src/Paragraph.cpp

   1 /**
   2  * \file Paragraph.cpp
   3  * This file is part of LyX, the document processor.
   4  * Licence details can be found in the file COPYING.
   5  *
   6  * \author Asger Alstrup
   7  * \author Lars Gullik Bjønnes
   8  * \author Richard Heck (XHTML output)
   9  * \author Jean-Marc Lasgouttes
  10  * \author Angus Leeming
  11  * \author John Levon
  12  * \author André Pönitz
  13  * \author Dekel Tsur
  14  * \author Jürgen Vigna
  15  *
  16  * Full author contact details are available in file CREDITS.
  17  */
  18
  19 #include <config.h>
  20
  21 #include "Paragraph.h"
  22
  23 #include "LayoutFile.h"
  24 #include "Buffer.h"
  25 #include "BufferParams.h"
  26 #include "Changes.h"
  27 #include "Counters.h"
  28 #include "BufferEncodings.h"
  29 #include "InsetList.h"
  30 #include "Language.h"
  31 #include "LaTeXFeatures.h"
  32 #include "Layout.h"
  33 #include "Length.h"
  34 #include "Font.h"
  35 #include "FontList.h"
  36 #include "LyXRC.h"
  37 #include "OutputParams.h"
  38 #include "output_latex.h"
  39 #include "output_xhtml.h"
  40 #include "output_docbook.h"
  41 #include "ParagraphParameters.h"
  42 #include "SpellChecker.h"
  43 #include "xml.h"
  44 #include "texstream.h"
  45 #include "TextClass.h"
  46 #include "TexRow.h"
  47 #include "Text.h"
  48 #include "WordLangTuple.h"
  49 #include "WordList.h"
  50
  51 #include "frontends/alert.h"
  52
  53 #include "insets/InsetBibitem.h"
  54 #include "insets/InsetLabel.h"
  55 #include "insets/InsetSpecialChar.h"
  56 #include "insets/InsetText.h"
  57
  58 #include "mathed/InsetMathHull.h"
  59
  60 #include "support/debug.h"
  61 #include "support/docstring_list.h"
  62 #include "support/ExceptionMessage.h"
  63 #include "support/gettext.h"
  64 #include "support/lassert.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 namespace {
  88
  89 /// Inset identifier (above 0x10ffff, for ucs-4)
  90 char_type const META_INSET = 0x200001;
  91
  92 } // namespace
  93
  94
  95 /////////////////////////////////////////////////////////////////////
  96 //
  97 // SpellResultRange
  98 //
  99 /////////////////////////////////////////////////////////////////////
 100
 101 class SpellResultRange {
 102 public:
 103         SpellResultRange(FontSpan range, SpellChecker::Result result)
 104         : range_(range), result_(result)
 105         {}
 106         ///
 107         FontSpan const & range() const { return range_; }
 108         ///
 109         void range(FontSpan const & r) { range_ = r; }
 110         ///
 111         SpellChecker::Result result() const { return result_; }
 112         ///
 113         void result(SpellChecker::Result r) { result_ = r; }
 114         ///
 115         bool contains(pos_type pos) const { return range_.contains(pos); }
 116         ///
 117         bool covered(FontSpan const & r) const
 118         {
 119                 // 1. first of new range inside current range or
 120                 // 2. last of new range inside current range or
 121                 // 3. first of current range inside new range or
 122                 // 4. last of current range inside new range
 123                 //FIXME: is this the same as !range_.intersect(r).empty() ?
 124                 return range_.contains(r.first) || range_.contains(r.last) ||
 125                         r.contains(range_.first) || r.contains(range_.last);
 126         }
 127         ///
 128         void shift(pos_type pos, int offset)
 129         {
 130                 if (range_.first > pos) {
 131                         range_.first += offset;
 132                         range_.last += offset;
 133                 } else if (range_.last >= pos) {
 134                         range_.last += offset;
 135                 }
 136         }
 137 private:
 138         FontSpan range_ ;
 139         SpellChecker::Result result_ ;
 140 };
 141
 142
 143 /////////////////////////////////////////////////////////////////////
 144 //
 145 // SpellCheckerState
 146 //
 147 /////////////////////////////////////////////////////////////////////
 148
 149 class SpellCheckerState {
 150 public:
 151         SpellCheckerState()
 152         {
 153                 needs_refresh_ = true;
 154                 current_change_number_ = 0;
 155         }
 156
 157         void setRange(FontSpan const & fp, SpellChecker::Result state)
 158         {
 159                 Ranges result;
 160                 RangesIterator et = ranges_.end();
 161                 RangesIterator it = ranges_.begin();
 162                 for (; it != et; ++it) {
 163                         if (!it->covered(fp))
 164                                 result.push_back(SpellResultRange(it->range(), it->result()));
 165                         else if (state == SpellChecker::WORD_OK) {
 166                                 // trim or split the current misspelled range
 167                                 // store misspelled ranges only
 168                                 FontSpan range = it->range();
 169                                 if (fp.first > range.first) {
 170                                         // misspelled area in front of WORD_OK
 171                                         range.last = fp.first - 1;
 172                                         result.push_back(SpellResultRange(range, it->result()));
 173                                         range = it->range();
 174                                 }
 175                                 if (fp.last < range.last) {
 176                                         // misspelled area after WORD_OK range
 177                                         range.first = fp.last + 1;
 178                                         result.push_back(SpellResultRange(range, it->result()));
 179                                 }
 180                         }
 181                 }
 182                 ranges_ = result;
 183                 if (state != SpellChecker::WORD_OK)
 184                         ranges_.push_back(SpellResultRange(fp, state));
 185         }
 186
 187         void increasePosAfterPos(pos_type pos)
 188         {
 189                 correctRangesAfterPos(pos, 1);
 190                 needsRefresh(pos);
 191         }
 192
 193         void decreasePosAfterPos(pos_type pos)
 194         {
 195                 correctRangesAfterPos(pos, -1);
 196                 needsRefresh(pos);
 197         }
 198
 199         void refreshLast(pos_type pos)
 200         {
 201                 if (pos < refresh_.last)
 202                         refresh_.last = pos;
 203         }
 204
 205         SpellChecker::Result getState(pos_type pos) const
 206         {
 207                 SpellChecker::Result result = SpellChecker::WORD_OK;
 208                 RangesIterator et = ranges_.end();
 209                 RangesIterator it = ranges_.begin();
 210                 for (; it != et; ++it) {
 211                         if(it->contains(pos)) {
 212                                 return it->result();
 213                         }
 214                 }
 215                 return result;
 216         }
 217
 218         FontSpan const & getRange(pos_type pos) const
 219         {
 220                 /// empty span to indicate mismatch
 221                 static FontSpan empty_;
 222                 RangesIterator et = ranges_.end();
 223                 RangesIterator it = ranges_.begin();
 224                 for (; it != et; ++it) {
 225                         if(it->contains(pos)) {
 226                                 return it->range();
 227                         }
 228                 }
 229                 return empty_;
 230         }
 231
 232         bool needsRefresh() const
 233         {
 234                 return needs_refresh_;
 235         }
 236
 237         SpellChecker::ChangeNumber currentChangeNumber() const
 238         {
 239                 return current_change_number_;
 240         }
 241
 242         void refreshRange(pos_type & first, pos_type & last) const
 243         {
 244                 first = refresh_.first;
 245                 last = refresh_.last;
 246         }
 247
 248         void needsRefresh(pos_type pos)
 249         {
 250                 if (needs_refresh_ && pos != -1) {
 251                         if (pos < refresh_.first)
 252                                 refresh_.first = pos;
 253                         if (pos > refresh_.last)
 254                                 refresh_.last = pos;
 255                 } else if (pos != -1) {
 256                         // init request check for neighbour positions too
 257                         refresh_.first = pos > 0 ? pos - 1 : 0;
 258                         // no need for special end of paragraph check
 259                         refresh_.last = pos + 1;
 260                 }
 261                 needs_refresh_ = pos != -1;
 262         }
 263
 264         void needsCompleteRefresh(SpellChecker::ChangeNumber change_number)
 265         {
 266                 needs_refresh_ = true;
 267                 refresh_.first = 0;
 268                 refresh_.last = -1;
 269                 current_change_number_ = change_number;
 270         }
 271 private:
 272         typedef vector<SpellResultRange> Ranges;
 273         typedef Ranges::const_iterator RangesIterator;
 274         Ranges ranges_;
 275         /// the area of the paragraph with pending spell check
 276         FontSpan refresh_;
 277         bool needs_refresh_;
 278         /// spell state cache version number
 279         SpellChecker::ChangeNumber current_change_number_;
 280
 281
 282         void correctRangesAfterPos(pos_type pos, int offset)
 283         {
 284                 RangesIterator et = ranges_.end();
 285                 Ranges::iterator it = ranges_.begin();
 286                 for (; it != et; ++it) {
 287                         it->shift(pos, offset);
 288                 }
 289         }
 290
 291 };
 292
 293 /////////////////////////////////////////////////////////////////////
 294 //
 295 // Paragraph::Private
 296 //
 297 /////////////////////////////////////////////////////////////////////
 298
 299 class Paragraph::Private
 300 {
 301         // Enforce our own "copy" constructor
 302         Private(Private const &) = delete;
 303         Private & operator=(Private const &) = delete;
 304         // Unique ID generator
 305         static int make_id();
 306 public:
 307         ///
 308         Private(Paragraph * owner, Layout const & layout);
 309         /// "Copy constructor"
 310         Private(Private const &, Paragraph * owner);
 311         /// Copy constructor from \p beg  to \p end
 312         Private(Private const &, Paragraph * owner, pos_type beg, pos_type end);
 313
 314         ///
 315         void insertChar(pos_type pos, char_type c, Change const & change);
 316
 317         /// Output the surrogate pair formed by \p c and \p next to \p os.
 318         /// \return the number of characters written.
 319         int latexSurrogatePair(BufferParams const &, otexstream & os,
 320                                char_type c, char_type next,
 321                                OutputParams const &);
 322
 323         /// Output a space in appropriate formatting (or a surrogate pair
 324         /// if the next character is a combining character).
 325         /// \return whether a surrogate pair was output.
 326         bool simpleTeXBlanks(BufferParams const &,
 327                              OutputParams const &,
 328                              otexstream &,
 329                              pos_type i,
 330                              unsigned int & column,
 331                              Font const & font,
 332                              Layout const & style);
 333
 334         /// This could go to ParagraphParameters if we want to.
 335         int startTeXParParams(BufferParams const &, otexstream &,
 336                               OutputParams const &) const;
 337
 338         /// This could go to ParagraphParameters if we want to.
 339         bool endTeXParParams(BufferParams const &, otexstream &,
 340                              OutputParams const &) const;
 341
 342         ///
 343         void latexInset(BufferParams const &,
 344                                    otexstream &,
 345                                    OutputParams &,
 346                                    Font & running_font,
 347                                    Font & basefont,
 348                                    Font const & outerfont,
 349                                    bool & open_font,
 350                                    Change & running_change,
 351                                    Layout const & style,
 352                                    pos_type & i,
 353                                    unsigned int & column);
 354
 355         ///
 356         void latexSpecialChar(
 357                                    otexstream & os,
 358                                    BufferParams const & bparams,
 359                                    OutputParams const & runparams,
 360                                    Font const & running_font,
 361                                    string & alien_script,
 362                                    Layout const & style,
 363                                    pos_type & i,
 364                                    pos_type end_pos,
 365                                    unsigned int & column);
 366
 367         ///
 368         bool latexSpecialT1(
 369                 char_type const c,
 370                 otexstream & os,
 371                 pos_type i,
 372                 unsigned int & column);
 373         ///
 374         bool latexSpecialTU(
 375                 char_type const c,
 376                 otexstream & os,
 377                 pos_type i,
 378                 unsigned int & column);
 379         ///
 380         bool latexSpecialT3(
 381                 char_type const c,
 382                 otexstream & os,
 383                 pos_type i,
 384                 unsigned int & column);
 385
 386         ///
 387         void validate(LaTeXFeatures & features) const;
 388
 389         /// Checks if the paragraph contains only text and no inset or font change.
 390         bool onlyText(Buffer const & buf, Font const & outerfont,
 391                       pos_type initial) const;
 392
 393         /// a vector of speller skip positions
 394         typedef vector<FontSpan> SkipPositions;
 395         typedef SkipPositions::const_iterator SkipPositionsIterator;
 396
 397         void appendSkipPosition(SkipPositions & skips, pos_type const pos) const;
 398
 399         Language * getSpellLanguage(pos_type const from) const;
 400
 401         Language * locateSpellRange(pos_type & from, pos_type & to,
 402                                     SkipPositions & skips) const;
 403
 404         bool hasSpellerChange() const
 405         {
 406                 SpellChecker::ChangeNumber speller_change_number = 0;
 407                 if (theSpellChecker())
 408                         speller_change_number = theSpellChecker()->changeNumber();
 409                 return speller_change_number > speller_state_.currentChangeNumber();
 410         }
 411
 412         bool ignoreWord(docstring const & word) const ;
 413
 414         void setMisspelled(pos_type from, pos_type to, SpellChecker::Result state)
 415         {
 416                 pos_type textsize = owner_->size();
 417                 // check for sane arguments
 418                 if (to <= from || from >= textsize)
 419                         return;
 420                 FontSpan fp = FontSpan(from, to - 1);
 421                 speller_state_.setRange(fp, state);
 422         }
 423
 424         void requestSpellCheck(pos_type pos)
 425         {
 426                 if (pos == -1)
 427                         speller_state_.needsCompleteRefresh(speller_state_.currentChangeNumber());
 428                 else
 429                         speller_state_.needsRefresh(pos);
 430         }
 431
 432         void readySpellCheck()
 433         {
 434                 speller_state_.needsRefresh(-1);
 435         }
 436
 437         bool needsSpellCheck() const
 438         {
 439                 return speller_state_.needsRefresh();
 440         }
 441
 442         void rangeOfSpellCheck(pos_type & first, pos_type & last) const
 443         {
 444                 speller_state_.refreshRange(first, last);
 445                 if (last == -1) {
 446                         last = owner_->size();
 447                         return;
 448                 }
 449                 pos_type endpos = last;
 450                 owner_->locateWord(first, endpos, WHOLE_WORD, true);
 451                 if (endpos < last) {
 452                         endpos = last;
 453                         owner_->locateWord(last, endpos, WHOLE_WORD, true);
 454                 }
 455                 last = endpos;
 456         }
 457
 458         int countSkips(SkipPositionsIterator & it, SkipPositionsIterator const et,
 459                             int & start) const
 460         {
 461                 int numskips = 0;
 462                 while (it != et && it->first < start) {
 463                         int skip = it->last - it->first + 1;
 464                         start += skip;
 465                         numskips += skip;
 466                         ++it;
 467                 }
 468                 return numskips;
 469         }
 470
 471         void markMisspelledWords(pos_type const & first, pos_type const & last,
 472                                                          SpellChecker::Result result,
 473                                                          docstring const & word,
 474                                                          SkipPositions const & skips);
 475
 476         InsetCode ownerCode() const
 477         {
 478                 return inset_owner_ ? inset_owner_->lyxCode() : NO_CODE;
 479         }
 480
 481         /// Which Paragraph owns us?
 482         Paragraph * owner_;
 483
 484         /// In which Inset?
 485         Inset const * inset_owner_;
 486
 487         ///
 488         FontList fontlist_;
 489
 490         ///
 491         int id_;
 492
 493         ///
 494         ParagraphParameters params_;
 495
 496         /// for recording and looking up changes
 497         Changes changes_;
 498
 499         ///
 500         InsetList insetlist_;
 501
 502         /// end of label
 503         pos_type begin_of_body_;
 504
 505         typedef docstring TextContainer;
 506         ///
 507         TextContainer text_;
 508
 509         typedef set<docstring> Words;
 510         typedef map<string, Words> LangWordsMap;
 511         ///
 512         LangWordsMap words_;
 513         ///
 514         Layout const * layout_;
 515         ///
 516         SpellCheckerState speller_state_;
 517 };
 518
 519
 520 Paragraph::Private::Private(Paragraph * owner, Layout const & layout)
 521         : owner_(owner), inset_owner_(nullptr), id_(-1), begin_of_body_(0), layout_(&layout)
 522 {
 523         text_.reserve(100);
 524 }
 525
 526
 527 //static
 528 int Paragraph::Private::make_id()
 529 {
 530         // The id is unique per session across buffers because it is used in
 531         // LFUN_PARAGRAPH_GOTO to switch to a different buffer, for instance in the
 532         // outliner.
 533         // (thread-safe)
 534         static atomic_uint next_id(0);
 535         return next_id++;
 536 }
 537
 538
 539 Paragraph::Private::Private(Private const & p, Paragraph * owner)
 540         : owner_(owner), inset_owner_(p.inset_owner_), fontlist_(p.fontlist_),
 541           id_(make_id()),
 542           params_(p.params_), changes_(p.changes_), insetlist_(p.insetlist_),
 543           begin_of_body_(p.begin_of_body_), text_(p.text_), words_(p.words_),
 544           layout_(p.layout_)
 545 {
 546         requestSpellCheck(p.text_.size());
 547 }
 548
 549
 550 Paragraph::Private::Private(Private const & p, Paragraph * owner,
 551         pos_type beg, pos_type end)
 552         : owner_(owner), inset_owner_(p.inset_owner_), id_(make_id()),
 553           params_(p.params_), changes_(p.changes_),
 554           insetlist_(p.insetlist_, beg, end),
 555           begin_of_body_(p.begin_of_body_), words_(p.words_),
 556           layout_(p.layout_)
 557 {
 558         if (beg >= pos_type(p.text_.size()))
 559                 return;
 560         text_ = p.text_.substr(beg, end - beg);
 561
 562         FontList::const_iterator fcit = fontlist_.begin();
 563         FontList::const_iterator fend = fontlist_.end();
 564         for (; fcit != fend; ++fcit) {
 565                 if (fcit->pos() < beg)
 566                         continue;
 567                 if (fcit->pos() >= end) {
 568                         // Add last entry in the fontlist_.
 569                         fontlist_.set(text_.size() - 1, fcit->font());
 570                         break;
 571                 }
 572                 // Add a new entry in the fontlist_.
 573                 fontlist_.set(fcit->pos() - beg, fcit->font());
 574         }
 575         requestSpellCheck(p.text_.size());
 576 }
 577
 578
 579 void Paragraph::addChangesToToc(DocIterator const & cdit, Buffer const & buf,
 580                                 bool output_active, TocBackend & backend) const
 581 {
 582         d->changes_.addToToc(cdit, buf, output_active, backend);
 583 }
 584
 585
 586 bool Paragraph::isDeleted(pos_type start, pos_type end) const
 587 {
 588         LASSERT(start >= 0 && start <= size(), return false);
 589         LASSERT(end > start && end <= size() + 1, return false);
 590
 591         return d->changes_.isDeleted(start, end);
 592 }
 593
 594
 595 bool Paragraph::isChanged(pos_type start, pos_type end) const
 596 {
 597         LASSERT(start >= 0 && start <= size(), return false);
 598         LASSERT(end > start && end <= size() + 1, return false);
 599
 600         return d->changes_.isChanged(start, end);
 601 }
 602
 603 // FIXME: Ideally the diverse isChanged() methods should account for that!
 604 bool Paragraph::hasChangedInsets(pos_type start, pos_type end) const
 605 {
 606         LASSERT(start >= 0 && start <= size(), return false);
 607         LASSERT(end > start && end <= size() + 1, return false);
 608
 609         for (auto const & icit : d->insetlist_) {
 610                 if (icit.pos < start)
 611                         continue;
 612                 if (icit.pos >= end)
 613                         break;
 614                 if (icit.inset && icit.inset->isChanged())
 615                         return true;
 616         }
 617         return false;
 618 }
 619
 620 bool Paragraph::isChanged() const
 621 {
 622         return d->changes_.isChanged();
 623 }
 624
 625
 626 bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const
 627 {
 628         // keep the logic here in sync with the logic of eraseChars()
 629         if (!trackChanges)
 630                 return true;
 631
 632         Change const & change = d->changes_.lookup(size());
 633         return change.inserted() && change.currentAuthor();
 634 }
 635
 636 Change Paragraph::parEndChange() const
 637 {
 638         return d->changes_.lookup(size());
 639 }
 640
 641
 642 void Paragraph::setChange(Change const & change)
 643 {
 644         // beware of the imaginary end-of-par character!
