654         }
 655 }
 656
 657
 658 void Paragraph::setChange(pos_type pos, Change const & change)
 659 {
 660         LASSERT(pos >= 0 && pos <= size(), return);
 661         d->changes_.set(change, pos);
 662
 663         // see comment in setChange(Change const &) above
 664         if (!change.deleted() && pos < size())
 665                 if (Inset * inset = getInset(pos))
 666                         inset->setChange(change);
 667 }
 668
 669
 670 Change const & Paragraph::lookupChange(pos_type pos) const
 671 {
 672         LBUFERR(pos >= 0 && pos <= size());
 673         return d->changes_.lookup(pos);
 674 }
 675
 676
 677 void Paragraph::acceptChanges(pos_type start, pos_type end)
 678 {
 679         LASSERT(start >= 0 && start <= size(), return);
 680         LASSERT(end > start && end <= size() + 1, return);
 681
 682         for (pos_type pos = start; pos < end; ++pos) {
 683                 switch (lookupChange(pos).type) {
 684                         case Change::UNCHANGED:
 685                                 // accept changes in nested inset
 686                                 if (Inset * inset = getInset(pos))
 687                                         inset->acceptChanges();
 688                                 break;
 689
 690                         case Change::INSERTED:
 691                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 692                                 // also accept changes in nested inset
 693                                 if (Inset * inset = getInset(pos))
 694                                         inset->acceptChanges();
 695                                 break;
 696
 697                         case Change::DELETED:
 698                                 // Suppress access to non-existent
 699                                 // "end-of-paragraph char"
 700                                 if (pos < size()) {
 701                                         eraseChar(pos, false);
 702                                         --end;
 703                                         --pos;
 704                                 }
 705                                 break;
 706                 }
 707
 708         }
 709 }
 710
 711
 712 void Paragraph::rejectChanges(pos_type start, pos_type end)
 713 {
 714         LASSERT(start >= 0 && start <= size(), return);
 715         LASSERT(end > start && end <= size() + 1, return);
 716
 717         for (pos_type pos = start; pos < end; ++pos) {
 718                 switch (lookupChange(pos).type) {
 719                         case Change::UNCHANGED:
 720                                 // reject changes in nested inset
 721                                 if (Inset * inset = getInset(pos))
 722                                                 inset->rejectChanges();
 723                                 break;
 724
 725                         case Change::INSERTED:
 726                                 // Suppress access to non-existent
 727                                 // "end-of-paragraph char"
 728                                 if (pos < size()) {
 729                                         eraseChar(pos, false);
 730                                         --end;
 731                                         --pos;
 732                                 }
 733                                 break;
 734
 735                         case Change::DELETED:
 736                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 737
 738                                 // Do NOT reject changes within a deleted inset!
 739                                 // There may be insertions of a co-author inside of it!
 740
 741                                 break;
 742                 }
 743         }
 744 }
 745
 746
 747 void Paragraph::Private::insertChar(pos_type pos, char_type c,
 748                 Change const & change)
 749 {
 750         LASSERT(pos >= 0 && pos <= int(text_.size()), return);
 751
 752         // track change
 753         changes_.insert(change, pos);
 754
 755         // This is actually very common when parsing buffers (and
 756         // maybe inserting ascii text)
 757         if (pos == pos_type(text_.size())) {
 758                 // when appending characters, no need to update tables
 759                 text_.push_back(c);
 760                 // but we want spell checking
 761                 requestSpellCheck(pos);
 762                 return;
 763         }
 764
 765         text_.insert(text_.begin() + pos, c);
 766
 767         // Update the font table.
 768         fontlist_.increasePosAfterPos(pos);
 769
 770         // Update the insets
 771         insetlist_.increasePosAfterPos(pos);
 772
 773         // Update list of misspelled positions
 774         speller_state_.increasePosAfterPos(pos);
 775 }
 776
 777
 778 bool Paragraph::insertInset(pos_type pos, Inset * inset,
 779                                    Font const & font, Change const & change)
 780 {
 781         LASSERT(inset, return false);
 782         LASSERT(pos >= 0 && pos <= size(), return false);
 783
 784         // Paragraph::insertInset() can be used in cut/copy/paste operation where
 785         // d->inset_owner_ is not set yet.
 786         if (d->inset_owner_ && !d->inset_owner_->insetAllowed(inset->lyxCode()))
 787                 return false;
 788
 789         d->insertChar(pos, META_INSET, change);
 790         LASSERT(d->text_[pos] == META_INSET, return false);
 791
 792         // Add a new entry in the insetlist_.
 793         d->insetlist_.insert(inset, pos);
 794
 795         // Some insets require run of spell checker
 796         requestSpellCheck(pos);
 797         setFont(pos, font);
 798         return true;
 799 }
 800
 801
 802 bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
 803 {
 804         LASSERT(pos >= 0 && pos <= size(), return false);
 805
 806         // keep the logic here in sync with the logic of isMergedOnEndOfParDeletion()
 807
 808         if (trackChanges) {
 809                 Change change = d->changes_.lookup(pos);
 810
 811                 // set the character to DELETED if
 812                 //  a) it was previously unchanged or
 813                 //  b) it was inserted by a co-author
 814
 815                 if (!change.changed() ||
 816                       (change.inserted() && !change.currentAuthor())) {
 817                         setChange(pos, Change(Change::DELETED));
 818                         // request run of spell checker
 819                         requestSpellCheck(pos);
 820                         return false;
 821                 }
 822
 823                 if (change.deleted())
 824                         return false;
 825         }
 826
 827         // Don't physically access the imaginary end-of-paragraph character.
 828         // eraseChar() can only mark it as DELETED. A physical deletion of
 829         // end-of-par must be handled externally.
 830         if (pos == size()) {
 831                 return false;
 832         }
 833
 834         // track change
 835         d->changes_.erase(pos);
 836
 837         // if it is an inset, delete the inset entry
 838         if (d->text_[pos] == META_INSET)
 839                 d->insetlist_.erase(pos);
 840
 841         d->text_.erase(d->text_.begin() + pos);
 842
 843         // Update the fontlist_
 844         d->fontlist_.erase(pos);
 845
 846         // Update the insetlist_
 847         d->insetlist_.decreasePosAfterPos(pos);
 848
 849         // Update list of misspelled positions
 850         d->speller_state_.decreasePosAfterPos(pos);
 851         d->speller_state_.refreshLast(size());
 852
 853         return true;
 854 }
 855
 856
 857 int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
 858 {
 859         LASSERT(start >= 0 && start <= size(), return 0);
 860         LASSERT(end >= start && end <= size() + 1, return 0);
 861
 862         pos_type i = start;
 863         for (pos_type count = end - start; count; --count) {
 864                 if (!eraseChar(i, trackChanges))
 865                         ++i;
 866         }
 867         return end - i;
 868 }
 869
 870 // Handle combining characters
 871 int Paragraph::Private::latexSurrogatePair(BufferParams const & bparams,
 872                 otexstream & os, char_type c, char_type next,
 873                 OutputParams const & runparams)
 874 {
 875         // Writing next here may circumvent a possible font change between
 876         // c and next. Since next is only output if it forms a surrogate pair
 877         // with c we can ignore this:
 878         // A font change inside a surrogate pair does not make sense and is
 879         // hopefully impossible to input.
 880         // FIXME: change tracking
 881         // Is this correct WRT change tracking?
 882         Encoding const & encoding = *(runparams.encoding);
 883         docstring latex1 = encoding.latexChar(next).first;
 884         if (runparams.inIPA) {
 885                 string const tipashortcut = Encodings::TIPAShortcut(next);
 886                 if (!tipashortcut.empty()) {
 887                         latex1 = from_ascii(tipashortcut);
 888                 }
 889         }
 890         docstring latex2 = encoding.latexChar(c).first;
 891
 892         if (bparams.useNonTeXFonts || docstring(1, next) == latex1) {
 893                 // Encoding supports the combination:
 894                 // output as is (combining char after base char).
 895                 os << latex2 << latex1;
 896                 return latex1.length() + latex2.length();
 897         }
 898
 899         os << latex1 << "{" << latex2 << "}";
 900         return latex1.length() + latex2.length() + 2;
 901 }
 902
 903
 904 bool Paragraph::Private::simpleTeXBlanks(BufferParams const & bparams,
 905                                        OutputParams const & runparams,
 906                                        otexstream & os,
 907                                        pos_type i,
 908                                        unsigned int & column,
 909                                        Font const & font,
 910                                        Layout const & style)
 911 {
 912         if (style.pass_thru || runparams.pass_thru)
 913                 return false;
 914
 915         if (i + 1 < int(text_.size())) {
 916                 char_type next = text_[i + 1];
 917                 if (Encodings::isCombiningChar(next)) {
 918                         // This space has an accent, so we must always output it.
 919                         column += latexSurrogatePair(bparams, os, ' ', next, runparams) - 1;
 920                         return true;
 921                 }
 922         }
 923
 924         if (runparams.linelen > 0
 925             && column > runparams.linelen
 926             && i
 927             && text_[i - 1] != ' '
 928             && (i + 1 < int(text_.size()))
 929             // same in FreeSpacing mode
 930             && !owner_->isFreeSpacing()
 931             // In typewriter mode, we want to avoid
 932             // ! . ? : at the end of a line
 933             && !(font.fontInfo().family() == TYPEWRITER_FAMILY
 934                  && (text_[i - 1] == '.'
 935                      || text_[i - 1] == '?'
 936                      || text_[i - 1] == ':'
 937                      || text_[i - 1] == '!'))) {
 938                 os << '\n';
 939                 os.texrow().start(owner_->id(), i + 1);
 940                 column = 0;
 941         } else if (style.free_spacing) {
 942                 os << '~';
 943         } else {
 944                 os << ' ';
 945         }
 946         return false;
 947 }
 948
 949
 950 void Paragraph::Private::latexInset(BufferParams const & bparams,
 951                                     otexstream & os,
 952                                     OutputParams & runparams,
 953                                     Font & running_font,
 954                                     Font & basefont,
 955                                     Font const & outerfont,
 956                                     bool & open_font,
 957                                     Change & running_change,
 958                                     Layout const & style,
 959                                     pos_type & i,
 960                                     unsigned int & column)
 961 {
 962         Inset * inset = owner_->getInset(i);
 963         LBUFERR(inset);
 964
 965         if (style.pass_thru) {
 966                 odocstringstream ods;
 967                 inset->plaintext(ods, runparams);
 968                 os << ods.str();
 969                 return;
 970         }
 971
 972         // FIXME: move this to InsetNewline::latex
 973         if (inset->lyxCode() == NEWLINE_CODE || inset->lyxCode() == SEPARATOR_CODE) {
 974                 // newlines are handled differently here than
 975                 // the default in simpleTeXSpecialChars().
 976                 if (!style.newline_allowed) {
 977                         os << '\n';
 978                 } else {
 979                         if (open_font) {
 980                                 bool needPar = false;
 981                                 column += running_font.latexWriteEndChanges(
 982                                         os, bparams, runparams,
 983                                         basefont, basefont, needPar);
 984                                 open_font = false;
 985                         }
 986
 987                         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY)
 988                                 os << '~';
 989
 990                         basefont = owner_->getLayoutFont(bparams, outerfont);
 991                         running_font = basefont;
 992
 993                         if (runparams.moving_arg)
 994                                 os << "\\protect ";
 995
 996                 }
 997                 os.texrow().start(owner_->id(), i + 1);
 998                 column = 0;
 999         }
1000
1001         if (owner_->isDeleted(i)) {
1002                 if( ++runparams.inDeletedInset == 1)
1003                         runparams.changeOfDeletedInset = owner_->lookupChange(i);
1004         }
1005
1006         if (inset->canTrackChanges()) {
1007                 column += Changes::latexMarkChange(os, bparams, running_change,
1008                         Change(Change::UNCHANGED), runparams);
1009                 running_change = Change(Change::UNCHANGED);
1010         }
1011
1012         bool close = false;
1013         odocstream::pos_type const len = os.os().tellp();
1014
1015         if (inset->forceLTR(runparams)
1016             && running_font.isRightToLeft()
1017             // ERT is an exception, it should be output with no
1018             // decorations at all
1019             && inset->lyxCode() != ERT_CODE) {
1020                 if (runparams.use_polyglossia) {
1021                         os << "\\LRE{";
1022                 } else if (running_font.language()->lang() == "farsi"
1023                            || running_font.language()->lang() == "arabic_arabi")
1024                         os << "\\textLR{" << termcmd;
1025                 else
1026                         os << "\\L{";
1027                 close = true;
1028         }
1029
1030         if (open_font && (!inset->inheritFont() || inset->allowMultiPar())) {
1031                 // Some insets cannot be inside a font change command.
1032                 // However, even such insets *can* be placed in \L or \R
1033                 // or their equivalents (for RTL language switches),
1034                 // so we don't close the language in those cases
1035                 // (= differing isRightToLeft()).
1036                 // ArabTeX, though, doesn't seem to handle this special behavior.
1037                 bool const inRLSwitch =
1038                                 basefont.isRightToLeft() != running_font.isRightToLeft()
1039                                 && basefont.language()->lang() != "arabic_arabtex"
1040                                 && running_font.language()->lang() != "arabic_arabtex";
1041                 // Having said that, PassThru insets must be inside a font change command,
1042                 // as we do not re-open the font inside. So:
1043                 bool const closeLanguage = !inset->isPassThru() && !inRLSwitch;
1044                 bool lang_closed = false;
1045                 bool lang_switched_at_inset = false;
1046                 // Close language if needed
1047                 if (closeLanguage) {
1048                         // We need prev_font here as language changes directly at inset
1049                         // will only be started inside the inset.
1050                         Font const prev_font = (i > 0) ?
1051                                                 owner_->getFont(bparams, i - 1, outerfont)
1052                                               : running_font;
1053                         Font tmpfont(basefont);
1054                         tmpfont.setLanguage(prev_font.language());
1055                         bool needPar = false;
1056                         unsigned int count = tmpfont.latexWriteEndChanges(os, bparams, runparams,
1057                                                                           basefont, basefont,
1058                                                                           needPar, closeLanguage);
1059                         column += count;
1060                         lang_closed = count > 0;
1061                         lang_switched_at_inset = prev_font.language() != running_font.language();
1062                 }
1063                 // Update the running_font, making sure, however,
1064                 // to leave the language as it was.
1065                 // FIXME: probably a better way to keep track of the old
1066                 // language, than copying the entire font?
1067                 Font const copy_font(running_font);
1068                 basefont = owner_->getLayoutFont(bparams, outerfont);
1069                 running_font = basefont;
1070                 if (!closeLanguage)
1071                         running_font.setLanguage(copy_font.language());
1072                 // For these, we use switches, so they should be taken as
1073                 // base inside the inset.
1074                 basefont.fontInfo().setSize(copy_font.fontInfo().size());
1075                 basefont.fontInfo().setFamily(copy_font.fontInfo().family());
1076                 basefont.fontInfo().setSeries(copy_font.fontInfo().series());
1077                 // Now re-do font changes in a way needed here
1078                 // (using switches with multi-par insets)
1079                 InsetText const * textinset = inset->asInsetText();
1080                 bool const cprotect = textinset
1081                         ? textinset->hasCProtectContent(runparams.moving_arg)
1082                           && !textinset->text().isMainText()
1083                         : false;
1084                 unsigned int count2 = basefont.latexWriteStartChanges(os, bparams,
1085                                                       runparams, running_font,
1086                                                       basefont, true,
1087                                                       cprotect);
1088                 column += count2;
1089                 if (count2 == 0 && (lang_closed || lang_switched_at_inset))
1090                         // All fonts closed
1091                         open_font = false;
1092                 if (closeLanguage)
1093                         runparams.local_font = &basefont;
1094         }
1095
1096         size_t const previous_row_count = os.texrow().rows();
1097
1098         try {
1099                 runparams.lastid = id_;
1100                 runparams.lastpos = i;
1101                 inset->latex(os, runparams);
1102         } catch (EncodingException & e) {
1103                 // add location information and throw again.
1104                 e.par_id = id_;
1105                 e.pos = i;
1106                 throw(e);
1107         }
1108
1109         if (close)
1110                 os << '}';
1111
1112         if (os.texrow().rows() > previous_row_count) {
1113                 os.texrow().start(owner_->id(), i + 1);
1114                 column = 0;
1115         } else {
1116                 column += (unsigned int)(os.os().tellp() - len);
1117         }
1118
1119         if (owner_->isDeleted(i))
1120                 --runparams.inDeletedInset;
1121 }
1122
1123
1124 void Paragraph::Private::latexSpecialChar(otexstream & os,
1125                                           BufferParams const & bparams,
1126                                           OutputParams const & runparams,
1127                                           Font const & running_font,
1128                                           string & alien_script,
1129                                           Layout const & style,
1130                                           pos_type & i,
1131                                           pos_type end_pos,
1132                                           unsigned int & column)
1133 {
1134         char_type const c = owner_->getUChar(bparams, runparams, i);
1135
1136         if (style.pass_thru || runparams.pass_thru
1137             || contains(style.pass_thru_chars, c)
1138             || contains(runparams.pass_thru_chars, c)) {
1139                 if (c != '\0') {
1140                         Encoding const * const enc = runparams.encoding;
1141                         if (enc && !enc->encodable(c))
1142                                 throw EncodingException(c);
1143                         os.put(c);
1144                 }
1145                 return;
1146         }
1147
1148         // TIPA uses its own T3 encoding
1149         if (runparams.inIPA && latexSpecialT3(c, os, i, column))
1150                 return;
1151         // If T1 font encoding is used, use the special
1152         // characters it provides.
1153         // NOTE: Some languages reset the font encoding internally to a
1154         //       non-standard font encoding. If we are using such a language,
1155         //       we do not output special T1 chars.
