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 /////////////////////////////////////////////////////////////////////
  88 //
  89 // SpellResultRange
  90 //
  91 /////////////////////////////////////////////////////////////////////
  92
  93 class SpellResultRange {
  94 public:
  95         SpellResultRange(FontSpan range, SpellChecker::Result result)
  96         : range_(range), result_(result)
  97         {}
  98         ///
  99         FontSpan const & range() const { return range_; }
 100         ///
 101         void range(FontSpan const & r) { range_ = r; }
 102         ///
 103         SpellChecker::Result result() const { return result_; }
 104         ///
 105         void result(SpellChecker::Result r) { result_ = r; }
 106         ///
 107         bool contains(pos_type pos) const { return range_.contains(pos); }
 108         ///
 109         bool covered(FontSpan const & r) const
 110         {
 111                 // 1. first of new range inside current range or
 112                 // 2. last of new range inside current range or
 113                 // 3. first of current range inside new range or
 114                 // 4. last of current range inside new range
 115                 //FIXME: is this the same as !range_.intersect(r).empty() ?
 116                 return range_.contains(r.first) || range_.contains(r.last) ||
 117                         r.contains(range_.first) || r.contains(range_.last);
 118         }
 119         ///
 120         void shift(pos_type pos, int offset)
 121         {
 122                 if (range_.first > pos) {
 123                         range_.first += offset;
 124                         range_.last += offset;
 125                 } else if (range_.last >= pos) {
 126                         range_.last += offset;
 127                 }
 128         }
 129 private:
 130         FontSpan range_ ;
 131         SpellChecker::Result result_ ;
 132 };
 133
 134
 135 /////////////////////////////////////////////////////////////////////
 136 //
 137 // SpellCheckerState
 138 //
 139 /////////////////////////////////////////////////////////////////////
 140
 141 class SpellCheckerState {
 142 public:
 143         SpellCheckerState()
 144         {
 145                 needs_refresh_ = true;
 146                 current_change_number_ = 0;
 147         }
 148
 149         void setRange(FontSpan const & fp, SpellChecker::Result state)
 150         {
 151                 Ranges result;
 152                 RangesIterator et = ranges_.end();
 153                 RangesIterator it = ranges_.begin();
 154                 for (; it != et; ++it) {
 155                         if (!it->covered(fp))
 156                                 result.push_back(SpellResultRange(it->range(), it->result()));
 157                         else if (state == SpellChecker::WORD_OK) {
 158                                 // trim or split the current misspelled range
 159                                 // store misspelled ranges only
 160                                 FontSpan range = it->range();
 161                                 if (fp.first > range.first) {
 162                                         // misspelled area in front of WORD_OK
 163                                         range.last = fp.first - 1;
 164                                         result.push_back(SpellResultRange(range, it->result()));
 165                                         range = it->range();
 166                                 }
 167                                 if (fp.last < range.last) {
 168                                         // misspelled area after WORD_OK range
 169                                         range.first = fp.last + 1;
 170                                         result.push_back(SpellResultRange(range, it->result()));
 171                                 }
 172                         }
 173                 }
 174                 ranges_ = result;
 175                 if (state != SpellChecker::WORD_OK)
 176                         ranges_.push_back(SpellResultRange(fp, state));
 177         }
 178
 179         void increasePosAfterPos(pos_type pos)
 180         {
 181                 correctRangesAfterPos(pos, 1);
 182                 needsRefresh(pos);
 183         }
 184
 185         void decreasePosAfterPos(pos_type pos)
 186         {
 187                 correctRangesAfterPos(pos, -1);
 188                 needsRefresh(pos);
 189         }
 190
 191         void refreshLast(pos_type pos)
 192         {
 193                 if (pos < refresh_.last)
 194                         refresh_.last = pos;
 195         }
 196
 197         SpellChecker::Result getState(pos_type pos) const
 198         {
 199                 SpellChecker::Result result = SpellChecker::WORD_OK;
 200                 RangesIterator et = ranges_.end();
 201                 RangesIterator it = ranges_.begin();
 202                 for (; it != et; ++it) {
 203                         if(it->contains(pos)) {
 204                                 return it->result();
 205                         }
 206                 }
 207                 return result;
 208         }
 209
 210         FontSpan const & getRange(pos_type pos) const
 211         {
 212                 /// empty span to indicate mismatch
 213                 static FontSpan empty_;
 214                 RangesIterator et = ranges_.end();
 215                 RangesIterator it = ranges_.begin();
 216                 for (; it != et; ++it) {
 217                         if(it->contains(pos)) {
 218                                 return it->range();
 219                         }
 220                 }
 221                 return empty_;
 222         }
 223
 224         bool needsRefresh() const
 225         {
 226                 return needs_refresh_;
 227         }
 228
 229         SpellChecker::ChangeNumber currentChangeNumber() const
 230         {
 231                 return current_change_number_;
 232         }
 233
 234         void refreshRange(pos_type & first, pos_type & last) const
 235         {
 236                 first = refresh_.first;
 237                 last = refresh_.last;
 238         }
 239
 240         void needsRefresh(pos_type pos)
 241         {
 242                 if (needs_refresh_ && pos != -1) {
 243                         if (pos < refresh_.first)
 244                                 refresh_.first = pos;
 245                         if (pos > refresh_.last)
 246                                 refresh_.last = pos;
 247                 } else if (pos != -1) {
 248                         // init request check for neighbour positions too
 249                         refresh_.first = pos > 0 ? pos - 1 : 0;
 250                         // no need for special end of paragraph check
 251                         refresh_.last = pos + 1;
 252                 }
 253                 needs_refresh_ = pos != -1;
 254         }
 255
 256         void needsCompleteRefresh(SpellChecker::ChangeNumber change_number)
 257         {
 258                 needs_refresh_ = true;
 259                 refresh_.first = 0;
 260                 refresh_.last = -1;
 261                 current_change_number_ = change_number;
 262         }
 263 private:
 264         typedef vector<SpellResultRange> Ranges;
 265         typedef Ranges::const_iterator RangesIterator;
 266         Ranges ranges_;
 267         /// the area of the paragraph with pending spell check
 268         FontSpan refresh_;
 269         /// spell state cache version number
 270         SpellChecker::ChangeNumber current_change_number_;
 271         bool needs_refresh_;
 272
 273
 274         void correctRangesAfterPos(pos_type pos, int offset)
 275         {
 276                 RangesIterator et = ranges_.end();
 277                 Ranges::iterator it = ranges_.begin();
 278                 for (; it != et; ++it) {
 279                         it->shift(pos, offset);
 280                 }
 281         }
 282
 283 };
 284
 285 /////////////////////////////////////////////////////////////////////
 286 //
 287 // Paragraph::Private
 288 //
 289 /////////////////////////////////////////////////////////////////////
 290
 291 class Paragraph::Private
 292 {
 293         // Enforce our own "copy" constructor
 294         Private(Private const &) = delete;
 295         Private & operator=(Private const &) = delete;
 296         // Unique ID generator
 297         static int make_id();
 298 public:
 299         ///
 300         Private(Paragraph * owner, Layout const & layout);
 301         /// "Copy constructor"
 302         Private(Private const &, Paragraph * owner);
 303         /// Copy constructor from \p beg  to \p end
 304         Private(Private const &, Paragraph * owner, pos_type beg, pos_type end);
 305
 306         ///
 307         void insertChar(pos_type pos, char_type c, Change const & change);
 308
 309         /// Output the surrogate pair formed by \p c and \p next to \p os.
 310         /// \return the number of characters written.
 311         int latexSurrogatePair(BufferParams const &, otexstream & os,
 312                                char_type c, char_type next,
 313                                OutputParams const &);
 314
 315         /// Output a space in appropriate formatting (or a surrogate pair
 316         /// if the next character is a combining character).
 317         /// \return whether a surrogate pair was output.
 318         bool simpleTeXBlanks(BufferParams const &,
 319                              OutputParams const &,
 320                              otexstream &,
 321                              pos_type i,
 322                              unsigned int & column,
 323                              Font const & font,
 324                              Layout const & style);
 325
 326         /// This could go to ParagraphParameters if we want to.
 327         int startTeXParParams(BufferParams const &, otexstream &,
 328                               OutputParams const &) const;
 329
 330         /// This could go to ParagraphParameters if we want to.
 331         bool endTeXParParams(BufferParams const &, otexstream &,
 332                              OutputParams const &) const;
 333
 334         ///
 335         void latexInset(BufferParams const &,
 336                                    otexstream &,
 337                                    OutputParams &,
 338                                    Font & running_font,
 339                                    Font & basefont,
 340                                    Font const & outerfont,
 341                                    bool & open_font,
 342                                    Change & running_change,
 343                                    Layout const & style,
 344                                    pos_type & i,
 345                                    unsigned int & column);
 346
 347         ///
 348         void latexSpecialChar(
 349                                    otexstream & os,
 350                                    BufferParams const & bparams,
 351                                    OutputParams const & runparams,
 352                                    Font const & running_font,
 353                                    string & alien_script,
 354                                    Layout const & style,
 355                                    pos_type & i,
 356                                    pos_type end_pos,
 357                                    unsigned int & column);
 358
 359         ///
 360         bool latexSpecialT1(
 361                 char_type const c,
 362                 otexstream & os,
 363                 pos_type i,
 364                 unsigned int & column);
 365         ///
 366         bool latexSpecialTU(
 367                 char_type const c,
 368                 otexstream & os,
 369                 pos_type i,
 370                 unsigned int & column);
 371         ///
 372         bool latexSpecialT3(
 373                 char_type const c,
 374                 otexstream & os,
 375                 pos_type i,
 376                 unsigned int & column);
 377
 378         ///
 379         void validate(LaTeXFeatures & features) const;
 380
 381         /// Checks if the paragraph contains only text and no inset or font change.