 645         d->changes_.set(change, 0, size() + 1);
 646
 647         /*
 648          * Propagate the change recursively - but not in case of DELETED!
 649          *
 650          * Imagine that your co-author makes changes in an existing inset. He
 651          * sends your document to you and you come to the conclusion that the
 652          * inset should go completely. If you erase it, LyX must not delete all
 653          * text within the inset. Otherwise, the change tracked insertions of
 654          * your co-author get lost and there is no way to restore them later.
 655          *
 656          * Conclusion: An inset's content should remain untouched if you delete it
 657          */
 658
 659         if (!change.deleted()) {
 660                 for (pos_type pos = 0; pos < size(); ++pos) {
 661                         if (Inset * inset = getInset(pos))
 662                                 inset->setChange(change);
 663                 }
 664         }
 665 }
 666
 667
 668 void Paragraph::setChange(pos_type pos, Change const & change)
 669 {
 670         LASSERT(pos >= 0 && pos <= size(), return);
 671         d->changes_.set(change, pos);
 672
 673         // see comment in setChange(Change const &) above
 674         if (!change.deleted() && pos < size())
 675                 if (Inset * inset = getInset(pos))
 676                         inset->setChange(change);
 677 }
 678
 679
 680 Change const & Paragraph::lookupChange(pos_type pos) const
 681 {
 682         LBUFERR(pos >= 0 && pos <= size());
 683         return d->changes_.lookup(pos);
 684 }
 685
 686
 687 void Paragraph::acceptChanges(pos_type start, pos_type end)
 688 {
 689         LASSERT(start >= 0 && start <= size(), return);
 690         LASSERT(end > start && end <= size() + 1, return);
 691
 692         for (pos_type pos = start; pos < end; ++pos) {
 693                 switch (lookupChange(pos).type) {
 694                         case Change::UNCHANGED:
 695                                 // accept changes in nested inset
 696                                 if (Inset * inset = getInset(pos))
 697                                         inset->acceptChanges();
 698                                 break;
 699
 700                         case Change::INSERTED:
 701                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 702                                 // also accept changes in nested inset
 703                                 if (Inset * inset = getInset(pos))
 704                                         inset->acceptChanges();
 705                                 break;
 706
 707                         case Change::DELETED:
 708                                 // Suppress access to non-existent
 709                                 // "end-of-paragraph char"
 710                                 if (pos < size()) {
 711                                         eraseChar(pos, false);
 712                                         --end;
 713                                         --pos;
 714                                 }
 715                                 break;
 716                 }
 717
 718         }
 719 }
 720
 721
 722 void Paragraph::rejectChanges(pos_type start, pos_type end)
 723 {
 724         LASSERT(start >= 0 && start <= size(), return);
 725         LASSERT(end > start && end <= size() + 1, return);
 726
 727         for (pos_type pos = start; pos < end; ++pos) {
 728                 switch (lookupChange(pos).type) {
 729                         case Change::UNCHANGED:
 730                                 // reject changes in nested inset
 731                                 if (Inset * inset = getInset(pos))
 732                                                 inset->rejectChanges();
 733                                 break;
 734
 735                         case Change::INSERTED:
 736                                 // Suppress access to non-existent
 737                                 // "end-of-paragraph char"
 738                                 if (pos < size()) {
 739                                         eraseChar(pos, false);
 740                                         --end;
 741                                         --pos;
 742                                 }
 743                                 break;
 744
 745                         case Change::DELETED:
 746                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 747
 748                                 // Do NOT reject changes within a deleted inset!
 749                                 // There may be insertions of a co-author inside of it!
 750
 751                                 break;
 752                 }
 753         }
 754 }
 755
 756
 757 void Paragraph::Private::insertChar(pos_type pos, char_type c,
 758                 Change const & change)
 759 {
 760         LASSERT(pos >= 0 && pos <= int(text_.size()), return);
 761
 762         // track change
 763         changes_.insert(change, pos);
 764
 765         // This is actually very common when parsing buffers (and
 766         // maybe inserting ascii text)
 767         if (pos == pos_type(text_.size())) {
 768                 // when appending characters, no need to update tables
 769                 text_.push_back(c);
 770                 // but we want spell checking
 771                 requestSpellCheck(pos);
 772                 return;
 773         }
 774
 775         text_.insert(text_.begin() + pos, c);
 776
 777         // Update the font table.
 778         fontlist_.increasePosAfterPos(pos);
 779
 780         // Update the insets
 781         insetlist_.increasePosAfterPos(pos);
 782
 783         // Update list of misspelled positions
 784         speller_state_.increasePosAfterPos(pos);
 785 }
 786
 787
 788 bool Paragraph::insertInset(pos_type pos, Inset * inset,
 789                                    Font const & font, Change const & change)
 790 {
 791         LASSERT(inset, return false);
 792         LASSERT(pos >= 0 && pos <= size(), return false);
 793
 794         // Paragraph::insertInset() can be used in cut/copy/paste operation where
 795         // d->inset_owner_ is not set yet.
 796         if (d->inset_owner_ && !d->inset_owner_->insetAllowed(inset->lyxCode()))
 797                 return false;
 798
 799         d->insertChar(pos, META_INSET, change);
 800         LASSERT(d->text_[pos] == META_INSET, return false);
 801
 802         // Add a new entry in the insetlist_.
 803         d->insetlist_.insert(inset, pos);
 804
 805         // Some insets require run of spell checker
 806         requestSpellCheck(pos);
 807         setFont(pos, font);
 808         return true;
 809 }
 810
 811
 812 bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
 813 {
 814         LASSERT(pos >= 0 && pos <= size(), return false);
 815
 816         // keep the logic here in sync with the logic of isMergedOnEndOfParDeletion()
 817
 818         if (trackChanges) {
 819                 Change change = d->changes_.lookup(pos);
 820
 821                 // set the character to DELETED if
 822                 //  a) it was previously unchanged or
 823                 //  b) it was inserted by a co-author
 824
 825                 if (!change.changed() ||
 826                       (change.inserted() && !change.currentAuthor())) {
 827                         setChange(pos, Change(Change::DELETED));
 828                         // request run of spell checker
 829                         requestSpellCheck(pos);
 830                         return false;
 831                 }
 832
 833                 if (change.deleted())
 834                         return false;
 835         }
 836
 837         // Don't physically access the imaginary end-of-paragraph character.
 838         // eraseChar() can only mark it as DELETED. A physical deletion of
 839         // end-of-par must be handled externally.
 840         if (pos == size()) {
 841                 return false;
 842         }
 843
 844         // track change
 845         d->changes_.erase(pos);
 846
 847         // if it is an inset, delete the inset entry
 848         if (d->text_[pos] == META_INSET)
 849                 d->insetlist_.erase(pos);
 850
 851         d->text_.erase(d->text_.begin() + pos);
 852
 853         // Update the fontlist_
 854         d->fontlist_.erase(pos);
 855
 856         // Update the insetlist_
 857         d->insetlist_.decreasePosAfterPos(pos);
 858
 859         // Update list of misspelled positions
 860         d->speller_state_.decreasePosAfterPos(pos);
 861         d->speller_state_.refreshLast(size());
 862
 863         return true;
 864 }
 865
 866
 867 int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
 868 {
 869         LASSERT(start >= 0 && start <= size(), return 0);
 870         LASSERT(end >= start && end <= size() + 1, return 0);
 871
 872         pos_type i = start;
 873         for (pos_type count = end - start; count; --count) {
 874                 if (!eraseChar(i, trackChanges))
 875                         ++i;
 876         }
 877         return end - i;
 878 }
 879
 880 // Handle combining characters
 881 int Paragraph::Private::latexSurrogatePair(BufferParams const & bparams,
 882                 otexstream & os, char_type c, char_type next,
 883                 OutputParams const & runparams)
 884 {
 885         // Writing next here may circumvent a possible font change between
 886         // c and next. Since next is only output if it forms a surrogate pair
 887         // with c we can ignore this:
 888         // A font change inside a surrogate pair does not make sense and is
 889         // hopefully impossible to input.
 890         // FIXME: change tracking
 891         // Is this correct WRT change tracking?
 892         Encoding const & encoding = *(runparams.encoding);
 893         docstring latex1 = encoding.latexChar(next).first;
 894         if (runparams.inIPA) {
 895                 string const tipashortcut = Encodings::TIPAShortcut(next);
 896                 if (!tipashortcut.empty()) {
 897                         latex1 = from_ascii(tipashortcut);
 898                 }
 899         }
 900         docstring latex2 = encoding.latexChar(c).first;
 901
 902         if (bparams.useNonTeXFonts || docstring(1, next) == latex1) {
 903                 // Encoding supports the combination:
 904                 // output as is (combining char after base char).
 905                 os << latex2 << latex1;
 906                 return latex1.length() + latex2.length();
 907         }
 908
 909         os << latex1 << "{" << latex2 << "}";
 910         return latex1.length() + latex2.length() + 2;
 911 }
 912
 913
 914 bool Paragraph::Private::simpleTeXBlanks(BufferParams const & bparams,
 915                                        OutputParams const & runparams,
 916                                        otexstream & os,
 917                                        pos_type i,
 918                                        unsigned int & column,
 919                                        Font const & font,
 920                                        Layout const & style)
 921 {
 922         if (style.pass_thru || runparams.pass_thru)
 923                 return false;
 924
 925         if (i + 1 < int(text_.size())) {
 926                 char_type next = text_[i + 1];
 927                 if (Encodings::isCombiningChar(next)) {
 928                         // This space has an accent, so we must always output it.
 929                         column += latexSurrogatePair(bparams, os, ' ', next, runparams) - 1;
 930                         return true;
 931                 }
 932         }
 933
 934         if (runparams.linelen > 0
 935             && column > runparams.linelen
 936             && i
 937             && text_[i - 1] != ' '
 938             && (i + 1 < int(text_.size()))
 939             // same in FreeSpacing mode
 940             && !owner_->isFreeSpacing()
 941             // In typewriter mode, we want to avoid
 942             // ! . ? : at the end of a line
 943             && !(font.fontInfo().family() == TYPEWRITER_FAMILY
 944                  && (text_[i - 1] == '.'
 945                      || text_[i - 1] == '?'
 946                      || text_[i - 1] == ':'
 947                      || text_[i - 1] == '!'))) {
 948                 os << '\n';
 949                 os.texrow().start(owner_->id(), i + 1);
 950                 column = 0;
 951         } else if (style.free_spacing) {
 952                 os << '~';
 953         } else {
 954                 os << ' ';
 955         }
 956         return false;
 957 }
 958
 959
 960 void Paragraph::Private::latexInset(BufferParams const & bparams,
 961                                     otexstream & os,
 962                                     OutputParams & runparams,
 963                                     Font & running_font,
 964                                     Font & basefont,
 965                                     Font const & outerfont,
 966                                     bool & open_font,
 967                                     Change & running_change,
 968                                     Layout const & style,
 969                                     pos_type & i,
 970                                     unsigned int & column)
 971 {
 972         Inset * inset = owner_->getInset(i);
 973         LBUFERR(inset);
 974
 975         if (style.pass_thru) {
 976                 odocstringstream ods;
 977                 inset->plaintext(ods, runparams);
 978                 os << ods.str();
 979                 return;
 980         }
 981
 982         // FIXME: move this to InsetNewline::latex
 983         if (inset->lyxCode() == NEWLINE_CODE || inset->lyxCode() == SEPARATOR_CODE) {
 984                 // newlines are handled differently here than
 985                 // the default in simpleTeXSpecialChars().
 986                 if (!style.newline_allowed) {
 987                         os << '\n';
 988                 } else {
 989                         if (open_font) {
 990                                 bool needPar = false;
 991                                 column += running_font.latexWriteEndChanges(
 992                                         os, bparams, runparams,
 993                                         basefont, basefont, needPar);
 994                                 open_font = false;
 995                         }
 996
 997                         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY)
 998                                 os << '~';
 999
1000                         basefont = owner_->getLayoutFont(bparams, outerfont);
1001                         running_font = basefont;
1002
1003                         if (runparams.moving_arg)
1004                                 os << "\\protect ";
1005
1006                 }
1007                 os.texrow().start(owner_->id(), i + 1);
1008                 column = 0;
1009         }
1010
1011         if (owner_->isDeleted(i)) {
1012                 if( ++runparams.inDeletedInset == 1)
1013                         runparams.changeOfDeletedInset = owner_->lookupChange(i);
1014         }
1015
1016         if (inset->canTrackChanges()) {
1017                 column += Changes::latexMarkChange(os, bparams, running_change,
1018                         Change(Change::UNCHANGED), runparams);
1019                 running_change = Change(Change::UNCHANGED);
1020         }
1021
1022         bool close = false;
1023         odocstream::pos_type const len = os.os().tellp();
1024
1025         if (inset->forceLTR(runparams)
1026             && running_font.isRightToLeft()
1027             // ERT is an exception, it should be output with no
1028             // decorations at all
1029             && inset->lyxCode() != ERT_CODE) {
1030                 if (runparams.use_polyglossia) {
1031                         os << "\\LRE{";
1032                 } else if (running_font.language()->lang() == "farsi"
1033                            || running_font.language()->lang() == "arabic_arabi")
1034                         os << "\\textLR{" << termcmd;
1035                 else
1036                         os << "\\L{";
1037                 close = true;
1038         }
1039
1040         // FIXME: Bug: we can have an empty font change here!
1041         // if there has just been a font change, we are going to close it
1042         // right now, which means stupid latex code like \textsf{}. AFAIK,
1043         // this does not harm dvi output. A minor bug, thus (JMarc)
1044
1045         // Some insets cannot be inside a font change command.
1046         // However, even such insets *can* be placed in \L or \R
1047         // or their equivalents (for RTL language switches), so we don't
1048         // close the language in those cases.
1049         // ArabTeX, though, cannot handle this special behavior, it seems.
1050         bool arabtex = basefont.language()->lang() == "arabic_arabtex"
1051                 || running_font.language()->lang() == "arabic_arabtex";
1052         if (open_font && !inset->inheritFont()) {
1053                 bool needPar = false;
1054                 bool closeLanguage = arabtex
1055                         || basefont.isRightToLeft() == running_font.isRightToLeft();
1056                 // We pass non_inherit_inset = true here since size switches
1057                 // ought not to be terminated here (#8384).
1058                 unsigned int count = running_font.latexWriteEndChanges(os,
1059                                         bparams, runparams, basefont, basefont,
1060                                         needPar, closeLanguage, true);
1061                 column += count;
1062                 // if any font properties were closed, update the running_font,
1063                 // making sure, however, to leave the language as it was
1064                 if (count > 0) {
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 no need to close
1073                         basefont.fontInfo().setSize(copy_font.fontInfo().size());
1074                         basefont.fontInfo().setFamily(copy_font.fontInfo().family());
1075                         basefont.fontInfo().setSeries(copy_font.fontInfo().series());
1076                         // leave font open if language or any of the switches is still open
1077                         open_font = (running_font.language() == basefont.language()
1078                                      || running_font.fontInfo().size() == basefont.fontInfo().size()
1079                                      || running_font.fontInfo().family() == basefont.fontInfo().family()
1080                                      || running_font.fontInfo().series() == basefont.fontInfo().series());
1081                         if (closeLanguage)
1082                                 runparams.local_font = &basefont;
1083                 }
1084         }
1085
1086         size_t const previous_row_count = os.texrow().rows();
1087
1088         try {
1089                 runparams.lastid = id_;
1090                 runparams.lastpos = i;
1091                 inset->latex(os, runparams);
1092         } catch (EncodingException & e) {
1093                 // add location information and throw again.
1094                 e.par_id = id_;
1095                 e.pos = i;
1096                 throw(e);
1097         }
1098
1099         if (close)
1100                 os << '}';
1101
1102         if (os.texrow().rows() > previous_row_count) {
1103                 os.texrow().start(owner_->id(), i + 1);
1104                 column = 0;
1105         } else {
1106                 column += (unsigned int)(os.os().tellp() - len);
1107         }
1108
1109         if (owner_->isDeleted(i))
1110                 --runparams.inDeletedInset;
1111 }
1112
1113
1114 void Paragraph::Private::latexSpecialChar(otexstream & os,
1115                                           BufferParams const & bparams,
1116                                           OutputParams const & runparams,
1117                                           Font const & running_font,
1118                                           string & alien_script,
1119                                           Layout const & style,
1120                                           pos_type & i,
1121                                           pos_type end_pos,
1122                                           unsigned int & column)
1123 {
1124         char_type const c = owner_->getUChar(bparams, runparams, i);
1125
1126         if (style.pass_thru || runparams.pass_thru
1127             || contains(style.pass_thru_chars, c)
1128             || contains(runparams.pass_thru_chars, c)) {
1129                 if (c != '\0') {
1130                         Encoding const * const enc = runparams.encoding;
1131                         if (enc && !enc->encodable(c))
1132                                 throw EncodingException(c);
1133                         os.put(c);
1134                 }
1135                 return;
1136         }
1137
1138         // TIPA uses its own T3 encoding
1139         if (runparams.inIPA && latexSpecialT3(c, os, i, column))
1140                 return;
1141         // If T1 font encoding is used, use the special
1142         // characters it provides.
1143         // NOTE: Some languages reset the font encoding internally to a
1144         //       non-standard font encoding. If we are using such a language,
1145         //       we do not output special T1 chars.
1146         if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1147             && !runparams.isFullUnicode() && bparams.main_font_encoding() == "T1"
1148             && latexSpecialT1(c, os, i, column))
1149                 return;
1150         // NOTE: "fontspec" (non-TeX fonts) sets the font encoding to "TU" (untill 2017 "EU1" or "EU2")
1151         else if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1152                  && runparams.isFullUnicode() && latexSpecialTU(c, os, i, column))
1153                      return;
1154
1155         // Otherwise, we use what LaTeX provides us.
1156         switch (c) {
1157         case '\\':
1158                 os << "\\textbackslash" << termcmd;
1159                 column += 15;
1160                 break;
1161         case '<':
1162                 os << "\\textless" << termcmd;
1163                 column += 10;
1164                 break;
1165         case '>':
1166                 os << "\\textgreater" << termcmd;
1167                 column += 13;
1168                 break;
1169         case '|':
1170                 os << "\\textbar" << termcmd;
1171                 column += 9;
1172                 break;
1173         case '-':
1174                 os << '-';
1175                 if (i + 1 < static_cast<pos_type>(text_.size()) &&
1176                     (end_pos == -1 || i + 1 < end_pos) &&
1177                     text_[i+1] == '-') {
1178                         // Prevent "--" becoming an en dash and "---" an em dash.
1179                         // (Within \ttfamily, "---" is merged to en dash + hyphen.)
1180                         os << "{}";
1181                         column += 2;
1182                 }
1183                 break;
1184         case '\"':
1185                 os << "\\textquotedbl" << termcmd;
1186                 column += 14;
1187                 break;
1188
1189         case '$': case '&':
1190         case '%': case '#': case '{':
1191         case '}': case '_':
1192                 os << '\\';
1193                 os.put(c);
1194                 column += 1;
1195                 break;
1196
1197         case '~':
1198                 os << "\\textasciitilde" << termcmd;
1199                 column += 16;
1200                 break;
1201
1202         case '^':
1203                 os << "\\textasciicircum" << termcmd;
1204                 column += 17;
1205                 break;
1206
1207         case '*':
1208         case '[':
1209         case ']':
1210                 // avoid being mistaken for optional arguments
1211                 os << '{';
1212                 os.put(c);
1213                 os << '}';
1214                 column += 2;
1215                 break;
1216
1217         case ' ':
1218                 // Blanks are printed before font switching.