1156         if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1157             && !runparams.isFullUnicode() && bparams.main_font_encoding() == "T1"
1158             && latexSpecialT1(c, os, i, column))
1159                 return;
1160         // NOTE: "fontspec" (non-TeX fonts) sets the font encoding to "TU" (untill 2017 "EU1" or "EU2")
1161         else if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1162                  && runparams.isFullUnicode() && latexSpecialTU(c, os, i, column))
1163                      return;
1164
1165         // Otherwise, we use what LaTeX provides us.
1166         switch (c) {
1167         case '\\':
1168                 os << "\\textbackslash" << termcmd;
1169                 column += 15;
1170                 break;
1171         case '<':
1172                 os << "\\textless" << termcmd;
1173                 column += 10;
1174                 break;
1175         case '>':
1176                 os << "\\textgreater" << termcmd;
1177                 column += 13;
1178                 break;
1179         case '|':
1180                 os << "\\textbar" << termcmd;
1181                 column += 9;
1182                 break;
1183         case '-':
1184                 os << '-';
1185                 if (i + 1 < static_cast<pos_type>(text_.size()) &&
1186                     (end_pos == -1 || i + 1 < end_pos) &&
1187                     text_[i+1] == '-') {
1188                         // Prevent "--" becoming an en dash and "---" an em dash.
1189                         // (Within \ttfamily, "---" is merged to en dash + hyphen.)
1190                         os << "{}";
1191                         column += 2;
1192                 }
1193                 break;
1194         case '\"':
1195                 os << "\\textquotedbl" << termcmd;
1196                 column += 14;
1197                 break;
1198
1199         case '$': case '&':
1200         case '%': case '#': case '{':
1201         case '}': case '_':
1202                 os << '\\';
1203                 os.put(c);
1204                 column += 1;
1205                 break;
1206
1207         case '~':
1208                 os << "\\textasciitilde" << termcmd;
1209                 column += 16;
1210                 break;
1211
1212         case '^':
1213                 os << "\\textasciicircum" << termcmd;
1214                 column += 17;
1215                 break;
1216
1217         case '*':
1218         case '[':
1219         case ']':
1220                 // avoid being mistaken for optional arguments
1221                 os << '{';
1222                 os.put(c);
1223                 os << '}';
1224                 column += 2;
1225                 break;
1226
1227         case ' ':
1228                 // Blanks are printed before font switching.
1229                 // Sure? I am not! (try nice-latex)
1230                 // I am sure it's correct. LyX might be smarter
1231                 // in the future, but for now, nothing wrong is
1232                 // written. (Asger)
1233                 break;
1234
1235         case 0x2013:
1236         case 0x2014:
1237                 // XeTeX's dash behaviour is determined via a global setting
1238                 if (bparams.use_dash_ligatures
1239                     && owner_->getFontSettings(bparams, i).fontInfo().family() != TYPEWRITER_FAMILY
1240                     && !runparams.inIPA
1241                         // TODO #10961: && not in inset Flex Code
1242                         // TODO #10961: && not in layout LyXCode
1243                     && (!bparams.useNonTeXFonts || runparams.flavor != OutputParams::XETEX)) {
1244                         if (c == 0x2013) {
1245                                 // en-dash
1246                                 os << "--";
1247                                 column +=2;
1248                         } else {
1249                                 // em-dash
1250                                 os << "---";
1251                                 column +=3;
1252                         }
1253                         break;
1254                 }
1255                 // fall through
1256         default:
1257                 if (c == '\0')
1258                         return;
1259
1260                 Encoding const & encoding = *(runparams.encoding);
1261                 char_type next = '\0';
1262                 if (i + 1 < int(text_.size())) {
1263                         next = text_[i + 1];
1264                         if (Encodings::isCombiningChar(next)) {
1265                                 column += latexSurrogatePair(bparams, os, c, next, runparams) - 1;
1266                                 ++i;
1267                                 break;
1268                         }
1269                 }
1270                 pair<docstring, bool> latex = encoding.latexChar(c);
1271                 docstring nextlatex;
1272                 bool nexttipas = false;
1273                 string nexttipashortcut;
1274                 if (next != '\0' && next != META_INSET && !encoding.encodable(next)) {
1275                         nextlatex = encoding.latexChar(next).first;
1276                         if (runparams.inIPA) {
1277                                 nexttipashortcut = Encodings::TIPAShortcut(next);
1278                                 nexttipas = !nexttipashortcut.empty();
1279                         }
1280                 }
1281                 bool tipas = false;
1282                 if (runparams.inIPA) {
1283                         string const tipashortcut = Encodings::TIPAShortcut(c);
1284                         if (!tipashortcut.empty()) {
1285                                 latex.first = from_ascii(tipashortcut);
1286                                 latex.second = false;
1287                                 tipas = true;
1288                         }
1289                 }
1290                 // eventually close "script wrapper" command (see `Paragraph::latex`)
1291                 if (!alien_script.empty()
1292                         && alien_script != Encodings::isKnownScriptChar(next)) {
1293                         column += latex.first.length();
1294                         alien_script.clear();
1295                         os << latex.first << "}";
1296                         break;
1297                 }
1298                 if (latex.second
1299                          && ((!prefixIs(nextlatex, '\\')
1300                                && !prefixIs(nextlatex, '{')
1301                                && !prefixIs(nextlatex, '}'))
1302                              || (nexttipas
1303                                  && !prefixIs(from_ascii(nexttipashortcut), '\\')))
1304                          && !tipas) {
1305                         // Prevent eating of a following space or command corruption by
1306                         // following characters
1307                         if (next == ' ' || next == '\0') {
1308                                 column += latex.first.length() + 1;
1309                                 os << latex.first << "{}";
1310                         } else {
1311                                 column += latex.first.length();
1312                                 os << latex.first << " ";
1313                         }
1314                 } else {
1315                         column += latex.first.length() - 1;
1316                         os << latex.first;
1317                 }
1318                 break;
1319         }
1320 }
1321
1322
1323 bool Paragraph::Private::latexSpecialT1(char_type const c, otexstream & os,
1324         pos_type i, unsigned int & column)
1325 {
1326         switch (c) {
1327         case '>':
1328         case '<':
1329                 os.put(c);
1330                 // In T1 encoding, these characters exist
1331                 // but we should avoid ligatures
1332                 if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
1333                         return true;
1334                 os << "\\textcompwordmark" << termcmd;
1335                 column += 19;
1336                 return true;
1337         case '|':
1338                 os.put(c);
1339                 return true;
1340         case '\"':
1341                 // soul.sty breaks with \char`\"
1342                 os << "\\textquotedbl" << termcmd;
1343                 column += 14;
1344                 return true;
1345         default:
1346                 return false;
1347         }
1348 }
1349
1350
1351 bool Paragraph::Private::latexSpecialTU(char_type const c, otexstream & os,
1352         pos_type i, unsigned int & column)
1353 {
1354         // TU encoding is currently on par with T1.
1355         return latexSpecialT1(c, os, i, column);
1356 }
1357
1358
1359 bool Paragraph::Private::latexSpecialT3(char_type const c, otexstream & os,
1360         pos_type /*i*/, unsigned int & column)
1361 {
1362         switch (c) {
1363         case '*':
1364         case '[':
1365         case ']':
1366         case '\"':
1367                 os.put(c);
1368                 return true;
1369         case '|':
1370                 os << "\\textvertline" << termcmd;
1371                 column += 14;
1372                 return true;
1373         default:
1374                 return false;
1375         }
1376 }
1377
1378
1379 void Paragraph::Private::validate(LaTeXFeatures & features) const
1380 {
1381         if (layout_->inpreamble && inset_owner_) {
1382                 // FIXME: Using a string stream here circumvents the encoding
1383                 // switching machinery of odocstream. Therefore the
1384                 // output is wrong if this paragraph contains content
1385                 // that needs to switch encoding.
1386                 Buffer const & buf = inset_owner_->buffer();
1387                 otexstringstream os;
1388                 os << layout_->preamble();
1389                 size_t const length = os.length();
1390                 TeXOnePar(buf, *inset_owner_->getText(int(buf.getParFromID(owner_->id()).idx())),
1391                           buf.getParFromID(owner_->id()).pit(), os,
1392                           features.runparams(), string(), 0, -1, true);
1393                 if (os.length() > length)
1394                         features.addPreambleSnippet(os.release(), true);
1395         }
1396
1397         if (features.runparams().flavor == OutputParams::HTML
1398             && layout_->htmltitle()) {
1399                 features.setHTMLTitle(owner_->asString(AS_STR_INSETS | AS_STR_SKIPDELETE));
1400         }
1401
1402         // check the params.
1403         if (!params_.spacing().isDefault())
1404                 features.require("setspace");
1405
1406         // then the layouts
1407         features.useLayout(layout_->name());
1408
1409         // then the fonts
1410         fontlist_.validate(features);
1411
1412         // then the indentation
1413         if (!params_.leftIndent().zero())
1414                 features.require("ParagraphLeftIndent");
1415
1416         // then the insets
1417         InsetList::const_iterator icit = insetlist_.begin();
1418         InsetList::const_iterator iend = insetlist_.end();
1419         for (; icit != iend; ++icit) {
1420                 if (icit->inset) {
1421                         features.inDeletedInset(owner_->isDeleted(icit->pos));
1422                         if (icit->inset->lyxCode() == FOOT_CODE) {
1423                                 // FIXME: an item inset would make things much easier.
1424                                 if ((layout_->latextype == LATEX_LIST_ENVIRONMENT
1425                                      || (layout_->latextype == LATEX_ITEM_ENVIRONMENT
1426                                          && layout_->margintype == MARGIN_FIRST_DYNAMIC))
1427                                     && (icit->pos < begin_of_body_
1428                                         || (icit->pos == begin_of_body_
1429                                             && (icit->pos == 0 || text_[icit->pos - 1] != ' '))))
1430                                         features.saveNoteEnv("description");
1431                         }
1432                         icit->inset->validate(features);
1433                         features.inDeletedInset(false);
1434                         if (layout_->needprotect &&
1435                             icit->inset->lyxCode() == FOOT_CODE)
1436                                 features.require("NeedLyXFootnoteCode");
1437                 }
1438         }
1439
1440         // then the contents
1441         BufferParams const bp = features.runparams().is_child
1442                 ? features.buffer().masterParams() : features.buffer().params();
1443         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1444                 char_type c = text_[i];
1445                 CharInfo const & ci = Encodings::unicodeCharInfo(c);
1446                 if (c == 0x0022) {
1447                         if (features.runparams().isFullUnicode() && bp.useNonTeXFonts)
1448                                 features.require("textquotedblp");
1449                         else if (bp.main_font_encoding() != "T1"
1450                                  || ((&owner_->getFontSettings(bp, i))->language()->internalFontEncoding()))
1451                                 features.require("textquotedbl");
1452                 } else if (ci.textfeature() && contains(ci.textpreamble(), '=')) {
1453                         // features that depend on the font or input encoding
1454                         string feats = ci.textpreamble();
1455                         string fontenc = (&owner_->getFontSettings(bp, i))->language()->fontenc(bp);
1456                         if (fontenc.empty())
1457                                 fontenc = features.runparams().main_fontenc;
1458                         while (!feats.empty()) {
1459                                 string feat;
1460                                 feats = split(feats, feat, ',');
1461                                 if (contains(feat, "!=")) {
1462                                         // a feature that is required except for the spcified
1463                                         // font or input encodings
1464                                         string realfeature;
1465                                         string const contexts = ltrim(split(feat, realfeature, '!'), "=");
1466                                         // multiple encodings are separated by semicolon
1467                                         vector<string> context = getVectorFromString(contexts, ";");
1468                                         // require feature if the context matches neither current font
1469                                         // nor input encoding
1470                                         if (std::find(context.begin(), context.end(), fontenc) == context.end()
1471                                             && std::find(context.begin(), context.end(),
1472                                                          features.runparams().encoding->name()) == context.end())
1473                                                 features.require(realfeature);
1474                                 } else if (contains(feat, '=')) {
1475                                         // a feature that is required only for the spcified
1476                                         // font or input encodings
1477                                         string realfeature;
1478                                         string const contexts = split(feat, realfeature, '=');
1479                                         // multiple encodings are separated by semicolon
1480                                         vector<string> context = getVectorFromString(contexts, ";");
1481                                         // require feature if the context matches either current font
1482                                         // or input encoding
1483                                         if (std::find(context.begin(), context.end(), fontenc) != context.end()
1484                                             || std::find(context.begin(), context.end(),
1485                                                          features.runparams().encoding->name()) != context.end())
1486                                                 features.require(realfeature);
1487                                 }
1488                         }
1489                 } else if (!bp.use_dash_ligatures
1490                            && (c == 0x2013 || c == 0x2014)
1491                            && bp.useNonTeXFonts
1492                            && features.runparams().flavor == OutputParams::XETEX)
1493                         // XeTeX's dash behaviour is determined via a global setting
1494                         features.require("xetexdashbreakstate");
1495                 BufferEncodings::validate(c, features);
1496         }
1497 }
1498
1499 /////////////////////////////////////////////////////////////////////
1500 //
1501 // Paragraph
1502 //
1503 /////////////////////////////////////////////////////////////////////
1504
1505 namespace {
1506         Layout const emptyParagraphLayout;
1507 }
1508
1509 Paragraph::Paragraph()
1510         : d(new Paragraph::Private(this, emptyParagraphLayout))
1511 {
1512         itemdepth = 0;
1513         d->params_.clear();
1514 }
1515
1516
1517 Paragraph::Paragraph(Paragraph const & par)
1518         : itemdepth(par.itemdepth),
1519         d(new Paragraph::Private(*par.d, this))
1520 {
1521         registerWords();
1522 }
1523
1524
1525 Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
1526         : itemdepth(par.itemdepth),
1527         d(new Paragraph::Private(*par.d, this, beg, end))
1528 {
1529         registerWords();
1530 }
1531
1532
1533 Paragraph & Paragraph::operator=(Paragraph const & par)
1534 {
1535         // needed as we will destroy the private part before copying it
1536         if (&par != this) {
1537                 itemdepth = par.itemdepth;
1538
1539                 deregisterWords();
1540                 delete d;
1541                 d = new Private(*par.d, this);
1542                 registerWords();
1543         }
1544         return *this;
1545 }
1546
1547
1548 Paragraph::~Paragraph()
1549 {
1550         deregisterWords();
1551         delete d;
1552 }
1553
1554
1555 namespace {
1556
1557 // this shall be called just before every "os << ..." action.
1558 void flushString(ostream & os, docstring & s)
1559 {
1560         os << to_utf8(s);
1561         s.erase();
1562 }
1563
1564 } // namespace
1565
1566
1567 void Paragraph::write(ostream & os, BufferParams const & bparams,
1568         depth_type & dth) const
1569 {
1570         // The beginning or end of a deeper (i.e. nested) area?
1571         if (dth != d->params_.depth()) {
1572                 if (d->params_.depth() > dth) {
1573                         while (d->params_.depth() > dth) {
1574                                 os << "\n\\begin_deeper";
1575                                 ++dth;
1576                         }
1577                 } else {
1578                         while (d->params_.depth() < dth) {
1579                                 os << "\n\\end_deeper";
1580                                 --dth;
1581                         }
1582                 }
1583         }
1584
1585         // First write the layout
1586         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1587
1588         d->params_.write(os);
1589
1590         Font font1(inherit_font, bparams.language);
1591
1592         Change running_change = Change(Change::UNCHANGED);
1593
1594         // this string is used as a buffer to avoid repetitive calls
1595         // to to_utf8(), which turn out to be expensive (JMarc)
1596         docstring write_buffer;
1597
1598         int column = 0;
1599         for (pos_type i = 0; i <= size(); ++i) {
1600
1601                 Change const & change = lookupChange(i);
1602                 if (change != running_change)
1603                         flushString(os, write_buffer);
1604                 Changes::lyxMarkChange(os, bparams, column, running_change, change);
1605                 running_change = change;
1606
1607                 if (i == size())
1608                         break;
1609
1610                 // Write font changes
1611                 Font font2 = getFontSettings(bparams, i);
1612                 if (font2 != font1) {
1613                         flushString(os, write_buffer);
1614                         font2.lyxWriteChanges(font1, os);
1615                         column = 0;
1616                         font1 = font2;
1617                 }
1618
1619                 char_type const c = d->text_[i];
1620                 switch (c) {
1621                 case META_INSET:
1622                         if (Inset const * inset = getInset(i)) {
1623                                 flushString(os, write_buffer);
1624                                 if (inset->directWrite()) {
1625                                         // international char, let it write
1626                                         // code directly so it's shorter in
1627                                         // the file
1628                                         inset->write(os);
1629                                 } else {
1630                                         if (i)
1631                                                 os << '\n';
1632                                         os << "\\begin_inset ";
1633                                         inset->write(os);
1634                                         os << "\n\\end_inset\n\n";
1635                                         column = 0;
1636                                 }
1637                                 // FIXME This can be removed again once the mystery
1638                                 // crash has been resolved.
1639                                 os << flush;
1640                         }
1641                         break;
1642                 case '\\':
1643                         flushString(os, write_buffer);
1644                         os << "\n\\backslash\n";
1645                         column = 0;
1646                         break;
1647                 case '.':
1648                         flushString(os, write_buffer);
1649                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1650                                 os << ".\n";
1651                                 column = 0;
1652                         } else
1653                                 os << '.';
1654                         break;
1655                 default:
1656                         if ((column > 70 && c == ' ')
1657                             || column > 79) {
1658                                 flushString(os, write_buffer);
1659                                 os << '\n';
1660                                 column = 0;
1661                         }
1662                         // this check is to amend a bug. LyX sometimes
1663                         // inserts '\0' this could cause problems.
1664                         if (c != '\0')
1665                                 write_buffer.push_back(c);
1666                         else
1667                                 LYXERR0("NUL char in structure.");
1668                         ++column;
1669                         break;
1670                 }
1671         }
1672
1673         flushString(os, write_buffer);
1674         os << "\n\\end_layout\n";
1675         // FIXME This can be removed again once the mystery
1676         // crash has been resolved.