 382         bool onlyText(Buffer const & buf, Font const & outerfont,
 383                       pos_type initial) const;
 384
 385         /// a vector of speller skip positions
 386         typedef vector<FontSpan> SkipPositions;
 387         typedef SkipPositions::const_iterator SkipPositionsIterator;
 388
 389         void appendSkipPosition(SkipPositions & skips, pos_type const pos) const;
 390
 391         Language * getSpellLanguage(pos_type const from) const;
 392
 393         Language * locateSpellRange(pos_type & from, pos_type & to,
 394                                     SkipPositions & skips) const;
 395
 396         bool hasSpellerChange() const
 397         {
 398                 SpellChecker::ChangeNumber speller_change_number = 0;
 399                 if (theSpellChecker())
 400                         speller_change_number = theSpellChecker()->changeNumber();
 401                 return speller_change_number > speller_state_.currentChangeNumber();
 402         }
 403
 404         bool ignoreWord(docstring const & word) const ;
 405
 406         void setMisspelled(pos_type from, pos_type to, SpellChecker::Result state)
 407         {
 408                 pos_type textsize = owner_->size();
 409                 // check for sane arguments
 410                 if (to <= from || from >= textsize)
 411                         return;
 412                 FontSpan fp = FontSpan(from, to - 1);
 413                 speller_state_.setRange(fp, state);
 414         }
 415
 416         void requestSpellCheck(pos_type pos)
 417         {
 418                 if (pos == -1)
 419                         speller_state_.needsCompleteRefresh(speller_state_.currentChangeNumber());
 420                 else
 421                         speller_state_.needsRefresh(pos);
 422         }
 423
 424         void readySpellCheck()
 425         {
 426                 speller_state_.needsRefresh(-1);
 427         }
 428
 429         bool needsSpellCheck() const
 430         {
 431                 return speller_state_.needsRefresh();
 432         }
 433
 434         void rangeOfSpellCheck(pos_type & first, pos_type & last) const
 435         {
 436                 speller_state_.refreshRange(first, last);
 437                 if (last == -1) {
 438                         last = owner_->size();
 439                         return;
 440                 }
 441                 pos_type endpos = last;
 442                 owner_->locateWord(first, endpos, WHOLE_WORD, true);
 443                 if (endpos < last) {
 444                         endpos = last;
 445                         owner_->locateWord(last, endpos, WHOLE_WORD, true);
 446                 }
 447                 last = endpos;
 448         }
 449
 450         int countSkips(SkipPositionsIterator & it, SkipPositionsIterator const et,
 451                             int & start) const
 452         {
 453                 int numskips = 0;
 454                 while (it != et && it->first < start) {
 455                         long skip = it->last - it->first + 1;
 456                         start += skip;
 457                         numskips += skip;
 458                         ++it;
 459                 }
 460                 return numskips;
 461         }
 462
 463         void markMisspelledWords(pos_type const & first, pos_type const & last,
 464                                                          SpellChecker::Result result,
 465                                                          docstring const & word,
 466                                                          SkipPositions const & skips);
 467
 468         InsetCode ownerCode() const
 469         {
 470                 return inset_owner_ ? inset_owner_->lyxCode() : NO_CODE;
 471         }
 472
 473         /// Which Paragraph owns us?
 474         Paragraph * owner_;
 475
 476         /// In which Inset?
 477         Inset const * inset_owner_;
 478
 479         ///
 480         FontList fontlist_;
 481
 482         ///
 483         ParagraphParameters params_;
 484
 485         /// for recording and looking up changes
 486         Changes changes_;
 487
 488         ///
 489         InsetList insetlist_;
 490
 491         /// end of label
 492         pos_type begin_of_body_;
 493
 494         typedef docstring TextContainer;
 495         ///
 496         TextContainer text_;
 497
 498         typedef set<docstring> Words;
 499         typedef map<string, Words> LangWordsMap;
 500         ///
 501         LangWordsMap words_;
 502         ///
 503         Layout const * layout_;
 504         ///
 505         SpellCheckerState speller_state_;
 506         ///
 507         int id_;
 508 };
 509
 510
 511 Paragraph::Private::Private(Paragraph * owner, Layout const & layout)
 512         : owner_(owner), inset_owner_(nullptr), begin_of_body_(0), layout_(&layout), id_(-1)
 513 {
 514         text_.reserve(100);
 515 }
 516
 517
 518 //static
 519 int Paragraph::Private::make_id()
 520 {
 521         // The id is unique per session across buffers because it is used in
 522         // LFUN_PARAGRAPH_GOTO to switch to a different buffer, for instance in the
 523         // outliner.
 524         // (thread-safe)
 525         static int next_id(0);
 526         return next_id++;
 527 }
 528
 529
 530 Paragraph::Private::Private(Private const & p, Paragraph * owner)
 531         : owner_(owner), inset_owner_(p.inset_owner_), fontlist_(p.fontlist_),
 532           params_(p.params_), changes_(p.changes_), insetlist_(p.insetlist_),
 533           begin_of_body_(p.begin_of_body_), text_(p.text_), words_(p.words_),
 534           layout_(p.layout_), id_(make_id())
 535 {
 536         requestSpellCheck(p.text_.size());
 537 }
 538
 539
 540 Paragraph::Private::Private(Private const & p, Paragraph * owner,
 541         pos_type beg, pos_type end)
 542         : owner_(owner), inset_owner_(p.inset_owner_),
 543           params_(p.params_), changes_(p.changes_),
 544           insetlist_(p.insetlist_, beg, end),
 545           begin_of_body_(p.begin_of_body_), words_(p.words_),
 546           layout_(p.layout_), id_(make_id())
 547 {
 548         if (beg >= pos_type(p.text_.size()))
 549                 return;
 550         text_ = p.text_.substr(beg, end - beg);
 551
 552         FontList::const_iterator fcit = fontlist_.begin();
 553         FontList::const_iterator fend = fontlist_.end();
 554         for (; fcit != fend; ++fcit) {
 555                 if (fcit->pos() < beg)
 556                         continue;
 557                 if (fcit->pos() >= end) {
 558                         // Add last entry in the fontlist_.
 559                         fontlist_.set(text_.size() - 1, fcit->font());
 560                         break;
 561                 }
 562                 // Add a new entry in the fontlist_.
 563                 fontlist_.set(fcit->pos() - beg, fcit->font());
 564         }
 565         requestSpellCheck(p.text_.size());
 566 }
 567
 568
 569 void Paragraph::addChangesToToc(DocIterator const & cdit, Buffer const & buf,
 570                                 bool output_active, TocBackend & backend) const
 571 {
 572         d->changes_.addToToc(cdit, buf, output_active, backend);
 573 }
 574
 575
 576 bool Paragraph::isDeleted(pos_type start, pos_type end) const
 577 {
 578         LASSERT(start >= 0 && start <= size(), return false);
 579         LASSERT(end > start && end <= size() + 1, return false);
 580
 581         return d->changes_.isDeleted(start, end);
 582 }
 583
 584
 585 bool Paragraph::isChanged(pos_type start, pos_type end) const
 586 {
 587         LASSERT(start >= 0 && start <= size(), return false);
 588         LASSERT(end > start && end <= size() + 1, return false);
 589
 590         return d->changes_.isChanged(start, end);
 591 }
 592
 593 // FIXME: Ideally the diverse isChanged() methods should account for that!
 594 bool Paragraph::hasChangedInsets(pos_type start, pos_type end) const
 595 {
 596         LASSERT(start >= 0 && start <= size(), return false);
 597         LASSERT(end > start && end <= size() + 1, return false);
 598
 599         for (auto const & icit : d->insetlist_) {
 600                 if (icit.pos < start)
 601                         continue;
 602                 if (icit.pos >= end)
 603                         break;
 604                 if (icit.inset && icit.inset->isChanged())
 605                         return true;
 606         }
 607         return false;
 608 }
 609
 610 bool Paragraph::isChanged() const
 611 {
 612         return d->changes_.isChanged();
 613 }
 614
 615
 616 bool Paragraph::isMergedOnEndOfParDeletion(bool trackChanges) const
 617 {
 618         // keep the logic here in sync with the logic of eraseChars()
 619         if (!trackChanges)
 620                 return true;
 621
 622         Change const & change = d->changes_.lookup(size());
 623         return change.inserted() && change.currentAuthor();
 624 }
 625
 626 Change Paragraph::parEndChange() const
 627 {
 628         return d->changes_.lookup(size());
 629 }
 630
 631
 632 void Paragraph::setChange(Change const & change)
 633 {
 634         // beware of the imaginary end-of-par character!