1219                 // Sure? I am not! (try nice-latex)
1220                 // I am sure it's correct. LyX might be smarter
1221                 // in the future, but for now, nothing wrong is
1222                 // written. (Asger)
1223                 break;
1224
1225         case 0x2013:
1226         case 0x2014:
1227                 // XeTeX's dash behaviour is determined via a global setting
1228                 if (bparams.use_dash_ligatures
1229                     && owner_->getFontSettings(bparams, i).fontInfo().family() != TYPEWRITER_FAMILY
1230                     && !runparams.inIPA
1231                         // TODO #10961: && not in inset Flex Code
1232                         // TODO #10961: && not in layout LyXCode
1233                     && (!bparams.useNonTeXFonts || runparams.flavor != OutputParams::XETEX)) {
1234                         if (c == 0x2013) {
1235                                 // en-dash
1236                                 os << "--";
1237                                 column +=2;
1238                         } else {
1239                                 // em-dash
1240                                 os << "---";
1241                                 column +=3;
1242                         }
1243                         break;
1244                 }
1245                 // fall through
1246         default:
1247                 if (c == '\0')
1248                         return;
1249
1250                 Encoding const & encoding = *(runparams.encoding);
1251                 char_type next = '\0';
1252                 if (i + 1 < int(text_.size())) {
1253                         next = text_[i + 1];
1254                         if (Encodings::isCombiningChar(next)) {
1255                                 column += latexSurrogatePair(bparams, os, c, next, runparams) - 1;
1256                                 ++i;
1257                                 break;
1258                         }
1259                 }
1260                 pair<docstring, bool> latex = encoding.latexChar(c);
1261                 docstring nextlatex;
1262                 bool nexttipas = false;
1263                 string nexttipashortcut;
1264                 if (next != '\0' && next != META_INSET && !encoding.encodable(next)) {
1265                         nextlatex = encoding.latexChar(next).first;
1266                         if (runparams.inIPA) {
1267                                 nexttipashortcut = Encodings::TIPAShortcut(next);
1268                                 nexttipas = !nexttipashortcut.empty();
1269                         }
1270                 }
1271                 bool tipas = false;
1272                 if (runparams.inIPA) {
1273                         string const tipashortcut = Encodings::TIPAShortcut(c);
1274                         if (!tipashortcut.empty()) {
1275                                 latex.first = from_ascii(tipashortcut);
1276                                 latex.second = false;
1277                                 tipas = true;
1278                         }
1279                 }
1280                 // eventually close "script wrapper" command (see `Paragraph::latex`)
1281                 if (!alien_script.empty()
1282                         && alien_script != Encodings::isKnownScriptChar(next)) {
1283                         column += latex.first.length();
1284                         alien_script.clear();
1285                         os << latex.first << "}";
1286                         break;
1287                 }
1288                 if (latex.second
1289                          && ((!prefixIs(nextlatex, '\\')
1290                                && !prefixIs(nextlatex, '{')
1291                                && !prefixIs(nextlatex, '}'))
1292                              || (nexttipas
1293                                  && !prefixIs(from_ascii(nexttipashortcut), '\\')))
1294                          && !tipas) {
1295                         // Prevent eating of a following space or command corruption by
1296                         // following characters
1297                         if (next == ' ' || next == '\0') {
1298                                 column += latex.first.length() + 1;
1299                                 os << latex.first << "{}";
1300                         } else {
1301                                 column += latex.first.length();
1302                                 os << latex.first << " ";
1303                         }
1304                 } else {
1305                         column += latex.first.length() - 1;
1306                         os << latex.first;
1307                 }
1308                 break;
1309         }
1310 }
1311
1312
1313 bool Paragraph::Private::latexSpecialT1(char_type const c, otexstream & os,
1314         pos_type i, unsigned int & column)
1315 {
1316         switch (c) {
1317         case '>':
1318         case '<':
1319                 os.put(c);
1320                 // In T1 encoding, these characters exist
1321                 // but we should avoid ligatures
1322                 if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
1323                         return true;
1324                 os << "\\textcompwordmark" << termcmd;
1325                 column += 19;
1326                 return true;
1327         case '|':
1328                 os.put(c);
1329                 return true;
1330         case '\"':
1331                 // soul.sty breaks with \char`\"
1332                 os << "\\textquotedbl" << termcmd;
1333                 column += 14;
1334                 return true;
1335         default:
1336                 return false;
1337         }
1338 }
1339
1340
1341 bool Paragraph::Private::latexSpecialTU(char_type const c, otexstream & os,
1342         pos_type i, unsigned int & column)
1343 {
1344         // TU encoding is currently on par with T1.
1345         return latexSpecialT1(c, os, i, column);
1346 }
1347
1348
1349 bool Paragraph::Private::latexSpecialT3(char_type const c, otexstream & os,
1350         pos_type /*i*/, unsigned int & column)
1351 {
1352         switch (c) {
1353         case '*':
1354         case '[':
1355         case ']':
1356         case '\"':
1357                 os.put(c);
1358                 return true;
1359         case '|':
1360                 os << "\\textvertline" << termcmd;
1361                 column += 14;
1362                 return true;
1363         default:
1364                 return false;
1365         }
1366 }
1367
1368
1369 void Paragraph::Private::validate(LaTeXFeatures & features) const
1370 {
1371         if (layout_->inpreamble && inset_owner_) {
1372                 // FIXME: Using a string stream here circumvents the encoding
1373                 // switching machinery of odocstream. Therefore the
1374                 // output is wrong if this paragraph contains content
1375                 // that needs to switch encoding.
1376                 Buffer const & buf = inset_owner_->buffer();
1377                 otexstringstream os;
1378                 os << layout_->preamble();
1379                 size_t const length = os.length();
1380                 TeXOnePar(buf, *inset_owner_->getText(int(buf.getParFromID(owner_->id()).idx())),
1381                           buf.getParFromID(owner_->id()).pit(), os,
1382                           features.runparams(), string(), 0, -1, true);
1383                 if (os.length() > length)
1384                         features.addPreambleSnippet(os.release(), true);
1385         }
1386
1387         if (features.runparams().flavor == OutputParams::HTML
1388             && layout_->htmltitle()) {
1389                 features.setHTMLTitle(owner_->asString(AS_STR_INSETS | AS_STR_SKIPDELETE));
1390         }
1391
1392         // check the params.
1393         if (!params_.spacing().isDefault())
1394                 features.require("setspace");
1395
1396         // then the layouts
1397         features.useLayout(layout_->name());
1398
1399         // then the fonts
1400         fontlist_.validate(features);
1401
1402         // then the indentation
1403         if (!params_.leftIndent().zero())
1404                 features.require("ParagraphLeftIndent");
1405
1406         // then the insets
1407         InsetList::const_iterator icit = insetlist_.begin();
1408         InsetList::const_iterator iend = insetlist_.end();
1409         for (; icit != iend; ++icit) {
1410                 if (icit->inset) {
1411                         features.inDeletedInset(owner_->isDeleted(icit->pos));
1412                         if (icit->inset->lyxCode() == FOOT_CODE) {
1413                                 // FIXME: an item inset would make things much easier.
1414                                 if ((layout_->latextype == LATEX_LIST_ENVIRONMENT
1415                                      || (layout_->latextype == LATEX_ITEM_ENVIRONMENT
1416                                          && layout_->margintype == MARGIN_FIRST_DYNAMIC))
1417                                     && (icit->pos < begin_of_body_
1418                                         || (icit->pos == begin_of_body_
1419                                             && (icit->pos == 0 || text_[icit->pos - 1] != ' '))))
1420                                         features.saveNoteEnv("description");
1421                         }
1422                         icit->inset->validate(features);
1423                         features.inDeletedInset(false);
1424                         if (layout_->needprotect &&
1425                             icit->inset->lyxCode() == FOOT_CODE)
1426                                 features.require("NeedLyXFootnoteCode");
1427                 }
1428         }
1429
1430         // then the contents
1431         BufferParams const bp = features.runparams().is_child
1432                 ? features.buffer().masterParams() : features.buffer().params();
1433         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1434                 char_type c = text_[i];
1435                 CharInfo const & ci = Encodings::unicodeCharInfo(c);
1436                 if (c == 0x0022) {
1437                         if (features.runparams().isFullUnicode() && bp.useNonTeXFonts)
1438                                 features.require("textquotedblp");
1439                         else if (bp.main_font_encoding() != "T1"
1440                                  || ((&owner_->getFontSettings(bp, i))->language()->internalFontEncoding()))
1441                                 features.require("textquotedbl");
1442                 } else if (ci.textfeature() && contains(ci.textpreamble(), '=')) {
1443                         // features that depend on the font or input encoding
1444                         string feats = ci.textpreamble();
1445                         string fontenc = (&owner_->getFontSettings(bp, i))->language()->fontenc(bp);
1446                         if (fontenc.empty())
1447                                 fontenc = features.runparams().main_fontenc;
1448                         while (!feats.empty()) {
1449                                 string feat;
1450                                 feats = split(feats, feat, ',');
1451                                 if (contains(feat, "!=")) {
1452                                         // a feature that is required except for the spcified
1453                                         // font or input encodings
1454                                         string realfeature;
1455                                         string const contexts = ltrim(split(feat, realfeature, '!'), "=");
1456                                         // multiple encodings are separated by semicolon
1457                                         vector<string> context = getVectorFromString(contexts, ";");
1458                                         // require feature if the context matches neither current font
1459                                         // nor input encoding
1460                                         if (std::find(context.begin(), context.end(), fontenc) == context.end()
1461                                             && std::find(context.begin(), context.end(),
1462                                                          features.runparams().encoding->name()) == context.end())
1463                                                 features.require(realfeature);
1464                                 } else if (contains(feat, '=')) {
1465                                         // a feature that is required only for the spcified
1466                                         // font or input encodings
1467                                         string realfeature;
1468                                         string const contexts = split(feat, realfeature, '=');
1469                                         // multiple encodings are separated by semicolon
1470                                         vector<string> context = getVectorFromString(contexts, ";");
1471                                         // require feature if the context matches either current font
1472                                         // or input encoding
1473                                         if (std::find(context.begin(), context.end(), fontenc) != context.end()
1474                                             || std::find(context.begin(), context.end(),
1475                                                          features.runparams().encoding->name()) != context.end())
1476                                                 features.require(realfeature);
1477                                 }
1478                         }
1479                 } else if (!bp.use_dash_ligatures
1480                            && (c == 0x2013 || c == 0x2014)
1481                            && bp.useNonTeXFonts
1482                            && features.runparams().flavor == OutputParams::XETEX)
1483                         // XeTeX's dash behaviour is determined via a global setting
1484                         features.require("xetexdashbreakstate");
1485                 BufferEncodings::validate(c, features);
1486         }
1487 }
1488
1489 /////////////////////////////////////////////////////////////////////
1490 //
1491 // Paragraph
1492 //
1493 /////////////////////////////////////////////////////////////////////
1494
1495 namespace {
1496         Layout const emptyParagraphLayout;
1497 }
1498
1499 Paragraph::Paragraph()
1500         : d(new Paragraph::Private(this, emptyParagraphLayout))
1501 {
1502         itemdepth = 0;
1503         d->params_.clear();
1504 }
1505
1506
1507 Paragraph::Paragraph(Paragraph const & par)
1508         : itemdepth(par.itemdepth),
1509         d(new Paragraph::Private(*par.d, this))
1510 {
1511         registerWords();
1512 }
1513
1514
1515 Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
1516         : itemdepth(par.itemdepth),
1517         d(new Paragraph::Private(*par.d, this, beg, end))
1518 {
1519         registerWords();
1520 }
1521
1522
1523 Paragraph & Paragraph::operator=(Paragraph const & par)
1524 {
1525         // needed as we will destroy the private part before copying it
1526         if (&par != this) {
1527                 itemdepth = par.itemdepth;
1528
1529                 deregisterWords();
1530                 delete d;
1531                 d = new Private(*par.d, this);
1532                 registerWords();
1533         }
1534         return *this;
1535 }
1536
1537
1538 Paragraph::~Paragraph()
1539 {
1540         deregisterWords();
1541         delete d;
1542 }
1543
1544
1545 namespace {
1546
1547 // this shall be called just before every "os << ..." action.
1548 void flushString(ostream & os, docstring & s)
1549 {
1550         os << to_utf8(s);
1551         s.erase();
1552 }
1553
1554 } // namespace
1555
1556
1557 void Paragraph::write(ostream & os, BufferParams const & bparams,
1558         depth_type & dth) const
1559 {
1560         // The beginning or end of a deeper (i.e. nested) area?
1561         if (dth != d->params_.depth()) {
1562                 if (d->params_.depth() > dth) {
1563                         while (d->params_.depth() > dth) {
1564                                 os << "\n\\begin_deeper";
1565                                 ++dth;
1566                         }
1567                 } else {
1568                         while (d->params_.depth() < dth) {
1569                                 os << "\n\\end_deeper";
1570                                 --dth;
1571                         }
1572                 }
1573         }
1574
1575         // First write the layout
1576         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1577
1578         d->params_.write(os);
1579
1580         Font font1(inherit_font, bparams.language);
1581
1582         Change running_change = Change(Change::UNCHANGED);
1583
1584         // this string is used as a buffer to avoid repetitive calls
1585         // to to_utf8(), which turn out to be expensive (JMarc)
1586         docstring write_buffer;
1587
1588         int column = 0;
1589         for (pos_type i = 0; i <= size(); ++i) {
1590
1591                 Change const & change = lookupChange(i);
1592                 if (change != running_change)
1593                         flushString(os, write_buffer);
1594                 Changes::lyxMarkChange(os, bparams, column, running_change, change);
1595                 running_change = change;
1596
1597                 if (i == size())
1598                         break;
1599
1600                 // Write font changes
1601                 Font font2 = getFontSettings(bparams, i);
1602                 if (font2 != font1) {
1603                         flushString(os, write_buffer);
1604                         font2.lyxWriteChanges(font1, os);
1605                         column = 0;
1606                         font1 = font2;
1607                 }
1608
1609                 char_type const c = d->text_[i];
1610                 switch (c) {
1611                 case META_INSET:
1612                         if (Inset const * inset = getInset(i)) {
1613                                 flushString(os, write_buffer);
1614                                 if (inset->directWrite()) {
1615                                         // international char, let it write
1616                                         // code directly so it's shorter in
1617                                         // the file
1618                                         inset->write(os);
1619                                 } else {
1620                                         if (i)
1621                                                 os << '\n';
1622                                         os << "\\begin_inset ";
1623                                         inset->write(os);
1624                                         os << "\n\\end_inset\n\n";
1625                                         column = 0;
1626                                 }
1627                                 // FIXME This can be removed again once the mystery
1628                                 // crash has been resolved.
1629                                 os << flush;
1630                         }
1631                         break;
1632                 case '\\':
1633                         flushString(os, write_buffer);
1634                         os << "\n\\backslash\n";
1635                         column = 0;
1636                         break;
1637                 case '.':
1638                         flushString(os, write_buffer);
1639                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1640                                 os << ".\n";
1641                                 column = 0;
1642                         } else
1643                                 os << '.';
1644                         break;
1645                 default:
1646                         if ((column > 70 && c == ' ')
1647                             || column > 79) {
1648                                 flushString(os, write_buffer);
1649                                 os << '\n';
1650                                 column = 0;
1651                         }
1652                         // this check is to amend a bug. LyX sometimes
1653                         // inserts '\0' this could cause problems.
1654                         if (c != '\0')
1655                                 write_buffer.push_back(c);
1656                         else
1657                                 LYXERR0("NUL char in structure.");
1658                         ++column;
1659                         break;
1660                 }
1661         }
1662
1663         flushString(os, write_buffer);
1664         os << "\n\\end_layout\n";
1665         // FIXME This can be removed again once the mystery
1666         // crash has been resolved.
1667         os << flush;
1668 }
1669
1670
1671 void Paragraph::validate(LaTeXFeatures & features) const
1672 {
1673         d->validate(features);
1674         bool fragile = features.runparams().moving_arg;
1675         fragile |= layout().needprotect;
1676         if (inInset().getLayout().isNeedProtect())
1677                 fragile = true;
1678         if (needsCProtection(fragile))
1679                 features.require("cprotect");
1680 }
1681
1682
1683 void Paragraph::insert(pos_type start, docstring const & str,
1684                        Font const & font, Change const & change)
1685 {
1686         for (size_t i = 0, n = str.size(); i != n ; ++i)
1687                 insertChar(start + i, str[i], font, change);
1688 }
1689
1690
1691 void Paragraph::appendChar(char_type c, Font const & font,
1692                 Change const & change)
1693 {
1694         // track change
1695         d->changes_.insert(change, d->text_.size());
1696         // when appending characters, no need to update tables
1697         d->text_.push_back(c);
1698         setFont(d->text_.size() - 1, font);
1699         d->requestSpellCheck(d->text_.size() - 1);
1700 }
1701
1702
1703 void Paragraph::appendString(docstring const & s, Font const & font,
1704                 Change const & change)
1705 {
1706         pos_type end = s.size();
1707         size_t oldsize = d->text_.size();
1708         size_t newsize = oldsize + end;
1709         size_t capacity = d->text_.capacity();
1710         if (newsize >= capacity)
1711                 d->text_.reserve(max(capacity + 100, newsize));
1712
1713         // when appending characters, no need to update tables
1714         d->text_.append(s);
1715
1716         // FIXME: Optimize this!
1717         for (size_t i = oldsize; i != newsize; ++i) {
1718                 // track change
1719                 d->changes_.insert(change, i);
1720                 d->requestSpellCheck(i);
1721         }
1722         d->fontlist_.set(oldsize, font);
1723         d->fontlist_.set(newsize - 1, font);
1724 }
1725
1726
1727 void Paragraph::insertChar(pos_type pos, char_type c,
1728                            bool trackChanges)
1729 {
1730         d->insertChar(pos, c, Change(trackChanges ?
1731                            Change::INSERTED : Change::UNCHANGED));
1732 }
1733
1734
1735 void Paragraph::insertChar(pos_type pos, char_type c,
1736                            Font const & font, bool trackChanges)
1737 {
1738         d->insertChar(pos, c, Change(trackChanges ?
1739                            Change::INSERTED : Change::UNCHANGED));
1740         setFont(pos, font);
1741 }
1742
1743
1744 void Paragraph::insertChar(pos_type pos, char_type c,
1745                            Font const & font, Change const & change)
1746 {
1747         d->insertChar(pos, c, change);
1748         setFont(pos, font);
1749 }
1750
1751
1752 void Paragraph::resetFonts(Font const & font)
1753 {
1754         d->fontlist_.clear();
1755         d->fontlist_.set(0, font);
1756         d->fontlist_.set(d->text_.size() - 1, font);
1757 }
1758
1759 // Gets uninstantiated font setting at position.
1760 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1761                                          pos_type pos) const
1762 {
1763         if (pos > size()) {
1764                 LYXERR0("pos: " << pos << " size: " << size());
1765                 LBUFERR(false);
1766         }
1767
1768         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1769         if (cit != d->fontlist_.end())
1770                 return cit->font();
1771
1772         if (pos == size() && !empty())
1773                 return getFontSettings(bparams, pos - 1);
1774
1775         // Optimisation: avoid a full font instantiation if there is no
1776         // language change from previous call.
1777         static Font previous_font;
1778         static Language const * previous_lang = nullptr;
1779         Language const * lang = getParLanguage(bparams);
1780         if (lang != previous_lang) {
1781                 previous_lang = lang;
1782                 previous_font = Font(inherit_font, lang);
1783         }
1784         return previous_font;
1785 }
1786
1787
1788 FontSpan Paragraph::fontSpan(pos_type pos) const
1789 {
1790         LBUFERR(pos <= size());
1791
1792         if (pos == size())
1793                 return FontSpan(pos, pos);
1794
1795         pos_type start = 0;
1796         FontList::const_iterator cit = d->fontlist_.begin();
1797         FontList::const_iterator end = d->fontlist_.end();
1798         for (; cit != end; ++cit) {
1799                 if (cit->pos() >= pos) {
1800                         if (pos >= beginOfBody())
1801                                 return FontSpan(max(start, beginOfBody()),
1802                                                 cit->pos());
1803                         else
1804                                 return FontSpan(start,
1805                                                 min(beginOfBody() - 1,
1806                                                          cit->pos()));
1807                 }
1808                 start = cit->pos() + 1;
1809         }
1810
1811         // This should not happen, but if so, we take no chances.