1677         os << flush;
1678 }
1679
1680
1681 void Paragraph::validate(LaTeXFeatures & features) const
1682 {
1683         d->validate(features);
1684         bool fragile = features.runparams().moving_arg;
1685         fragile |= layout().needprotect;
1686         if (inInset().getLayout().isNeedProtect())
1687                 fragile = true;
1688         if (needsCProtection(fragile))
1689                 features.require("cprotect");
1690 }
1691
1692
1693 void Paragraph::insert(pos_type start, docstring const & str,
1694                        Font const & font, Change const & change)
1695 {
1696         for (size_t i = 0, n = str.size(); i != n ; ++i)
1697                 insertChar(start + i, str[i], font, change);
1698 }
1699
1700
1701 void Paragraph::appendChar(char_type c, Font const & font,
1702                 Change const & change)
1703 {
1704         // track change
1705         d->changes_.insert(change, d->text_.size());
1706         // when appending characters, no need to update tables
1707         d->text_.push_back(c);
1708         setFont(d->text_.size() - 1, font);
1709         d->requestSpellCheck(d->text_.size() - 1);
1710 }
1711
1712
1713 void Paragraph::appendString(docstring const & s, Font const & font,
1714                 Change const & change)
1715 {
1716         pos_type end = s.size();
1717         size_t oldsize = d->text_.size();
1718         size_t newsize = oldsize + end;
1719         size_t capacity = d->text_.capacity();
1720         if (newsize >= capacity)
1721                 d->text_.reserve(max(capacity + 100, newsize));
1722
1723         // when appending characters, no need to update tables
1724         d->text_.append(s);
1725
1726         // FIXME: Optimize this!
1727         for (size_t i = oldsize; i != newsize; ++i) {
1728                 // track change
1729                 d->changes_.insert(change, i);
1730                 d->requestSpellCheck(i);
1731         }
1732         d->fontlist_.set(oldsize, font);
1733         d->fontlist_.set(newsize - 1, font);
1734 }
1735
1736
1737 void Paragraph::insertChar(pos_type pos, char_type c,
1738                            bool trackChanges)
1739 {
1740         d->insertChar(pos, c, Change(trackChanges ?
1741                            Change::INSERTED : Change::UNCHANGED));
1742 }
1743
1744
1745 void Paragraph::insertChar(pos_type pos, char_type c,
1746                            Font const & font, bool trackChanges)
1747 {
1748         d->insertChar(pos, c, Change(trackChanges ?
1749                            Change::INSERTED : Change::UNCHANGED));
1750         setFont(pos, font);
1751 }
1752
1753
1754 void Paragraph::insertChar(pos_type pos, char_type c,
1755                            Font const & font, Change const & change)
1756 {
1757         d->insertChar(pos, c, change);
1758         setFont(pos, font);
1759 }
1760
1761
1762 void Paragraph::resetFonts(Font const & font)
1763 {
1764         d->fontlist_.clear();
1765         d->fontlist_.set(0, font);
1766         d->fontlist_.set(d->text_.size() - 1, font);
1767 }
1768
1769 // Gets uninstantiated font setting at position.
1770 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1771                                          pos_type pos) const
1772 {
1773         if (pos > size()) {
1774                 LYXERR0("pos: " << pos << " size: " << size());
1775                 LBUFERR(false);
1776         }
1777
1778         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1779         if (cit != d->fontlist_.end())
1780                 return cit->font();
1781
1782         if (pos == size() && !empty())
1783                 return getFontSettings(bparams, pos - 1);
1784
1785         // Optimisation: avoid a full font instantiation if there is no
1786         // language change from previous call.
1787         static Font previous_font;
1788         static Language const * previous_lang = nullptr;
1789         Language const * lang = getParLanguage(bparams);
1790         if (lang != previous_lang) {
1791                 previous_lang = lang;
1792                 previous_font = Font(inherit_font, lang);
1793         }
1794         return previous_font;
1795 }
1796
1797
1798 FontSpan Paragraph::fontSpan(pos_type pos) const
1799 {
1800         LBUFERR(pos <= size());
1801
1802         if (pos == size())
1803                 return FontSpan(pos, pos);
1804
1805         pos_type start = 0;
1806         FontList::const_iterator cit = d->fontlist_.begin();
1807         FontList::const_iterator end = d->fontlist_.end();
1808         for (; cit != end; ++cit) {
1809                 if (cit->pos() >= pos) {
1810                         if (pos >= beginOfBody())
1811                                 return FontSpan(max(start, beginOfBody()),
1812                                                 cit->pos());
1813                         else
1814                                 return FontSpan(start,
1815                                                 min(beginOfBody() - 1,
1816                                                          cit->pos()));
1817                 }
1818                 start = cit->pos() + 1;
1819         }
1820
1821         // This should not happen, but if so, we take no chances.
1822         LYXERR0("Paragraph::fontSpan: position not found in fontinfo table!");
1823         LASSERT(false, return FontSpan(pos, pos));
1824 }
1825
1826
1827 // Gets uninstantiated font setting at position 0
1828 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1829 {
1830         if (!empty() && !d->fontlist_.empty())
1831                 return d->fontlist_.begin()->font();
1832
1833         // Optimisation: avoid a full font instantiation if there is no
1834         // language change from previous call.
1835         static Font previous_font;
1836         static Language const * previous_lang = nullptr;
1837         if (bparams.language != previous_lang) {
1838                 previous_lang = bparams.language;
1839                 previous_font = Font(inherit_font, bparams.language);
1840         }
1841
1842         return previous_font;
1843 }
1844
1845
1846 // Gets the fully instantiated font at a given position in a paragraph
1847 // This is basically the same function as Text::GetFont() in text2.cpp.
1848 // The difference is that this one is used for generating the LaTeX file,
1849 // and thus cosmetic "improvements" are disallowed: This has to deliver
1850 // the true picture of the buffer. (Asger)
1851 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1852                                  Font const & outerfont) const
1853 {
1854         LBUFERR(pos >= 0);
1855
1856         Font font = getFontSettings(bparams, pos);
1857
1858         pos_type const body_pos = beginOfBody();
1859         FontInfo & fi = font.fontInfo();
1860         if (pos < body_pos)
1861                 fi.realize(d->layout_->labelfont);
1862         else
1863                 fi.realize(d->layout_->font);
1864
1865         fi.realize(outerfont.fontInfo());
1866         fi.realize(bparams.getFont().fontInfo());
1867
1868         return font;
1869 }
1870
1871
1872 Font const Paragraph::getLabelFont
1873         (BufferParams const & bparams, Font const & outerfont) const
1874 {
1875         FontInfo tmpfont = d->layout_->labelfont;
1876         tmpfont.realize(outerfont.fontInfo());
1877         tmpfont.realize(bparams.getFont().fontInfo());
1878         return Font(tmpfont, getParLanguage(bparams));
1879 }
1880
1881
1882 Font const Paragraph::getLayoutFont
1883         (BufferParams const & bparams, Font const & outerfont) const
1884 {
1885         FontInfo tmpfont = d->layout_->font;
1886         tmpfont.realize(outerfont.fontInfo());
1887         tmpfont.realize(bparams.getFont().fontInfo());
1888         return Font(tmpfont, getParLanguage(bparams));
1889 }
1890
1891
1892 char_type Paragraph::getUChar(BufferParams const & bparams,
1893                               OutputParams const & rp,
1894                               pos_type pos) const
1895 {
1896         char_type c = d->text_[pos];
1897
1898         // Return unchanged character in LTR languages
1899         // or if we use poylglossia/bidi (XeTeX).
1900         if (rp.useBidiPackage()
1901             || !getFontSettings(bparams, pos).isRightToLeft())
1902                 return c;
1903
1904         // Without polyglossia/bidi, we need to account for some special cases.
1905         // FIXME This needs to be audited!
1906         // Check if:
1907         // * The input is as expected for all delimiters
1908         //   => checked for Hebrew!
1909         // * The output matches the display in the LyX workarea
1910         //   => checked for Hebrew!
1911         // * The special cases below are really necessary
1912         //   => checked for Hebrew!
1913         // * In arabic_arabi, brackets are transformed to Arabic
1914         //   Ornate Parentheses. Is this is really wanted?
1915
1916         string const & lang = getFontSettings(bparams, pos).language()->lang();
1917         char_type uc = c;
1918
1919         // 1. In the following languages, parentheses need to be reversed.
1920         //    Also with polyglodia/luabidi
1921         bool const reverseparens = (lang == "hebrew" || rp.use_polyglossia);
1922
1923         // 2. In the following languages, brackets don't need to be reversed.
1924         bool const reversebrackets = lang != "arabic_arabtex"
1925                         && lang != "arabic_arabi"
1926                         && lang != "farsi";
1927
1928         // Now swap delimiters if needed.
1929         switch (c) {
1930         case '(':
1931                 if (reverseparens)
1932                         uc = ')';
1933                 break;
1934         case ')':
1935                 if (reverseparens)
1936                         uc = '(';
1937                 break;
1938         case '[':
1939                 if (reversebrackets)
1940                         uc = ']';
1941                 break;
1942         case ']':
1943                 if (reversebrackets)
1944                         uc = '[';
1945                 break;
1946         case '{':
1947                 uc = '}';
1948                 break;
1949         case '}':
1950                 uc = '{';
1951                 break;
1952         case '<':
1953                 uc = '>';
1954                 break;
1955         case '>':
1956                 uc = '<';
1957                 break;
1958         }
1959
1960         return uc;
1961 }
1962
1963
1964 void Paragraph::setFont(pos_type pos, Font const & font)
1965 {
1966         LASSERT(pos <= size(), return);
1967
1968         // First, reduce font against layout/label font
1969         // Update: The setCharFont() routine in text2.cpp already
1970         // reduces font, so we don't need to do that here. (Asger)
1971
1972         d->fontlist_.set(pos, font);
1973 }
1974
1975
1976 void Paragraph::makeSameLayout(Paragraph const & par)
1977 {
1978         d->layout_ = par.d->layout_;
1979         d->params_ = par.d->params_;
1980 }
1981
1982
1983 bool Paragraph::stripLeadingSpaces(bool trackChanges)
1984 {
1985         if (isFreeSpacing())
1986                 return false;
1987
1988         int pos = 0;
1989         int count = 0;
1990
1991         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
1992                 if (eraseChar(pos, trackChanges))
1993                         ++count;
1994                 else
1995                         ++pos;
1996         }
1997
1998         return count > 0 || pos > 0;
1999 }
2000
2001
2002 bool Paragraph::hasSameLayout(Paragraph const & par) const
2003 {
2004         return par.d->layout_ == d->layout_
2005                 && d->params_.sameLayout(par.d->params_);
2006 }
2007
2008
2009 depth_type Paragraph::getDepth() const
2010 {
2011         return d->params_.depth();
2012 }
2013
2014
2015 depth_type Paragraph::getMaxDepthAfter() const
2016 {
2017         if (d->layout_->isEnvironment())
2018                 return d->params_.depth() + 1;
2019         else
2020                 return d->params_.depth();
2021 }
2022
2023
2024 LyXAlignment Paragraph::getAlign(BufferParams const & bparams) const
2025 {
2026         if (d->params_.align() == LYX_ALIGN_LAYOUT)
2027                 return getDefaultAlign(bparams);
2028         else
2029                 return d->params_.align();
2030 }
2031
2032
2033 LyXAlignment Paragraph::getDefaultAlign(BufferParams const & bparams) const
2034 {
2035         LyXAlignment res = layout().align;
2036         if (isRTL(bparams)) {
2037                 // Swap sides
2038                 if (res == LYX_ALIGN_LEFT)
2039                         res = LYX_ALIGN_RIGHT;
2040                 else if  (res == LYX_ALIGN_RIGHT)
2041                         res = LYX_ALIGN_LEFT;
2042         }
2043         return res;
2044 }
2045
2046
2047 docstring const & Paragraph::labelString() const
2048 {
2049         return d->params_.labelString();
2050 }
2051
2052
2053 // the next two functions are for the manual labels
2054 docstring const Paragraph::getLabelWidthString() const
2055 {
2056         if (d->layout_->margintype == MARGIN_MANUAL
2057             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
2058                 return d->params_.labelWidthString();
2059         else
2060                 return _("Senseless with this layout!");
2061 }
2062
2063
2064 void Paragraph::setLabelWidthString(docstring const & s)
2065 {
2066         d->params_.labelWidthString(s);
2067 }
2068
2069
2070 docstring Paragraph::expandLabel(Layout const & layout,
2071                 BufferParams const & bparams) const
2072 {
2073         return expandParagraphLabel(layout, bparams, true);
2074 }
2075
2076
2077 docstring Paragraph::expandParagraphLabel(Layout const & layout,
2078                 BufferParams const & bparams, bool process_appendix) const
2079 {
2080         DocumentClass const & tclass = bparams.documentClass();
2081         string const & lang = getParLanguage(bparams)->code();
2082         bool const in_appendix = process_appendix && d->params_.appendix();
2083         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
2084
2085         if (fmt.empty() && !layout.counter.empty())
2086                 return tclass.counters().theCounter(layout.counter, lang);
2087
2088         // handle 'inherited level parts' in 'fmt',
2089         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
2090         size_t const i = fmt.find('@', 0);
2091         if (i != docstring::npos) {
2092                 size_t const j = fmt.find('@', i + 1);
2093                 if (j != docstring::npos) {
2094                         docstring parent(fmt, i + 1, j - i - 1);
2095                         docstring label = from_ascii("??");
2096                         if (tclass.hasLayout(parent))
2097                                 label = expandParagraphLabel(tclass[parent], bparams,
2098                                                       process_appendix);
2099                         fmt = docstring(fmt, 0, i) + label
2100                                 + docstring(fmt, j + 1, docstring::npos);
2101                 }
2102         }
2103
2104         return tclass.counters().counterLabel(fmt, lang);
2105 }
2106
2107
2108 void Paragraph::applyLayout(Layout const & new_layout)
2109 {
2110         d->layout_ = &new_layout;
2111         LyXAlignment const oldAlign = d->params_.align();
2112
2113         if (!(oldAlign & d->layout_->alignpossible)) {
2114                 frontend::Alert::warning(_("Alignment not permitted"),
2115                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
2116                 d->params_.align(LYX_ALIGN_LAYOUT);
2117         }
2118 }
2119
2120
2121 pos_type Paragraph::beginOfBody() const
2122 {
2123         return d->begin_of_body_;
2124 }
2125
2126
2127 void Paragraph::setBeginOfBody()
2128 {
2129         if (d->layout_->labeltype != LABEL_MANUAL) {
2130                 d->begin_of_body_ = 0;
2131                 return;
2132         }
2133
2134         // Unroll the first two cycles of the loop
2135         // and remember the previous character to
2136         // remove unnecessary getChar() calls
2137         pos_type i = 0;
2138         pos_type end = size();
2139         bool prev_char_deleted = false;
2140         if (i < end && (!(isNewline(i) || isEnvSeparator(i)) || isDeleted(i))) {
2141                 ++i;
2142                 if (i < end) {
2143                         char_type previous_char = d->text_[i];
2144                         if (!(isNewline(i) || isEnvSeparator(i))) {
2145                                 ++i;
2146                                 while (i < end && (previous_char != ' ' || prev_char_deleted)) {
2147                                         char_type temp = d->text_[i];
2148                                         prev_char_deleted = isDeleted(i);
2149                                         if (!isDeleted(i) && (isNewline(i) || isEnvSeparator(i)))
2150                                                 break;
2151                                         ++i;
2152                                         previous_char = temp;
2153                                 }
2154                         }
2155                 }
2156         }
2157
2158         d->begin_of_body_ = i;
2159 }
2160
2161
2162 bool Paragraph::allowParagraphCustomization() const
2163 {
2164         return inInset().allowParagraphCustomization();
2165 }
2166
2167
2168 bool Paragraph::usePlainLayout() const
2169 {
2170         return inInset().usePlainLayout();
2171 }
2172
2173
2174 bool Paragraph::isPassThru() const
2175 {
2176         return inInset().isPassThru() || d->layout_->pass_thru;
2177 }
2178
2179 namespace {
2180
2181 // paragraphs inside floats need different alignment tags to avoid
2182 // unwanted space
2183
2184 bool noTrivlistCentering(InsetCode code)
2185 {
2186         return code == FLOAT_CODE
2187                || code == WRAP_CODE
2188                || code == CELL_CODE;
2189 }
2190
2191
2192 string correction(string const & orig)
2193 {
2194         if (orig == "flushleft")
2195                 return "raggedright";
2196         if (orig == "flushright")
2197                 return "raggedleft";
2198         if (orig == "center")
2199                 return "centering";
2200         return orig;
2201 }
2202
2203
2204 bool corrected_env(otexstream & os, string const & suffix, string const & env,
2205         InsetCode code, bool const lastpar, int & col)
2206 {
2207         string macro = suffix + "{";
2208         if (noTrivlistCentering(code)) {
2209                 if (lastpar) {
2210                         // the last paragraph in non-trivlist-aligned
2211                         // context is special (to avoid unwanted whitespace)
2212                         if (suffix == "\\begin") {
2213                                 macro = "\\" + correction(env) + "{}";
2214                                 os << from_ascii(macro);
2215                                 col += macro.size();
2216                                 return true;
2217                         }
2218                         return false;
2219                 }
2220                 macro += correction(env);
2221         } else
2222                 macro += env;
2223         macro += "}";
2224         if (suffix == "\\par\\end") {
2225                 os << breakln;
2226                 col = 0;
2227         }
2228         os << from_ascii(macro);
2229         col += macro.size();
2230         if (suffix == "\\begin") {
2231                 os << breakln;
2232                 col = 0;
2233         }
2234         return true;
2235 }
2236
2237 } // namespace
2238
2239
2240 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
2241                         otexstream & os, OutputParams const & runparams) const
2242 {
2243         int column = 0;
2244
2245         bool canindent =
2246                 (bparams.paragraph_separation == BufferParams::ParagraphIndentSeparation) ?