 635         d->changes_.set(change, 0, size() + 1);
 636
 637         /*
 638          * Propagate the change recursively - but not in case of DELETED!
 639          *
 640          * Imagine that your co-author makes changes in an existing inset. He
 641          * sends your document to you and you come to the conclusion that the
 642          * inset should go completely. If you erase it, LyX must not delete all
 643          * text within the inset. Otherwise, the change tracked insertions of
 644          * your co-author get lost and there is no way to restore them later.
 645          *
 646          * Conclusion: An inset's content should remain untouched if you delete it
 647          */
 648
 649         if (!change.deleted()) {
 650                 for (pos_type pos = 0; pos < size(); ++pos) {
 651                         if (Inset * inset = getInset(pos))
 652                                 inset->setChange(change);
 653                 }
 654         }
 655 }
 656
 657
 658 void Paragraph::setChange(pos_type pos, Change const & change)
 659 {
 660         LASSERT(pos >= 0 && pos <= size(), return);
 661         d->changes_.set(change, pos);
 662
 663         // see comment in setChange(Change const &) above
 664         if (!change.deleted() && pos < size())
 665                 if (Inset * inset = getInset(pos))
 666                         inset->setChange(change);
 667 }
 668
 669
 670 Change const & Paragraph::lookupChange(pos_type pos) const
 671 {
 672         LBUFERR(pos >= 0 && pos <= size());
 673         return d->changes_.lookup(pos);
 674 }
 675
 676
 677 void Paragraph::acceptChanges(pos_type start, pos_type end)
 678 {
 679         LASSERT(start >= 0 && start <= size(), return);
 680         LASSERT(end > start && end <= size() + 1, return);
 681
 682         for (pos_type pos = start; pos < end; ++pos) {
 683                 switch (lookupChange(pos).type) {
 684                         case Change::UNCHANGED:
 685                                 // accept changes in nested inset
 686                                 if (Inset * inset = getInset(pos))
 687                                         inset->acceptChanges();
 688                                 break;
 689
 690                         case Change::INSERTED:
 691                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 692                                 // also accept changes in nested inset
 693                                 if (Inset * inset = getInset(pos))
 694                                         inset->acceptChanges();
 695                                 break;
 696
 697                         case Change::DELETED:
 698                                 // Suppress access to non-existent
 699                                 // "end-of-paragraph char"
 700                                 if (pos < size()) {
 701                                         eraseChar(pos, false);
 702                                         --end;
 703                                         --pos;
 704                                 }
 705                                 break;
 706                 }
 707
 708         }
 709 }
 710
 711
 712 void Paragraph::rejectChanges(pos_type start, pos_type end)
 713 {
 714         LASSERT(start >= 0 && start <= size(), return);
 715         LASSERT(end > start && end <= size() + 1, return);
 716
 717         for (pos_type pos = start; pos < end; ++pos) {
 718                 switch (lookupChange(pos).type) {
 719                         case Change::UNCHANGED:
 720                                 // reject changes in nested inset
 721                                 if (Inset * inset = getInset(pos))
 722                                                 inset->rejectChanges();
 723                                 break;
 724
 725                         case Change::INSERTED:
 726                                 // Suppress access to non-existent
 727                                 // "end-of-paragraph char"
 728                                 if (pos < size()) {
 729                                         eraseChar(pos, false);
 730                                         --end;
 731                                         --pos;
 732                                 }
 733                                 break;
 734
 735                         case Change::DELETED:
 736                                 d->changes_.set(Change(Change::UNCHANGED), pos);
 737
 738                                 // Do NOT reject changes within a deleted inset!
 739                                 // There may be insertions of a co-author inside of it!
 740
 741                                 break;
 742                 }
 743         }
 744 }
 745
 746
 747 void Paragraph::Private::insertChar(pos_type pos, char_type c,
 748                 Change const & change)
 749 {
 750         LASSERT(pos >= 0 && pos <= int(text_.size()), return);
 751
 752         // track change
 753         changes_.insert(change, pos);
 754
 755         // This is actually very common when parsing buffers (and
 756         // maybe inserting ascii text)
 757         if (pos == pos_type(text_.size())) {
 758                 // when appending characters, no need to update tables
 759                 text_.push_back(c);
 760                 // but we want spell checking
 761                 requestSpellCheck(pos);
 762                 return;
 763         }
 764
 765         text_.insert(text_.begin() + pos, c);
 766
 767         // Update the font table.
 768         fontlist_.increasePosAfterPos(pos);
 769
 770         // Update the insets
 771         insetlist_.increasePosAfterPos(pos);
 772
 773         // Update list of misspelled positions
 774         speller_state_.increasePosAfterPos(pos);
 775 }
 776
 777
 778 bool Paragraph::insertInset(pos_type pos, Inset * inset,
 779                                    Font const & font, Change const & change)
 780 {
 781         LASSERT(inset, return false);
 782         LASSERT(pos >= 0 && pos <= size(), return false);
 783
 784         // Paragraph::insertInset() can be used in cut/copy/paste operation where
 785         // d->inset_owner_ is not set yet.
 786         if (d->inset_owner_ && !d->inset_owner_->insetAllowed(inset->lyxCode()))
 787                 return false;
 788
 789         d->insertChar(pos, META_INSET, change);
 790         LASSERT(d->text_[pos] == META_INSET, return false);
 791
 792         // Add a new entry in the insetlist_.
 793         d->insetlist_.insert(inset, pos);
 794
 795         // Some insets require run of spell checker
 796         requestSpellCheck(pos);
 797         setFont(pos, font);
 798         return true;
 799 }
 800
 801
 802 bool Paragraph::eraseChar(pos_type pos, bool trackChanges)
 803 {
 804         LASSERT(pos >= 0 && pos <= size(), return false);
 805
 806         // keep the logic here in sync with the logic of isMergedOnEndOfParDeletion()
 807
 808         if (trackChanges) {
 809                 Change change = d->changes_.lookup(pos);
 810
 811                 // set the character to DELETED if
 812                 //  a) it was previously unchanged or
 813                 //  b) it was inserted by a co-author
 814
 815                 if (!change.changed() ||
 816                       (change.inserted() && !change.currentAuthor())) {
 817                         setChange(pos, Change(Change::DELETED));
 818                         // request run of spell checker
 819                         requestSpellCheck(pos);
 820                         return false;
 821                 }
 822
 823                 if (change.deleted())
 824                         return false;
 825         }
 826
 827         // Don't physically access the imaginary end-of-paragraph character.
 828         // eraseChar() can only mark it as DELETED. A physical deletion of
 829         // end-of-par must be handled externally.
 830         if (pos == size()) {
 831                 return false;
 832         }
 833
 834         // track change
 835         d->changes_.erase(pos);
 836
 837         // if it is an inset, delete the inset entry
 838         if (d->text_[pos] == META_INSET)
 839                 d->insetlist_.erase(pos);
 840
 841         d->text_.erase(d->text_.begin() + pos);
 842
 843         // Update the fontlist_
 844         d->fontlist_.erase(pos);
 845
 846         // Update the insetlist_
 847         d->insetlist_.decreasePosAfterPos(pos);
 848
 849         // Update list of misspelled positions
 850         d->speller_state_.decreasePosAfterPos(pos);
 851         d->speller_state_.refreshLast(size());
 852
 853         return true;
 854 }
 855
 856
 857 int Paragraph::eraseChars(pos_type start, pos_type end, bool trackChanges)
 858 {
 859         LASSERT(start >= 0 && start <= size(), return 0);
 860         LASSERT(end >= start && end <= size() + 1, return 0);
 861
 862         pos_type i = start;
 863         for (pos_type count = end - start; count; --count) {
 864                 if (!eraseChar(i, trackChanges))
 865                         ++i;
 866         }
 867         return end - i;
 868 }
 869
 870 // Handle combining characters
 871 int Paragraph::Private::latexSurrogatePair(BufferParams const & bparams,
 872                 otexstream & os, char_type c, char_type next,
 873                 OutputParams const & runparams)
 874 {
 875         // Writing next here may circumvent a possible font change between
 876         // c and next. Since next is only output if it forms a surrogate pair
 877         // with c we can ignore this:
 878         // A font change inside a surrogate pair does not make sense and is
 879         // hopefully impossible to input.
 880         // FIXME: change tracking
 881         // Is this correct WRT change tracking?
 882         Encoding const & encoding = *(runparams.encoding);
 883         docstring latex1 = encoding.latexChar(next).first;
 884         if (runparams.inIPA) {
 885                 string const tipashortcut = Encodings::TIPAShortcut(next);
 886                 if (!tipashortcut.empty()) {
 887                         latex1 = from_ascii(tipashortcut);
 888                 }
 889         }
 890         docstring latex2 = encoding.latexChar(c).first;
 891
 892         if (bparams.useNonTeXFonts || docstring(1, next) == latex1) {
 893                 // Encoding supports the combination:
 894                 // output as is (combining char after base char).