1812         LYXERR0("Paragraph::fontSpan: position not found in fontinfo table!");
1813         LASSERT(false, return FontSpan(pos, pos));
1814 }
1815
1816
1817 // Gets uninstantiated font setting at position 0
1818 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1819 {
1820         if (!empty() && !d->fontlist_.empty())
1821                 return d->fontlist_.begin()->font();
1822
1823         // Optimisation: avoid a full font instantiation if there is no
1824         // language change from previous call.
1825         static Font previous_font;
1826         static Language const * previous_lang = nullptr;
1827         if (bparams.language != previous_lang) {
1828                 previous_lang = bparams.language;
1829                 previous_font = Font(inherit_font, bparams.language);
1830         }
1831
1832         return previous_font;
1833 }
1834
1835
1836 // Gets the fully instantiated font at a given position in a paragraph
1837 // This is basically the same function as Text::GetFont() in text2.cpp.
1838 // The difference is that this one is used for generating the LaTeX file,
1839 // and thus cosmetic "improvements" are disallowed: This has to deliver
1840 // the true picture of the buffer. (Asger)
1841 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1842                                  Font const & outerfont) const
1843 {
1844         LBUFERR(pos >= 0);
1845
1846         Font font = getFontSettings(bparams, pos);
1847
1848         pos_type const body_pos = beginOfBody();
1849         FontInfo & fi = font.fontInfo();
1850         if (pos < body_pos)
1851                 fi.realize(d->layout_->labelfont);
1852         else
1853                 fi.realize(d->layout_->font);
1854
1855         fi.realize(outerfont.fontInfo());
1856         fi.realize(bparams.getFont().fontInfo());
1857
1858         return font;
1859 }
1860
1861
1862 Font const Paragraph::getLabelFont
1863         (BufferParams const & bparams, Font const & outerfont) const
1864 {
1865         FontInfo tmpfont = d->layout_->labelfont;
1866         tmpfont.realize(outerfont.fontInfo());
1867         tmpfont.realize(bparams.getFont().fontInfo());
1868         return Font(tmpfont, getParLanguage(bparams));
1869 }
1870
1871
1872 Font const Paragraph::getLayoutFont
1873         (BufferParams const & bparams, Font const & outerfont) const
1874 {
1875         FontInfo tmpfont = d->layout_->font;
1876         tmpfont.realize(outerfont.fontInfo());
1877         tmpfont.realize(bparams.getFont().fontInfo());
1878         return Font(tmpfont, getParLanguage(bparams));
1879 }
1880
1881
1882 char_type Paragraph::getUChar(BufferParams const & bparams,
1883                               OutputParams const & rp,
1884                               pos_type pos) const
1885 {
1886         char_type c = d->text_[pos];
1887
1888         // Return unchanged character in LTR languages
1889         // or if we use poylglossia/bidi (XeTeX).
1890         if (rp.useBidiPackage()
1891             || !getFontSettings(bparams, pos).isRightToLeft())
1892                 return c;
1893
1894         // Without polyglossia/bidi, we need to account for some special cases.
1895         // FIXME This needs to be audited!
1896         // Check if:
1897         // * The input is as expected for all delimiters
1898         //   => checked for Hebrew!
1899         // * The output matches the display in the LyX workarea
1900         //   => checked for Hebrew!
1901         // * The special cases below are really necessary
1902         //   => checked for Hebrew!
1903         // * In arabic_arabi, brackets are transformed to Arabic
1904         //   Ornate Parentheses. Is this is really wanted?
1905
1906         string const & lang = getFontSettings(bparams, pos).language()->lang();
1907         char_type uc = c;
1908
1909         // 1. In the following languages, parentheses need to be reversed.
1910         //    Also with polyglodia/luabidi
1911         bool const reverseparens = (lang == "hebrew" || rp.use_polyglossia);
1912
1913         // 2. In the following languages, brackets don't need to be reversed.
1914         bool const reversebrackets = lang != "arabic_arabtex"
1915                         && lang != "arabic_arabi"
1916                         && lang != "farsi";
1917
1918         // Now swap delimiters if needed.
1919         switch (c) {
1920         case '(':
1921                 if (reverseparens)
1922                         uc = ')';
1923                 break;
1924         case ')':
1925                 if (reverseparens)
1926                         uc = '(';
1927                 break;
1928         case '[':
1929                 if (reversebrackets)
1930                         uc = ']';
1931                 break;
1932         case ']':
1933                 if (reversebrackets)
1934                         uc = '[';
1935                 break;
1936         case '{':
1937                 uc = '}';
1938                 break;
1939         case '}':
1940                 uc = '{';
1941                 break;
1942         case '<':
1943                 uc = '>';
1944                 break;
1945         case '>':
1946                 uc = '<';
1947                 break;
1948         }
1949
1950         return uc;
1951 }
1952
1953
1954 void Paragraph::setFont(pos_type pos, Font const & font)
1955 {
1956         LASSERT(pos <= size(), return);
1957
1958         // First, reduce font against layout/label font
1959         // Update: The setCharFont() routine in text2.cpp already
1960         // reduces font, so we don't need to do that here. (Asger)
1961
1962         d->fontlist_.set(pos, font);
1963 }
1964
1965
1966 void Paragraph::makeSameLayout(Paragraph const & par)
1967 {
1968         d->layout_ = par.d->layout_;
1969         d->params_ = par.d->params_;
1970 }
1971
1972
1973 bool Paragraph::stripLeadingSpaces(bool trackChanges)
1974 {
1975         if (isFreeSpacing())
1976                 return false;
1977
1978         int pos = 0;
1979         int count = 0;
1980
1981         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
1982                 if (eraseChar(pos, trackChanges))
1983                         ++count;
1984                 else
1985                         ++pos;
1986         }
1987
1988         return count > 0 || pos > 0;
1989 }
1990
1991
1992 bool Paragraph::hasSameLayout(Paragraph const & par) const
1993 {
1994         return par.d->layout_ == d->layout_
1995                 && d->params_.sameLayout(par.d->params_);
1996 }
1997
1998
1999 depth_type Paragraph::getDepth() const
2000 {
2001         return d->params_.depth();
2002 }
2003
2004
2005 depth_type Paragraph::getMaxDepthAfter() const
2006 {
2007         if (d->layout_->isEnvironment())
2008                 return d->params_.depth() + 1;
2009         else
2010                 return d->params_.depth();
2011 }
2012
2013
2014 LyXAlignment Paragraph::getAlign(BufferParams const & bparams) const
2015 {
2016         if (d->params_.align() == LYX_ALIGN_LAYOUT)
2017                 return getDefaultAlign(bparams);
2018         else
2019                 return d->params_.align();
2020 }
2021
2022
2023 LyXAlignment Paragraph::getDefaultAlign(BufferParams const & bparams) const
2024 {
2025         LyXAlignment res = layout().align;
2026         if (isRTL(bparams)) {
2027                 // Swap sides
2028                 if (res == LYX_ALIGN_LEFT)
2029                         res = LYX_ALIGN_RIGHT;
2030                 else if  (res == LYX_ALIGN_RIGHT)
2031                         res = LYX_ALIGN_LEFT;
2032         }
2033         return res;
2034 }
2035
2036
2037 docstring const & Paragraph::labelString() const
2038 {
2039         return d->params_.labelString();
2040 }
2041
2042
2043 // the next two functions are for the manual labels
2044 docstring const Paragraph::getLabelWidthString() const
2045 {
2046         if (d->layout_->margintype == MARGIN_MANUAL
2047             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
2048                 return d->params_.labelWidthString();
2049         else
2050                 return _("Senseless with this layout!");
2051 }
2052
2053
2054 void Paragraph::setLabelWidthString(docstring const & s)
2055 {
2056         d->params_.labelWidthString(s);
2057 }
2058
2059
2060 docstring Paragraph::expandLabel(Layout const & layout,
2061                 BufferParams const & bparams) const
2062 {
2063         return expandParagraphLabel(layout, bparams, true);
2064 }
2065
2066
2067 docstring Paragraph::expandParagraphLabel(Layout const & layout,
2068                 BufferParams const & bparams, bool process_appendix) const
2069 {
2070         DocumentClass const & tclass = bparams.documentClass();
2071         string const & lang = getParLanguage(bparams)->code();
2072         bool const in_appendix = process_appendix && d->params_.appendix();
2073         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
2074
2075         if (fmt.empty() && !layout.counter.empty())
2076                 return tclass.counters().theCounter(layout.counter, lang);
2077
2078         // handle 'inherited level parts' in 'fmt',
2079         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
2080         size_t const i = fmt.find('@', 0);
2081         if (i != docstring::npos) {
2082                 size_t const j = fmt.find('@', i + 1);
2083                 if (j != docstring::npos) {
2084                         docstring parent(fmt, i + 1, j - i - 1);
2085                         docstring label = from_ascii("??");
2086                         if (tclass.hasLayout(parent))
2087                                 label = expandParagraphLabel(tclass[parent], bparams,
2088                                                       process_appendix);
2089                         fmt = docstring(fmt, 0, i) + label
2090                                 + docstring(fmt, j + 1, docstring::npos);
2091                 }
2092         }
2093
2094         return tclass.counters().counterLabel(fmt, lang);
2095 }
2096
2097
2098 void Paragraph::applyLayout(Layout const & new_layout)
2099 {
2100         d->layout_ = &new_layout;
2101         LyXAlignment const oldAlign = d->params_.align();
2102
2103         if (!(oldAlign & d->layout_->alignpossible)) {
2104                 frontend::Alert::warning(_("Alignment not permitted"),
2105                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
2106                 d->params_.align(LYX_ALIGN_LAYOUT);
2107         }
2108 }
2109
2110
2111 pos_type Paragraph::beginOfBody() const
2112 {
2113         return d->begin_of_body_;
2114 }
2115
2116
2117 void Paragraph::setBeginOfBody()
2118 {
2119         if (d->layout_->labeltype != LABEL_MANUAL) {
2120                 d->begin_of_body_ = 0;
2121                 return;
2122         }
2123
2124         // Unroll the first two cycles of the loop
2125         // and remember the previous character to
2126         // remove unnecessary getChar() calls
2127         pos_type i = 0;
2128         pos_type end = size();
2129         bool prev_char_deleted = false;
2130         if (i < end && (!(isNewline(i) || isEnvSeparator(i)) || isDeleted(i))) {
2131                 ++i;
2132                 if (i < end) {
2133                         char_type previous_char = d->text_[i];
2134                         if (!(isNewline(i) || isEnvSeparator(i))) {
2135                                 ++i;
2136                                 while (i < end && (previous_char != ' ' || prev_char_deleted)) {
2137                                         char_type temp = d->text_[i];
2138                                         prev_char_deleted = isDeleted(i);
2139                                         if (!isDeleted(i) && (isNewline(i) || isEnvSeparator(i)))
2140                                                 break;
2141                                         ++i;
2142                                         previous_char = temp;
2143                                 }
2144                         }
2145                 }
2146         }
2147
2148         d->begin_of_body_ = i;
2149 }
2150
2151
2152 bool Paragraph::allowParagraphCustomization() const
2153 {
2154         return inInset().allowParagraphCustomization();
2155 }
2156
2157
2158 bool Paragraph::usePlainLayout() const
2159 {
2160         return inInset().usePlainLayout();
2161 }
2162
2163
2164 bool Paragraph::isPassThru() const
2165 {
2166         return inInset().isPassThru() || d->layout_->pass_thru;
2167 }
2168
2169 namespace {
2170
2171 // paragraphs inside floats need different alignment tags to avoid
2172 // unwanted space
2173
2174 bool noTrivlistCentering(InsetCode code)
2175 {
2176         return code == FLOAT_CODE
2177                || code == WRAP_CODE
2178                || code == CELL_CODE;
2179 }
2180
2181
2182 string correction(string const & orig)
2183 {
2184         if (orig == "flushleft")
2185                 return "raggedright";
2186         if (orig == "flushright")
2187                 return "raggedleft";
2188         if (orig == "center")
2189                 return "centering";
2190         return orig;
2191 }
2192
2193
2194 bool corrected_env(otexstream & os, string const & suffix, string const & env,
2195         InsetCode code, bool const lastpar, int & col)
2196 {
2197         string macro = suffix + "{";
2198         if (noTrivlistCentering(code)) {
2199                 if (lastpar) {
2200                         // the last paragraph in non-trivlist-aligned
2201                         // context is special (to avoid unwanted whitespace)
2202                         if (suffix == "\\begin") {
2203                                 macro = "\\" + correction(env) + "{}";
2204                                 os << from_ascii(macro);
2205                                 col += macro.size();
2206                                 return true;
2207                         }
2208                         return false;
2209                 }
2210                 macro += correction(env);
2211         } else
2212                 macro += env;
2213         macro += "}";
2214         if (suffix == "\\par\\end") {
2215                 os << breakln;
2216                 col = 0;
2217         }
2218         os << from_ascii(macro);
2219         col += macro.size();
2220         if (suffix == "\\begin") {
2221                 os << breakln;
2222                 col = 0;
2223         }
2224         return true;
2225 }
2226
2227 } // namespace
2228
2229
2230 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
2231                         otexstream & os, OutputParams const & runparams) const
2232 {
2233         int column = 0;
2234
2235         bool canindent =
2236                 (bparams.paragraph_separation == BufferParams::ParagraphIndentSeparation) ?
2237                         (layout_->toggle_indent != ITOGGLE_NEVER) :
2238                         (layout_->toggle_indent == ITOGGLE_ALWAYS);
2239
2240         if (canindent && params_.noindent() && !layout_->pass_thru) {
2241                 os << "\\noindent ";
2242                 column += 10;
2243         }
2244
2245         LyXAlignment const curAlign = params_.align();
2246
2247         if (curAlign == layout_->align)
2248                 return column;
2249
2250         switch (curAlign) {
2251         case LYX_ALIGN_NONE:
2252         case LYX_ALIGN_BLOCK:
2253         case LYX_ALIGN_LAYOUT:
2254         case LYX_ALIGN_SPECIAL:
2255         case LYX_ALIGN_DECIMAL:
2256                 break;
2257         case LYX_ALIGN_LEFT:
2258         case LYX_ALIGN_RIGHT:
2259         case LYX_ALIGN_CENTER:
2260                 if (runparams.moving_arg) {
2261                         os << "\\protect";
2262                         column += 8;
2263                 }
2264                 break;
2265         }
2266
2267         string const begin_tag = "\\begin";
2268         InsetCode code = ownerCode();
2269         bool const lastpar = runparams.isLastPar;
2270         // RTL in classic (PDF)LaTeX (without the Bidi package)
2271         // Luabibdi (used by LuaTeX) behaves like classic
2272         bool const rtl_classic = owner_->getParLanguage(bparams)->rightToLeft()
2273                 && !runparams.useBidiPackage();
2274
2275         switch (curAlign) {
2276         case LYX_ALIGN_NONE:
2277         case LYX_ALIGN_BLOCK:
2278         case LYX_ALIGN_LAYOUT:
2279         case LYX_ALIGN_SPECIAL:
2280         case LYX_ALIGN_DECIMAL:
2281                 break;
2282         case LYX_ALIGN_LEFT: {
2283                 if (rtl_classic)
2284                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2285                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2286                 else
2287                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2288                 break;
2289         } case LYX_ALIGN_RIGHT: {
2290                 if (rtl_classic)
2291                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2292                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2293                 else
2294                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2295                 break;
2296         } case LYX_ALIGN_CENTER: {
2297                 corrected_env(os, begin_tag, "center", code, lastpar, column);
2298                 break;
2299         }
2300         }
2301
2302         return column;
2303 }
2304
2305
2306 bool Paragraph::Private::endTeXParParams(BufferParams const & bparams,
2307                         otexstream & os, OutputParams const & runparams) const
2308 {
2309         LyXAlignment const curAlign = params_.align();
2310
2311         if (curAlign == layout_->align)
2312                 return false;
2313
2314         switch (curAlign) {
2315         case LYX_ALIGN_NONE:
2316         case LYX_ALIGN_BLOCK:
2317         case LYX_ALIGN_LAYOUT:
2318         case LYX_ALIGN_SPECIAL:
2319         case LYX_ALIGN_DECIMAL:
2320                 break;
2321         case LYX_ALIGN_LEFT:
2322         case LYX_ALIGN_RIGHT:
2323         case LYX_ALIGN_CENTER:
2324                 if (runparams.moving_arg)
2325                         os << "\\protect";
2326                 break;
2327         }
2328
2329         bool output = false;
2330         int col = 0;
2331         string const end_tag = "\\par\\end";
2332         InsetCode code = ownerCode();
2333         bool const lastpar = runparams.isLastPar;
2334         // RTL in classic (PDF)LaTeX (without the Bidi package)
2335         // Luabibdi (used by LuaTeX) behaves like classic
2336         bool const rtl_classic = owner_->getParLanguage(bparams)->rightToLeft()
2337                 && !runparams.useBidiPackage();
2338
2339         switch (curAlign) {
2340         case LYX_ALIGN_NONE:
2341         case LYX_ALIGN_BLOCK:
2342         case LYX_ALIGN_LAYOUT:
2343         case LYX_ALIGN_SPECIAL:
2344         case LYX_ALIGN_DECIMAL:
2345                 break;
2346         case LYX_ALIGN_LEFT: {
2347                 if (rtl_classic)
2348                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2349                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2350                 else
2351                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2352                 break;
2353         } case LYX_ALIGN_RIGHT: {
2354                 if (rtl_classic)
2355                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2356                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2357                 else
2358                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2359                 break;
2360         } case LYX_ALIGN_CENTER: {
2361                 corrected_env(os, end_tag, "center", code, lastpar, col);
2362                 break;
2363         }
2364         }
2365
2366         return output || lastpar;
2367 }
2368
2369
2370 // This one spits out the text of the paragraph
2371 void Paragraph::latex(BufferParams const & bparams,
2372         Font const & outerfont,
2373         otexstream & os,
2374         OutputParams const & runparams,
2375         int start_pos, int end_pos, bool force) const
2376 {
2377         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
2378
2379         // FIXME This check should not be needed. Perhaps issue an
2380         // error if it triggers.
2381         Layout const & style = inInset().forcePlainLayout() ?
2382                 bparams.documentClass().plainLayout() : *d->layout_;
2383
2384         if (!force && style.inpreamble)
2385                 return;
2386
2387         bool const allowcust = allowParagraphCustomization();
2388
2389         // Current base font for all inherited font changes, without any
2390         // change caused by an individual character, except for the language:
2391         // It is set to the language of the first character.
2392         // As long as we are in the label, this font is the base font of the
2393         // label. Before the first body character it is set to the base font
2394         // of the body.
2395         Font basefont;
2396
2397         // If there is an open font-encoding changing command (script wrapper),
2398         // alien_script is set to its name
2399         string alien_script;
2400         string script;
2401
2402         // Maybe we have to create a optional argument.
2403         pos_type body_pos = beginOfBody();
2404         unsigned int column = 0;
2405
2406         // If we are inside an inset, the real outerfont is local_font
2407         Font const real_outerfont = (runparams.local_font != nullptr)
2408                         ? Font(runparams.local_font->fontInfo()) : outerfont;
2409
2410         if (body_pos > 0) {
2411                 // the optional argument is kept in curly brackets in
2412                 // case it contains a ']'
2413                 // This is not strictly needed, but if this is changed it
2414                 // would be a file format change, and tex2lyx would need
2415                 // to be adjusted, since it unconditionally removes the
2416                 // braces when it parses \item.
2417                 os << "[{";
2418                 column += 2;
2419                 basefont = getLabelFont(bparams, real_outerfont);
2420         } else {
2421                 basefont = getLayoutFont(bparams, real_outerfont);
2422         }
2423
2424         // Which font is currently active?
2425         Font running_font(basefont);
2426         // Do we have an open font change?
2427         bool open_font = false;
2428
2429         Change runningChange = Change(Change::UNCHANGED);
2430
2431         Encoding const * const prev_encoding = runparams.encoding;
2432
2433         os.texrow().start(id(), 0);
2434
2435         // if the paragraph is empty, the loop will not be entered at all
2436         if (empty()) {
2437                 // For InTitle commands, we have already opened a group
2438                 // in output_latex::TeXOnePar.