2247                         (layout_->toggle_indent != ITOGGLE_NEVER) :
2248                         (layout_->toggle_indent == ITOGGLE_ALWAYS);
2249
2250         if (canindent && params_.noindent() && !layout_->pass_thru) {
2251                 os << "\\noindent ";
2252                 column += 10;
2253         }
2254
2255         LyXAlignment const curAlign = params_.align();
2256
2257         if (curAlign == layout_->align)
2258                 return column;
2259
2260         switch (curAlign) {
2261         case LYX_ALIGN_NONE:
2262         case LYX_ALIGN_BLOCK:
2263         case LYX_ALIGN_LAYOUT:
2264         case LYX_ALIGN_SPECIAL:
2265         case LYX_ALIGN_DECIMAL:
2266                 break;
2267         case LYX_ALIGN_LEFT:
2268         case LYX_ALIGN_RIGHT:
2269         case LYX_ALIGN_CENTER:
2270                 if (runparams.moving_arg) {
2271                         os << "\\protect";
2272                         column += 8;
2273                 }
2274                 break;
2275         }
2276
2277         string const begin_tag = "\\begin";
2278         InsetCode code = ownerCode();
2279         bool const lastpar = runparams.isLastPar;
2280         // RTL in classic (PDF)LaTeX (without the Bidi package)
2281         // Luabibdi (used by LuaTeX) behaves like classic
2282         bool const rtl_classic = owner_->getParLanguage(bparams)->rightToLeft()
2283                 && !runparams.useBidiPackage();
2284
2285         switch (curAlign) {
2286         case LYX_ALIGN_NONE:
2287         case LYX_ALIGN_BLOCK:
2288         case LYX_ALIGN_LAYOUT:
2289         case LYX_ALIGN_SPECIAL:
2290         case LYX_ALIGN_DECIMAL:
2291                 break;
2292         case LYX_ALIGN_LEFT: {
2293                 if (rtl_classic)
2294                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2295                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2296                 else
2297                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2298                 break;
2299         } case LYX_ALIGN_RIGHT: {
2300                 if (rtl_classic)
2301                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2302                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2303                 else
2304                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2305                 break;
2306         } case LYX_ALIGN_CENTER: {
2307                 corrected_env(os, begin_tag, "center", code, lastpar, column);
2308                 break;
2309         }
2310         }
2311
2312         return column;
2313 }
2314
2315
2316 bool Paragraph::Private::endTeXParParams(BufferParams const & bparams,
2317                         otexstream & os, OutputParams const & runparams) const
2318 {
2319         LyXAlignment const curAlign = params_.align();
2320
2321         if (curAlign == layout_->align)
2322                 return false;
2323
2324         switch (curAlign) {
2325         case LYX_ALIGN_NONE:
2326         case LYX_ALIGN_BLOCK:
2327         case LYX_ALIGN_LAYOUT:
2328         case LYX_ALIGN_SPECIAL:
2329         case LYX_ALIGN_DECIMAL:
2330                 break;
2331         case LYX_ALIGN_LEFT:
2332         case LYX_ALIGN_RIGHT:
2333         case LYX_ALIGN_CENTER:
2334                 if (runparams.moving_arg)
2335                         os << "\\protect";
2336                 break;
2337         }
2338
2339         bool output = false;
2340         int col = 0;
2341         string const end_tag = "\\par\\end";
2342         InsetCode code = ownerCode();
2343         bool const lastpar = runparams.isLastPar;
2344         // RTL in classic (PDF)LaTeX (without the Bidi package)
2345         // Luabibdi (used by LuaTeX) behaves like classic
2346         bool const rtl_classic = owner_->getParLanguage(bparams)->rightToLeft()
2347                 && !runparams.useBidiPackage();
2348
2349         switch (curAlign) {
2350         case LYX_ALIGN_NONE:
2351         case LYX_ALIGN_BLOCK:
2352         case LYX_ALIGN_LAYOUT:
2353         case LYX_ALIGN_SPECIAL:
2354         case LYX_ALIGN_DECIMAL:
2355                 break;
2356         case LYX_ALIGN_LEFT: {
2357                 if (rtl_classic)
2358                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2359                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2360                 else
2361                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2362                 break;
2363         } case LYX_ALIGN_RIGHT: {
2364                 if (rtl_classic)
2365                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2366                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2367                 else
2368                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2369                 break;
2370         } case LYX_ALIGN_CENTER: {
2371                 corrected_env(os, end_tag, "center", code, lastpar, col);
2372                 break;
2373         }
2374         }
2375
2376         return output || lastpar;
2377 }
2378
2379
2380 // This one spits out the text of the paragraph
2381 void Paragraph::latex(BufferParams const & bparams,
2382         Font const & outerfont,
2383         otexstream & os,
2384         OutputParams const & runparams,
2385         int start_pos, int end_pos, bool force) const
2386 {
2387         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
2388
2389         // FIXME This check should not be needed. Perhaps issue an
2390         // error if it triggers.
2391         Layout const & style = inInset().forcePlainLayout() ?
2392                 bparams.documentClass().plainLayout() : *d->layout_;
2393
2394         if (!force && style.inpreamble)
2395                 return;
2396
2397         bool const allowcust = allowParagraphCustomization();
2398
2399         // Current base font for all inherited font changes, without any
2400         // change caused by an individual character, except for the language:
2401         // It is set to the language of the first character.
2402         // As long as we are in the label, this font is the base font of the
2403         // label. Before the first body character it is set to the base font
2404         // of the body.
2405         Font basefont;
2406
2407         // If there is an open font-encoding changing command (script wrapper),
2408         // alien_script is set to its name
2409         string alien_script;
2410         string script;
2411
2412         // Maybe we have to create a optional argument.
2413         pos_type body_pos = beginOfBody();
2414         unsigned int column = 0;
2415
2416         // If we are inside an non inheritFont() inset, the real outerfont is local_font
2417         Font const real_outerfont = (!inInset().inheritFont()
2418                                      && runparams.local_font != nullptr)
2419                         ? Font(runparams.local_font->fontInfo()) : outerfont;
2420
2421         if (body_pos > 0) {
2422                 // the optional argument is kept in curly brackets in
2423                 // case it contains a ']'
2424                 // This is not strictly needed, but if this is changed it
2425                 // would be a file format change, and tex2lyx would need
2426                 // to be adjusted, since it unconditionally removes the
2427                 // braces when it parses \item.
2428                 os << "[{";
2429                 column += 2;
2430                 basefont = getLabelFont(bparams, real_outerfont);
2431         } else {
2432                 basefont = getLayoutFont(bparams, real_outerfont);
2433         }
2434
2435         // Which font is currently active?
2436         Font running_font(basefont);
2437         // Do we have an open font change?
2438         bool open_font = false;
2439
2440         Change runningChange = Change(Change::UNCHANGED);
2441
2442         Encoding const * const prev_encoding = runparams.encoding;
2443
2444         os.texrow().start(id(), 0);
2445
2446         // if the paragraph is empty, the loop will not be entered at all
2447         if (empty()) {
2448                 // For InTitle commands, we have already opened a group
2449                 // in output_latex::TeXOnePar.
2450                 if (style.isCommand() && !style.intitle) {
2451                         os << '{';
2452                         ++column;
2453                 }
2454                 if (!style.leftdelim().empty()) {
2455                         os << style.leftdelim();
2456                         column += style.leftdelim().size();
2457                 }
2458                 if (allowcust)
2459                         column += d->startTeXParParams(bparams, os, runparams);
2460         }
2461
2462         // Whether a \par can be issued for insets typeset inline with text.
2463         // Yes if greater than 0. This has to be static.
2464         THREAD_LOCAL_STATIC int parInline = 0;
2465
2466         for (pos_type i = 0; i < size(); ++i) {
2467                 // First char in paragraph or after label?
2468                 if (i == body_pos) {
2469                         if (body_pos > 0) {
2470                                 if (open_font) {
2471                                         bool needPar = false;
2472                                         column += running_font.latexWriteEndChanges(
2473                                                 os, bparams, runparams,
2474                                                 basefont, basefont, needPar);
2475                                         open_font = false;
2476                                 }
2477                                 basefont = getLayoutFont(bparams, real_outerfont);
2478                                 running_font = basefont;
2479
2480                                 column += Changes::latexMarkChange(os, bparams,
2481                                                 runningChange, Change(Change::UNCHANGED),
2482                                                 runparams);
2483                                 runningChange = Change(Change::UNCHANGED);
2484
2485                                 os << "}] ";
2486                                 column +=3;
2487                         }
2488                         // For InTitle commands, we have already opened a group
2489                         // in output_latex::TeXOnePar.
2490                         if (style.isCommand() && !style.intitle) {
2491                                 os << '{';
2492                                 ++column;
2493                         }
2494
2495                         if (!style.leftdelim().empty()) {
2496                                 os << style.leftdelim();
2497                                 column += style.leftdelim().size();
2498                         }
2499
2500                         if (allowcust)
2501                                 column += d->startTeXParParams(bparams, os,
2502                                                             runparams);
2503                 }
2504
2505                 runparams.wasDisplayMath = runparams.inDisplayMath;
2506                 runparams.inDisplayMath = false;
2507                 bool deleted_display_math = false;
2508                 Change const & change = runparams.inDeletedInset
2509                         ? runparams.changeOfDeletedInset : lookupChange(i);
2510
2511                 char_type const c = d->text_[i];
2512
2513                 // Check whether a display math inset follows
2514                 if (c == META_INSET
2515                     && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2516                         if (isDeleted(i))
2517                                 runparams.ctObject = getInset(i)->CtObject(runparams);
2518
2519                         InsetMath const * im = getInset(i)->asInsetMath();
2520                         if (im && im->asHullInset()
2521                             && im->asHullInset()->outerDisplay()) {
2522                                 runparams.inDisplayMath = true;
2523                                 // runparams.inDeletedInset will be set by
2524                                 // latexInset later, but we need this info
2525                                 // before it is called. On the other hand, we
2526                                 // cannot set it here because it is a counter.
2527                                 deleted_display_math = isDeleted(i);
2528                         }
2529                         if (bparams.output_changes && deleted_display_math
2530                             && runningChange == change
2531                             && change.type == Change::DELETED
2532                             && !os.afterParbreak()) {
2533                                 // A display math in the same paragraph follows.
2534                                 // We have to close and then reopen \lyxdeleted,
2535                                 // otherwise the math will be shifted up.
2536                                 OutputParams rp = runparams;
2537                                 if (open_font) {
2538                                         bool needPar = false;
2539                                         column += running_font.latexWriteEndChanges(
2540                                                 os, bparams, rp, basefont,
2541                                                 basefont, needPar);
2542                                         open_font = false;
2543                                 }
2544                                 basefont = (body_pos > i) ? getLabelFont(bparams, real_outerfont)
2545                                                           : getLayoutFont(bparams, real_outerfont);
2546                                 running_font = basefont;
2547                                 column += Changes::latexMarkChange(os, bparams,
2548                                         Change(Change::INSERTED), change, rp);
2549                         }
2550                 }
2551
2552                 if (bparams.output_changes && runningChange != change) {
2553                         if (!alien_script.empty()) {
2554                                 column += 1;
2555                                 os << "}";
2556                                 alien_script.clear();
2557                         }
2558                         if (open_font) {
2559                                 bool needPar = false;
2560                                 column += running_font.latexWriteEndChanges(
2561                                                 os, bparams, runparams,
2562                                                 basefont, basefont, needPar);
2563                                 open_font = false;
2564                         }
2565                         basefont = (body_pos > i) ? getLabelFont(bparams, real_outerfont)
2566                                                   : getLayoutFont(bparams, real_outerfont);
2567                         running_font = basefont;
2568                         column += Changes::latexMarkChange(os, bparams, runningChange,
2569                                                            change, runparams);
2570                         runningChange = change;
2571                 }
2572
2573                 // do not output text which is marked deleted
2574                 // if change tracking output is disabled
2575                 if (!bparams.output_changes && change.deleted()) {
2576                         continue;
2577                 }
2578
2579                 ++column;
2580
2581                 // Fully instantiated font
2582                 Font const current_font = getFont(bparams, i, outerfont);
2583                 // Previous font
2584                 Font const prev_font = (i > 0) ?
2585                                         getFont(bparams, i - 1, outerfont)
2586                                       : current_font;
2587
2588                 Font const last_font = running_font;
2589                 bool const in_ct_deletion = (bparams.output_changes
2590                                              && runningChange == change
2591                                              && change.type == Change::DELETED
2592                                              && !os.afterParbreak());
2593                 bool const multipar_inset =
2594                         (c == META_INSET && getInset(i) && getInset(i)->allowMultiPar());
2595
2596                 // Do we need to close the previous font?
2597                 if (open_font &&
2598                     ((current_font != running_font
2599                       || current_font.language() != running_font.language())
2600                      || (multipar_inset
2601                          && (current_font == prev_font
2602                              || current_font.language() == prev_font.language()))))
2603                 {
2604                         // ensure there is no open script-wrapper
2605                         if (!alien_script.empty()) {
2606                                 column += 1;
2607                                 os << "}";
2608                                 alien_script.clear();
2609                         }
2610                         bool needPar = false;
2611                         if (in_ct_deletion) {
2612                                 // We have to close and then reopen \lyxdeleted,
2613                                 // as strikeout needs to be on lowest level.
2614                                 os << '}';
2615                                 column += 1;
2616                         }
2617                         column += running_font.latexWriteEndChanges(
2618                                     os, bparams, runparams, basefont,
2619                                     (i == body_pos-1) ? basefont : current_font,
2620                                     needPar);
2621                         if (in_ct_deletion) {
2622                                 // We have to close and then reopen \lyxdeleted,
2623                                 // as strikeout needs to be on lowest level.
2624                                 OutputParams rp = runparams;
2625                                 column += Changes::latexMarkChange(os, bparams,
2626                                         Change(Change::UNCHANGED), Change(Change::DELETED), rp);
2627                         }
2628                         running_font = basefont;
2629                         open_font = false;
2630                 }
2631
2632                 // if necessary, close language environment before opening CJK
2633                 string const running_lang = running_font.language()->babel();
2634                 string const lang_end_command = lyxrc.language_command_end;
2635                 if (!lang_end_command.empty() && !bparams.useNonTeXFonts
2636                         && !running_lang.empty()
2637                         && running_lang == openLanguageName()
2638                         && current_font.language()->encoding()->package() == Encoding::CJK) {
2639                         string end_tag = subst(lang_end_command, "$$lang", running_lang);
2640                         os << from_ascii(end_tag);
2641                         column += end_tag.length();
2642                         popLanguageName();
2643                 }
2644
2645                 // Switch file encoding if necessary (and allowed)
2646                 if (!runparams.pass_thru && !style.pass_thru &&
2647                     runparams.encoding->package() != Encoding::none &&
2648                     current_font.language()->encoding()->package() != Encoding::none) {
2649                         pair<bool, int> const enc_switch =
2650                                 switchEncoding(os.os(), bparams, runparams,
2651                                         *(current_font.language()->encoding()));
2652                         if (enc_switch.first) {
2653                                 column += enc_switch.second;
2654                                 runparams.encoding = current_font.language()->encoding();
2655                         }
2656                 }
2657
2658                 // A display math inset inside an ulem command will be output
2659                 // as a box of width \linewidth, so we have to either disable
2660                 // indentation if the inset starts a paragraph, or start a new
2661                 // line to accommodate such box. This has to be done before
2662                 // writing any font changing commands.
2663                 if (runparams.inDisplayMath && !deleted_display_math
2664                     && runparams.inulemcmd) {
2665                         if (os.afterParbreak())
2666                                 os << "\\noindent";
2667                         else
2668                                 os << "\\\\\n";
2669                 }
2670
2671                 // Do we need to change font?
2672                 if ((current_font != running_font ||
2673                      current_font.language() != running_font.language())
2674                     && i != body_pos - 1)
2675                 {
2676                         if (in_ct_deletion) {
2677                                 // We have to close and then reopen \lyxdeleted,
2678                                 // as strikeout needs to be on lowest level.
2679                                 bool needPar = false;
2680                                 OutputParams rp = runparams;
2681                                 column += running_font.latexWriteEndChanges(
2682                                         os, bparams, rp, basefont,
2683                                         basefont, needPar);
2684                                 os << '}';
2685                                 column += 1;
2686                         }
2687                         otexstringstream ots;
2688                         if (!multipar_inset) {
2689                                 InsetText const * textinset = inInset().asInsetText();
2690                                 bool const cprotect = textinset
2691                                         ? textinset->hasCProtectContent(runparams.moving_arg)
2692                                           && !textinset->text().isMainText()
2693                                         : false;
2694                                 column += current_font.latexWriteStartChanges(ots, bparams,
2695                                                                               runparams, basefont, last_font, false,
2696                                                                               cprotect);
2697                         }
2698                         // Check again for display math in ulem commands as a
2699                         // font change may also occur just before a math inset.
2700                         if (runparams.inDisplayMath && !deleted_display_math
2701                             && runparams.inulemcmd) {
2702                                 if (os.afterParbreak())
2703                                         os << "\\noindent";
2704                                 else
2705                                         os << "\\\\\n";
2706                         }
2707                         running_font = current_font;
2708                         open_font = true;
2709                         docstring fontchange = ots.str();
2710                         os << fontchange;
2711                         // check whether the fontchange ends with a \\textcolor
2712                         // modifier and the text starts with a space. If so we
2713                         // need to add } in order to prevent \\textcolor from gobbling
2714                         // the space (bug 4473).
2715                         docstring const last_modifier = rsplit(fontchange, '\\');
2716                         if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2717                                 os << from_ascii("{}");
2718                         else if (ots.terminateCommand())
2719                                 os << termcmd;
2720                         if (in_ct_deletion) {
2721                                 // We have to close and then reopen \lyxdeleted,
2722                                 // as strikeout needs to be on lowest level.
2723                                 OutputParams rp = runparams;
2724                                 column += Changes::latexMarkChange(os, bparams,
2725                                         Change(Change::UNCHANGED), change, rp);
2726                         }
2727                 }
2728
2729                 // FIXME: think about end_pos implementation...