 895                 os << latex2 << latex1;
 896                 return latex1.length() + latex2.length();
 897         }
 898
 899         os << latex1 << "{" << latex2 << "}";
 900         return latex1.length() + latex2.length() + 2;
 901 }
 902
 903
 904 bool Paragraph::Private::simpleTeXBlanks(BufferParams const & bparams,
 905                                        OutputParams const & runparams,
 906                                        otexstream & os,
 907                                        pos_type i,
 908                                        unsigned int & column,
 909                                        Font const & font,
 910                                        Layout const & style)
 911 {
 912         if (style.pass_thru || runparams.pass_thru)
 913                 return false;
 914
 915         if (i + 1 < int(text_.size())) {
 916                 char_type next = text_[i + 1];
 917                 if (Encodings::isCombiningChar(next)) {
 918                         // This space has an accent, so we must always output it.
 919                         column += latexSurrogatePair(bparams, os, ' ', next, runparams) - 1;
 920                         return true;
 921                 }
 922         }
 923
 924         if (runparams.linelen > 0
 925             && column > runparams.linelen
 926             && i
 927             && text_[i - 1] != ' '
 928             && (i + 1 < int(text_.size()))
 929             // same in FreeSpacing mode
 930             && !owner_->isFreeSpacing()
 931             // In typewriter mode, we want to avoid
 932             // ! . ? : at the end of a line
 933             && !(font.fontInfo().family() == TYPEWRITER_FAMILY
 934                  && (text_[i - 1] == '.'
 935                      || text_[i - 1] == '?'
 936                      || text_[i - 1] == ':'
 937                      || text_[i - 1] == '!'))) {
 938                 os << '\n';
 939                 os.texrow().start(owner_->id(), i + 1);
 940                 column = 0;
 941         } else if (style.free_spacing) {
 942                 os << '~';
 943         } else {
 944                 os << ' ';
 945         }
 946         return false;
 947 }
 948
 949
 950 void Paragraph::Private::latexInset(BufferParams const & bparams,
 951                                     otexstream & os,
 952                                     OutputParams & runparams,
 953                                     Font & running_font,
 954                                     Font & basefont,
 955                                     Font const & outerfont,
 956                                     bool & open_font,
 957                                     Change & running_change,
 958                                     Layout const & style,
 959                                     pos_type & i,
 960                                     unsigned int & column)
 961 {
 962         Inset * inset = owner_->getInset(i);
 963         LBUFERR(inset);
 964
 965         if (style.pass_thru) {
 966                 odocstringstream ods;
 967                 inset->plaintext(ods, runparams);
 968                 os << ods.str();
 969                 return;
 970         }
 971
 972         // FIXME: move this to InsetNewline::latex
 973         if (inset->lyxCode() == NEWLINE_CODE || inset->lyxCode() == SEPARATOR_CODE) {
 974                 // newlines are handled differently here than
 975                 // the default in simpleTeXSpecialChars().
 976                 if (!style.newline_allowed) {
 977                         os << '\n';
 978                 } else {
 979                         if (open_font) {
 980                                 bool needPar = false;
 981                                 column += running_font.latexWriteEndChanges(
 982                                         os, bparams, runparams,
 983                                         basefont, basefont, needPar);
 984                                 open_font = false;
 985                         }
 986
 987                         if (running_font.fontInfo().family() == TYPEWRITER_FAMILY)
 988                                 os << '~';
 989
 990                         basefont = owner_->getLayoutFont(bparams, outerfont);
 991                         running_font = basefont;
 992
 993                         if (runparams.moving_arg)
 994                                 os << "\\protect ";
 995
 996                 }
 997                 os.texrow().start(owner_->id(), i + 1);
 998                 column = 0;
 999         }
1000
1001         if (owner_->isDeleted(i)) {
1002                 if( ++runparams.inDeletedInset == 1)
1003                         runparams.changeOfDeletedInset = owner_->lookupChange(i);
1004         }
1005
1006         if (inset->canTrackChanges()) {
1007                 column += Changes::latexMarkChange(os, bparams, running_change,
1008                         Change(Change::UNCHANGED), runparams);
1009                 running_change = Change(Change::UNCHANGED);
1010         }
1011
1012         bool close = false;
1013         odocstream::pos_type const len = os.os().tellp();
1014
1015         if (inset->forceLTR(runparams)
1016             && running_font.isRightToLeft()
1017             // ERT is an exception, it should be output with no
1018             // decorations at all
1019             && inset->lyxCode() != ERT_CODE) {
1020                 if (runparams.use_polyglossia) {
1021                         os << "\\LRE{";
1022                 } else if (running_font.language()->lang() == "farsi"
1023                            || running_font.language()->lang() == "arabic_arabi")
1024                         os << "\\textLR{" << termcmd;
1025                 else
1026                         os << "\\L{";
1027                 close = true;
1028         }
1029
1030         // FIXME: Bug: we can have an empty font change here!
1031         // if there has just been a font change, we are going to close it
1032         // right now, which means stupid latex code like \textsf{}. AFAIK,
1033         // this does not harm dvi output. A minor bug, thus (JMarc)
1034
1035         // Some insets cannot be inside a font change command.
1036         // However, even such insets *can* be placed in \L or \R
1037         // or their equivalents (for RTL language switches), so we don't
1038         // close the language in those cases.
1039         // ArabTeX, though, cannot handle this special behavior, it seems.
1040         bool arabtex = basefont.language()->lang() == "arabic_arabtex"
1041                 || running_font.language()->lang() == "arabic_arabtex";
1042         if (open_font && !inset->inheritFont()) {
1043                 bool needPar = false;
1044                 bool closeLanguage = arabtex
1045                         || basefont.isRightToLeft() == running_font.isRightToLeft();
1046                 // We pass non_inherit_inset = true here since size switches
1047                 // ought not to be terminated here (#8384).
1048                 unsigned int count = running_font.latexWriteEndChanges(os,
1049                                         bparams, runparams, basefont, basefont,
1050                                         needPar, closeLanguage, true);
1051                 column += count;
1052                 // if any font properties were closed, update the running_font,
1053                 // making sure, however, to leave the language as it was
1054                 if (count > 0) {
1055                         // FIXME: probably a better way to keep track of the old
1056                         // language, than copying the entire font?
1057                         Font const copy_font(running_font);
1058                         basefont = owner_->getLayoutFont(bparams, outerfont);
1059                         running_font = basefont;
1060                         if (!closeLanguage)
1061                                 running_font.setLanguage(copy_font.language());
1062                         // For these, we use switches, so no need to close
1063                         basefont.fontInfo().setSize(copy_font.fontInfo().size());
1064                         basefont.fontInfo().setFamily(copy_font.fontInfo().family());
1065                         basefont.fontInfo().setSeries(copy_font.fontInfo().series());
1066                         // leave font open if language or any of the switches is still open
1067                         open_font = (running_font.language() == basefont.language()
1068                                      || running_font.fontInfo().size() == basefont.fontInfo().size()
1069                                      || running_font.fontInfo().family() == basefont.fontInfo().family()
1070                                      || running_font.fontInfo().series() == basefont.fontInfo().series());
1071                         if (closeLanguage)
1072                                 runparams.local_font = &basefont;
1073                 }
1074         }
1075
1076         size_t const previous_row_count = os.texrow().rows();
1077
1078         try {
1079                 runparams.lastid = id_;
1080                 runparams.lastpos = i;
1081                 inset->latex(os, runparams);
1082         } catch (EncodingException & e) {
1083                 // add location information and throw again.
1084                 e.par_id = id_;
1085                 e.pos = i;
1086                 throw(e);
1087         }
1088
1089         if (close)
1090                 os << '}';
1091
1092         if (os.texrow().rows() > previous_row_count) {
1093                 os.texrow().start(owner_->id(), i + 1);
1094                 column = 0;
1095         } else {
1096                 column += (unsigned int)(os.os().tellp() - len);
1097         }
1098
1099         if (owner_->isDeleted(i))
1100                 --runparams.inDeletedInset;
1101 }
1102
1103
1104 void Paragraph::Private::latexSpecialChar(otexstream & os,
1105                                           BufferParams const & bparams,
1106                                           OutputParams const & runparams,
1107                                           Font const & running_font,
1108                                           string & alien_script,
1109                                           Layout const & style,
1110                                           pos_type & i,
1111                                           pos_type end_pos,
1112                                           unsigned int & column)
1113 {
1114         char_type const c = owner_->getUChar(bparams, runparams, i);
1115
1116         if (style.pass_thru || runparams.pass_thru
1117             || contains(style.pass_thru_chars, c)
1118             || contains(runparams.pass_thru_chars, c)) {
1119                 if (c != '\0') {
1120                         Encoding const * const enc = runparams.encoding;
1121                         if (enc && !enc->encodable(c))
1122                                 throw EncodingException(c);
1123                         os.put(c);
1124                 }
1125                 return;
1126         }
1127
1128         // TIPA uses its own T3 encoding
1129         if (runparams.inIPA && latexSpecialT3(c, os, i, column))
1130                 return;
1131         // If T1 font encoding is used, use the special
1132         // characters it provides.