2439                 if (style.isCommand() && !style.intitle) {
2440                         os << '{';
2441                         ++column;
2442                 }
2443                 if (!style.leftdelim().empty()) {
2444                         os << style.leftdelim();
2445                         column += style.leftdelim().size();
2446                 }
2447                 if (allowcust)
2448                         column += d->startTeXParParams(bparams, os, runparams);
2449         }
2450
2451         // Whether a \par can be issued for insets typeset inline with text.
2452         // Yes if greater than 0. This has to be static.
2453         THREAD_LOCAL_STATIC int parInline = 0;
2454
2455         for (pos_type i = 0; i < size(); ++i) {
2456                 // First char in paragraph or after label?
2457                 if (i == body_pos) {
2458                         if (body_pos > 0) {
2459                                 if (open_font) {
2460                                         bool needPar = false;
2461                                         column += running_font.latexWriteEndChanges(
2462                                                 os, bparams, runparams,
2463                                                 basefont, basefont, needPar);
2464                                         open_font = false;
2465                                 }
2466                                 basefont = getLayoutFont(bparams, real_outerfont);
2467                                 running_font = basefont;
2468
2469                                 column += Changes::latexMarkChange(os, bparams,
2470                                                 runningChange, Change(Change::UNCHANGED),
2471                                                 runparams);
2472                                 runningChange = Change(Change::UNCHANGED);
2473
2474                                 os << "}] ";
2475                                 column +=3;
2476                         }
2477                         // For InTitle commands, we have already opened a group
2478                         // in output_latex::TeXOnePar.
2479                         if (style.isCommand() && !style.intitle) {
2480                                 os << '{';
2481                                 ++column;
2482                         }
2483
2484                         if (!style.leftdelim().empty()) {
2485                                 os << style.leftdelim();
2486                                 column += style.leftdelim().size();
2487                         }
2488
2489                         if (allowcust)
2490                                 column += d->startTeXParParams(bparams, os,
2491                                                             runparams);
2492                 }
2493
2494                 runparams.wasDisplayMath = runparams.inDisplayMath;
2495                 runparams.inDisplayMath = false;
2496                 bool deleted_display_math = false;
2497                 Change const & change = runparams.inDeletedInset
2498                         ? runparams.changeOfDeletedInset : lookupChange(i);
2499
2500                 char_type const c = d->text_[i];
2501
2502                 // Check whether a display math inset follows
2503                 if (c == META_INSET
2504                     && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2505                         if (isDeleted(i))
2506                                 runparams.ctObject = getInset(i)->CtObject(runparams);
2507
2508                         InsetMath const * im = getInset(i)->asInsetMath();
2509                         if (im && im->asHullInset()
2510                             && im->asHullInset()->outerDisplay()) {
2511                                 runparams.inDisplayMath = true;
2512                                 // runparams.inDeletedInset will be set by
2513                                 // latexInset later, but we need this info
2514                                 // before it is called. On the other hand, we
2515                                 // cannot set it here because it is a counter.
2516                                 deleted_display_math = isDeleted(i);
2517                         }
2518                         if (bparams.output_changes && deleted_display_math
2519                             && runningChange == change
2520                             && change.type == Change::DELETED
2521                             && !os.afterParbreak()) {
2522                                 // A display math in the same paragraph follows.
2523                                 // We have to close and then reopen \lyxdeleted,
2524                                 // otherwise the math will be shifted up.
2525                                 OutputParams rp = runparams;
2526                                 if (open_font) {
2527                                         bool needPar = false;
2528                                         column += running_font.latexWriteEndChanges(
2529                                                 os, bparams, rp, basefont,
2530                                                 basefont, needPar);
2531                                         open_font = false;
2532                                 }
2533                                 basefont = (body_pos > i) ? getLabelFont(bparams, real_outerfont)
2534                                                           : getLayoutFont(bparams, real_outerfont);
2535                                 running_font = basefont;
2536                                 column += Changes::latexMarkChange(os, bparams,
2537                                         Change(Change::INSERTED), change, rp);
2538                         }
2539                 }
2540
2541                 if (bparams.output_changes && runningChange != change) {
2542                         if (!alien_script.empty()) {
2543                                 column += 1;
2544                                 os << "}";
2545                                 alien_script.clear();
2546                         }
2547                         if (open_font) {
2548                                 bool needPar = false;
2549                                 column += running_font.latexWriteEndChanges(
2550                                                 os, bparams, runparams,
2551                                                 basefont, basefont, needPar);
2552                                 open_font = false;
2553                         }
2554                         basefont = (body_pos > i) ? getLabelFont(bparams, real_outerfont)
2555                                                   : getLayoutFont(bparams, real_outerfont);
2556                         running_font = basefont;
2557                         column += Changes::latexMarkChange(os, bparams, runningChange,
2558                                                            change, runparams);
2559                         runningChange = change;
2560                 }
2561
2562                 // do not output text which is marked deleted
2563                 // if change tracking output is disabled
2564                 if (!bparams.output_changes && change.deleted()) {
2565                         continue;
2566                 }
2567
2568                 ++column;
2569
2570                 // Fully instantiated font
2571                 Font const current_font = getFont(bparams, i, real_outerfont);
2572
2573                 Font const last_font = running_font;
2574                 bool const in_ct_deletion = (bparams.output_changes
2575                                              && runningChange == change
2576                                              && change.type == Change::DELETED
2577                                              && !os.afterParbreak());
2578
2579                 // Do we need to close the previous font?
2580                 if (open_font &&
2581                     (current_font != running_font ||
2582                      current_font.language() != running_font.language()))
2583                 {
2584                         // ensure there is no open script-wrapper
2585                         if (!alien_script.empty()) {
2586                                 column += 1;
2587                                 os << "}";
2588                                 alien_script.clear();
2589                         }
2590                         bool needPar = false;
2591                         if (in_ct_deletion) {
2592                                 // We have to close and then reopen \lyxdeleted,
2593                                 // as strikeout needs to be on lowest level.
2594                                 os << '}';
2595                                 column += 1;
2596                         }
2597                         column += running_font.latexWriteEndChanges(
2598                                     os, bparams, runparams, basefont,
2599                                     (i == body_pos-1) ? basefont : current_font,
2600                                     needPar);
2601                         if (in_ct_deletion) {
2602                                 // We have to close and then reopen \lyxdeleted,
2603                                 // as strikeout needs to be on lowest level.
2604                                 OutputParams rp = runparams;
2605                                 column += Changes::latexMarkChange(os, bparams,
2606                                         Change(Change::UNCHANGED), Change(Change::DELETED), rp);
2607                         }
2608                         running_font = basefont;
2609                         open_font = false;
2610                 }
2611
2612                 // if necessary, close language environment before opening CJK
2613                 string const running_lang = running_font.language()->babel();
2614                 string const lang_end_command = lyxrc.language_command_end;
2615                 if (!lang_end_command.empty() && !bparams.useNonTeXFonts
2616                         && !running_lang.empty()
2617                         && running_lang == openLanguageName()
2618                         && current_font.language()->encoding()->package() == Encoding::CJK) {
2619                         string end_tag = subst(lang_end_command, "$$lang", running_lang);
2620                         os << from_ascii(end_tag);
2621                         column += end_tag.length();
2622                         popLanguageName();
2623                 }
2624
2625                 // Switch file encoding if necessary (and allowed)
2626                 if (!runparams.pass_thru && !style.pass_thru &&
2627                     runparams.encoding->package() != Encoding::none &&
2628                     current_font.language()->encoding()->package() != Encoding::none) {
2629                         pair<bool, int> const enc_switch =
2630                                 switchEncoding(os.os(), bparams, runparams,
2631                                         *(current_font.language()->encoding()));
2632                         if (enc_switch.first) {
2633                                 column += enc_switch.second;
2634                                 runparams.encoding = current_font.language()->encoding();
2635                         }
2636                 }
2637
2638                 // A display math inset inside an ulem command will be output
2639                 // as a box of width \linewidth, so we have to either disable
2640                 // indentation if the inset starts a paragraph, or start a new
2641                 // line to accommodate such box. This has to be done before
2642                 // writing any font changing commands.
2643                 if (runparams.inDisplayMath && !deleted_display_math
2644                     && runparams.inulemcmd) {
2645                         if (os.afterParbreak())
2646                                 os << "\\noindent";
2647                         else
2648                                 os << "\\\\\n";
2649                 }
2650
2651                 // Do we need to change font?
2652                 if ((current_font != running_font ||
2653                      current_font.language() != running_font.language())
2654                     && i != body_pos - 1)
2655                 {
2656                         if (in_ct_deletion) {
2657                                 // We have to close and then reopen \lyxdeleted,
2658                                 // as strikeout needs to be on lowest level.
2659                                 bool needPar = false;
2660                                 OutputParams rp = runparams;
2661                                 column += running_font.latexWriteEndChanges(
2662                                         os, bparams, rp, basefont,
2663                                         basefont, needPar);
2664                                 os << '}';
2665                                 column += 1;
2666                         }
2667                         otexstringstream ots;
2668                         bool const non_inherit_inset = (c == META_INSET && getInset(i) && !getInset(i)->inheritFont());
2669                         column += current_font.latexWriteStartChanges(ots, bparams,
2670                                                               runparams, basefont,
2671                                                               last_font, non_inherit_inset);
2672                         // Check again for display math in ulem commands as a
2673                         // font change may also occur just before a math inset.
2674                         if (runparams.inDisplayMath && !deleted_display_math
2675                             && runparams.inulemcmd) {
2676                                 if (os.afterParbreak())
2677                                         os << "\\noindent";
2678                                 else
2679                                         os << "\\\\\n";
2680                         }
2681                         running_font = current_font;
2682                         open_font = true;
2683                         docstring fontchange = ots.str();
2684                         os << fontchange;
2685                         // check whether the fontchange ends with a \\textcolor
2686                         // modifier and the text starts with a space. If so we
2687                         // need to add } in order to prevent \\textcolor from gobbling
2688                         // the space (bug 4473).
2689                         docstring const last_modifier = rsplit(fontchange, '\\');
2690                         if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2691                                 os << from_ascii("{}");
2692                         else if (ots.terminateCommand())
2693                                 os << termcmd;
2694                         if (in_ct_deletion) {
2695                                 // We have to close and then reopen \lyxdeleted,
2696                                 // as strikeout needs to be on lowest level.
2697                                 OutputParams rp = runparams;
2698                                 column += Changes::latexMarkChange(os, bparams,
2699                                         Change(Change::UNCHANGED), change, rp);
2700                         }
2701                 }
2702
2703                 // FIXME: think about end_pos implementation...
2704                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2705                         // FIXME: integrate this case in latexSpecialChar
2706                         // Do not print the separation of the optional argument
2707                         // if style.pass_thru is false. This works because
2708                         // latexSpecialChar ignores spaces if
2709                         // style.pass_thru is false.
2710                         if (i != body_pos - 1) {
2711                                 if (d->simpleTeXBlanks(bparams, runparams, os,
2712                                                 i, column, current_font, style)) {
2713                                         // A surrogate pair was output. We
2714                                         // must not call latexSpecialChar
2715                                         // in this iteration, since it would output
2716                                         // the combining character again.
2717                                         ++i;
2718                                         continue;
2719                                 }
2720                         }
2721                 }
2722
2723                 OutputParams rp = runparams;
2724                 rp.free_spacing = style.free_spacing;
2725                 rp.local_font = &current_font;
2726                 rp.intitle = style.intitle;
2727
2728                 // Two major modes:  LaTeX or plain
2729                 // Handle here those cases common to both modes
2730                 // and then split to handle the two modes separately.
2731                 if (c == META_INSET) {
2732                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2733                                 // Greyedout notes and, in general, all insets
2734                                 // with InsetLayout::isDisplay() == false,
2735                                 // are typeset inline with the text. So, we
2736                                 // can add a \par to the last paragraph of
2737                                 // such insets only if nothing else follows.
2738                                 bool incremented = false;
2739                                 Inset const * inset = getInset(i);
2740                                 InsetText const * textinset = inset
2741                                                         ? inset->asInsetText()
2742                                                         : nullptr;
2743                                 if (i + 1 == size() && textinset
2744                                     && !inset->getLayout().isDisplay()) {
2745                                         ParagraphList const & pars =
2746                                                 textinset->text().paragraphs();
2747                                         pit_type const pit = pars.size() - 1;
2748                                         Font const lastfont =
2749                                                 pit < 0 || pars[pit].empty()
2750                                                 ? pars[pit].getLayoutFont(
2751                                                                 bparams,
2752                                                                 real_outerfont)
2753                                                 : pars[pit].getFont(bparams,
2754                                                         pars[pit].size() - 1,
2755                                                         real_outerfont);
2756                                         if (lastfont.fontInfo().size() !=
2757                                             basefont.fontInfo().size()) {
2758                                                 ++parInline;
2759                                                 incremented = true;
2760                                         }
2761                                 }
2762                                 d->latexInset(bparams, os, rp, running_font,
2763                                                 basefont, real_outerfont, open_font,
2764                                                 runningChange, style, i, column);
2765                                 if (incremented)
2766                                         --parInline;
2767
2768                                 if (runparams.ctObject == OutputParams::CT_DISPLAYOBJECT
2769                                     || runparams.ctObject == OutputParams::CT_UDISPLAYOBJECT) {
2770                                         // Close \lyx*deleted and force its
2771                                         // reopening (if needed)
2772                                         os << '}';
2773                                         column++;
2774                                         runningChange = Change(Change::UNCHANGED);
2775                                         runparams.ctObject = OutputParams::CT_NORMAL;
2776                                 }
2777                         }
2778                 } else if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2779                         if (!bparams.useNonTeXFonts)
2780                           script = Encodings::isKnownScriptChar(c);
2781                         if (script != alien_script) {
2782                                 if (!alien_script.empty()) {
2783                                         os << "}";
2784                                         alien_script.clear();
2785                                 }
2786                                 string fontenc = running_font.language()->fontenc(bparams);
2787                                 if (!script.empty()
2788                                         && !Encodings::fontencSupportsScript(fontenc, script)) {
2789                                         column += script.length() + 2;
2790                                         os << "\\" << script << "{";
2791                                         alien_script = script;
2792                                 }
2793                         }
2794                         try {
2795                                 d->latexSpecialChar(os, bparams, rp, running_font,
2796                                                                         alien_script, style, i, end_pos, column);
2797                         } catch (EncodingException & e) {
2798                                 if (runparams.dryrun) {
2799                                         os << "<" << _("LyX Warning: ")
2800                                            << _("uncodable character") << " '";
2801                                         os.put(c);
2802                                         os << "'>";
2803                                 } else {
2804                                         // add location information and throw again.
2805                                         e.par_id = id();
2806                                         e.pos = i;
2807                                         throw(e);
2808                                 }
2809                         }
2810                 }
2811
2812                 // Set the encoding to that returned from latexSpecialChar (see
2813                 // comment for encoding member in OutputParams.h)
2814                 runparams.encoding = rp.encoding;
2815
2816                 // Also carry on the info on a closed ulem command for insets
2817                 // such as Note that do not produce any output, so that no
2818                 // command is ever executed but its opening was recorded.
2819                 runparams.inulemcmd = rp.inulemcmd;
2820
2821                 // These need to be passed upstream as well
2822                 runparams.need_maketitle = rp.need_maketitle;
2823                 runparams.have_maketitle = rp.have_maketitle;
2824
2825                 // And finally, pass the post_macros upstream
2826                 runparams.post_macro = rp.post_macro;
2827         }
2828
2829         // Close wrapper for alien script
2830         if (!alien_script.empty()) {
2831                 os << "}";
2832                 alien_script.clear();
2833         }
2834
2835         // If we have an open font definition, we have to close it
2836         if (open_font) {
2837                 // Make sure that \\par is done with the font of the last
2838                 // character if this has another size as the default.
2839                 // This is necessary because LaTeX (and LyX on the screen)
2840                 // calculates the space between the baselines according
2841                 // to this font. (Matthias)
2842                 //
2843                 // We must not change the font for the last paragraph
2844                 // of non-multipar insets, tabular cells or commands,
2845                 // since this produces unwanted whitespace.
2846
2847                 Font const font = empty()
2848                         ? getLayoutFont(bparams, real_outerfont)
2849                         : getFont(bparams, size() - 1, real_outerfont);
2850
2851                 InsetText const * textinset = inInset().asInsetText();
2852
2853                 bool const maintext = textinset
2854                         ? textinset->text().isMainText()
2855                         : false;
2856
2857                 size_t const numpars = textinset
2858                         ? textinset->text().paragraphs().size()
2859                         : 0;
2860
2861                 bool needPar = false;
2862
2863                 if (style.resfont.size() != font.fontInfo().size()
2864                     && (!runparams.isLastPar || maintext
2865                         || (numpars > 1 && d->ownerCode() != CELL_CODE
2866                             && (inInset().getLayout().isDisplay()
2867                                 || parInline)))
2868                     && !style.isCommand()) {
2869                         needPar = true;
2870                 }
2871 #ifdef FIXED_LANGUAGE_END_DETECTION
2872                 if (next_) {
2873                         running_font.latexWriteEndChanges(os, bparams,
2874                                         runparams, basefont,
2875                                         next_->getFont(bparams, 0, outerfont),
2876                                                        needPar);
2877                 } else {
2878                         running_font.latexWriteEndChanges(os, bparams,
2879                                         runparams, basefont, basefont, needPar);
2880                 }
2881 #else
2882 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2883 //FIXME: there as we start another \selectlanguage with the next paragraph if
2884 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2885                 running_font.latexWriteEndChanges(os, bparams, runparams,
2886                                 basefont, basefont, needPar);
2887 #endif
2888                 if (needPar) {
2889                         // The \par could not be inserted at the same nesting
2890                         // level of the font size change, so do it now.
2891                         os << "{\\" << font.latexSize() << "\\par}";
2892                 }
2893         }
2894
2895         column += Changes::latexMarkChange(os, bparams, runningChange,
2896                                            Change(Change::UNCHANGED), runparams);
2897
2898         // Needed if there is an optional argument but no contents.
2899         if (body_pos > 0 && body_pos == size()) {
2900                 os << "}]~";
2901         }
2902
2903         if (!style.rightdelim().empty()) {
2904                 os << style.rightdelim();
2905                 column += style.rightdelim().size();
2906         }
2907
2908         if (allowcust && d->endTeXParParams(bparams, os, runparams)
2909             && runparams.encoding != prev_encoding) {
2910                 runparams.encoding = prev_encoding;
2911                 os << setEncoding(prev_encoding->iconvName());
2912         }
2913
2914         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2915 }
2916
2917
2918 bool Paragraph::emptyTag() const
2919 {
2920         for (pos_type i = 0; i < size(); ++i) {
2921                 if (Inset const * inset = getInset(i)) {
2922                         InsetCode lyx_code = inset->lyxCode();
2923                         // FIXME testing like that is wrong. What is
2924                         // the intent?