2730                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2731                         // FIXME: integrate this case in latexSpecialChar
2732                         // Do not print the separation of the optional argument
2733                         // if style.pass_thru is false. This works because
2734                         // latexSpecialChar ignores spaces if
2735                         // style.pass_thru is false.
2736                         if (i != body_pos - 1) {
2737                                 if (d->simpleTeXBlanks(bparams, runparams, os,
2738                                                 i, column, current_font, style)) {
2739                                         // A surrogate pair was output. We
2740                                         // must not call latexSpecialChar
2741                                         // in this iteration, since it would output
2742                                         // the combining character again.
2743                                         ++i;
2744                                         continue;
2745                                 }
2746                         }
2747                 }
2748
2749                 OutputParams rp = runparams;
2750                 rp.free_spacing = style.free_spacing;
2751                 rp.local_font = &current_font;
2752                 rp.intitle = style.intitle;
2753
2754                 // Two major modes:  LaTeX or plain
2755                 // Handle here those cases common to both modes
2756                 // and then split to handle the two modes separately.
2757                 if (c == META_INSET) {
2758                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2759                                 // Greyedout notes and, in general, all insets
2760                                 // with InsetLayout::isDisplay() == false,
2761                                 // are typeset inline with the text. So, we
2762                                 // can add a \par to the last paragraph of
2763                                 // such insets only if nothing else follows.
2764                                 bool incremented = false;
2765                                 Inset const * inset = getInset(i);
2766                                 InsetText const * textinset = inset
2767                                                         ? inset->asInsetText()
2768                                                         : nullptr;
2769                                 if (i + 1 == size() && textinset
2770                                     && !inset->getLayout().isDisplay()) {
2771                                         ParagraphList const & pars =
2772                                                 textinset->text().paragraphs();
2773                                         pit_type const pit = pars.size() - 1;
2774                                         Font const lastfont =
2775                                                 pit < 0 || pars[pit].empty()
2776                                                 ? pars[pit].getLayoutFont(
2777                                                                 bparams,
2778                                                                 real_outerfont)
2779                                                 : pars[pit].getFont(bparams,
2780                                                         pars[pit].size() - 1,
2781                                                         real_outerfont);
2782                                         if (lastfont.fontInfo().size() !=
2783                                             basefont.fontInfo().size()) {
2784                                                 ++parInline;
2785                                                 incremented = true;
2786                                         }
2787                                 }
2788                                 // We need to restore parts of this after insets with
2789                                 // allowMultiPar() true
2790                                 Font const save_basefont = basefont;
2791                                 d->latexInset(bparams, os, rp, running_font,
2792                                                 basefont, real_outerfont, open_font,
2793                                                 runningChange, style, i, column);
2794                                 if (multipar_inset) {
2795                                         if (open_font) {
2796                                                 bool needPar = false;
2797                                                 column += running_font.latexWriteEndChanges(
2798                                                         os, bparams, runparams,
2799                                                         basefont, basefont, needPar);
2800                                                 open_font = false;
2801                                         }
2802                                         basefont.fontInfo().setSize(save_basefont.fontInfo().size());
2803                                         basefont.fontInfo().setFamily(save_basefont.fontInfo().family());
2804                                         basefont.fontInfo().setSeries(save_basefont.fontInfo().series());
2805                                 }
2806                                 if (incremented)
2807                                         --parInline;
2808
2809                                 if (runparams.ctObject == OutputParams::CT_DISPLAYOBJECT
2810                                     || runparams.ctObject == OutputParams::CT_UDISPLAYOBJECT) {
2811                                         // Close \lyx*deleted and force its
2812                                         // reopening (if needed)
2813                                         os << '}';
2814                                         column++;
2815                                         runningChange = Change(Change::UNCHANGED);
2816                                         runparams.ctObject = OutputParams::CT_NORMAL;
2817                                 }
2818                         }
2819                 } else if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2820                         if (!bparams.useNonTeXFonts)
2821                           script = Encodings::isKnownScriptChar(c);
2822                         if (script != alien_script) {
2823                                 if (!alien_script.empty()) {
2824                                         os << "}";
2825                                         alien_script.clear();
2826                                 }
2827                                 string fontenc = running_font.language()->fontenc(bparams);
2828                                 if (!script.empty()
2829                                         && !Encodings::fontencSupportsScript(fontenc, script)) {
2830                                         column += script.length() + 2;
2831                                         os << "\\" << script << "{";
2832                                         alien_script = script;
2833                                 }
2834                         }
2835                         try {
2836                                 d->latexSpecialChar(os, bparams, rp, running_font,
2837                                                                         alien_script, style, i, end_pos, column);
2838                         } catch (EncodingException & e) {
2839                                 if (runparams.dryrun) {
2840                                         os << "<" << _("LyX Warning: ")
2841                                            << _("uncodable character") << " '";
2842                                         os.put(c);
2843                                         os << "'>";
2844                                 } else {
2845                                         // add location information and throw again.
2846                                         e.par_id = id();
2847                                         e.pos = i;
2848                                         throw(e);
2849                                 }
2850                         }
2851                 }
2852
2853                 // Set the encoding to that returned from latexSpecialChar (see
2854                 // comment for encoding member in OutputParams.h)
2855                 runparams.encoding = rp.encoding;
2856
2857                 // Also carry on the info on a closed ulem command for insets
2858                 // such as Note that do not produce any output, so that no
2859                 // command is ever executed but its opening was recorded.
2860                 runparams.inulemcmd = rp.inulemcmd;
2861
2862                 // These need to be passed upstream as well
2863                 runparams.need_maketitle = rp.need_maketitle;
2864                 runparams.have_maketitle = rp.have_maketitle;
2865
2866                 // And finally, pass the post_macros upstream
2867                 runparams.post_macro = rp.post_macro;
2868         }
2869
2870         // Close wrapper for alien script
2871         if (!alien_script.empty()) {
2872                 os << "}";
2873                 alien_script.clear();
2874         }
2875
2876         // If we have an open font definition, we have to close it
2877         if (open_font) {
2878                 // Make sure that \\par is done with the font of the last
2879                 // character if this has another size as the default.
2880                 // This is necessary because LaTeX (and LyX on the screen)
2881                 // calculates the space between the baselines according
2882                 // to this font. (Matthias)
2883                 //
2884                 // We must not change the font for the last paragraph
2885                 // of non-multipar insets, tabular cells or commands,
2886                 // since this produces unwanted whitespace.
2887
2888                 Font const font = empty()
2889                         ? getLayoutFont(bparams, real_outerfont)
2890                         : getFont(bparams, size() - 1, real_outerfont);
2891
2892                 InsetText const * textinset = inInset().asInsetText();
2893
2894                 bool const maintext = textinset
2895                         ? textinset->text().isMainText()
2896                         : false;
2897
2898                 size_t const numpars = textinset
2899                         ? textinset->text().paragraphs().size()
2900                         : 0;
2901
2902                 bool needPar = false;
2903
2904                 if (style.resfont.size() != font.fontInfo().size()
2905                     && (!runparams.isLastPar || maintext
2906                         || (numpars > 1 && d->ownerCode() != CELL_CODE
2907                             && (inInset().getLayout().isDisplay()
2908                                 || parInline)))
2909                     && !style.isCommand()) {
2910                         needPar = true;
2911                 }
2912 #ifdef FIXED_LANGUAGE_END_DETECTION
2913                 if (next_) {
2914                         running_font.latexWriteEndChanges(os, bparams,
2915                                         runparams, basefont,
2916                                         next_->getFont(bparams, 0, outerfont),
2917                                                        needPar);
2918                 } else {
2919                         running_font.latexWriteEndChanges(os, bparams,
2920                                         runparams, basefont, basefont, needPar);
2921                 }
2922 #else
2923 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2924 //FIXME: there as we start another \selectlanguage with the next paragraph if
2925 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2926                 running_font.latexWriteEndChanges(os, bparams, runparams,
2927                                 basefont, basefont, needPar);
2928 #endif
2929                 if (needPar) {
2930                         // The \par could not be inserted at the same nesting
2931                         // level of the font size change, so do it now.
2932                         os << "{\\" << font.latexSize() << "\\par}";
2933                 }
2934         }
2935
2936         column += Changes::latexMarkChange(os, bparams, runningChange,
2937                                            Change(Change::UNCHANGED), runparams);
2938
2939         // Needed if there is an optional argument but no contents.
2940         if (body_pos > 0 && body_pos == size()) {
2941                 os << "}]~";
2942         }
2943
2944         if (!style.rightdelim().empty()) {
2945                 os << style.rightdelim();
2946                 column += style.rightdelim().size();
2947         }
2948
2949         if (allowcust && d->endTeXParParams(bparams, os, runparams)
2950             && runparams.encoding != prev_encoding) {
2951                 runparams.encoding = prev_encoding;
2952                 os << setEncoding(prev_encoding->iconvName());
2953         }
2954
2955         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2956 }
2957
2958
2959 bool Paragraph::emptyTag() const
2960 {
2961         for (pos_type i = 0; i < size(); ++i) {
2962                 if (Inset const * inset = getInset(i)) {
2963                         InsetCode lyx_code = inset->lyxCode();
2964                         // FIXME testing like that is wrong. What is
2965                         // the intent?
2966                         if (lyx_code != TOC_CODE &&
2967                             lyx_code != INCLUDE_CODE &&
2968                             lyx_code != GRAPHICS_CODE &&
2969                             lyx_code != ERT_CODE &&
2970                             lyx_code != LISTINGS_CODE &&
2971                             lyx_code != FLOAT_CODE &&
2972                             lyx_code != TABULAR_CODE) {
2973                                 return false;
2974                         }
2975                 } else {
2976                         char_type c = d->text_[i];
2977                         if (c != ' ' && c != '\t')
2978                                 return false;
2979                 }
2980         }
2981         return true;
2982 }
2983
2984
2985 string Paragraph::getID(Buffer const &, OutputParams const &)
2986         const
2987 {
2988         for (pos_type i = 0; i < size(); ++i) {
2989                 if (Inset const * inset = getInset(i)) {
2990                         InsetCode lyx_code = inset->lyxCode();
2991                         if (lyx_code == LABEL_CODE) {
2992                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2993                                 docstring const & id = il->getParam("name");
2994                                 return "id='" + to_utf8(xml::cleanID(id)) + "'";
2995                         }
2996                 }
2997         }
2998         return string();
2999 }
3000
3001
3002 pos_type Paragraph::firstWordDocBook(XMLStream & xs, OutputParams const & runparams) const
3003 {
3004         pos_type i;
3005         for (i = 0; i < size(); ++i) {
3006                 if (Inset const * inset = getInset(i)) {
3007                         inset->docbook(xs, runparams);
3008                 } else {
3009                         char_type c = d->text_[i];
3010                         if (c == ' ')
3011                                 break;
3012                         xs << c;
3013                 }
3014         }
3015         return i;
3016 }
3017
3018
3019 pos_type Paragraph::firstWordLyXHTML(XMLStream & xs, OutputParams const & runparams)
3020         const
3021 {
3022         pos_type i;
3023         for (i = 0; i < size(); ++i) {
3024                 if (Inset const * inset = getInset(i)) {
3025                         inset->xhtml(xs, runparams);
3026                 } else {
3027                         char_type c = d->text_[i];
3028                         if (c == ' ')
3029                                 break;
3030                         xs << c;
3031                 }
3032         }
3033         return i;
3034 }
3035
3036
3037 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
3038 {
3039         Font font_old;
3040         pos_type size = text_.size();
3041         for (pos_type i = initial; i < size; ++i) {
3042                 Font font = owner_->getFont(buf.params(), i, outerfont);
3043                 if (text_[i] == META_INSET)
3044                         return false;
3045                 if (i != initial && font != font_old)
3046                         return false;
3047                 font_old = font;
3048         }
3049
3050         return true;
3051 }
3052
3053
3054 namespace {
3055
3056 void doFontSwitchDocBook(vector<xml::FontTag> & tagsToOpen,
3057                   vector<xml::EndFontTag> & tagsToClose,
3058                   bool & flag, FontState curstate, xml::FontTypes type)
3059 {
3060         if (curstate == FONT_ON) {
3061                 tagsToOpen.push_back(docbookStartFontTag(type));
3062                 flag = true;
3063         } else if (flag) {
3064                 tagsToClose.push_back(docbookEndFontTag(type));
3065                 flag = false;
3066         }
3067 }
3068
3069 class OptionalFontType {
3070 public:
3071         xml::FontTypes ft;
3072         bool has_value;
3073
3074         OptionalFontType(): ft(xml::FT_EMPH), has_value(false) {} // A possible value at random for ft.
3075         OptionalFontType(xml::FontTypes ft): ft(ft), has_value(true) {}
3076 };
3077
3078 OptionalFontType fontShapeToXml(FontShape fs)
3079 {
3080         switch (fs) {
3081         case ITALIC_SHAPE:
3082                 return {xml::FT_ITALIC};
3083         case SLANTED_SHAPE:
3084                 return {xml::FT_SLANTED};
3085         case SMALLCAPS_SHAPE:
3086                 return {xml::FT_SMALLCAPS};
3087         case UP_SHAPE:
3088         case INHERIT_SHAPE:
3089                 return {};
3090         default:
3091                 // the other tags are for internal use
3092                 LATTEST(false);
3093                 return {};
3094         }
3095 }
3096
3097 OptionalFontType fontFamilyToXml(FontFamily fm)
3098 {
3099         switch (fm) {
3100         case ROMAN_FAMILY:
3101                 return {xml::FT_ROMAN};
3102         case SANS_FAMILY:
3103                 return {xml::FT_SANS};
3104         case TYPEWRITER_FAMILY:
3105                 return {xml::FT_TYPE};
3106         case INHERIT_FAMILY:
3107                 return {};
3108         default:
3109                 // the other tags are for internal use
3110                 LATTEST(false);
3111                 return {};
3112         }
3113 }
3114
3115 OptionalFontType fontSizeToXml(FontSize fs)
3116 {
3117         switch (fs) {
3118         case TINY_SIZE:
3119                 return {xml::FT_SIZE_TINY};
3120         case SCRIPT_SIZE:
3121                 return {xml::FT_SIZE_SCRIPT};
3122         case FOOTNOTE_SIZE:
3123                 return {xml::FT_SIZE_FOOTNOTE};
3124         case SMALL_SIZE:
3125                 return {xml::FT_SIZE_SMALL};
3126         case LARGE_SIZE:
3127                 return {xml::FT_SIZE_LARGE};
3128         case LARGER_SIZE:
3129                 return {xml::FT_SIZE_LARGER};
3130         case LARGEST_SIZE:
3131                 return {xml::FT_SIZE_LARGEST};
3132         case HUGE_SIZE:
3133                 return {xml::FT_SIZE_HUGE};
3134         case HUGER_SIZE:
3135                 return {xml::FT_SIZE_HUGER};
3136         case INCREASE_SIZE:
3137                 return {xml::FT_SIZE_INCREASE};
3138         case DECREASE_SIZE:
3139                 return {xml::FT_SIZE_DECREASE};
3140         case INHERIT_SIZE:
3141         case NORMAL_SIZE:
3142                 return {};
3143         default:
3144                 // the other tags are for internal use
3145                 LATTEST(false);
3146                 return {};
3147         }
3148 }
3149
3150 struct DocBookFontState
3151 {
3152         FontShape  curr_fs   = INHERIT_SHAPE;
3153         FontFamily curr_fam  = INHERIT_FAMILY;
3154         FontSize   curr_size = INHERIT_SIZE;
3155
3156         // track whether we have opened these tags
3157         bool emph_flag = false;
3158         bool bold_flag = false;
3159         bool noun_flag = false;
3160         bool ubar_flag = false;
3161         bool dbar_flag = false;
3162         bool sout_flag = false;
3163         bool xout_flag = false;
3164         bool wave_flag = false;
3165         // shape tags
3166         bool shap_flag = false;
3167         // family tags
3168         bool faml_flag = false;
3169         // size tags
3170         bool size_flag = false;
3171 };
3172
3173 std::tuple<vector<xml::FontTag>, vector<xml::EndFontTag>> computeDocBookFontSwitch(FontInfo const & font_old,
3174                                                                                            Font const & font,
3175                                                                                            std::string const & default_family,
3176                                                                                            DocBookFontState & fs)
3177 {
3178         vector<xml::FontTag> tagsToOpen;
3179         vector<xml::EndFontTag> tagsToClose;
3180
3181         // emphasis
3182         FontState curstate = font.fontInfo().emph();
3183         if (font_old.emph() != curstate)
3184                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.emph_flag, curstate, xml::FT_EMPH);
3185
3186         // noun
3187         curstate = font.fontInfo().noun();
3188         if (font_old.noun() != curstate)
3189                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.noun_flag, curstate, xml::FT_NOUN);
3190
3191         // underbar
3192         curstate = font.fontInfo().underbar();
3193         if (font_old.underbar() != curstate)
3194                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.ubar_flag, curstate, xml::FT_UBAR);
3195
3196         // strikeout
3197         curstate = font.fontInfo().strikeout();
3198         if (font_old.strikeout() != curstate)
3199                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.sout_flag, curstate, xml::FT_SOUT);
3200
3201         // xout
3202         curstate = font.fontInfo().xout();
3203         if (font_old.xout() != curstate)
3204                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.xout_flag, curstate, xml::FT_XOUT);
3205
3206         // double underbar
3207         curstate = font.fontInfo().uuline();
3208         if (font_old.uuline() != curstate)
3209                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.dbar_flag, curstate, xml::FT_DBAR);
3210
3211         // wavy line
3212         curstate = font.fontInfo().uwave();
3213         if (font_old.uwave() != curstate)
3214                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.wave_flag, curstate, xml::FT_WAVE);
3215
3216         // bold
3217         // a little hackish, but allows us to reuse what we have.