1133         // NOTE: Some languages reset the font encoding internally to a
1134         //       non-standard font encoding. If we are using such a language,
1135         //       we do not output special T1 chars.
1136         if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1137             && !runparams.isFullUnicode() && bparams.main_font_encoding() == "T1"
1138             && latexSpecialT1(c, os, i, column))
1139                 return;
1140         // NOTE: "fontspec" (non-TeX fonts) sets the font encoding to "TU" (untill 2017 "EU1" or "EU2")
1141         else if (!runparams.inIPA && !running_font.language()->internalFontEncoding()
1142                  && runparams.isFullUnicode() && latexSpecialTU(c, os, i, column))
1143                      return;
1144
1145         // Otherwise, we use what LaTeX provides us.
1146         switch (c) {
1147         case '\\':
1148                 os << "\\textbackslash" << termcmd;
1149                 column += 15;
1150                 break;
1151         case '<':
1152                 os << "\\textless" << termcmd;
1153                 column += 10;
1154                 break;
1155         case '>':
1156                 os << "\\textgreater" << termcmd;
1157                 column += 13;
1158                 break;
1159         case '|':
1160                 os << "\\textbar" << termcmd;
1161                 column += 9;
1162                 break;
1163         case '-':
1164                 os << '-';
1165                 if (i + 1 < static_cast<pos_type>(text_.size()) &&
1166                     (end_pos == -1 || i + 1 < end_pos) &&
1167                     text_[i+1] == '-') {
1168                         // Prevent "--" becoming an en dash and "---" an em dash.
1169                         // (Within \ttfamily, "---" is merged to en dash + hyphen.)
1170                         os << "{}";
1171                         column += 2;
1172                 }
1173                 break;
1174         case '\"':
1175                 os << "\\textquotedbl" << termcmd;
1176                 column += 14;
1177                 break;
1178
1179         case '$': case '&':
1180         case '%': case '#': case '{':
1181         case '}': case '_':
1182                 os << '\\';
1183                 os.put(c);
1184                 column += 1;
1185                 break;
1186
1187         case '~':
1188                 os << "\\textasciitilde" << termcmd;
1189                 column += 16;
1190                 break;
1191
1192         case '^':
1193                 os << "\\textasciicircum" << termcmd;
1194                 column += 17;
1195                 break;
1196
1197         case '*':
1198         case '[':
1199         case ']':
1200                 // avoid being mistaken for optional arguments
1201                 os << '{';
1202                 os.put(c);
1203                 os << '}';
1204                 column += 2;
1205                 break;
1206
1207         case ' ':
1208                 // Blanks are printed before font switching.
1209                 // Sure? I am not! (try nice-latex)
1210                 // I am sure it's correct. LyX might be smarter
1211                 // in the future, but for now, nothing wrong is
1212                 // written. (Asger)
1213                 break;
1214
1215         case 0x2013:
1216         case 0x2014:
1217                 // XeTeX's dash behaviour is determined via a global setting
1218                 if (bparams.use_dash_ligatures
1219                     && owner_->getFontSettings(bparams, i).fontInfo().family() != TYPEWRITER_FAMILY
1220                     && !runparams.inIPA
1221                         // TODO #10961: && not in inset Flex Code
1222                         // TODO #10961: && not in layout LyXCode
1223                     && (!bparams.useNonTeXFonts || runparams.flavor != OutputParams::XETEX)) {
1224                         if (c == 0x2013) {
1225                                 // en-dash
1226                                 os << "--";
1227                                 column +=2;
1228                         } else {
1229                                 // em-dash
1230                                 os << "---";
1231                                 column +=3;
1232                         }
1233                         break;
1234                 }
1235                 // fall through
1236         default:
1237                 if (c == '\0')
1238                         return;
1239
1240                 Encoding const & encoding = *(runparams.encoding);
1241                 char_type next = '\0';
1242                 if (i + 1 < int(text_.size())) {
1243                         next = text_[i + 1];
1244                         if (Encodings::isCombiningChar(next)) {
1245                                 column += latexSurrogatePair(bparams, os, c, next, runparams) - 1;
1246                                 ++i;
1247                                 break;
1248                         }
1249                 }
1250                 pair<docstring, bool> latex = encoding.latexChar(c);
1251                 docstring nextlatex;
1252                 bool nexttipas = false;
1253                 string nexttipashortcut;
1254                 if (next != '\0' && next != META_INSET && !encoding.encodable(next)) {
1255                         nextlatex = encoding.latexChar(next).first;
1256                         if (runparams.inIPA) {
1257                                 nexttipashortcut = Encodings::TIPAShortcut(next);
1258                                 nexttipas = !nexttipashortcut.empty();
1259                         }
1260                 }
1261                 bool tipas = false;
1262                 if (runparams.inIPA) {
1263                         string const tipashortcut = Encodings::TIPAShortcut(c);
1264                         if (!tipashortcut.empty()) {
1265                                 latex.first = from_ascii(tipashortcut);
1266                                 latex.second = false;
1267                                 tipas = true;
1268                         }
1269                 }
1270                 // eventually close "script wrapper" command (see `Paragraph::latex`)
1271                 if (!alien_script.empty()
1272                         && alien_script != Encodings::isKnownScriptChar(next)) {
1273                         column += latex.first.length();
1274                         alien_script.clear();
1275                         os << latex.first << "}";
1276                         break;
1277                 }
1278                 if (latex.second
1279                          && ((!prefixIs(nextlatex, '\\')
1280                                && !prefixIs(nextlatex, '{')
1281                                && !prefixIs(nextlatex, '}'))
1282                              || (nexttipas
1283                                  && !prefixIs(from_ascii(nexttipashortcut), '\\')))
1284                          && !tipas) {
1285                         // Prevent eating of a following space or command corruption by
1286                         // following characters
1287                         if (next == ' ' || next == '\0') {
1288                                 column += latex.first.length() + 1;
1289                                 os << latex.first << "{}";
1290                         } else {
1291                                 column += latex.first.length();
1292                                 os << latex.first << " ";
1293                         }
1294                 } else {
1295                         column += latex.first.length() - 1;
1296                         os << latex.first;
1297                 }
1298                 break;
1299         }
1300 }
1301
1302
1303 bool Paragraph::Private::latexSpecialT1(char_type const c, otexstream & os,
1304         pos_type i, unsigned int & column)
1305 {
1306         switch (c) {
1307         case '>':
1308         case '<':
1309                 os.put(c);
1310                 // In T1 encoding, these characters exist
1311                 // but we should avoid ligatures
1312                 if (i + 1 >= int(text_.size()) || text_[i + 1] != c)
1313                         return true;
1314                 os << "\\textcompwordmark" << termcmd;
1315                 column += 19;
1316                 return true;
1317         case '|':
1318                 os.put(c);
1319                 return true;
1320         case '\"':
1321                 // soul.sty breaks with \char`\"
1322                 os << "\\textquotedbl" << termcmd;
1323                 column += 14;
1324                 return true;
1325         default:
1326                 return false;
1327         }
1328 }
1329
1330
1331 bool Paragraph::Private::latexSpecialTU(char_type const c, otexstream & os,
1332         pos_type i, unsigned int & column)
1333 {
1334         // TU encoding is currently on par with T1.
1335         return latexSpecialT1(c, os, i, column);
1336 }
1337
1338
1339 bool Paragraph::Private::latexSpecialT3(char_type const c, otexstream & os,
1340         pos_type /*i*/, unsigned int & column)
1341 {
1342         switch (c) {
1343         case '*':
1344         case '[':
1345         case ']':
1346         case '\"':
1347                 os.put(c);
1348                 return true;
1349         case '|':
1350                 os << "\\textvertline" << termcmd;
1351                 column += 14;
1352                 return true;
1353         default:
1354                 return false;
1355         }
1356 }
1357
1358
1359 void Paragraph::Private::validate(LaTeXFeatures & features) const
1360 {
1361         if (layout_->inpreamble && inset_owner_) {
1362                 // FIXME: Using a string stream here circumvents the encoding
1363                 // switching machinery of odocstream. Therefore the
1364                 // output is wrong if this paragraph contains content
1365                 // that needs to switch encoding.
1366                 Buffer const & buf = inset_owner_->buffer();
1367                 otexstringstream os;
1368                 os << layout_->preamble();
1369                 size_t const length = os.length();
1370                 TeXOnePar(buf, *inset_owner_->getText(int(buf.getParFromID(owner_->id()).idx())),
1371                           buf.getParFromID(owner_->id()).pit(), os,
1372                           features.runparams(), string(), 0, -1, true);
1373                 if (os.length() > length)
1374                         features.addPreambleSnippet(os.release(), true);
1375         }
1376
1377         if (features.runparams().flavor == OutputParams::HTML
1378             && layout_->htmltitle()) {
1379                 features.setHTMLTitle(owner_->asString(AS_STR_INSETS | AS_STR_SKIPDELETE));
1380         }
1381
1382         // check the params.