2925                         if (lyx_code != TOC_CODE &&
2926                             lyx_code != INCLUDE_CODE &&
2927                             lyx_code != GRAPHICS_CODE &&
2928                             lyx_code != ERT_CODE &&
2929                             lyx_code != LISTINGS_CODE &&
2930                             lyx_code != FLOAT_CODE &&
2931                             lyx_code != TABULAR_CODE) {
2932                                 return false;
2933                         }
2934                 } else {
2935                         char_type c = d->text_[i];
2936                         if (c != ' ' && c != '\t')
2937                                 return false;
2938                 }
2939         }
2940         return true;
2941 }
2942
2943
2944 string Paragraph::getID(Buffer const &, OutputParams const &)
2945         const
2946 {
2947         for (pos_type i = 0; i < size(); ++i) {
2948                 if (Inset const * inset = getInset(i)) {
2949                         InsetCode lyx_code = inset->lyxCode();
2950                         if (lyx_code == LABEL_CODE) {
2951                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2952                                 docstring const & id = il->getParam("name");
2953                                 return "id='" + to_utf8(xml::cleanID(id)) + "'";
2954                         }
2955                 }
2956         }
2957         return string();
2958 }
2959
2960
2961 pos_type Paragraph::firstWordDocBook(XMLStream & xs, OutputParams const & runparams) const
2962 {
2963         pos_type i;
2964         for (i = 0; i < size(); ++i) {
2965                 if (Inset const * inset = getInset(i)) {
2966                         inset->docbook(xs, runparams);
2967                 } else {
2968                         char_type c = d->text_[i];
2969                         if (c == ' ')
2970                                 break;
2971                         xs << c;
2972                 }
2973         }
2974         return i;
2975 }
2976
2977
2978 pos_type Paragraph::firstWordLyXHTML(XMLStream & xs, OutputParams const & runparams)
2979         const
2980 {
2981         pos_type i;
2982         for (i = 0; i < size(); ++i) {
2983                 if (Inset const * inset = getInset(i)) {
2984                         inset->xhtml(xs, runparams);
2985                 } else {
2986                         char_type c = d->text_[i];
2987                         if (c == ' ')
2988                                 break;
2989                         xs << c;
2990                 }
2991         }
2992         return i;
2993 }
2994
2995
2996 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2997 {
2998         Font font_old;
2999         pos_type size = text_.size();
3000         for (pos_type i = initial; i < size; ++i) {
3001                 Font font = owner_->getFont(buf.params(), i, outerfont);
3002                 if (text_[i] == META_INSET)
3003                         return false;
3004                 if (i != initial && font != font_old)
3005                         return false;
3006                 font_old = font;
3007         }
3008
3009         return true;
3010 }
3011
3012
3013 namespace {
3014
3015 void doFontSwitchDocBook(vector<xml::FontTag> & tagsToOpen,
3016                   vector<xml::EndFontTag> & tagsToClose,
3017                   bool & flag, FontState curstate, xml::FontTypes type)
3018 {
3019         if (curstate == FONT_ON) {
3020                 tagsToOpen.push_back(docbookStartFontTag(type));
3021                 flag = true;
3022         } else if (flag) {
3023                 tagsToClose.push_back(docbookEndFontTag(type));
3024                 flag = false;
3025         }
3026 }
3027
3028 class OptionalFontType {
3029 public:
3030         bool has_value;
3031         xml::FontTypes ft;
3032
3033         OptionalFontType(): has_value(false), ft(xml::FT_EMPH) {} // A possible value at random for ft.
3034         OptionalFontType(xml::FontTypes ft): has_value(true), ft(ft) {}
3035 };
3036
3037 OptionalFontType fontShapeToXml(FontShape fs)
3038 {
3039         switch (fs) {
3040         case ITALIC_SHAPE:
3041                 return {xml::FT_ITALIC};
3042         case SLANTED_SHAPE:
3043                 return {xml::FT_SLANTED};
3044         case SMALLCAPS_SHAPE:
3045                 return {xml::FT_SMALLCAPS};
3046         case UP_SHAPE:
3047         case INHERIT_SHAPE:
3048                 return {};
3049         default:
3050                 // the other tags are for internal use
3051                 LATTEST(false);
3052                 return {};
3053         }
3054 }
3055
3056 OptionalFontType fontFamilyToXml(FontFamily fm)
3057 {
3058         switch (fm) {
3059         case ROMAN_FAMILY:
3060                 return {xml::FT_ROMAN};
3061                 break;
3062         case SANS_FAMILY:
3063                 return {xml::FT_SANS};
3064                 break;
3065         case TYPEWRITER_FAMILY:
3066                 return {xml::FT_TYPE};
3067                 break;
3068         case INHERIT_FAMILY:
3069                 return {};
3070         default:
3071                 // the other tags are for internal use
3072                 LATTEST(false);
3073                 return {};
3074         }
3075 }
3076
3077 OptionalFontType fontSizeToXml(FontSize fs)
3078 {
3079         switch (fs) {
3080         case TINY_SIZE:
3081                 return {xml::FT_SIZE_TINY};
3082         case SCRIPT_SIZE:
3083                 return {xml::FT_SIZE_SCRIPT};
3084         case FOOTNOTE_SIZE:
3085                 return {xml::FT_SIZE_FOOTNOTE};
3086         case SMALL_SIZE:
3087                 return {xml::FT_SIZE_SMALL};
3088         case LARGE_SIZE:
3089                 return {xml::FT_SIZE_LARGE};
3090         case LARGER_SIZE:
3091                 return {xml::FT_SIZE_LARGER};
3092         case LARGEST_SIZE:
3093                 return {xml::FT_SIZE_LARGEST};
3094         case HUGE_SIZE:
3095                 return {xml::FT_SIZE_HUGE};
3096         case HUGER_SIZE:
3097                 return {xml::FT_SIZE_HUGER};
3098         case INCREASE_SIZE:
3099                 return {xml::FT_SIZE_INCREASE};
3100         case DECREASE_SIZE:
3101                 return {xml::FT_SIZE_DECREASE};
3102         case INHERIT_SIZE:
3103         case NORMAL_SIZE:
3104                 return {};
3105         default:
3106                 // the other tags are for internal use
3107                 LATTEST(false);
3108                 return {};
3109         }
3110 }
3111
3112 struct DocBookFontState
3113 {
3114         // track whether we have opened these tags
3115         bool emph_flag = false;
3116         bool bold_flag = false;
3117         bool noun_flag = false;
3118         bool ubar_flag = false;
3119         bool dbar_flag = false;
3120         bool sout_flag = false;
3121         bool xout_flag = false;
3122         bool wave_flag = false;
3123         // shape tags
3124         bool shap_flag = false;
3125         // family tags
3126         bool faml_flag = false;
3127         // size tags
3128         bool size_flag = false;
3129
3130         FontShape  curr_fs   = INHERIT_SHAPE;
3131         FontFamily curr_fam  = INHERIT_FAMILY;
3132         FontSize   curr_size = INHERIT_SIZE;
3133 };
3134
3135 std::tuple<vector<xml::FontTag>, vector<xml::EndFontTag>> computeDocBookFontSwitch(FontInfo const & font_old,
3136                                                                                            Font const & font,
3137                                                                                            std::string const & default_family,
3138                                                                                            DocBookFontState fs)
3139 {
3140         vector<xml::FontTag> tagsToOpen;
3141         vector<xml::EndFontTag> tagsToClose;
3142
3143         // emphasis
3144         FontState curstate = font.fontInfo().emph();
3145         if (font_old.emph() != curstate)
3146                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.emph_flag, curstate, xml::FT_EMPH);
3147
3148         // noun
3149         curstate = font.fontInfo().noun();
3150         if (font_old.noun() != curstate)
3151                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.noun_flag, curstate, xml::FT_NOUN);
3152
3153         // underbar
3154         curstate = font.fontInfo().underbar();
3155         if (font_old.underbar() != curstate)
3156                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.ubar_flag, curstate, xml::FT_UBAR);
3157
3158         // strikeout
3159         curstate = font.fontInfo().strikeout();
3160         if (font_old.strikeout() != curstate)
3161                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.sout_flag, curstate, xml::FT_SOUT);
3162
3163         // double underbar
3164         curstate = font.fontInfo().uuline();
3165         if (font_old.uuline() != curstate)
3166                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.dbar_flag, curstate, xml::FT_DBAR);
3167
3168         // wavy line
3169         curstate = font.fontInfo().uwave();
3170         if (font_old.uwave() != curstate)
3171                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.wave_flag, curstate, xml::FT_WAVE);
3172
3173         // bold
3174         // a little hackish, but allows us to reuse what we have.
3175         curstate = (font.fontInfo().series() == BOLD_SERIES ? FONT_ON : FONT_OFF);
3176         if (font_old.series() != font.fontInfo().series())
3177                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.bold_flag, curstate, xml::FT_BOLD);
3178
3179         // Font shape
3180         fs.curr_fs = font.fontInfo().shape();
3181         FontShape old_fs = font_old.shape();
3182         if (old_fs != fs.curr_fs) {
3183                 if (fs.shap_flag) {
3184                         OptionalFontType tag = fontShapeToXml(old_fs);
3185                         if (tag.has_value) {
3186                                 tagsToClose.push_back(docbookEndFontTag(tag.ft));
3187                         }
3188                         fs.shap_flag = false;
3189                 }
3190
3191                 OptionalFontType tag = fontShapeToXml(fs.curr_fs);
3192                 if (tag.has_value) {
3193                         tagsToOpen.push_back(docbookStartFontTag(tag.ft));
3194                 }
3195         }
3196
3197         // Font family
3198         fs.curr_fam = font.fontInfo().family();
3199         FontFamily old_fam = font_old.family();
3200         if (old_fam != fs.curr_fam) {
3201                 if (fs.faml_flag) {
3202                         OptionalFontType tag = fontFamilyToXml(old_fam);
3203                         if (tag.has_value) {
3204                                 tagsToClose.push_back(docbookEndFontTag(tag.ft));
3205                         }
3206                         fs.faml_flag = false;
3207                 }
3208                 switch (fs.curr_fam) {
3209                         case ROMAN_FAMILY:
3210                                 // we will treat a "default" font family as roman, since we have
3211                                 // no other idea what to do.
3212                                 if (default_family != "rmdefault" && default_family != "default") {
3213                                         tagsToOpen.push_back(docbookStartFontTag(xml::FT_ROMAN));
3214                                         fs.faml_flag = true;
3215                                 }
3216                                 break;
3217                         case SANS_FAMILY:
3218                                 if (default_family != "sfdefault") {
3219                                         tagsToOpen.push_back(docbookStartFontTag(xml::FT_SANS));
3220                                         fs.faml_flag = true;
3221                                 }
3222                                 break;
3223                         case TYPEWRITER_FAMILY:
3224                                 if (default_family != "ttdefault") {
3225                                         tagsToOpen.push_back(docbookStartFontTag(xml::FT_TYPE));
3226                                         fs.faml_flag = true;
3227                                 }
3228                                 break;
3229                         case INHERIT_FAMILY:
3230                                 break;
3231                         default:
3232                                 // the other tags are for internal use
3233                                 LATTEST(false);
3234                                 break;
3235                 }
3236         }
3237
3238         // Font size
3239         fs.curr_size = font.fontInfo().size();
3240         FontSize old_size = font_old.size();
3241         if (old_size != fs.curr_size) {
3242                 if (fs.size_flag) {
3243                         OptionalFontType tag = fontSizeToXml(old_size);
3244                         if (tag.has_value) {
3245                                 tagsToClose.push_back(docbookEndFontTag(tag.ft));
3246                         }
3247                         fs.size_flag = false;
3248                 }
3249
3250                 OptionalFontType tag = fontSizeToXml(fs.curr_size);
3251                 if (tag.has_value) {
3252                         tagsToOpen.push_back(docbookStartFontTag(tag.ft));
3253                         fs.size_flag = true;
3254                 }
3255         }
3256
3257         return std::tuple(tagsToOpen, tagsToClose);
3258 }
3259
3260 }// anonymous namespace
3261
3262
3263 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
3264                                     XMLStream & xs,
3265                                     OutputParams const & runparams,
3266                                     Font const & outerfont,
3267                                     bool start_paragraph, bool close_paragraph,
3268                                     pos_type initial) const
3269 {
3270         // track whether we have opened these tags
3271         DocBookFontState fs;
3272
3273         if (start_paragraph)
3274                 xs.startDivision(allowEmpty());
3275
3276         Layout const & style = *d->layout_;
3277         FontInfo font_old =
3278                         style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
3279
3280         string const default_family =
3281                         buf.masterBuffer()->params().fonts_default_family;
3282
3283         vector<xml::FontTag> tagsToOpen;
3284         vector<xml::EndFontTag> tagsToClose;
3285
3286         // parsing main loop
3287         for (pos_type i = initial; i < size(); ++i) {
3288                 // let's not show deleted material in the output
3289                 if (isDeleted(i))
3290                         continue;
3291
3292                 Font const font = getFont(buf.masterBuffer()->params(), i, outerfont);
3293
3294                 if (start_paragraph) {
3295                         // Determine which tags should be opened or closed.
3296                         tie(tagsToOpen, tagsToClose) = computeDocBookFontSwitch(font_old, font, default_family, fs);
3297
3298                         // FIXME XHTML
3299                         // Other such tags? What about the other text ranges?
3300
3301                         vector<xml::EndFontTag>::const_iterator cit = tagsToClose.begin();
3302                         vector<xml::EndFontTag>::const_iterator cen = tagsToClose.end();
3303                         for (; cit != cen; ++cit)
3304                                 xs << *cit;
3305
3306                         vector<xml::FontTag>::const_iterator sit = tagsToOpen.begin();
3307                         vector<xml::FontTag>::const_iterator sen = tagsToOpen.end();
3308                         for (; sit != sen; ++sit)
3309                                 xs << *sit;
3310
3311                         tagsToClose.clear();
3312                         tagsToOpen.clear();
3313                 }
3314
3315                 if (Inset const * inset = getInset(i)) {
3316                         if (!runparams.for_toc || inset->isInToc()) {
3317                                 OutputParams np = runparams;
3318                                 np.local_font = &font;
3319                                 // If the paragraph has size 1, then we are in the "special
3320                                 // case" where we do not output the containing paragraph info.
3321                                 // This "special case" is defined in more details in output_docbook.cpp, makeParagraphs. The results
3322                                 // of that brittle logic is passed to this function through open_par.
3323                                 if (!inset->getLayout().htmlisblock() && size() != 1) // TODO: htmlisblock here too!
3324                                         np.docbook_in_par = true;
3325                                 inset->docbook(xs, np);
3326                         }
3327                 } else {
3328                         char_type c = getUChar(buf.masterBuffer()->params(), runparams, i);
3329                         xs << c;
3330                 }
3331                 font_old = font.fontInfo();
3332         }
3333
3334         // FIXME, this code is just imported from XHTML
3335         // I'm worried about what happens if a branch, say, is itself
3336         // wrapped in some font stuff. I think that will not work.
3337         xs.closeFontTags();
3338         if (runparams.docbook_in_listing)
3339                 xs << xml::CR();
3340         if (close_paragraph)
3341                 xs.endDivision();
3342 }
3343
3344
3345 namespace {
3346
3347 void doFontSwitchXHTML(vector<xml::FontTag> & tagsToOpen,
3348                   vector<xml::EndFontTag> & tagsToClose,
3349                   bool & flag, FontState curstate, xml::FontTypes type)
3350 {
3351         if (curstate == FONT_ON) {
3352                 tagsToOpen.push_back(xhtmlStartFontTag(type));
3353                 flag = true;
3354         } else if (flag) {
3355                 tagsToClose.push_back(xhtmlEndFontTag(type));
3356                 flag = false;
3357         }
3358 }
3359
3360 } // anonymous namespace
3361
3362
3363 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
3364                                     XMLStream & xs,
3365                                     OutputParams const & runparams,
3366                                     Font const & outerfont,
3367                                     bool start_paragraph, bool close_paragraph,
3368                                     pos_type initial) const
3369 {
3370         docstring retval;
3371
3372         // track whether we have opened these tags
3373         bool emph_flag = false;
3374         bool bold_flag = false;
3375         bool noun_flag = false;
3376         bool ubar_flag = false;
3377         bool dbar_flag = false;
3378         bool sout_flag = false;
3379         bool xout_flag = false;
3380         bool wave_flag = false;
3381         // shape tags
3382         bool shap_flag = false;
3383         // family tags
3384         bool faml_flag = false;
3385         // size tags
3386         bool size_flag = false;
3387
3388         Layout const & style = *d->layout_;
3389
3390         if (start_paragraph)
3391                 xs.startDivision(allowEmpty());
3392
3393         FontInfo font_old =
3394                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
3395
3396         FontShape  curr_fs   = INHERIT_SHAPE;
3397         FontFamily curr_fam  = INHERIT_FAMILY;
3398         FontSize   curr_size = INHERIT_SIZE;
3399
3400         string const default_family =
3401                 buf.masterBuffer()->params().fonts_default_family;
3402
3403         vector<xml::FontTag> tagsToOpen;
3404         vector<xml::EndFontTag> tagsToClose;
3405
3406         // parsing main loop
3407         for (pos_type i = initial; i < size(); ++i) {
3408                 // let's not show deleted material in the output
3409                 if (isDeleted(i))
3410                         continue;
3411
3412                 Font const font = getFont(buf.masterBuffer()->params(), i, outerfont);
3413
3414                 // emphasis
3415                 FontState curstate = font.fontInfo().emph();
3416                 if (font_old.emph() != curstate)
3417                         doFontSwitchXHTML(tagsToOpen, tagsToClose, emph_flag, curstate, xml::FT_EMPH);
3418
3419                 // noun
3420                 curstate = font.fontInfo().noun();
3421                 if (font_old.noun() != curstate)
3422                         doFontSwitchXHTML(tagsToOpen, tagsToClose, noun_flag, curstate, xml::FT_NOUN);
3423
3424                 // underbar
3425                 curstate = font.fontInfo().underbar();
3426                 if (font_old.underbar() != curstate)
3427                         doFontSwitchXHTML(tagsToOpen, tagsToClose, ubar_flag, curstate, xml::FT_UBAR);
3428
3429                 // strikeout
3430                 curstate = font.fontInfo().strikeout();
3431                 if (font_old.strikeout() != curstate)
3432                         doFontSwitchXHTML(tagsToOpen, tagsToClose, sout_flag, curstate, xml::FT_SOUT);
3433
3434                 // xout
3435                 curstate = font.fontInfo().xout();
3436                 if (font_old.xout() != curstate)
3437                         doFontSwitchXHTML(tagsToOpen, tagsToClose, xout_flag, curstate, xml::FT_XOUT);
3438
3439                 // double underbar
3440                 curstate = font.fontInfo().uuline();
3441                 if (font_old.uuline() != curstate)
3442                         doFontSwitchXHTML(tagsToOpen, tagsToClose, dbar_flag, curstate, xml::FT_DBAR);
3443
3444                 // wavy line
3445                 curstate = font.fontInfo().uwave();
3446                 if (font_old.uwave() != curstate)
3447                         doFontSwitchXHTML(tagsToOpen, tagsToClose, wave_flag, curstate, xml::FT_WAVE);
3448
3449                 // bold
3450                 // a little hackish, but allows us to reuse what we have.
3451                 curstate = (font.fontInfo().series() == BOLD_SERIES ? FONT_ON : FONT_OFF);
3452                 if (font_old.series() != font.fontInfo().series())
3453                         doFontSwitchXHTML(tagsToOpen, tagsToClose, bold_flag, curstate, xml::FT_BOLD);
3454
3455                 // Font shape
3456                 curr_fs = font.fontInfo().shape();
3457                 FontShape old_fs = font_old.shape();
3458                 if (old_fs != curr_fs) {
3459                         if (shap_flag) {
3460                                 switch (old_fs) {
3461                                 case ITALIC_SHAPE:
3462                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_ITALIC));
3463                                         break;
3464                                 case SLANTED_SHAPE:
3465                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SLANTED));
3466                                         break;
3467                                 case SMALLCAPS_SHAPE:
3468                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SMALLCAPS));
3469                                         break;
3470                                 case UP_SHAPE:
3471                                 case INHERIT_SHAPE:
3472                                         break;
3473                                 default:
3474                                         // the other tags are for internal use
3475                                         LATTEST(false);
3476                                         break;
3477                                 }
3478                                 shap_flag = false;
3479                         }
3480                         switch (curr_fs) {
3481                         case ITALIC_SHAPE:
3482                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_ITALIC));
3483                                 shap_flag = true;
3484                                 break;
3485                         case SLANTED_SHAPE:
3486                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SLANTED));
3487                                 shap_flag = true;
3488                                 break;
3489                         case SMALLCAPS_SHAPE:
3490                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SMALLCAPS));
3491                                 shap_flag = true;
3492                                 break;
3493                         case UP_SHAPE:
3494                         case INHERIT_SHAPE:
3495                                 break;
3496                         default:
3497                                 // the other tags are for internal use
3498                                 LATTEST(false);
3499                                 break;
3500                         }
3501                 }
3502
3503                 // Font family
3504                 curr_fam = font.fontInfo().family();
3505                 FontFamily old_fam = font_old.family();
3506                 if (old_fam != curr_fam) {
3507                         if (faml_flag) {
3508                                 switch (old_fam) {
3509                                 case ROMAN_FAMILY:
3510                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_ROMAN));
3511                                         break;
3512                                 case SANS_FAMILY:
3513                                     tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SANS));
3514                                     break;
3515                                 case TYPEWRITER_FAMILY:
3516                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_TYPE));
3517                                         break;
3518                                 case INHERIT_FAMILY:
3519                                         break;
3520                                 default:
3521                                         // the other tags are for internal use
3522                                         LATTEST(false);
3523                                         break;
3524                                 }
3525                                 faml_flag = false;
3526                         }
3527                         switch (curr_fam) {
3528                         case ROMAN_FAMILY:
3529                                 // we will treat a "default" font family as roman, since we have
3530                                 // no other idea what to do.