3218         curstate = (font.fontInfo().series() == BOLD_SERIES ? FONT_ON : FONT_OFF);
3219         if (font_old.series() != font.fontInfo().series())
3220                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.bold_flag, curstate, xml::FT_BOLD);
3221
3222         // Font shape
3223         fs.curr_fs = font.fontInfo().shape();
3224         FontShape old_fs = font_old.shape();
3225         if (old_fs != fs.curr_fs) {
3226                 if (fs.shap_flag) {
3227                         OptionalFontType tag = fontShapeToXml(old_fs);
3228                         if (tag.has_value)
3229                                 tagsToClose.push_back(docbookEndFontTag(tag.ft));
3230                         fs.shap_flag = false;
3231                 }
3232
3233                 OptionalFontType tag = fontShapeToXml(fs.curr_fs);
3234                 if (tag.has_value)
3235                         tagsToOpen.push_back(docbookStartFontTag(tag.ft));
3236         }
3237
3238         // Font family
3239         fs.curr_fam = font.fontInfo().family();
3240         FontFamily old_fam = font_old.family();
3241         if (old_fam != fs.curr_fam) {
3242                 if (fs.faml_flag) {
3243                         OptionalFontType tag = fontFamilyToXml(old_fam);
3244                         if (tag.has_value)
3245                                 tagsToClose.push_back(docbookEndFontTag(tag.ft));
3246                         fs.faml_flag = false;
3247                 }
3248                 switch (fs.curr_fam) {
3249                         case ROMAN_FAMILY:
3250                                 // we will treat a "default" font family as roman, since we have
3251                                 // no other idea what to do.
3252                                 if (default_family != "rmdefault" && default_family != "default") {
3253                                         tagsToOpen.push_back(docbookStartFontTag(xml::FT_ROMAN));
3254                                         fs.faml_flag = true;
3255                                 }
3256                                 break;
3257                         case SANS_FAMILY:
3258                                 if (default_family != "sfdefault") {
3259                                         tagsToOpen.push_back(docbookStartFontTag(xml::FT_SANS));
3260                                         fs.faml_flag = true;
3261                                 }
3262                                 break;
3263                         case TYPEWRITER_FAMILY:
3264                                 if (default_family != "ttdefault") {
3265                                         tagsToOpen.push_back(docbookStartFontTag(xml::FT_TYPE));
3266                                         fs.faml_flag = true;
3267                                 }
3268                                 break;
3269                         case INHERIT_FAMILY:
3270                                 break;
3271                         default:
3272                                 // the other tags are for internal use
3273                                 LATTEST(false);
3274                                 break;
3275                 }
3276         }
3277
3278         // Font size
3279         fs.curr_size = font.fontInfo().size();
3280         FontSize old_size = font_old.size();
3281         if (old_size != fs.curr_size) {
3282                 if (fs.size_flag) {
3283                         OptionalFontType tag = fontSizeToXml(old_size);
3284                         if (tag.has_value)
3285                                 tagsToClose.push_back(docbookEndFontTag(tag.ft));
3286                         fs.size_flag = false;
3287                 }
3288
3289                 OptionalFontType tag = fontSizeToXml(fs.curr_size);
3290                 if (tag.has_value) {
3291                         tagsToOpen.push_back(docbookStartFontTag(tag.ft));
3292                         fs.size_flag = true;
3293                 }
3294         }
3295
3296         return std::tuple<vector<xml::FontTag>, vector<xml::EndFontTag>>(tagsToOpen, tagsToClose);
3297 }
3298
3299 } // anonymous namespace
3300
3301
3302 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
3303                                     XMLStream & xs,
3304                                     OutputParams const & runparams,
3305                                     Font const & outerfont,
3306                                     bool start_paragraph, bool close_paragraph,
3307                                     pos_type initial) const
3308 {
3309         // track whether we have opened these tags
3310         DocBookFontState fs;
3311
3312         if (start_paragraph)
3313                 xs.startDivision(allowEmpty());
3314
3315         Layout const & style = *d->layout_;
3316         FontInfo font_old =
3317                         style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
3318
3319         string const default_family =
3320                         buf.masterBuffer()->params().fonts_default_family;
3321
3322         vector<xml::FontTag> tagsToOpen;
3323         vector<xml::EndFontTag> tagsToClose;
3324
3325         // parsing main loop
3326         for (pos_type i = initial; i < size(); ++i) {
3327                 // let's not show deleted material in the output
3328                 if (isDeleted(i))
3329                         continue;
3330
3331                 Font const font = getFont(buf.masterBuffer()->params(), i, outerfont);
3332
3333                 // Determine which tags should be opened or closed.
3334                 tie(tagsToOpen, tagsToClose) = computeDocBookFontSwitch(font_old, font, default_family, fs);
3335
3336                 // FIXME XHTML
3337                 // Other such tags? What about the other text ranges?
3338
3339                 vector<xml::EndFontTag>::const_iterator cit = tagsToClose.begin();
3340                 vector<xml::EndFontTag>::const_iterator cen = tagsToClose.end();
3341                 for (; cit != cen; ++cit)
3342                         xs << *cit;
3343
3344                 vector<xml::FontTag>::const_iterator sit = tagsToOpen.begin();
3345                 vector<xml::FontTag>::const_iterator sen = tagsToOpen.end();
3346                 for (; sit != sen; ++sit)
3347                         xs << *sit;
3348
3349                 tagsToClose.clear();
3350                 tagsToOpen.clear();
3351
3352                 if (Inset const * inset = getInset(i)) {
3353                         if (!runparams.for_toc || inset->isInToc()) {
3354                                 OutputParams np = runparams;
3355                                 np.local_font = &font;
3356                                 // If the paragraph has size 1, then we are in the "special
3357                                 // case" where we do not output the containing paragraph info.
3358                                 // This "special case" is defined in more details in output_docbook.cpp, makeParagraphs. The results
3359                                 // of that brittle logic is passed to this function through open_par.
3360                                 if (!inset->getLayout().htmlisblock() && size() != 1) // TODO: htmlisblock here too!
3361                                         np.docbook_in_par = true;
3362                                 inset->docbook(xs, np);
3363                         }
3364                 } else {
3365                         char_type c = getUChar(buf.masterBuffer()->params(), runparams, i);
3366                         xs << c;
3367                 }
3368                 font_old = font.fontInfo();
3369         }
3370
3371         // FIXME, this code is just imported from XHTML
3372         // I'm worried about what happens if a branch, say, is itself
3373         // wrapped in some font stuff. I think that will not work.
3374         xs.closeFontTags();
3375         if (runparams.docbook_in_listing)
3376                 xs << xml::CR();
3377         if (close_paragraph)
3378                 xs.endDivision();
3379 }
3380
3381
3382 namespace {
3383
3384 void doFontSwitchXHTML(vector<xml::FontTag> & tagsToOpen,
3385                   vector<xml::EndFontTag> & tagsToClose,
3386                   bool & flag, FontState curstate, xml::FontTypes type)
3387 {
3388         if (curstate == FONT_ON) {
3389                 tagsToOpen.push_back(xhtmlStartFontTag(type));
3390                 flag = true;
3391         } else if (flag) {
3392                 tagsToClose.push_back(xhtmlEndFontTag(type));
3393                 flag = false;
3394         }
3395 }
3396
3397 } // anonymous namespace
3398
3399
3400 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
3401                                     XMLStream & xs,
3402                                     OutputParams const & runparams,
3403                                     Font const & outerfont,
3404                                     bool start_paragraph, bool close_paragraph,
3405                                     pos_type initial) const
3406 {
3407         docstring retval;
3408
3409         // track whether we have opened these tags
3410         bool emph_flag = false;
3411         bool bold_flag = false;
3412         bool noun_flag = false;
3413         bool ubar_flag = false;
3414         bool dbar_flag = false;
3415         bool sout_flag = false;
3416         bool xout_flag = false;
3417         bool wave_flag = false;
3418         // shape tags
3419         bool shap_flag = false;
3420         // family tags
3421         bool faml_flag = false;
3422         // size tags
3423         bool size_flag = false;
3424
3425         Layout const & style = *d->layout_;
3426
3427         if (start_paragraph)
3428                 xs.startDivision(allowEmpty());
3429
3430         FontInfo font_old =
3431                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
3432
3433         FontShape  curr_fs   = INHERIT_SHAPE;
3434         FontFamily curr_fam  = INHERIT_FAMILY;
3435         FontSize   curr_size = INHERIT_SIZE;
3436
3437         string const default_family =
3438                 buf.masterBuffer()->params().fonts_default_family;
3439
3440         vector<xml::FontTag> tagsToOpen;
3441         vector<xml::EndFontTag> tagsToClose;
3442
3443         // parsing main loop
3444         for (pos_type i = initial; i < size(); ++i) {
3445                 // let's not show deleted material in the output
3446                 if (isDeleted(i))
3447                         continue;
3448
3449                 Font const font = getFont(buf.masterBuffer()->params(), i, outerfont);
3450
3451                 // emphasis
3452                 FontState curstate = font.fontInfo().emph();
3453                 if (font_old.emph() != curstate)
3454                         doFontSwitchXHTML(tagsToOpen, tagsToClose, emph_flag, curstate, xml::FT_EMPH);
3455
3456                 // noun
3457                 curstate = font.fontInfo().noun();
3458                 if (font_old.noun() != curstate)
3459                         doFontSwitchXHTML(tagsToOpen, tagsToClose, noun_flag, curstate, xml::FT_NOUN);
3460
3461                 // underbar
3462                 curstate = font.fontInfo().underbar();
3463                 if (font_old.underbar() != curstate)
3464                         doFontSwitchXHTML(tagsToOpen, tagsToClose, ubar_flag, curstate, xml::FT_UBAR);
3465
3466                 // strikeout
3467                 curstate = font.fontInfo().strikeout();
3468                 if (font_old.strikeout() != curstate)
3469                         doFontSwitchXHTML(tagsToOpen, tagsToClose, sout_flag, curstate, xml::FT_SOUT);
3470
3471                 // xout
3472                 curstate = font.fontInfo().xout();
3473                 if (font_old.xout() != curstate)
3474                         doFontSwitchXHTML(tagsToOpen, tagsToClose, xout_flag, curstate, xml::FT_XOUT);
3475
3476                 // double underbar
3477                 curstate = font.fontInfo().uuline();
3478                 if (font_old.uuline() != curstate)
3479                         doFontSwitchXHTML(tagsToOpen, tagsToClose, dbar_flag, curstate, xml::FT_DBAR);
3480
3481                 // wavy line
3482                 curstate = font.fontInfo().uwave();
3483                 if (font_old.uwave() != curstate)
3484                         doFontSwitchXHTML(tagsToOpen, tagsToClose, wave_flag, curstate, xml::FT_WAVE);
3485
3486                 // bold
3487                 // a little hackish, but allows us to reuse what we have.
3488                 curstate = (font.fontInfo().series() == BOLD_SERIES ? FONT_ON : FONT_OFF);
3489                 if (font_old.series() != font.fontInfo().series())
3490                         doFontSwitchXHTML(tagsToOpen, tagsToClose, bold_flag, curstate, xml::FT_BOLD);
3491
3492                 // Font shape
3493                 curr_fs = font.fontInfo().shape();
3494                 FontShape old_fs = font_old.shape();
3495                 if (old_fs != curr_fs) {
3496                         if (shap_flag) {
3497                                 switch (old_fs) {
3498                                 case ITALIC_SHAPE:
3499                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_ITALIC));
3500                                         break;
3501                                 case SLANTED_SHAPE:
3502                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SLANTED));
3503                                         break;
3504                                 case SMALLCAPS_SHAPE:
3505                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SMALLCAPS));
3506                                         break;
3507                                 case UP_SHAPE:
3508                                 case INHERIT_SHAPE:
3509                                         break;
3510                                 default:
3511                                         // the other tags are for internal use
3512                                         LATTEST(false);
3513                                         break;
3514                                 }
3515                                 shap_flag = false;
3516                         }
3517                         switch (curr_fs) {
3518                         case ITALIC_SHAPE:
3519                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_ITALIC));
3520                                 shap_flag = true;
3521                                 break;
3522                         case SLANTED_SHAPE:
3523                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SLANTED));
3524                                 shap_flag = true;
3525                                 break;
3526                         case SMALLCAPS_SHAPE:
3527                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SMALLCAPS));
3528                                 shap_flag = true;
3529                                 break;
3530                         case UP_SHAPE:
3531                         case INHERIT_SHAPE:
3532                                 break;
3533                         default:
3534                                 // the other tags are for internal use
3535                                 LATTEST(false);
3536                                 break;
3537                         }
3538                 }
3539
3540                 // Font family
3541                 curr_fam = font.fontInfo().family();
3542                 FontFamily old_fam = font_old.family();
3543                 if (old_fam != curr_fam) {
3544                         if (faml_flag) {
3545                                 switch (old_fam) {
3546                                 case ROMAN_FAMILY:
3547                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_ROMAN));
3548                                         break;
3549                                 case SANS_FAMILY:
3550                                     tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SANS));
3551                                     break;
3552                                 case TYPEWRITER_FAMILY:
3553                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_TYPE));
3554                                         break;
3555                                 case INHERIT_FAMILY:
3556                                         break;
3557                                 default:
3558                                         // the other tags are for internal use
3559                                         LATTEST(false);
3560                                         break;
3561                                 }
3562                                 faml_flag = false;
3563                         }
3564                         switch (curr_fam) {
3565                         case ROMAN_FAMILY:
3566                                 // we will treat a "default" font family as roman, since we have
3567                                 // no other idea what to do.
3568                                 if (default_family != "rmdefault" && default_family != "default") {
3569                                         tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_ROMAN));
3570                                         faml_flag = true;
3571                                 }
3572                                 break;
3573                         case SANS_FAMILY:
3574                                 if (default_family != "sfdefault") {
3575                                         tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SANS));
3576                                         faml_flag = true;
3577                                 }
3578                                 break;
3579                         case TYPEWRITER_FAMILY:
3580                                 if (default_family != "ttdefault") {
3581                                         tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_TYPE));
3582                                         faml_flag = true;
3583                                 }
3584                                 break;
3585                         case INHERIT_FAMILY:
3586                                 break;
3587                         default:
3588                                 // the other tags are for internal use
3589                                 LATTEST(false);
3590                                 break;
3591                         }
3592                 }
3593
3594                 // Font size
3595                 curr_size = font.fontInfo().size();
3596                 FontSize old_size = font_old.size();
3597                 if (old_size != curr_size) {
3598                         if (size_flag) {
3599                                 switch (old_size) {
3600                                 case TINY_SIZE:
3601                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_TINY));
3602                                         break;
3603                                 case SCRIPT_SIZE:
3604                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_SCRIPT));
3605                                         break;
3606                                 case FOOTNOTE_SIZE:
3607                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_FOOTNOTE));
3608                                         break;
3609                                 case SMALL_SIZE:
3610                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_SMALL));
3611                                         break;
3612                                 case LARGE_SIZE:
3613                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_LARGE));
3614                                         break;
3615                                 case LARGER_SIZE:
3616                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_LARGER));
3617                                         break;
3618                                 case LARGEST_SIZE:
3619                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_LARGEST));
3620                                         break;
3621                                 case HUGE_SIZE:
3622                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_HUGE));
3623                                         break;
3624                                 case HUGER_SIZE:
3625                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_HUGER));
3626                                         break;
3627                                 case INCREASE_SIZE:
3628                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_INCREASE));
3629                                         break;
3630                                 case DECREASE_SIZE:
3631                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_DECREASE));
3632                                         break;
3633                                 case INHERIT_SIZE:
3634                                 case NORMAL_SIZE:
3635                                         break;
3636                                 default:
3637                                         // the other tags are for internal use
3638                                         LATTEST(false);
3639                                         break;
3640                                 }
3641                                 size_flag = false;
3642                         }
3643                         switch (curr_size) {
3644                         case TINY_SIZE:
3645                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_TINY));
3646                                 size_flag = true;
3647                                 break;
3648                         case SCRIPT_SIZE:
3649                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_SCRIPT));
3650                                 size_flag = true;
3651                                 break;
3652                         case FOOTNOTE_SIZE:
3653                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_FOOTNOTE));
3654                                 size_flag = true;
3655                                 break;
3656                         case SMALL_SIZE:
3657                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_SMALL));
3658                                 size_flag = true;
3659                                 break;
3660                         case LARGE_SIZE:
3661                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_LARGE));
3662                                 size_flag = true;
3663                                 break;
3664                         case LARGER_SIZE:
3665                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_LARGER));
3666                                 size_flag = true;
3667                                 break;
3668                         case LARGEST_SIZE:
3669                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_LARGEST));
3670                                 size_flag = true;
3671                                 break;
3672                         case HUGE_SIZE:
3673                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_HUGE));
3674                                 size_flag = true;
3675                                 break;
3676                         case HUGER_SIZE:
3677                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_HUGER));
3678                                 size_flag = true;
3679                                 break;
3680                         case INCREASE_SIZE:
3681                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_INCREASE));
3682                                 size_flag = true;
3683                                 break;
3684                         case DECREASE_SIZE:
3685                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_DECREASE));
3686                                 size_flag = true;
3687                                 break;
3688                         case INHERIT_SIZE:
3689                         case NORMAL_SIZE:
3690                                 break;
3691                         default:
3692                                 // the other tags are for internal use
3693                                 LATTEST(false);
3694                                 break;
3695                         }
3696                 }
3697
3698                 // FIXME XHTML
3699                 // Other such tags? What about the other text ranges?