1383         if (!params_.spacing().isDefault())
1384                 features.require("setspace");
1385
1386         // then the layouts
1387         features.useLayout(layout_->name());
1388
1389         // then the fonts
1390         fontlist_.validate(features);
1391
1392         // then the indentation
1393         if (!params_.leftIndent().zero())
1394                 features.require("ParagraphLeftIndent");
1395
1396         // then the insets
1397         InsetList::const_iterator icit = insetlist_.begin();
1398         InsetList::const_iterator iend = insetlist_.end();
1399         for (; icit != iend; ++icit) {
1400                 if (icit->inset) {
1401                         features.inDeletedInset(owner_->isDeleted(icit->pos));
1402                         if (icit->inset->lyxCode() == FOOT_CODE) {
1403                                 // FIXME: an item inset would make things much easier.
1404                                 if ((layout_->latextype == LATEX_LIST_ENVIRONMENT
1405                                      || (layout_->latextype == LATEX_ITEM_ENVIRONMENT
1406                                          && layout_->margintype == MARGIN_FIRST_DYNAMIC))
1407                                     && (icit->pos < begin_of_body_
1408                                         || (icit->pos == begin_of_body_
1409                                             && (icit->pos == 0 || text_[icit->pos - 1] != ' '))))
1410                                         features.saveNoteEnv("description");
1411                         }
1412                         icit->inset->validate(features);
1413                         features.inDeletedInset(false);
1414                         if (layout_->needprotect &&
1415                             icit->inset->lyxCode() == FOOT_CODE)
1416                                 features.require("NeedLyXFootnoteCode");
1417                 }
1418         }
1419
1420         // then the contents
1421         BufferParams const bp = features.runparams().is_child
1422                 ? features.buffer().masterParams() : features.buffer().params();
1423         for (pos_type i = 0; i < int(text_.size()) ; ++i) {
1424                 char_type c = text_[i];
1425                 CharInfo const & ci = Encodings::unicodeCharInfo(c);
1426                 if (c == 0x0022) {
1427                         if (features.runparams().isFullUnicode() && bp.useNonTeXFonts)
1428                                 features.require("textquotedblp");
1429                         else if (bp.main_font_encoding() != "T1"
1430                                  || ((&owner_->getFontSettings(bp, i))->language()->internalFontEncoding()))
1431                                 features.require("textquotedbl");
1432                 } else if (ci.textfeature() && contains(ci.textpreamble(), '=')) {
1433                         // features that depend on the font or input encoding
1434                         string feats = ci.textpreamble();
1435                         string fontenc = (&owner_->getFontSettings(bp, i))->language()->fontenc(bp);
1436                         if (fontenc.empty())
1437                                 fontenc = features.runparams().main_fontenc;
1438                         while (!feats.empty()) {
1439                                 string feat;
1440                                 feats = split(feats, feat, ',');
1441                                 if (contains(feat, "!=")) {
1442                                         // a feature that is required except for the spcified
1443                                         // font or input encodings
1444                                         string realfeature;
1445                                         string const contexts = ltrim(split(feat, realfeature, '!'), "=");
1446                                         // multiple encodings are separated by semicolon
1447                                         vector<string> context = getVectorFromString(contexts, ";");
1448                                         // require feature if the context matches neither current font
1449                                         // nor input encoding
1450                                         if (std::find(context.begin(), context.end(), fontenc) == context.end()
1451                                             && std::find(context.begin(), context.end(),
1452                                                          features.runparams().encoding->name()) == context.end())
1453                                                 features.require(realfeature);
1454                                 } else if (contains(feat, '=')) {
1455                                         // a feature that is required only for the spcified
1456                                         // font or input encodings
1457                                         string realfeature;
1458                                         string const contexts = split(feat, realfeature, '=');
1459                                         // multiple encodings are separated by semicolon
1460                                         vector<string> context = getVectorFromString(contexts, ";");
1461                                         // require feature if the context matches either current font
1462                                         // or input encoding
1463                                         if (std::find(context.begin(), context.end(), fontenc) != context.end()
1464                                             || std::find(context.begin(), context.end(),
1465                                                          features.runparams().encoding->name()) != context.end())
1466                                                 features.require(realfeature);
1467                                 }
1468                         }
1469                 } else if (!bp.use_dash_ligatures
1470                            && (c == 0x2013 || c == 0x2014)
1471                            && bp.useNonTeXFonts
1472                            && features.runparams().flavor == OutputParams::XETEX)
1473                         // XeTeX's dash behaviour is determined via a global setting
1474                         features.require("xetexdashbreakstate");
1475                 BufferEncodings::validate(c, features);
1476         }
1477 }
1478
1479 /////////////////////////////////////////////////////////////////////
1480 //
1481 // Paragraph
1482 //
1483 /////////////////////////////////////////////////////////////////////
1484
1485 namespace {
1486         Layout const emptyParagraphLayout;
1487 }
1488
1489 Paragraph::Paragraph()
1490         : d(new Paragraph::Private(this, emptyParagraphLayout))
1491 {
1492         itemdepth = 0;
1493         d->params_.clear();
1494 }
1495
1496
1497 Paragraph::Paragraph(Paragraph const & par)
1498         : itemdepth(par.itemdepth),
1499         d(new Paragraph::Private(*par.d, this))
1500 {
1501         registerWords();
1502 }
1503
1504
1505 Paragraph::Paragraph(Paragraph const & par, pos_type beg, pos_type end)
1506         : itemdepth(par.itemdepth),
1507         d(new Paragraph::Private(*par.d, this, beg, end))
1508 {
1509         registerWords();
1510 }
1511
1512
1513 Paragraph & Paragraph::operator=(Paragraph const & par)
1514 {
1515         // needed as we will destroy the private part before copying it
1516         if (&par != this) {
1517                 itemdepth = par.itemdepth;
1518
1519                 deregisterWords();
1520                 delete d;
1521                 d = new Private(*par.d, this);
1522                 registerWords();
1523         }
1524         return *this;
1525 }
1526
1527
1528 Paragraph::~Paragraph()
1529 {
1530         deregisterWords();
1531         delete d;
1532 }
1533
1534
1535 namespace {
1536
1537 // this shall be called just before every "os << ..." action.
1538 void flushString(ostream & os, docstring & s)
1539 {
1540         os << to_utf8(s);
1541         s.erase();
1542 }
1543
1544 } // namespace
1545
1546
1547 void Paragraph::write(ostream & os, BufferParams const & bparams,
1548         depth_type & dth) const
1549 {
1550         // The beginning or end of a deeper (i.e. nested) area?
1551         if (dth != d->params_.depth()) {
1552                 if (d->params_.depth() > dth) {
1553                         while (d->params_.depth() > dth) {
1554                                 os << "\n\\begin_deeper";
1555                                 ++dth;
1556                         }
1557                 } else {
1558                         while (d->params_.depth() < dth) {
1559                                 os << "\n\\end_deeper";
1560                                 --dth;
1561                         }
1562                 }
1563         }
1564
1565         // First write the layout
1566         os << "\n\\begin_layout " << to_utf8(d->layout_->name()) << '\n';
1567
1568         d->params_.write(os);
1569
1570         Font font1(inherit_font, bparams.language);
1571
1572         Change running_change = Change(Change::UNCHANGED);
1573
1574         // this string is used as a buffer to avoid repetitive calls
1575         // to to_utf8(), which turn out to be expensive (JMarc)
1576         docstring write_buffer;
1577
1578         int column = 0;
1579         for (pos_type i = 0; i <= size(); ++i) {
1580
1581                 Change const & change = lookupChange(i);
1582                 if (change != running_change)
1583                         flushString(os, write_buffer);
1584                 Changes::lyxMarkChange(os, bparams, column, running_change, change);
1585                 running_change = change;
1586
1587                 if (i == size())
1588                         break;
1589
1590                 // Write font changes
1591                 Font font2 = getFontSettings(bparams, i);
1592                 if (font2 != font1) {
1593                         flushString(os, write_buffer);
1594                         font2.lyxWriteChanges(font1, os);
1595                         column = 0;
1596                         font1 = font2;
1597                 }
1598
1599                 char_type const c = d->text_[i];
1600                 switch (c) {
1601                 case META_INSET:
1602                         if (Inset const * inset = getInset(i)) {
1603                                 flushString(os, write_buffer);
1604                                 if (inset->directWrite()) {
1605                                         // international char, let it write
1606                                         // code directly so it's shorter in
1607                                         // the file
1608                                         inset->write(os);
1609                                 } else {
1610                                         if (i)
1611                                                 os << '\n';
1612                                         os << "\\begin_inset ";
1613                                         inset->write(os);
1614                                         os << "\n\\end_inset\n\n";
1615                                         column = 0;
1616                                 }
1617                                 // FIXME This can be removed again once the mystery
1618                                 // crash has been resolved.