3531                                 if (default_family != "rmdefault" && default_family != "default") {
3532                                         tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_ROMAN));
3533                                         faml_flag = true;
3534                                 }
3535                                 break;
3536                         case SANS_FAMILY:
3537                                 if (default_family != "sfdefault") {
3538                                         tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SANS));
3539                                         faml_flag = true;
3540                                 }
3541                                 break;
3542                         case TYPEWRITER_FAMILY:
3543                                 if (default_family != "ttdefault") {
3544                                         tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_TYPE));
3545                                         faml_flag = true;
3546                                 }
3547                                 break;
3548                         case INHERIT_FAMILY:
3549                                 break;
3550                         default:
3551                                 // the other tags are for internal use
3552                                 LATTEST(false);
3553                                 break;
3554                         }
3555                 }
3556
3557                 // Font size
3558                 curr_size = font.fontInfo().size();
3559                 FontSize old_size = font_old.size();
3560                 if (old_size != curr_size) {
3561                         if (size_flag) {
3562                                 switch (old_size) {
3563                                 case TINY_SIZE:
3564                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_TINY));
3565                                         break;
3566                                 case SCRIPT_SIZE:
3567                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_SCRIPT));
3568                                         break;
3569                                 case FOOTNOTE_SIZE:
3570                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_FOOTNOTE));
3571                                         break;
3572                                 case SMALL_SIZE:
3573                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_SMALL));
3574                                         break;
3575                                 case LARGE_SIZE:
3576                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_LARGE));
3577                                         break;
3578                                 case LARGER_SIZE:
3579                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_LARGER));
3580                                         break;
3581                                 case LARGEST_SIZE:
3582                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_LARGEST));
3583                                         break;
3584                                 case HUGE_SIZE:
3585                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_HUGE));
3586                                         break;
3587                                 case HUGER_SIZE:
3588                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_HUGER));
3589                                         break;
3590                                 case INCREASE_SIZE:
3591                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_INCREASE));
3592                                         break;
3593                                 case DECREASE_SIZE:
3594                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_DECREASE));
3595                                         break;
3596                                 case INHERIT_SIZE:
3597                                 case NORMAL_SIZE:
3598                                         break;
3599                                 default:
3600                                         // the other tags are for internal use
3601                                         LATTEST(false);
3602                                         break;
3603                                 }
3604                                 size_flag = false;
3605                         }
3606                         switch (curr_size) {
3607                         case TINY_SIZE:
3608                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_TINY));
3609                                 size_flag = true;
3610                                 break;
3611                         case SCRIPT_SIZE:
3612                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_SCRIPT));
3613                                 size_flag = true;
3614                                 break;
3615                         case FOOTNOTE_SIZE:
3616                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_FOOTNOTE));
3617                                 size_flag = true;
3618                                 break;
3619                         case SMALL_SIZE:
3620                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_SMALL));
3621                                 size_flag = true;
3622                                 break;
3623                         case LARGE_SIZE:
3624                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_LARGE));
3625                                 size_flag = true;
3626                                 break;
3627                         case LARGER_SIZE:
3628                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_LARGER));
3629                                 size_flag = true;
3630                                 break;
3631                         case LARGEST_SIZE:
3632                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_LARGEST));
3633                                 size_flag = true;
3634                                 break;
3635                         case HUGE_SIZE:
3636                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_HUGE));
3637                                 size_flag = true;
3638                                 break;
3639                         case HUGER_SIZE:
3640                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_HUGER));
3641                                 size_flag = true;
3642                                 break;
3643                         case INCREASE_SIZE:
3644                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_INCREASE));
3645                                 size_flag = true;
3646                                 break;
3647                         case DECREASE_SIZE:
3648                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_DECREASE));
3649                                 size_flag = true;
3650                                 break;
3651                         case INHERIT_SIZE:
3652                         case NORMAL_SIZE:
3653                                 break;
3654                         default:
3655                                 // the other tags are for internal use
3656                                 LATTEST(false);
3657                                 break;
3658                         }
3659                 }
3660
3661                 // FIXME XHTML
3662                 // Other such tags? What about the other text ranges?
3663
3664                 vector<xml::EndFontTag>::const_iterator cit = tagsToClose.begin();
3665                 vector<xml::EndFontTag>::const_iterator cen = tagsToClose.end();
3666                 for (; cit != cen; ++cit)
3667                         xs << *cit;
3668
3669                 vector<xml::FontTag>::const_iterator sit = tagsToOpen.begin();
3670                 vector<xml::FontTag>::const_iterator sen = tagsToOpen.end();
3671                 for (; sit != sen; ++sit)
3672                         xs << *sit;
3673
3674                 tagsToClose.clear();
3675                 tagsToOpen.clear();
3676
3677                 Inset const * inset = getInset(i);
3678                 if (inset) {
3679                         if (!runparams.for_toc || inset->isInToc()) {
3680                                 OutputParams np = runparams;
3681                                 np.local_font = &font;
3682                                 // If the paragraph has size 1, then we are in the "special
3683                                 // case" where we do not output the containing paragraph info
3684                                 if (!inset->getLayout().htmlisblock() && size() != 1)
3685                                         np.html_in_par = true;
3686                                 retval += inset->xhtml(xs, np);
3687                         }
3688                 } else {
3689                         char_type c = getUChar(buf.masterBuffer()->params(),
3690                                                runparams, i);
3691                         if (c == ' ' && (style.free_spacing || runparams.free_spacing))
3692                                 xs << XMLStream::ESCAPE_NONE << "&nbsp;";
3693                         else
3694                                 xs << c;
3695                 }
3696                 font_old = font.fontInfo();
3697         }
3698
3699         // FIXME XHTML
3700         // I'm worried about what happens if a branch, say, is itself
3701         // wrapped in some font stuff. I think that will not work.
3702         xs.closeFontTags();
3703         if (close_paragraph)
3704                 xs.endDivision();
3705
3706         return retval;
3707 }
3708
3709
3710 bool Paragraph::isHfill(pos_type pos) const
3711 {
3712         Inset const * inset = getInset(pos);
3713         return inset && inset->isHfill();
3714 }
3715
3716
3717 bool Paragraph::isNewline(pos_type pos) const
3718 {
3719         // U+2028 LINE SEPARATOR
3720         // U+2029 PARAGRAPH SEPARATOR
3721         char_type const c = d->text_[pos];
3722         if (c == 0x2028 || c == 0x2029)
3723                 return true;
3724         Inset const * inset = getInset(pos);
3725         return inset && inset->lyxCode() == NEWLINE_CODE;
3726 }
3727
3728
3729 bool Paragraph::isEnvSeparator(pos_type pos) const
3730 {
3731         Inset const * inset = getInset(pos);
3732         return inset && inset->lyxCode() == SEPARATOR_CODE;
3733 }
3734
3735
3736 bool Paragraph::isLineSeparator(pos_type pos) const
3737 {
3738         char_type const c = d->text_[pos];
3739         if (isLineSeparatorChar(c))
3740                 return true;
3741         Inset const * inset = getInset(pos);
3742         return inset && inset->isLineSeparator();
3743 }
3744
3745
3746 bool Paragraph::isWordSeparator(pos_type pos, bool const ignore_deleted) const
3747 {
3748         if (pos == size())
3749                 return true;
3750         if (ignore_deleted && isDeleted(pos))
3751                 return false;
3752         if (Inset const * inset = getInset(pos))
3753                 return !inset->isLetter();
3754         // if we have a hard hyphen (no en- or emdash) or apostrophe
3755         // we pass this to the spell checker
3756         // FIXME: this method is subject to change, visit
3757         // https://bugzilla.mozilla.org/show_bug.cgi?id=355178
3758         // to get an impression how complex this is.
3759         if (isHardHyphenOrApostrophe(pos))
3760                 return false;
3761         char_type const c = d->text_[pos];
3762         // We want to pass the escape chars to the spellchecker
3763         docstring const escape_chars = from_utf8(lyxrc.spellchecker_esc_chars);
3764         return !isLetterChar(c) && !isDigitASCII(c) && !contains(escape_chars, c);
3765 }
3766
3767
3768 bool Paragraph::isHardHyphenOrApostrophe(pos_type pos) const
3769 {
3770         pos_type const psize = size();
3771         if (pos >= psize)
3772                 return false;
3773         char_type const c = d->text_[pos];
3774         if (c != '-' && c != '\'')
3775                 return false;
3776         int nextpos = pos + 1;
3777         int prevpos = pos > 0 ? pos - 1 : 0;
3778         if ((nextpos == psize || isSpace(nextpos))
3779                 && (pos == 0 || isSpace(prevpos)))
3780                 return false;
3781         return true;
3782 }
3783
3784
3785 bool Paragraph::needsCProtection(bool const fragile) const
3786 {
3787         // first check the layout of the paragraph, but only in insets
3788         InsetText const * textinset = inInset().asInsetText();
3789         bool const maintext = textinset
3790                 ? textinset->text().isMainText()
3791                 : false;
3792
3793         if (!maintext && layout().needcprotect) {
3794                 // Environments need cprotection regardless the content
3795                 if (layout().latextype == LATEX_ENVIRONMENT)
3796                         return true;
3797
3798                 // Commands need cprotection if they contain specific chars
3799                 int const nchars_escape = 9;
3800                 static char_type const chars_escape[nchars_escape] = {
3801                         '&', '_', '$', '%', '#', '^', '{', '}', '\\'};
3802
3803                 docstring const pars = asString();
3804                 for (int k = 0; k < nchars_escape; k++) {
3805                         if (contains(pars, chars_escape[k]))
3806                                 return true;
3807                 }
3808         }
3809
3810         // now check whether we have insets that need cprotection
3811         pos_type size = pos_type(d->text_.size());
3812         for (pos_type i = 0; i < size; ++i) {
3813                 if (!isInset(i))
3814                         continue;
3815                 Inset const * ins = getInset(i);
3816                 if (ins->needsCProtection(maintext, fragile))
3817                         return true;
3818                 if (ins->getLayout().latextype() == InsetLayout::ENVIRONMENT)
3819                         // Environments need cprotection regardless the content
3820                         return true;
3821                 // Now check math environments
3822                 InsetMath const * im = getInset(i)->asInsetMath();
3823                 if (!im || im->cell(0).empty())
3824                         continue;
3825                 switch(im->cell(0)[0]->lyxCode()) {
3826                 case MATH_AMSARRAY_CODE:
3827                 case MATH_SUBSTACK_CODE:
3828                 case MATH_ENV_CODE:
3829                 case MATH_XYMATRIX_CODE:
3830                         // these need cprotection
3831                         return true;
3832                 default:
3833                         break;
3834                 }
3835         }
3836
3837         return false;
3838 }
3839
3840
3841 FontSpan const & Paragraph::getSpellRange(pos_type pos) const
3842 {
3843         return d->speller_state_.getRange(pos);
3844 }
3845
3846
3847 bool Paragraph::isChar(pos_type pos) const
3848 {
3849         if (Inset const * inset = getInset(pos))
3850                 return inset->isChar();
3851         char_type const c = d->text_[pos];
3852         return !isLetterChar(c) && !isDigitASCII(c) && !lyx::isSpace(c);
3853 }
3854
3855
3856 bool Paragraph::isSpace(pos_type pos) const
3857 {
3858         if (Inset const * inset = getInset(pos))
3859                 return inset->isSpace();
3860         char_type const c = d->text_[pos];
3861         return lyx::isSpace(c);
3862 }
3863
3864
3865 Language const *
3866 Paragraph::getParLanguage(BufferParams const & bparams) const
3867 {
3868         if (!empty())
3869                 return getFirstFontSettings(bparams).language();
3870         // FIXME: we should check the prev par as well (Lgb)
3871         return bparams.language;
3872 }
3873
3874
3875 bool Paragraph::isRTL(BufferParams const & bparams) const
3876 {
3877         return getParLanguage(bparams)->rightToLeft()
3878                 && !inInset().getLayout().forceLTR();
3879 }
3880
3881
3882 void Paragraph::changeLanguage(BufferParams const & bparams,
3883                                Language const * from, Language const * to)
3884 {
3885         // change language including dummy font change at the end
3886         for (pos_type i = 0; i <= size(); ++i) {
3887                 Font font = getFontSettings(bparams, i);
3888                 if (font.language() == from) {
3889                         font.setLanguage(to);
3890                         setFont(i, font);
3891                         d->requestSpellCheck(i);
3892                 }
3893         }
3894 }
3895
3896
3897 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
3898 {
3899         Language const * doc_language = bparams.language;
3900         for (auto const & f : d->fontlist_)
3901                 if (f.font().language() != ignore_language &&
3902                     f.font().language() != latex_language &&
3903                     f.font().language() != doc_language)
3904                         return true;
3905         return false;
3906 }
3907
3908
3909 void Paragraph::getLanguages(std::set<Language const *> & langs) const
3910 {
3911         for (auto const & f : d->fontlist_) {
3912                 Language const * lang = f.font().language();
3913                 if (lang != ignore_language &&
3914                     lang != latex_language)
3915                         langs.insert(lang);
3916         }
3917 }
3918
3919
3920 docstring Paragraph::asString(int options) const
3921 {
3922         return asString(0, size(), options);
3923 }
3924
3925
3926 docstring Paragraph::asString(pos_type beg, pos_type end, int options, const OutputParams *runparams) const
3927 {
3928         odocstringstream os;
3929
3930         if (beg == 0
3931             && options & AS_STR_LABEL
3932             && !d->params_.labelString().empty())
3933                 os << d->params_.labelString() << ' ';
3934
3935         for (pos_type i = beg; i < end; ++i) {
3936                 if ((options & AS_STR_SKIPDELETE) && isDeleted(i))
3937                         continue;
3938                 char_type const c = d->text_[i];
3939                 if (isPrintable(c) || c == '\t'
3940                     || (c == '\n' && (options & AS_STR_NEWLINES)))
3941                         os.put(c);
3942                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
3943                         if (c == META_INSET && (options & AS_STR_PLAINTEXT)) {
3944                                 LASSERT(runparams != nullptr, return docstring());
3945                                 getInset(i)->plaintext(os, *runparams);
3946                         } else {
3947                                 getInset(i)->toString(os);
3948                         }
3949                 }
3950         }
3951
3952         return os.str();
3953 }
3954
3955
3956 void Paragraph::forOutliner(docstring & os, size_t const maxlen,
3957                             bool const shorten, bool const label) const
3958 {
3959         size_t tmplen = shorten ? maxlen + 1 : maxlen;
3960         if (label && !labelString().empty())
3961                 os += labelString() + ' ';
3962         if (!layout().isTocCaption())
3963                 return;
3964         for (pos_type i = 0; i < size() && os.length() < tmplen; ++i) {
3965                 if (isDeleted(i))
3966                         continue;
3967                 char_type const c = d->text_[i];
3968                 if (isPrintable(c))
3969                         os += c;
3970                 else if (c == META_INSET)
3971                         getInset(i)->forOutliner(os, tmplen, false);
3972         }
3973         if (shorten)
3974                 Text::shortenForOutliner(os, maxlen);
3975 }
3976
3977
3978 void Paragraph::setInsetOwner(Inset const * inset)
3979 {
3980         d->inset_owner_ = inset;
3981 }
3982
3983
3984 int Paragraph::id() const
3985 {
3986         return d->id_;
3987 }
3988
3989
3990 void Paragraph::setId(int id)
3991 {
3992         d->id_ = id;
3993 }
3994
3995
3996 Layout const & Paragraph::layout() const
3997 {
3998         return *d->layout_;
3999 }
4000
4001
4002 void Paragraph::setLayout(Layout const & layout)
4003 {
4004         d->layout_ = &layout;
4005 }
4006
4007
4008 void Paragraph::setDefaultLayout(DocumentClass const & tc)
4009 {
4010         setLayout(tc.defaultLayout());
4011 }
4012
4013
4014 void Paragraph::setPlainLayout(DocumentClass const & tc)
4015 {
4016         setLayout(tc.plainLayout());
4017 }
4018
4019
4020 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
4021 {
4022         if (usePlainLayout())
4023                 setPlainLayout(tclass);
4024         else
4025                 setDefaultLayout(tclass);
4026 }
4027
4028
4029 Inset const & Paragraph::inInset() const
4030 {
4031         LBUFERR(d->inset_owner_);
4032         return *d->inset_owner_;
4033 }
4034
4035
4036 ParagraphParameters & Paragraph::params()
4037 {
4038         return d->params_;
4039 }
4040
4041
4042 ParagraphParameters const & Paragraph::params() const
4043 {
4044         return d->params_;
4045 }
4046
4047
4048 bool Paragraph::isFreeSpacing() const
4049 {
4050         if (d->layout_->free_spacing)
4051                 return true;
4052         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
4053 }
4054
4055
4056 bool Paragraph::allowEmpty() const
4057 {
4058         if (d->layout_->keepempty)
4059                 return true;
4060         return d->inset_owner_ && d->inset_owner_->allowEmpty();
4061 }
4062
4063
4064 bool Paragraph::brokenBiblio() const
4065 {
4066         // There is a problem if there is no bibitem at position 0 in
4067         // paragraphs that need one, if there is another bibitem in the
4068         // paragraph or if this paragraph is not supposed to have
4069         // a bibitem inset at all.
4070         return ((d->layout_->labeltype == LABEL_BIBLIO
4071                 && (d->insetlist_.find(BIBITEM_CODE) != 0
4072                     || d->insetlist_.find(BIBITEM_CODE, 1) > 0))
4073                 || (d->layout_->labeltype != LABEL_BIBLIO
4074                     && d->insetlist_.find(BIBITEM_CODE) != -1));
4075 }
4076
4077
4078 int Paragraph::fixBiblio(Buffer const & buffer)
4079 {
4080         // FIXME: when there was already an inset at 0, the return value is 1,
4081         // which does not tell whether another inset has been remove; the
4082         // cursor cannot be correctly updated.
4083
4084         bool const track_changes = buffer.params().track_changes;
4085         int bibitem_pos = d->insetlist_.find(BIBITEM_CODE);
4086
4087         // The case where paragraph is not BIBLIO
4088         if (d->layout_->labeltype != LABEL_BIBLIO) {
4089                 if (bibitem_pos == -1)
4090                         // No InsetBibitem => OK
4091                         return 0;
4092                 // There is an InsetBibitem: remove it!
4093                 d->insetlist_.release(bibitem_pos);
4094                 eraseChar(bibitem_pos, track_changes);
4095                 return (bibitem_pos == 0) ? -1 : -bibitem_pos;
4096         }
4097
4098         bool const hasbibitem0 = bibitem_pos == 0;
4099         if (hasbibitem0) {
4100                 bibitem_pos = d->insetlist_.find(BIBITEM_CODE, 1);
4101                 // There was an InsetBibitem at pos 0,
4102                 // and no other one => OK
4103                 if (bibitem_pos == -1)
4104                         return 0;
4105                 // there is a bibitem at the 0 position, but since
4106                 // there is a second one, we copy the second on the
4107                 // first. We're assuming there are at most two of
4108                 // these, which there should be.