3700
3701                 vector<xml::EndFontTag>::const_iterator cit = tagsToClose.begin();
3702                 vector<xml::EndFontTag>::const_iterator cen = tagsToClose.end();
3703                 for (; cit != cen; ++cit)
3704                         xs << *cit;
3705
3706                 vector<xml::FontTag>::const_iterator sit = tagsToOpen.begin();
3707                 vector<xml::FontTag>::const_iterator sen = tagsToOpen.end();
3708                 for (; sit != sen; ++sit)
3709                         xs << *sit;
3710
3711                 tagsToClose.clear();
3712                 tagsToOpen.clear();
3713
3714                 Inset const * inset = getInset(i);
3715                 if (inset) {
3716                         if (!runparams.for_toc || inset->isInToc()) {
3717                                 OutputParams np = runparams;
3718                                 np.local_font = &font;
3719                                 // If the paragraph has size 1, then we are in the "special
3720                                 // case" where we do not output the containing paragraph info
3721                                 if (!inset->getLayout().htmlisblock() && size() != 1)
3722                                         np.html_in_par = true;
3723                                 retval += inset->xhtml(xs, np);
3724                         }
3725                 } else {
3726                         char_type c = getUChar(buf.masterBuffer()->params(),
3727                                                runparams, i);
3728                         if (c == ' ' && (style.free_spacing || runparams.free_spacing))
3729                                 xs << XMLStream::ESCAPE_NONE << "&nbsp;";
3730                         else
3731                                 xs << c;
3732                 }
3733                 font_old = font.fontInfo();
3734         }
3735
3736         // FIXME XHTML
3737         // I'm worried about what happens if a branch, say, is itself
3738         // wrapped in some font stuff. I think that will not work.
3739         xs.closeFontTags();
3740         if (close_paragraph)
3741                 xs.endDivision();
3742
3743         return retval;
3744 }
3745
3746
3747 bool Paragraph::isHfill(pos_type pos) const
3748 {
3749         Inset const * inset = getInset(pos);
3750         return inset && inset->isHfill();
3751 }
3752
3753
3754 bool Paragraph::isNewline(pos_type pos) const
3755 {
3756         // U+2028 LINE SEPARATOR
3757         // U+2029 PARAGRAPH SEPARATOR
3758         char_type const c = d->text_[pos];
3759         if (c == 0x2028 || c == 0x2029)
3760                 return true;
3761         Inset const * inset = getInset(pos);
3762         return inset && inset->lyxCode() == NEWLINE_CODE;
3763 }
3764
3765
3766 bool Paragraph::isEnvSeparator(pos_type pos) const
3767 {
3768         Inset const * inset = getInset(pos);
3769         return inset && inset->lyxCode() == SEPARATOR_CODE;
3770 }
3771
3772
3773 bool Paragraph::isLineSeparator(pos_type pos) const
3774 {
3775         char_type const c = d->text_[pos];
3776         if (isLineSeparatorChar(c))
3777                 return true;
3778         Inset const * inset = getInset(pos);
3779         return inset && inset->isLineSeparator();
3780 }
3781
3782
3783 bool Paragraph::isWordSeparator(pos_type pos, bool const ignore_deleted) const
3784 {
3785         if (pos == size())
3786                 return true;
3787         if (ignore_deleted && isDeleted(pos))
3788                 return false;
3789         if (Inset const * inset = getInset(pos))
3790                 return !inset->isLetter();
3791         // if we have a hard hyphen (no en- or emdash) or apostrophe
3792         // we pass this to the spell checker
3793         // FIXME: this method is subject to change, visit
3794         // https://bugzilla.mozilla.org/show_bug.cgi?id=355178
3795         // to get an impression how complex this is.
3796         if (isHardHyphenOrApostrophe(pos))
3797                 return false;
3798         char_type const c = d->text_[pos];
3799         // We want to pass the escape chars to the spellchecker
3800         docstring const escape_chars = from_utf8(lyxrc.spellchecker_esc_chars);
3801         return !isLetterChar(c) && !isDigitASCII(c) && !contains(escape_chars, c);
3802 }
3803
3804
3805 bool Paragraph::isHardHyphenOrApostrophe(pos_type pos) const
3806 {
3807         pos_type const psize = size();
3808         if (pos >= psize)
3809                 return false;
3810         char_type const c = d->text_[pos];
3811         if (c != '-' && c != '\'')
3812                 return false;
3813         int nextpos = pos + 1;
3814         int prevpos = pos > 0 ? pos - 1 : 0;
3815         if ((nextpos == psize || isSpace(nextpos))
3816                 && (pos == 0 || isSpace(prevpos)))
3817                 return false;
3818         return true;
3819 }
3820
3821
3822 bool Paragraph::needsCProtection(bool const fragile) const
3823 {
3824         // first check the layout of the paragraph, but only in insets
3825         InsetText const * textinset = inInset().asInsetText();
3826         bool const maintext = textinset
3827                 ? textinset->text().isMainText()
3828                 : false;
3829
3830         if (!maintext && layout().needcprotect) {
3831                 // Environments need cprotection regardless the content
3832                 if (layout().latextype == LATEX_ENVIRONMENT)
3833                         return true;
3834
3835                 // Commands need cprotection if they contain specific chars
3836                 int const nchars_escape = 9;
3837                 static char_type const chars_escape[nchars_escape] = {
3838                         '&', '_', '$', '%', '#', '^', '{', '}', '\\'};
3839
3840                 docstring const pars = asString();
3841                 for (int k = 0; k < nchars_escape; k++) {
3842                         if (contains(pars, chars_escape[k]))
3843                                 return true;
3844                 }
3845         }
3846
3847         // now check whether we have insets that need cprotection
3848         pos_type size = pos_type(d->text_.size());
3849         for (pos_type i = 0; i < size; ++i) {
3850                 if (!isInset(i))
3851                         continue;
3852                 Inset const * ins = getInset(i);
3853                 if (ins->needsCProtection(maintext, fragile))
3854                         return true;
3855                 if (ins->getLayout().latextype() == InsetLayout::ENVIRONMENT)
3856                         // Environments need cprotection regardless the content
3857                         return true;
3858                 // Now check math environments
3859                 InsetMath const * im = getInset(i)->asInsetMath();
3860                 if (!im || im->cell(0).empty())
3861                         continue;
3862                 switch(im->cell(0)[0]->lyxCode()) {
3863                 case MATH_AMSARRAY_CODE:
3864                 case MATH_SUBSTACK_CODE:
3865                 case MATH_ENV_CODE:
3866                 case MATH_XYMATRIX_CODE:
3867                         // these need cprotection
3868                         return true;
3869                 default:
3870                         break;
3871                 }
3872         }
3873
3874         return false;
3875 }
3876
3877
3878 FontSpan const & Paragraph::getSpellRange(pos_type pos) const
3879 {
3880         return d->speller_state_.getRange(pos);
3881 }
3882
3883
3884 bool Paragraph::isChar(pos_type pos) const
3885 {
3886         if (Inset const * inset = getInset(pos))
3887                 return inset->isChar();
3888         char_type const c = d->text_[pos];
3889         return !isLetterChar(c) && !isDigitASCII(c) && !lyx::isSpace(c);
3890 }
3891
3892
3893 bool Paragraph::isSpace(pos_type pos) const
3894 {
3895         if (Inset const * inset = getInset(pos))
3896                 return inset->isSpace();
3897         char_type const c = d->text_[pos];
3898         return lyx::isSpace(c);
3899 }
3900
3901
3902 Language const *
3903 Paragraph::getParLanguage(BufferParams const & bparams) const
3904 {
3905         if (!empty())
3906                 return getFirstFontSettings(bparams).language();
3907         // FIXME: we should check the prev par as well (Lgb)
3908         return bparams.language;
3909 }
3910
3911
3912 bool Paragraph::isRTL(BufferParams const & bparams) const
3913 {
3914         return getParLanguage(bparams)->rightToLeft()
3915                 && !inInset().getLayout().forceLTR();
3916 }
3917
3918
3919 void Paragraph::changeLanguage(BufferParams const & bparams,
3920                                Language const * from, Language const * to)
3921 {
3922         // change language including dummy font change at the end
3923         for (pos_type i = 0; i <= size(); ++i) {
3924                 Font font = getFontSettings(bparams, i);
3925                 if (font.language() == from) {
3926                         font.setLanguage(to);
3927                         setFont(i, font);
3928                         d->requestSpellCheck(i);
3929                 }
3930         }
3931 }
3932
3933
3934 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
3935 {
3936         Language const * doc_language = bparams.language;
3937         for (auto const & f : d->fontlist_)
3938                 if (f.font().language() != ignore_language &&
3939                     f.font().language() != latex_language &&
3940                     f.font().language() != doc_language)
3941                         return true;
3942         return false;
3943 }
3944
3945
3946 void Paragraph::getLanguages(std::set<Language const *> & langs) const
3947 {
3948         for (auto const & f : d->fontlist_) {
3949                 Language const * lang = f.font().language();
3950                 if (lang != ignore_language &&
3951                     lang != latex_language)
3952                         langs.insert(lang);
3953         }
3954 }
3955
3956
3957 docstring Paragraph::asString(int options) const
3958 {
3959         return asString(0, size(), options);
3960 }
3961
3962
3963 docstring Paragraph::asString(pos_type beg, pos_type end, int options, const OutputParams *runparams) const
3964 {
3965         odocstringstream os;
3966
3967         if (beg == 0
3968             && options & AS_STR_LABEL
3969             && !d->params_.labelString().empty())
3970                 os << d->params_.labelString() << ' ';
3971
3972         for (pos_type i = beg; i < end; ++i) {
3973                 if ((options & AS_STR_SKIPDELETE) && isDeleted(i))
3974                         continue;
3975                 char_type const c = d->text_[i];
3976                 if (isPrintable(c) || c == '\t'
3977                     || (c == '\n' && (options & AS_STR_NEWLINES)))
3978                         os.put(c);
3979                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
3980                         if (c == META_INSET && (options & AS_STR_PLAINTEXT)) {
3981                                 LASSERT(runparams != nullptr, return docstring());
3982                                 getInset(i)->plaintext(os, *runparams);
3983                         } else {
3984                                 getInset(i)->toString(os);
3985                         }
3986                 }
3987         }
3988
3989         return os.str();
3990 }
3991
3992
3993 void Paragraph::forOutliner(docstring & os, size_t const maxlen,
3994                             bool const shorten, bool const label) const
3995 {
3996         size_t tmplen = shorten ? maxlen + 1 : maxlen;
3997         if (label && !labelString().empty())
3998                 os += labelString() + ' ';
3999         if (!layout().isTocCaption())
4000                 return;
4001         for (pos_type i = 0; i < size() && os.length() < tmplen; ++i) {
4002                 if (isDeleted(i))
4003                         continue;
4004                 char_type const c = d->text_[i];
4005                 if (isPrintable(c))
4006                         os += c;
4007                 else if (c == META_INSET)
4008                         getInset(i)->forOutliner(os, tmplen, false);
4009         }
4010         if (shorten)
4011                 Text::shortenForOutliner(os, maxlen);
4012 }
4013
4014
4015 void Paragraph::setInsetOwner(Inset const * inset)
4016 {
4017         d->inset_owner_ = inset;
4018 }
4019
4020
4021 int Paragraph::id() const
4022 {
4023         return d->id_;
4024 }
4025
4026
4027 void Paragraph::setId(int id)
4028 {
4029         d->id_ = id;
4030 }
4031
4032
4033 Layout const & Paragraph::layout() const
4034 {
4035         return *d->layout_;
4036 }
4037
4038
4039 void Paragraph::setLayout(Layout const & layout)
4040 {
4041         d->layout_ = &layout;
4042 }
4043
4044
4045 void Paragraph::setDefaultLayout(DocumentClass const & tc)
4046 {
4047         setLayout(tc.defaultLayout());
4048 }
4049
4050
4051 void Paragraph::setPlainLayout(DocumentClass const & tc)
4052 {
4053         setLayout(tc.plainLayout());
4054 }
4055
4056
4057 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
4058 {
4059         if (usePlainLayout())
4060                 setPlainLayout(tclass);
4061         else
4062                 setDefaultLayout(tclass);
4063 }
4064
4065
4066 Inset const & Paragraph::inInset() const
4067 {
4068         LBUFERR(d->inset_owner_);
4069         return *d->inset_owner_;
4070 }
4071
4072
4073 ParagraphParameters & Paragraph::params()
4074 {
4075         return d->params_;
4076 }
4077
4078
4079 ParagraphParameters const & Paragraph::params() const
4080 {
4081         return d->params_;
4082 }
4083
4084
4085 bool Paragraph::isFreeSpacing() const
4086 {
4087         if (d->layout_->free_spacing)
4088                 return true;
4089         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
4090 }
4091
4092
4093 bool Paragraph::allowEmpty() const
4094 {
4095         if (d->layout_->keepempty)
4096                 return true;
4097         return d->inset_owner_ && d->inset_owner_->allowEmpty();
4098 }
4099
4100
4101 bool Paragraph::brokenBiblio() const
4102 {
4103         // There is a problem if there is no bibitem at position 0 in
4104         // paragraphs that need one, if there is another bibitem in the
4105         // paragraph or if this paragraph is not supposed to have
4106         // a bibitem inset at all.
4107         return ((d->layout_->labeltype == LABEL_BIBLIO
4108                 && (d->insetlist_.find(BIBITEM_CODE) != 0
4109                     || d->insetlist_.find(BIBITEM_CODE, 1) > 0))
4110                 || (d->layout_->labeltype != LABEL_BIBLIO
4111                     && d->insetlist_.find(BIBITEM_CODE) != -1));
4112 }
4113
4114
4115 int Paragraph::fixBiblio(Buffer const & buffer)
4116 {
4117         // FIXME: when there was already an inset at 0, the return value is 1,
4118         // which does not tell whether another inset has been removed; the
4119         // cursor cannot be correctly updated.
4120
4121         bool const track_changes = buffer.params().track_changes;
4122         int bibitem_pos = d->insetlist_.find(BIBITEM_CODE);
4123
4124         // The case where paragraph is not BIBLIO
4125         if (d->layout_->labeltype != LABEL_BIBLIO) {
4126                 if (bibitem_pos == -1)
4127                         // No InsetBibitem => OK
4128                         return 0;
4129                 // There is an InsetBibitem: remove it!
4130                 d->insetlist_.release(bibitem_pos);
4131                 eraseChar(bibitem_pos, track_changes);
4132                 return (bibitem_pos == 0) ? -1 : -bibitem_pos;
4133         }
4134
4135         bool const hasbibitem0 = bibitem_pos == 0;
4136         if (hasbibitem0) {
4137                 bibitem_pos = d->insetlist_.find(BIBITEM_CODE, 1);
4138                 // There was an InsetBibitem at pos 0,
4139                 // and no other one => OK
4140                 if (bibitem_pos == -1)
4141                         return 0;
4142                 // there is a bibitem at the 0 position, but since
4143                 // there is a second one, we copy the second on the
4144                 // first. We're assuming there are at most two of
4145                 // these, which there should be.
4146                 // FIXME: why does it make sense to do that rather
4147                 // than keep the first? (JMarc)
4148                 Inset * inset = releaseInset(bibitem_pos);
4149                 d->insetlist_.begin()->inset = inset;
4150                 // This needs to be done to update the counter (#8499)
4151                 buffer.updateBuffer();
4152                 return -bibitem_pos;
4153         }
4154
4155         // We need to create an inset at the beginning
4156         Inset * inset = nullptr;
4157         if (bibitem_pos > 0) {
4158                 // there was one somewhere in the paragraph, let's move it
4159                 inset = d->insetlist_.release(bibitem_pos);
4160                 eraseChar(bibitem_pos, track_changes);
4161         } else
4162                 // make a fresh one
4163                 inset = new InsetBibitem(const_cast<Buffer *>(&buffer),
4164                                          InsetCommandParams(BIBITEM_CODE));
4165
4166         Font font(inherit_font, buffer.params().language);
4167         insertInset(0, inset, font, Change(track_changes ? Change::INSERTED
4168                                                    : Change::UNCHANGED));
4169
4170         // This is needed to get the counters right
4171         buffer.updateBuffer();
4172         return 1;
4173 }
4174
4175
4176 void Paragraph::checkAuthors(AuthorList const & authorList)
4177 {
4178         d->changes_.checkAuthors(authorList);
4179 }
4180
4181
4182 bool Paragraph::isChanged(pos_type pos) const
4183 {
4184         return lookupChange(pos).changed();
4185 }
4186
4187
4188 bool Paragraph::isInserted(pos_type pos) const
4189 {
4190         return lookupChange(pos).inserted();
4191 }
4192
4193
4194 bool Paragraph::isDeleted(pos_type pos) const
4195 {
4196         return lookupChange(pos).deleted();
4197 }
4198
4199
4200 InsetList const & Paragraph::insetList() const
4201 {
4202         return d->insetlist_;
4203 }
4204
4205
4206 void Paragraph::setInsetBuffers(Buffer & b)
4207 {
4208         d->insetlist_.setBuffer(b);
4209 }
4210
4211
4212 void Paragraph::resetBuffer()
4213 {
4214         d->insetlist_.resetBuffer();
4215 }
4216
4217
4218 Inset * Paragraph::releaseInset(pos_type pos)
4219 {
4220         Inset * inset = d->insetlist_.release(pos);
4221         /// does not honour change tracking!
4222         eraseChar(pos, false);
4223         return inset;
4224 }
4225
4226
4227 Inset * Paragraph::getInset(pos_type pos)
4228 {
4229         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
4230                  ? d->insetlist_.get(pos) : nullptr;
4231 }
4232
4233
4234 Inset const * Paragraph::getInset(pos_type pos) const
4235 {
4236         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
4237                  ? d->insetlist_.get(pos) : nullptr;
4238 }
4239
4240
4241 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
4242                 pos_type & right, TextCase action)
4243 {
4244         // process sequences of modified characters; in change
4245         // tracking mode, this approach results in much better
4246         // usability than changing case on a char-by-char basis
4247         // We also need to track the current font, since font
4248         // changes within sequences can occur.
4249         vector<pair<char_type, Font> > changes;
4250
4251         bool const trackChanges = bparams.track_changes;
4252
4253         bool capitalize = true;
4254
4255         for (; pos < right; ++pos) {
4256                 char_type oldChar = d->text_[pos];
4257                 char_type newChar = oldChar;
4258
4259                 // ignore insets and don't play with deleted text!