1619                                 os << flush;
1620                         }
1621                         break;
1622                 case '\\':
1623                         flushString(os, write_buffer);
1624                         os << "\n\\backslash\n";
1625                         column = 0;
1626                         break;
1627                 case '.':
1628                         flushString(os, write_buffer);
1629                         if (i + 1 < size() && d->text_[i + 1] == ' ') {
1630                                 os << ".\n";
1631                                 column = 0;
1632                         } else
1633                                 os << '.';
1634                         break;
1635                 default:
1636                         if ((column > 70 && c == ' ')
1637                             || column > 79) {
1638                                 flushString(os, write_buffer);
1639                                 os << '\n';
1640                                 column = 0;
1641                         }
1642                         // this check is to amend a bug. LyX sometimes
1643                         // inserts '\0' this could cause problems.
1644                         if (c != '\0')
1645                                 write_buffer.push_back(c);
1646                         else
1647                                 LYXERR0("NUL char in structure.");
1648                         ++column;
1649                         break;
1650                 }
1651         }
1652
1653         flushString(os, write_buffer);
1654         os << "\n\\end_layout\n";
1655         // FIXME This can be removed again once the mystery
1656         // crash has been resolved.
1657         os << flush;
1658 }
1659
1660
1661 void Paragraph::validate(LaTeXFeatures & features) const
1662 {
1663         d->validate(features);
1664         bool fragile = features.runparams().moving_arg;
1665         fragile |= layout().needprotect;
1666         if (inInset().getLayout().isNeedProtect())
1667                 fragile = true;
1668         if (needsCProtection(fragile))
1669                 features.require("cprotect");
1670 }
1671
1672
1673 void Paragraph::insert(pos_type start, docstring const & str,
1674                        Font const & font, Change const & change)
1675 {
1676         for (size_t i = 0, n = str.size(); i != n ; ++i)
1677                 insertChar(start + i, str[i], font, change);
1678 }
1679
1680
1681 void Paragraph::appendChar(char_type c, Font const & font,
1682                 Change const & change)
1683 {
1684         // track change
1685         d->changes_.insert(change, d->text_.size());
1686         // when appending characters, no need to update tables
1687         d->text_.push_back(c);
1688         setFont(d->text_.size() - 1, font);
1689         d->requestSpellCheck(d->text_.size() - 1);
1690 }
1691
1692
1693 void Paragraph::appendString(docstring const & s, Font const & font,
1694                 Change const & change)
1695 {
1696         pos_type end = s.size();
1697         size_t oldsize = d->text_.size();
1698         size_t newsize = oldsize + end;
1699         size_t capacity = d->text_.capacity();
1700         if (newsize >= capacity)
1701                 d->text_.reserve(max(capacity + 100, newsize));
1702
1703         // when appending characters, no need to update tables
1704         d->text_.append(s);
1705
1706         // FIXME: Optimize this!
1707         for (size_t i = oldsize; i != newsize; ++i) {
1708                 // track change
1709                 d->changes_.insert(change, i);
1710                 d->requestSpellCheck(i);
1711         }
1712         d->fontlist_.set(oldsize, font);
1713         d->fontlist_.set(newsize - 1, font);
1714 }
1715
1716
1717 void Paragraph::insertChar(pos_type pos, char_type c,
1718                            bool trackChanges)
1719 {
1720         d->insertChar(pos, c, Change(trackChanges ?
1721                            Change::INSERTED : Change::UNCHANGED));
1722 }
1723
1724
1725 void Paragraph::insertChar(pos_type pos, char_type c,
1726                            Font const & font, bool trackChanges)
1727 {
1728         d->insertChar(pos, c, Change(trackChanges ?
1729                            Change::INSERTED : Change::UNCHANGED));
1730         setFont(pos, font);
1731 }
1732
1733
1734 void Paragraph::insertChar(pos_type pos, char_type c,
1735                            Font const & font, Change const & change)
1736 {
1737         d->insertChar(pos, c, change);
1738         setFont(pos, font);
1739 }
1740
1741
1742 void Paragraph::resetFonts(Font const & font)
1743 {
1744         d->fontlist_.clear();
1745         d->fontlist_.set(0, font);
1746         d->fontlist_.set(d->text_.size() - 1, font);
1747 }
1748
1749 // Gets uninstantiated font setting at position.
1750 Font const & Paragraph::getFontSettings(BufferParams const & bparams,
1751                                          pos_type pos) const
1752 {
1753         if (pos > size()) {
1754                 LYXERR0("pos: " << pos << " size: " << size());
1755                 LBUFERR(false);
1756         }
1757
1758         FontList::const_iterator cit = d->fontlist_.fontIterator(pos);
1759         if (cit != d->fontlist_.end())
1760                 return cit->font();
1761
1762         if (pos == size() && !empty())
1763                 return getFontSettings(bparams, pos - 1);
1764
1765         // Optimisation: avoid a full font instantiation if there is no
1766         // language change from previous call.
1767         static Font previous_font;
1768         static Language const * previous_lang = nullptr;
1769         Language const * lang = getParLanguage(bparams);
1770         if (lang != previous_lang) {
1771                 previous_lang = lang;
1772                 previous_font = Font(inherit_font, lang);
1773         }
1774         return previous_font;
1775 }
1776
1777
1778 FontSpan Paragraph::fontSpan(pos_type pos) const
1779 {
1780         LBUFERR(pos <= size());
1781
1782         if (pos == size())
1783                 return FontSpan(pos, pos);
1784
1785         pos_type start = 0;
1786         FontList::const_iterator cit = d->fontlist_.begin();
1787         FontList::const_iterator end = d->fontlist_.end();
1788         for (; cit != end; ++cit) {
1789                 if (cit->pos() >= pos) {
1790                         if (pos >= beginOfBody())
1791                                 return FontSpan(max(start, beginOfBody()),
1792                                                 cit->pos());
1793                         else
1794                                 return FontSpan(start,
1795                                                 min(beginOfBody() - 1,
1796                                                          cit->pos()));
1797                 }
1798                 start = cit->pos() + 1;
1799         }
1800
1801         // This should not happen, but if so, we take no chances.
1802         LYXERR0("Paragraph::fontSpan: position not found in fontinfo table!");
1803         LASSERT(false, return FontSpan(pos, pos));
1804 }
1805
1806
1807 // Gets uninstantiated font setting at position 0
1808 Font const & Paragraph::getFirstFontSettings(BufferParams const & bparams) const
1809 {
1810         if (!empty() && !d->fontlist_.empty())
1811                 return d->fontlist_.begin()->font();
1812
1813         // Optimisation: avoid a full font instantiation if there is no
1814         // language change from previous call.
1815         static Font previous_font;
1816         static Language const * previous_lang = nullptr;
1817         if (bparams.language != previous_lang) {
1818                 previous_lang = bparams.language;
1819                 previous_font = Font(inherit_font, bparams.language);
1820         }
1821
1822         return previous_font;
1823 }
1824
1825
1826 // Gets the fully instantiated font at a given position in a paragraph
1827 // This is basically the same function as Text::GetFont() in text2.cpp.
1828 // The difference is that this one is used for generating the LaTeX file,
1829 // and thus cosmetic "improvements" are disallowed: This has to deliver
1830 // the true picture of the buffer. (Asger)
1831 Font const Paragraph::getFont(BufferParams const & bparams, pos_type pos,
1832                                  Font const & outerfont) const
1833 {
1834         LBUFERR(pos >= 0);
1835
1836         Font font = getFontSettings(bparams, pos);
1837
1838         pos_type const body_pos = beginOfBody();
1839         FontInfo & fi = font.fontInfo();
1840         if (pos < body_pos)
1841                 fi.realize(d->layout_->labelfont);
1842         else
1843                 fi.realize(d->layout_->font);
1844
1845         fi.realize(outerfont.fontInfo());
1846         fi.realize(bparams.getFont().fontInfo());
1847
1848         return font;
1849 }
1850
1851
1852 Font const Paragraph::getLabelFont
1853         (BufferParams const & bparams, Font const & outerfont) const
1854 {
1855         FontInfo tmpfont = d->layout_->labelfont;
1856         tmpfont.realize(outerfont.fontInfo());
1857         tmpfont.realize(bparams.getFont().fontInfo());
1858         return Font(tmpfont, getParLanguage(bparams));
1859 }
1860
1861
1862 Font const Paragraph::getLayoutFont
1863         (BufferParams const & bparams, Font const & outerfont) const
1864 {
1865         FontInfo tmpfont = d->layout_->font;
1866         tmpfont.realize(outerfont.fontInfo());
1867         tmpfont.realize(bparams.getFont().fontInfo());
1868         return Font(tmpfont, getParLanguage(bparams));
1869 }
1870
1871
1872 char_type Paragraph::getUChar(BufferParams const & bparams,
1873                               OutputParams const & rp,
1874                               pos_type pos) const
1875 {
1876         char_type c = d->text_[pos];
1877
1878         // Return unchanged character in LTR languages
1879         // or if we use poylglossia/bidi (XeTeX).
1880         if (rp.useBidiPackage()
1881             || !getFontSettings(bparams, pos).isRightToLeft())
1882                 return c;
1883
1884         // Without polyglossia/bidi, we need to account for some special cases.
1885         // FIXME This needs to be audited!
1886         // Check if:
1887         // * The input is as expected for all delimiters
1888         //   => checked for Hebrew!
1889         // * The output matches the display in the LyX workarea
1890         //   => checked for Hebrew!