4109                 // FIXME: why does it make sense to do that rather
4110                 // than keep the first? (JMarc)
4111                 Inset * inset = releaseInset(bibitem_pos);
4112                 d->insetlist_.begin()->inset = inset;
4113                 return -bibitem_pos;
4114         }
4115
4116         // We need to create an inset at the beginning
4117         Inset * inset = nullptr;
4118         if (bibitem_pos > 0) {
4119                 // there was one somewhere in the paragraph, let's move it
4120                 inset = d->insetlist_.release(bibitem_pos);
4121                 eraseChar(bibitem_pos, track_changes);
4122         } else
4123                 // make a fresh one
4124                 inset = new InsetBibitem(const_cast<Buffer *>(&buffer),
4125                                          InsetCommandParams(BIBITEM_CODE));
4126
4127         Font font(inherit_font, buffer.params().language);
4128         insertInset(0, inset, font, Change(track_changes ? Change::INSERTED
4129                                                    : Change::UNCHANGED));
4130
4131         // This is needed to get the counters right
4132         buffer.updateBuffer();
4133         return 1;
4134 }
4135
4136
4137 void Paragraph::checkAuthors(AuthorList const & authorList)
4138 {
4139         d->changes_.checkAuthors(authorList);
4140 }
4141
4142
4143 bool Paragraph::isChanged(pos_type pos) const
4144 {
4145         return lookupChange(pos).changed();
4146 }
4147
4148
4149 bool Paragraph::isInserted(pos_type pos) const
4150 {
4151         return lookupChange(pos).inserted();
4152 }
4153
4154
4155 bool Paragraph::isDeleted(pos_type pos) const
4156 {
4157         return lookupChange(pos).deleted();
4158 }
4159
4160
4161 InsetList const & Paragraph::insetList() const
4162 {
4163         return d->insetlist_;
4164 }
4165
4166
4167 void Paragraph::setInsetBuffers(Buffer & b)
4168 {
4169         d->insetlist_.setBuffer(b);
4170 }
4171
4172
4173 void Paragraph::resetBuffer()
4174 {
4175         d->insetlist_.resetBuffer();
4176 }
4177
4178
4179 Inset * Paragraph::releaseInset(pos_type pos)
4180 {
4181         Inset * inset = d->insetlist_.release(pos);
4182         /// does not honour change tracking!
4183         eraseChar(pos, false);
4184         return inset;
4185 }
4186
4187
4188 Inset * Paragraph::getInset(pos_type pos)
4189 {
4190         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
4191                  ? d->insetlist_.get(pos) : nullptr;
4192 }
4193
4194
4195 Inset const * Paragraph::getInset(pos_type pos) const
4196 {
4197         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
4198                  ? d->insetlist_.get(pos) : nullptr;
4199 }
4200
4201
4202 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
4203                 pos_type & right, TextCase action)
4204 {
4205         // process sequences of modified characters; in change
4206         // tracking mode, this approach results in much better
4207         // usability than changing case on a char-by-char basis
4208         // We also need to track the current font, since font
4209         // changes within sequences can occur.
4210         vector<pair<char_type, Font> > changes;
4211
4212         bool const trackChanges = bparams.track_changes;
4213
4214         bool capitalize = true;
4215
4216         for (; pos < right; ++pos) {
4217                 char_type oldChar = d->text_[pos];
4218                 char_type newChar = oldChar;
4219
4220                 // ignore insets and don't play with deleted text!
4221                 if (oldChar != META_INSET && !isDeleted(pos)) {
4222                         switch (action) {
4223                                 case text_lowercase:
4224                                         newChar = lowercase(oldChar);
4225                                         break;
4226                                 case text_capitalization:
4227                                         if (capitalize) {
4228                                                 newChar = uppercase(oldChar);
4229                                                 capitalize = false;
4230                                         }
4231                                         break;
4232                                 case text_uppercase:
4233                                         newChar = uppercase(oldChar);
4234                                         break;
4235                         }
4236                 }
4237
4238                 if (isWordSeparator(pos) || isDeleted(pos)) {
4239                         // permit capitalization again
4240                         capitalize = true;
4241                 }
4242
4243                 if (oldChar != newChar) {
4244                         changes.push_back(make_pair(newChar, getFontSettings(bparams, pos)));
4245                         if (pos != right - 1)
4246                                 continue;
4247                         // step behind the changing area
4248                         pos++;
4249                 }
4250
4251                 int erasePos = pos - changes.size();
4252                 for (size_t i = 0; i < changes.size(); i++) {
4253                         insertChar(pos, changes[i].first,
4254                                    changes[i].second,
4255                                    trackChanges);
4256                         if (!eraseChar(erasePos, trackChanges)) {
4257                                 ++erasePos;
4258                                 ++pos; // advance
4259                                 ++right; // expand selection
4260                         }
4261                 }
4262                 changes.clear();
4263         }
4264 }
4265
4266
4267 int Paragraph::find(docstring const & str, bool cs, bool mw,
4268                 pos_type start_pos, bool del) const
4269 {
4270         pos_type pos = start_pos;
4271         int const strsize = str.length();
4272         int i = 0;
4273         pos_type const parsize = d->text_.size();
4274         for (i = 0; i < strsize && pos < parsize; ++i, ++pos) {
4275                 // Ignore "invisible" letters such as ligature breaks
4276                 // and hyphenation chars while searching
4277                 while (pos < parsize - 1 && isInset(pos)) {
4278                         odocstringstream os;
4279                         getInset(pos)->toString(os);
4280                         if (!getInset(pos)->isLetter() || !os.str().empty())
4281                                 break;
4282                         pos++;
4283                 }
4284                 if (cs && str[i] != d->text_[pos])
4285                         break;
4286                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos]))
4287                         break;
4288                 if (!del && isDeleted(pos))
4289                         break;
4290         }
4291
4292         if (i != strsize)
4293                 return 0;
4294
4295         // if necessary, check whether string matches word
4296         if (mw) {
4297                 if (start_pos > 0 && !isWordSeparator(start_pos - 1))
4298                         return 0;
4299                 if (pos < parsize
4300                         && !isWordSeparator(pos))
4301                         return 0;
4302         }
4303
4304         return pos - start_pos;
4305 }
4306
4307
4308 char_type Paragraph::getChar(pos_type pos) const
4309 {
4310         return d->text_[pos];
4311 }
4312
4313
4314 pos_type Paragraph::size() const
4315 {
4316         return d->text_.size();
4317 }
4318
4319
4320 bool Paragraph::empty() const
4321 {
4322         return d->text_.empty();
4323 }
4324
4325
4326 bool Paragraph::isInset(pos_type pos) const
4327 {
4328         return d->text_[pos] == META_INSET;
4329 }
4330
4331
4332 bool Paragraph::isSeparator(pos_type pos) const
4333 {
4334         //FIXME: Are we sure this can be the only separator?
4335         return d->text_[pos] == ' ';
4336 }
4337
4338
4339 void Paragraph::deregisterWords()
4340 {
4341         Private::LangWordsMap::const_iterator itl = d->words_.begin();
4342         Private::LangWordsMap::const_iterator ite = d->words_.end();
4343         for (; itl != ite; ++itl) {
4344                 WordList & wl = theWordList(itl->first);
4345                 Private::Words::const_iterator it = (itl->second).begin();
4346                 Private::Words::const_iterator et = (itl->second).end();
4347                 for (; it != et; ++it)
4348                         wl.remove(*it);
4349         }
4350         d->words_.clear();
4351 }
4352
4353
4354 void Paragraph::locateWord(pos_type & from, pos_type & to,
4355         word_location const loc, bool const ignore_deleted) const
4356 {
4357         switch (loc) {
4358         case WHOLE_WORD_STRICT:
4359                 if (from == 0 || from == size()
4360                     || isWordSeparator(from, ignore_deleted)
4361                     || isWordSeparator(from - 1, ignore_deleted)) {
4362                         to = from;
4363                         return;
4364                 }
4365                 // fall through
4366
4367         case WHOLE_WORD:
4368                 // If we are already at the beginning of a word, do nothing
4369                 if (!from || isWordSeparator(from - 1, ignore_deleted))
4370                         break;
4371                 // fall through
4372
4373         case PREVIOUS_WORD:
4374                 // always move the cursor to the beginning of previous word
4375                 while (from && !isWordSeparator(from - 1, ignore_deleted))
4376                         --from;
4377                 break;
4378         case NEXT_WORD:
4379                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
4380                 break;
4381         case PARTIAL_WORD:
4382                 // no need to move the 'from' cursor
4383                 break;
4384         }
4385         to = from;
4386         while (to < size() && !isWordSeparator(to, ignore_deleted))
4387                 ++to;
4388 }
4389
4390
4391 void Paragraph::collectWords()
4392 {
4393         for (pos_type pos = 0; pos < size(); ++pos) {
4394                 if (isWordSeparator(pos))
4395                         continue;
4396                 pos_type from = pos;
4397                 locateWord(from, pos, WHOLE_WORD);
4398                 // Work around MSVC warning: The statement
4399                 // if (pos < from + lyxrc.completion_minlength)
4400                 // triggers a signed vs. unsigned warning.
4401                 // I don't know why this happens, it could be a MSVC bug, or
4402                 // related to LLP64 (windows) vs. LP64 (unix) programming
4403                 // model, or the C++ standard might be ambigous in the section
4404                 // defining the "usual arithmetic conversions". However, using
4405                 // a temporary variable is safe and works on all compilers.
4406                 pos_type const endpos = from + lyxrc.completion_minlength;
4407                 if (pos < endpos)
4408                         continue;
4409                 FontList::const_iterator cit = d->fontlist_.fontIterator(from);
4410                 if (cit == d->fontlist_.end())
4411                         return;
4412                 Language const * lang = cit->font().language();
4413                 docstring const word = asString(from, pos, AS_STR_NONE);
4414                 d->words_[lang->lang()].insert(word);
4415         }
4416 }
4417
4418
4419 void Paragraph::registerWords()
4420 {
4421         Private::LangWordsMap::const_iterator itl = d->words_.begin();
4422         Private::LangWordsMap::const_iterator ite = d->words_.end();
4423         for (; itl != ite; ++itl) {
4424                 WordList & wl = theWordList(itl->first);
4425                 Private::Words::const_iterator it = (itl->second).begin();
4426                 Private::Words::const_iterator et = (itl->second).end();
4427                 for (; it != et; ++it)
4428                         wl.insert(*it);
4429         }
4430 }
4431
4432
4433 void Paragraph::updateWords()
4434 {
4435         deregisterWords();
4436         collectWords();
4437         registerWords();
4438 }
4439
4440
4441 void Paragraph::Private::appendSkipPosition(SkipPositions & skips, pos_type const pos) const
4442 {
4443         SkipPositionsIterator begin = skips.begin();
4444         SkipPositions::iterator end = skips.end();
4445         if (pos > 0 && begin < end) {
4446                 --end;
4447                 if (end->last == pos - 1) {
4448                         end->last = pos;
4449                         return;
4450                 }
4451         }
4452         skips.insert(end, FontSpan(pos, pos));
4453 }
4454
4455
4456 Language * Paragraph::Private::locateSpellRange(
4457         pos_type & from, pos_type & to,
4458         SkipPositions & skips) const
4459 {
4460         // skip leading white space
4461         while (from < to && owner_->isWordSeparator(from))
4462                 ++from;
4463         // don't check empty range
4464         if (from >= to)
4465                 return nullptr;
4466         // get current language
4467         Language * lang = getSpellLanguage(from);
4468         pos_type last = from;
4469         bool samelang = true;
4470         bool sameinset = true;
4471         while (last < to && samelang && sameinset) {
4472                 // hop to end of word
4473                 while (last < to && !owner_->isWordSeparator(last)) {
4474                         if (owner_->getInset(last)) {
4475                                 appendSkipPosition(skips, last);
4476                         } else if (owner_->isDeleted(last)) {
4477                                 appendSkipPosition(skips, last);
4478                         }
4479                         ++last;
4480                 }
4481                 // hop to next word while checking for insets
4482                 while (sameinset && last < to && owner_->isWordSeparator(last)) {
4483                         if (Inset const * inset = owner_->getInset(last))
4484                                 sameinset = inset->isChar() && inset->isLetter();
4485                         if (sameinset && owner_->isDeleted(last)) {
4486                                 appendSkipPosition(skips, last);
4487                         }
4488                         if (sameinset)
4489                                 last++;
4490                 }
4491                 if (sameinset && last < to) {
4492                         // now check for language change
4493                         samelang = lang == getSpellLanguage(last);
4494                 }
4495         }
4496         // if language change detected backstep is needed
4497         if (!samelang)
4498                 --last;
4499         to = last;
4500         return lang;
4501 }
4502
4503
4504 Language * Paragraph::Private::getSpellLanguage(pos_type const from) const
4505 {
4506         Language * lang =
4507                 const_cast<Language *>(owner_->getFontSettings(
4508                         inset_owner_->buffer().params(), from).language());
4509         if (lang == inset_owner_->buffer().params().language
4510                 && !lyxrc.spellchecker_alt_lang.empty()) {
4511                 string lang_code;
4512                 string const lang_variety =
4513                         split(lyxrc.spellchecker_alt_lang, lang_code, '-');
4514                 lang->setCode(lang_code);
4515                 lang->setVariety(lang_variety);
4516         }
4517         return lang;
4518 }
4519
4520
4521 void Paragraph::requestSpellCheck(pos_type pos)
4522 {
4523         d->requestSpellCheck(pos);
4524 }
4525
4526
4527 bool Paragraph::needsSpellCheck() const
4528 {
4529         SpellChecker::ChangeNumber speller_change_number = 0;
4530         if (theSpellChecker())
4531                 speller_change_number = theSpellChecker()->changeNumber();
4532         if (speller_change_number > d->speller_state_.currentChangeNumber()) {
4533                 d->speller_state_.needsCompleteRefresh(speller_change_number);
4534         }
4535         return d->needsSpellCheck();
4536 }
4537
4538
4539 bool Paragraph::Private::ignoreWord(docstring const & word) const
4540 {
4541         // Ignore words with digits
4542         // FIXME: make this customizable
4543         // (note that some checkers ignore words with digits by default)
4544         docstring::const_iterator cit = word.begin();
4545         docstring::const_iterator const end = word.end();
4546         for (; cit != end; ++cit) {
4547                 if (isNumber((*cit)))
4548                         return true;
4549         }
4550         return false;
4551 }
4552
4553
4554 SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
4555         WordLangTuple & wl, docstring_list & suggestions,
4556         bool do_suggestion, bool check_learned) const
4557 {
4558         SpellChecker::Result result = SpellChecker::WORD_OK;
4559         SpellChecker * speller = theSpellChecker();
4560         if (!speller)
4561                 return result;
4562
4563         if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
4564                 return result;
4565
4566         locateWord(from, to, WHOLE_WORD, true);
4567         if (from == to || from >= size())
4568                 return result;
4569
4570         docstring word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
4571         Language * lang = d->getSpellLanguage(from);
4572
4573         if (getFontSettings(d->inset_owner_->buffer().params(), from).fontInfo().nospellcheck() == FONT_ON)
4574                 return result;
4575
4576         wl = WordLangTuple(word, lang);
4577
4578         if (word.empty())
4579                 return result;
4580
4581         if (needsSpellCheck() || check_learned) {
4582                 pos_type end = to;
4583                 if (!d->ignoreWord(word)) {
4584                         bool const trailing_dot = to < size() && d->text_[to] == '.';
4585                         result = speller->check(wl);
4586                         if (SpellChecker::misspelled(result) && trailing_dot) {
4587                                 wl = WordLangTuple(word.append(from_ascii(".")), lang);
4588                                 result = speller->check(wl);
4589                                 if (!SpellChecker::misspelled(result)) {
4590                                         LYXERR(Debug::GUI, "misspelled word is correct with dot: \"" <<
4591                                            word << "\" [" <<
4592                                            from << ".." << to << "]");
4593                                 } else {
4594                                         // spell check with dot appended failed too
4595                                         // restore original word/lang value
4596                                         word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
4597                                         wl = WordLangTuple(word, lang);
4598                                 }
4599                         }
4600                 }
4601                 if (!SpellChecker::misspelled(result)) {
4602                         // area up to the begin of the next word is not misspelled
4603                         while (end < size() && isWordSeparator(end))
4604                                 ++end;
4605                 }
4606                 d->setMisspelled(from, end, result);
4607         } else {
4608                 result = d->speller_state_.getState(from);
4609         }
4610
4611         if (do_suggestion)
4612                 suggestions.clear();
4613
4614         if (SpellChecker::misspelled(result)) {
4615                 LYXERR(Debug::GUI, "misspelled word: \"" <<
4616                            word << "\" [" <<
4617                            from << ".." << to << "]");
4618                 if (do_suggestion)
4619                         speller->suggest(wl, suggestions);
4620         }
4621         return result;
4622 }
4623
4624
4625 void Paragraph::anonymize()
4626 {
4627         // This is a very crude anonymization for now
4628         for (char_type & c : d->text_)
4629                 if (isLetterChar(c) || isNumber(c))
4630                         c = 'a';
4631 }
4632
4633
4634 void Paragraph::Private::markMisspelledWords(
4635         pos_type const & first, pos_type const & last,
4636         SpellChecker::Result result,
4637         docstring const & word,
4638         SkipPositions const & skips)
4639 {
4640         if (!SpellChecker::misspelled(result)) {
4641                 setMisspelled(first, last, SpellChecker::WORD_OK);
4642                 return;
4643         }
4644         int snext = first;
4645         SpellChecker * speller = theSpellChecker();
4646         // locate and enumerate the error positions
4647         int nerrors = speller->numMisspelledWords();
4648         int numskipped = 0;
4649         SkipPositionsIterator it = skips.begin();
4650         SkipPositionsIterator et = skips.end();
4651         for (int index = 0; index < nerrors; ++index) {
4652                 int wstart;
4653                 int wlen = 0;
4654                 speller->misspelledWord(index, wstart, wlen);
4655                 /// should not happen if speller supports range checks
4656                 if (!wlen) continue;
4657                 docstring const misspelled = word.substr(wstart, wlen);
4658                 wstart += first + numskipped;
4659                 if (snext < wstart) {
4660                         /// mark the range of correct spelling
4661                         numskipped += countSkips(it, et, wstart);
4662                         setMisspelled(snext,
4663                                 wstart - 1, SpellChecker::WORD_OK);
4664                 }
4665                 snext = wstart + wlen;
4666                 numskipped += countSkips(it, et, snext);
4667                 /// mark the range of misspelling
4668                 setMisspelled(wstart, snext, result);
4669                 LYXERR(Debug::GUI, "misspelled word: \"" <<
4670                            misspelled << "\" [" <<
4671                            wstart << ".." << (snext-1) << "]");
4672                 ++snext;
4673         }
4674         if (snext <= last) {
4675                 /// mark the range of correct spelling at end
4676                 setMisspelled(snext, last, SpellChecker::WORD_OK);
4677         }
4678 }
4679
4680
4681 void Paragraph::spellCheck() const
4682 {
4683         SpellChecker * speller = theSpellChecker();
4684         if (!speller || empty() ||!needsSpellCheck())
4685                 return;
4686         pos_type start;
4687         pos_type endpos;
4688         d->rangeOfSpellCheck(start, endpos);
4689         if (speller->canCheckParagraph()) {
4690                 // loop until we leave the range
4691                 for (pos_type first = start; first < endpos; ) {
4692                         pos_type last = endpos;
4693                         Private::SkipPositions skips;
4694                         Language * lang = d->locateSpellRange(first, last, skips);
4695                         if (first >= endpos)
4696                                 break;
4697                         // start the spell checker on the unit of meaning
4698                         docstring word = asString(first, last, AS_STR_INSETS + AS_STR_SKIPDELETE);
4699                         WordLangTuple wl = WordLangTuple(word, lang);
4700                         SpellChecker::Result result = word.size() ?
4701                                 speller->check(wl) : SpellChecker::WORD_OK;
4702                         d->markMisspelledWords(first, last, result, word, skips);
4703                         first = ++last;
4704                 }
4705         } else {
4706                 static docstring_list suggestions;
4707                 pos_type to = endpos;
4708                 while (start < endpos) {
4709                         WordLangTuple wl;
4710                         spellCheck(start, to, wl, suggestions, false);
4711                         start = to + 1;
4712                 }
4713         }
4714         d->readySpellCheck();
4715 }
4716
4717
4718 bool Paragraph::isMisspelled(pos_type pos, bool check_boundary) const
4719 {
4720         bool result = SpellChecker::misspelled(d->speller_state_.getState(pos));
4721         if (result || pos <= 0 || pos > size())
4722                 return result;
4723         if (check_boundary && (pos == size() || isWordSeparator(pos)))
4724                 result = SpellChecker::misspelled(d->speller_state_.getState(pos - 1));
4725         return result;
4726 }
4727
4728
4729 string Paragraph::magicLabel() const
4730 {
4731         stringstream ss;
4732         ss << "magicparlabel-" << id();
4733         return ss.str();
4734 }
4735
4736
4737 } // namespace lyx