4260                 if (oldChar != META_INSET && !isDeleted(pos)) {
4261                         switch (action) {
4262                                 case text_lowercase:
4263                                         newChar = lowercase(oldChar);
4264                                         break;
4265                                 case text_capitalization:
4266                                         if (capitalize) {
4267                                                 newChar = uppercase(oldChar);
4268                                                 capitalize = false;
4269                                         }
4270                                         break;
4271                                 case text_uppercase:
4272                                         newChar = uppercase(oldChar);
4273                                         break;
4274                         }
4275                 }
4276
4277                 if (isWordSeparator(pos) || isDeleted(pos)) {
4278                         // permit capitalization again
4279                         capitalize = true;
4280                 }
4281
4282                 if (oldChar != newChar) {
4283                         changes.push_back(make_pair(newChar, getFontSettings(bparams, pos)));
4284                         if (pos != right - 1)
4285                                 continue;
4286                         // step behind the changing area
4287                         pos++;
4288                 }
4289
4290                 int erasePos = pos - changes.size();
4291                 for (size_t i = 0; i < changes.size(); i++) {
4292                         insertChar(pos, changes[i].first,
4293                                    changes[i].second,
4294                                    trackChanges);
4295                         if (!eraseChar(erasePos, trackChanges)) {
4296                                 ++erasePos;
4297                                 ++pos; // advance
4298                                 ++right; // expand selection
4299                         }
4300                 }
4301                 changes.clear();
4302         }
4303 }
4304
4305
4306 int Paragraph::find(docstring const & str, bool cs, bool mw,
4307                 pos_type start_pos, bool del) const
4308 {
4309         pos_type pos = start_pos;
4310         int const strsize = str.length();
4311         int i = 0;
4312         pos_type const parsize = d->text_.size();
4313         for (i = 0; i < strsize && pos < parsize; ++i, ++pos) {
4314                 // Ignore "invisible" letters such as ligature breaks
4315                 // and hyphenation chars while searching
4316                 while (pos < parsize - 1 && isInset(pos)) {
4317                         Inset const * inset = getInset(pos);
4318                         if (!inset->isLetter())
4319                                 break;
4320                         odocstringstream os;
4321                         inset->toString(os);
4322                         if (!os.str().empty())
4323                                 break;
4324                         pos++;
4325                 }
4326                 if (cs && str[i] != d->text_[pos])
4327                         break;
4328                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos]))
4329                         break;
4330                 if (!del && isDeleted(pos))
4331                         break;
4332         }
4333
4334         if (i != strsize)
4335                 return 0;
4336
4337         // if necessary, check whether string matches word
4338         if (mw) {
4339                 if (start_pos > 0 && !isWordSeparator(start_pos - 1))
4340                         return 0;
4341                 if (pos < parsize
4342                         && !isWordSeparator(pos))
4343                         return 0;
4344         }
4345
4346         return pos - start_pos;
4347 }
4348
4349
4350 char_type Paragraph::getChar(pos_type pos) const
4351 {
4352         return d->text_[pos];
4353 }
4354
4355
4356 pos_type Paragraph::size() const
4357 {
4358         return d->text_.size();
4359 }
4360
4361
4362 bool Paragraph::empty() const
4363 {
4364         return d->text_.empty();
4365 }
4366
4367
4368 bool Paragraph::isInset(pos_type pos) const
4369 {
4370         return d->text_[pos] == META_INSET;
4371 }
4372
4373
4374 bool Paragraph::isSeparator(pos_type pos) const
4375 {
4376         //FIXME: Are we sure this can be the only separator?
4377         return d->text_[pos] == ' ';
4378 }
4379
4380
4381 void Paragraph::deregisterWords()
4382 {
4383         Private::LangWordsMap::const_iterator itl = d->words_.begin();
4384         Private::LangWordsMap::const_iterator ite = d->words_.end();
4385         for (; itl != ite; ++itl) {
4386                 WordList & wl = theWordList(itl->first);
4387                 Private::Words::const_iterator it = (itl->second).begin();
4388                 Private::Words::const_iterator et = (itl->second).end();
4389                 for (; it != et; ++it)
4390                         wl.remove(*it);
4391         }
4392         d->words_.clear();
4393 }
4394
4395
4396 void Paragraph::locateWord(pos_type & from, pos_type & to,
4397         word_location const loc, bool const ignore_deleted) const
4398 {
4399         switch (loc) {
4400         case WHOLE_WORD_STRICT:
4401                 if (from == 0 || from == size()
4402                     || isWordSeparator(from, ignore_deleted)
4403                     || isWordSeparator(from - 1, ignore_deleted)) {
4404                         to = from;
4405                         return;
4406                 }
4407                 // fall through
4408
4409         case WHOLE_WORD:
4410                 // If we are already at the beginning of a word, do nothing
4411                 if (!from || isWordSeparator(from - 1, ignore_deleted))
4412                         break;
4413                 // fall through
4414
4415         case PREVIOUS_WORD:
4416                 // always move the cursor to the beginning of previous word
4417                 while (from && !isWordSeparator(from - 1, ignore_deleted))
4418                         --from;
4419                 break;
4420         case NEXT_WORD:
4421                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
4422                 break;
4423         case PARTIAL_WORD:
4424                 // no need to move the 'from' cursor
4425                 break;
4426         }
4427         to = from;
4428         while (to < size() && !isWordSeparator(to, ignore_deleted))
4429                 ++to;
4430 }
4431
4432
4433 void Paragraph::collectWords()
4434 {
4435         for (pos_type pos = 0; pos < size(); ++pos) {
4436                 if (isWordSeparator(pos))
4437                         continue;
4438                 pos_type from = pos;
4439                 locateWord(from, pos, WHOLE_WORD);
4440                 // Work around MSVC warning: The statement
4441                 // if (pos < from + lyxrc.completion_minlength)
4442                 // triggers a signed vs. unsigned warning.
4443                 // I don't know why this happens, it could be a MSVC bug, or
4444                 // related to LLP64 (windows) vs. LP64 (unix) programming
4445                 // model, or the C++ standard might be ambigous in the section
4446                 // defining the "usual arithmetic conversions". However, using
4447                 // a temporary variable is safe and works on all compilers.
4448                 pos_type const endpos = from + lyxrc.completion_minlength;
4449                 if (pos < endpos)
4450                         continue;
4451                 FontList::const_iterator cit = d->fontlist_.fontIterator(from);
4452                 if (cit == d->fontlist_.end())
4453                         return;
4454                 Language const * lang = cit->font().language();
4455                 docstring const word = asString(from, pos, AS_STR_NONE);
4456                 d->words_[lang->lang()].insert(word);
4457         }
4458 }
4459
4460
4461 void Paragraph::registerWords()
4462 {
4463         Private::LangWordsMap::const_iterator itl = d->words_.begin();
4464         Private::LangWordsMap::const_iterator ite = d->words_.end();
4465         for (; itl != ite; ++itl) {
4466                 WordList & wl = theWordList(itl->first);
4467                 Private::Words::const_iterator it = (itl->second).begin();
4468                 Private::Words::const_iterator et = (itl->second).end();
4469                 for (; it != et; ++it)
4470                         wl.insert(*it);
4471         }
4472 }
4473
4474
4475 void Paragraph::updateWords()
4476 {
4477         deregisterWords();
4478         collectWords();
4479         registerWords();
4480 }
4481
4482
4483 void Paragraph::Private::appendSkipPosition(SkipPositions & skips, pos_type const pos) const
4484 {
4485         SkipPositionsIterator begin = skips.begin();
4486         SkipPositions::iterator end = skips.end();
4487         if (pos > 0 && begin < end) {
4488                 --end;
4489                 if (end->last == pos - 1) {
4490                         end->last = pos;
4491                         return;
4492                 }
4493         }
4494         skips.insert(end, FontSpan(pos, pos));
4495 }
4496
4497
4498 Language * Paragraph::Private::locateSpellRange(
4499         pos_type & from, pos_type & to,
4500         SkipPositions & skips) const
4501 {
4502         // skip leading white space
4503         while (from < to && owner_->isWordSeparator(from))
4504                 ++from;
4505         // don't check empty range
4506         if (from >= to)
4507                 return nullptr;
4508         // get current language
4509         Language * lang = getSpellLanguage(from);
4510         pos_type last = from;
4511         bool samelang = true;
4512         bool sameinset = true;
4513         while (last < to && samelang && sameinset) {
4514                 // hop to end of word
4515                 while (last < to && !owner_->isWordSeparator(last)) {
4516                         Inset const * inset = owner_->getInset(last);
4517                         if (inset && inset->lyxCode() == SPECIALCHAR_CODE) {
4518                                 // check for "invisible" letters such as ligature breaks
4519                                 odocstringstream os;
4520                                 inset->toString(os);
4521                                 if (os.str().length() != 0) {
4522                                         // avoid spell check of visible special char insets
4523                                         // stop the loop in front of the special char inset
4524                                         sameinset = false;
4525                                         break;
4526                                 }
4527                         } else if (inset) {
4528                                 appendSkipPosition(skips, last);
4529                         } else if (owner_->isDeleted(last)) {
4530                                 appendSkipPosition(skips, last);
4531                         }
4532                         ++last;
4533                 }
4534                 // hop to next word while checking for insets
4535                 while (sameinset && last < to && owner_->isWordSeparator(last)) {
4536                         if (Inset const * inset = owner_->getInset(last))
4537                                 sameinset = inset->isChar() && inset->isLetter();
4538                         if (sameinset && owner_->isDeleted(last)) {
4539                                 appendSkipPosition(skips, last);
4540                         }
4541                         if (sameinset)
4542                                 last++;
4543                 }
4544                 if (sameinset && last < to) {
4545                         // now check for language change
4546                         samelang = lang == getSpellLanguage(last);
4547                 }
4548         }
4549         // if language change detected backstep is needed
4550         if (!samelang)
4551                 --last;
4552         to = last;
4553         return lang;
4554 }
4555
4556
4557 Language * Paragraph::Private::getSpellLanguage(pos_type const from) const
4558 {
4559         Language * lang =
4560                 const_cast<Language *>(owner_->getFontSettings(
4561                         inset_owner_->buffer().params(), from).language());
4562         if (lang == inset_owner_->buffer().params().language
4563                 && !lyxrc.spellchecker_alt_lang.empty()) {
4564                 string lang_code;
4565                 string const lang_variety =
4566                         split(lyxrc.spellchecker_alt_lang, lang_code, '-');
4567                 lang->setCode(lang_code);
4568                 lang->setVariety(lang_variety);
4569         }
4570         return lang;
4571 }
4572
4573
4574 void Paragraph::requestSpellCheck(pos_type pos)
4575 {
4576         d->requestSpellCheck(pos);
4577 }
4578
4579
4580 bool Paragraph::needsSpellCheck() const
4581 {
4582         SpellChecker::ChangeNumber speller_change_number = 0;
4583         if (theSpellChecker())
4584                 speller_change_number = theSpellChecker()->changeNumber();
4585         if (speller_change_number > d->speller_state_.currentChangeNumber()) {
4586                 d->speller_state_.needsCompleteRefresh(speller_change_number);
4587         }
4588         return d->needsSpellCheck();
4589 }
4590
4591
4592 bool Paragraph::Private::ignoreWord(docstring const & word) const
4593 {
4594         // Ignore words with digits
4595         // FIXME: make this customizable
4596         // (note that some checkers ignore words with digits by default)
4597         docstring::const_iterator cit = word.begin();
4598         docstring::const_iterator const end = word.end();
4599         for (; cit != end; ++cit) {
4600                 if (isNumber((*cit)))
4601                         return true;
4602         }
4603         return false;
4604 }
4605
4606
4607 SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
4608         WordLangTuple & wl, docstring_list & suggestions,
4609         bool do_suggestion, bool check_learned) const
4610 {
4611         SpellChecker::Result result = SpellChecker::WORD_OK;
4612         SpellChecker * speller = theSpellChecker();
4613         if (!speller)
4614                 return result;
4615
4616         if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
4617                 return result;
4618
4619         locateWord(from, to, WHOLE_WORD, true);
4620         if (from == to || from >= size())
4621                 return result;
4622
4623         docstring word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
4624         Language * lang = d->getSpellLanguage(from);
4625
4626         if (getFontSettings(d->inset_owner_->buffer().params(), from).fontInfo().nospellcheck() == FONT_ON)
4627                 return result;
4628
4629         wl = WordLangTuple(word, lang);
4630
4631         if (word.empty())
4632                 return result;
4633
4634         if (needsSpellCheck() || check_learned) {
4635                 pos_type end = to;
4636                 if (!d->ignoreWord(word)) {
4637                         bool const trailing_dot = to < size() && d->text_[to] == '.';
4638                         result = speller->check(wl);
4639                         if (SpellChecker::misspelled(result) && trailing_dot) {
4640                                 wl = WordLangTuple(word.append(from_ascii(".")), lang);
4641                                 result = speller->check(wl);
4642                                 if (!SpellChecker::misspelled(result)) {
4643                                         LYXERR(Debug::GUI, "misspelled word is correct with dot: \"" <<
4644                                            word << "\" [" <<
4645                                            from << ".." << to << "]");
4646                                 } else {
4647                                         // spell check with dot appended failed too
4648                                         // restore original word/lang value
4649                                         word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
4650                                         wl = WordLangTuple(word, lang);
4651                                 }
4652                         }
4653                 }
4654                 if (!SpellChecker::misspelled(result)) {
4655                         // area up to the begin of the next word is not misspelled
4656                         while (end < size() && isWordSeparator(end))
4657                                 ++end;
4658                 }
4659                 d->setMisspelled(from, end, result);
4660         } else {
4661                 result = d->speller_state_.getState(from);
4662         }
4663
4664         if (do_suggestion)
4665                 suggestions.clear();
4666
4667         if (SpellChecker::misspelled(result)) {
4668                 LYXERR(Debug::GUI, "misspelled word: \"" <<
4669                            word << "\" [" <<
4670                            from << ".." << to << "]");
4671                 if (do_suggestion)
4672                         speller->suggest(wl, suggestions);
4673         }
4674         return result;
4675 }
4676
4677
4678 void Paragraph::anonymize()
4679 {
4680         // This is a very crude anonymization for now
4681         for (char_type & c : d->text_)
4682                 if (isLetterChar(c) || isNumber(c))
4683                         c = 'a';
4684 }
4685
4686
4687 void Paragraph::Private::markMisspelledWords(
4688         pos_type const & first, pos_type const & last,
4689         SpellChecker::Result result,
4690         docstring const & word,
4691         SkipPositions const & skips)
4692 {
4693         if (!SpellChecker::misspelled(result)) {
4694                 setMisspelled(first, last, SpellChecker::WORD_OK);
4695                 return;
4696         }
4697         int snext = first;
4698         SpellChecker * speller = theSpellChecker();
4699         // locate and enumerate the error positions
4700         int nerrors = speller->numMisspelledWords();
4701         int numskipped = 0;
4702         SkipPositionsIterator it = skips.begin();
4703         SkipPositionsIterator et = skips.end();
4704         for (int index = 0; index < nerrors; ++index) {
4705                 int wstart;
4706                 int wlen = 0;
4707                 speller->misspelledWord(index, wstart, wlen);
4708                 /// should not happen if speller supports range checks
4709                 if (!wlen) continue;
4710                 docstring const misspelled = word.substr(wstart, wlen);
4711                 wstart += first + numskipped;
4712                 if (snext < wstart) {
4713                         /// mark the range of correct spelling
4714                         numskipped += countSkips(it, et, wstart);
4715                         setMisspelled(snext,
4716                                 wstart - 1, SpellChecker::WORD_OK);
4717                 }
4718                 snext = wstart + wlen;
4719                 numskipped += countSkips(it, et, snext);
4720                 /// mark the range of misspelling
4721                 setMisspelled(wstart, snext, result);
4722                 LYXERR(Debug::GUI, "misspelled word: \"" <<
4723                            misspelled << "\" [" <<
4724                            wstart << ".." << (snext-1) << "]");
4725                 ++snext;
4726         }
4727         if (snext <= last) {
4728                 /// mark the range of correct spelling at end
4729                 setMisspelled(snext, last, SpellChecker::WORD_OK);
4730         }
4731 }
4732
4733
4734 void Paragraph::spellCheck() const
4735 {
4736         SpellChecker * speller = theSpellChecker();
4737         if (!speller || empty() ||!needsSpellCheck())
4738                 return;
4739         pos_type start;
4740         pos_type endpos;
4741         d->rangeOfSpellCheck(start, endpos);
4742         if (speller->canCheckParagraph()) {
4743                 // loop until we leave the range
4744                 for (pos_type first = start; first < endpos; ) {
4745                         pos_type last = endpos;
4746                         Private::SkipPositions skips;
4747                         Language * lang = d->locateSpellRange(first, last, skips);
4748                         if (first >= endpos)
4749                                 break;
4750                         // start the spell checker on the unit of meaning
4751                         docstring word = asString(first, last, AS_STR_INSETS + AS_STR_SKIPDELETE);
4752                         WordLangTuple wl = WordLangTuple(word, lang);
4753                         SpellChecker::Result result = word.size() ?
4754                                 speller->check(wl) : SpellChecker::WORD_OK;
4755                         d->markMisspelledWords(first, last, result, word, skips);
4756                         first = ++last;
4757                 }
4758         } else {
4759                 static docstring_list suggestions;
4760                 pos_type to = endpos;
4761                 while (start < endpos) {
4762                         WordLangTuple wl;
4763                         spellCheck(start, to, wl, suggestions, false);
4764                         start = to + 1;
4765                 }
4766         }
4767         d->readySpellCheck();
4768 }
4769
4770
4771 bool Paragraph::isMisspelled(pos_type pos, bool check_boundary) const
4772 {
4773         bool result = SpellChecker::misspelled(d->speller_state_.getState(pos));
4774         if (result || pos <= 0 || pos > size())
4775                 return result;
4776         if (check_boundary && (pos == size() || isWordSeparator(pos)))
4777                 result = SpellChecker::misspelled(d->speller_state_.getState(pos - 1));
4778         return result;
4779 }
4780
4781
4782 string Paragraph::magicLabel() const
4783 {
4784         stringstream ss;
4785         ss << "magicparlabel-" << id();
4786         return ss.str();
4787 }
4788
4789
4790 } // namespace lyx