1891         // * The special cases below are really necessary
1892         //   => checked for Hebrew!
1893         // * In arabic_arabi, brackets are transformed to Arabic
1894         //   Ornate Parentheses. Is this is really wanted?
1895
1896         string const & lang = getFontSettings(bparams, pos).language()->lang();
1897         char_type uc = c;
1898
1899         // 1. In the following languages, parentheses need to be reversed.
1900         //    Also with polyglodia/luabidi
1901         bool const reverseparens = (lang == "hebrew" || rp.use_polyglossia);
1902
1903         // 2. In the following languages, brackets don't need to be reversed.
1904         bool const reversebrackets = lang != "arabic_arabtex"
1905                         && lang != "arabic_arabi"
1906                         && lang != "farsi";
1907
1908         // Now swap delimiters if needed.
1909         switch (c) {
1910         case '(':
1911                 if (reverseparens)
1912                         uc = ')';
1913                 break;
1914         case ')':
1915                 if (reverseparens)
1916                         uc = '(';
1917                 break;
1918         case '[':
1919                 if (reversebrackets)
1920                         uc = ']';
1921                 break;
1922         case ']':
1923                 if (reversebrackets)
1924                         uc = '[';
1925                 break;
1926         case '{':
1927                 uc = '}';
1928                 break;
1929         case '}':
1930                 uc = '{';
1931                 break;
1932         case '<':
1933                 uc = '>';
1934                 break;
1935         case '>':
1936                 uc = '<';
1937                 break;
1938         }
1939
1940         return uc;
1941 }
1942
1943
1944 void Paragraph::setFont(pos_type pos, Font const & font)
1945 {
1946         LASSERT(pos <= size(), return);
1947
1948         // First, reduce font against layout/label font
1949         // Update: The setCharFont() routine in text2.cpp already
1950         // reduces font, so we don't need to do that here. (Asger)
1951
1952         d->fontlist_.set(pos, font);
1953 }
1954
1955
1956 void Paragraph::makeSameLayout(Paragraph const & par)
1957 {
1958         d->layout_ = par.d->layout_;
1959         d->params_ = par.d->params_;
1960 }
1961
1962
1963 bool Paragraph::stripLeadingSpaces(bool trackChanges)
1964 {
1965         if (isFreeSpacing())
1966                 return false;
1967
1968         int pos = 0;
1969         int count = 0;
1970
1971         while (pos < size() && (isNewline(pos) || isLineSeparator(pos))) {
1972                 if (eraseChar(pos, trackChanges))
1973                         ++count;
1974                 else
1975                         ++pos;
1976         }
1977
1978         return count > 0 || pos > 0;
1979 }
1980
1981
1982 bool Paragraph::hasSameLayout(Paragraph const & par) const
1983 {
1984         return par.d->layout_ == d->layout_
1985                 && d->params_.sameLayout(par.d->params_);
1986 }
1987
1988
1989 depth_type Paragraph::getDepth() const
1990 {
1991         return d->params_.depth();
1992 }
1993
1994
1995 depth_type Paragraph::getMaxDepthAfter() const
1996 {
1997         if (d->layout_->isEnvironment())
1998                 return d->params_.depth() + 1;
1999         else
2000                 return d->params_.depth();
2001 }
2002
2003
2004 LyXAlignment Paragraph::getAlign(BufferParams const & bparams) const
2005 {
2006         if (d->params_.align() == LYX_ALIGN_LAYOUT)
2007                 return getDefaultAlign(bparams);
2008         else
2009                 return d->params_.align();
2010 }
2011
2012
2013 LyXAlignment Paragraph::getDefaultAlign(BufferParams const & bparams) const
2014 {
2015         LyXAlignment res = layout().align;
2016         if (isRTL(bparams)) {
2017                 // Swap sides
2018                 if (res == LYX_ALIGN_LEFT)
2019                         res = LYX_ALIGN_RIGHT;
2020                 else if  (res == LYX_ALIGN_RIGHT)
2021                         res = LYX_ALIGN_LEFT;
2022         }
2023         return res;
2024 }
2025
2026
2027 docstring const & Paragraph::labelString() const
2028 {
2029         return d->params_.labelString();
2030 }
2031
2032
2033 // the next two functions are for the manual labels
2034 docstring const Paragraph::getLabelWidthString() const
2035 {
2036         if (d->layout_->margintype == MARGIN_MANUAL
2037             || d->layout_->latextype == LATEX_BIB_ENVIRONMENT)
2038                 return d->params_.labelWidthString();
2039         else
2040                 return _("Senseless with this layout!");
2041 }
2042
2043
2044 void Paragraph::setLabelWidthString(docstring const & s)
2045 {
2046         d->params_.labelWidthString(s);
2047 }
2048
2049
2050 docstring Paragraph::expandLabel(Layout const & layout,
2051                 BufferParams const & bparams) const
2052 {
2053         return expandParagraphLabel(layout, bparams, true);
2054 }
2055
2056
2057 docstring Paragraph::expandParagraphLabel(Layout const & layout,
2058                 BufferParams const & bparams, bool process_appendix) const
2059 {
2060         DocumentClass const & tclass = bparams.documentClass();
2061         string const & lang = getParLanguage(bparams)->code();
2062         bool const in_appendix = process_appendix && d->params_.appendix();
2063         docstring fmt = translateIfPossible(layout.labelstring(in_appendix), lang);
2064
2065         if (fmt.empty() && !layout.counter.empty())
2066                 return tclass.counters().theCounter(layout.counter, lang);
2067
2068         // handle 'inherited level parts' in 'fmt',
2069         // i.e. the stuff between '@' in   '@Section@.\arabic{subsection}'
2070         size_t const i = fmt.find('@', 0);
2071         if (i != docstring::npos) {
2072                 size_t const j = fmt.find('@', i + 1);
2073                 if (j != docstring::npos) {
2074                         docstring parent(fmt, i + 1, j - i - 1);
2075                         docstring label = from_ascii("??");
2076                         if (tclass.hasLayout(parent))
2077                                 label = expandParagraphLabel(tclass[parent], bparams,
2078                                                       process_appendix);
2079                         fmt = docstring(fmt, 0, i) + label
2080                                 + docstring(fmt, j + 1, docstring::npos);
2081                 }
2082         }
2083
2084         return tclass.counters().counterLabel(fmt, lang);
2085 }
2086
2087
2088 void Paragraph::applyLayout(Layout const & new_layout)
2089 {
2090         d->layout_ = &new_layout;
2091         LyXAlignment const oldAlign = d->params_.align();
2092
2093         if (!(oldAlign & d->layout_->alignpossible)) {
2094                 frontend::Alert::warning(_("Alignment not permitted"),
2095                         _("The new layout does not permit the alignment previously used.\nSetting to default."));
2096                 d->params_.align(LYX_ALIGN_LAYOUT);
2097         }
2098 }
2099
2100
2101 pos_type Paragraph::beginOfBody() const
2102 {
2103         return d->begin_of_body_;
2104 }
2105
2106
2107 void Paragraph::setBeginOfBody()
2108 {
2109         if (d->layout_->labeltype != LABEL_MANUAL) {
2110                 d->begin_of_body_ = 0;
2111                 return;
2112         }
2113
2114         // Unroll the first two cycles of the loop
2115         // and remember the previous character to
2116         // remove unnecessary getChar() calls
2117         pos_type i = 0;
2118         pos_type end = size();
2119         bool prev_char_deleted = false;
2120         if (i < end && (!(isNewline(i) || isEnvSeparator(i)) || isDeleted(i))) {
2121                 ++i;
2122                 if (i < end) {
2123                         char_type previous_char = d->text_[i];
2124                         if (!(isNewline(i) || isEnvSeparator(i))) {
2125                                 ++i;
2126                                 while (i < end && (previous_char != ' ' || prev_char_deleted)) {
2127                                         char_type temp = d->text_[i];
2128                                         prev_char_deleted = isDeleted(i);
2129                                         if (!isDeleted(i) && (isNewline(i) || isEnvSeparator(i)))
2130                                                 break;
2131                                         ++i;
2132                                         previous_char = temp;
2133                                 }
2134                         }
2135                 }
2136         }
2137
2138         d->begin_of_body_ = i;
2139 }
2140
2141
2142 bool Paragraph::allowParagraphCustomization() const
2143 {
2144         return inInset().allowParagraphCustomization();
2145 }
2146
2147
2148 bool Paragraph::usePlainLayout() const
2149 {
2150         return inInset().usePlainLayout();
2151 }
2152
2153
2154 bool Paragraph::isPassThru() const
2155 {
2156         return inInset().isPassThru() || d->layout_->pass_thru;
2157 }
2158
2159 namespace {
2160
2161 // paragraphs inside floats need different alignment tags to avoid
2162 // unwanted space
2163
2164 bool noTrivlistCentering(InsetCode code)
2165 {
2166         return code == FLOAT_CODE
2167                || code == WRAP_CODE
2168                || code == CELL_CODE;
2169 }
2170
2171
2172 string correction(string const & orig)
2173 {
2174         if (orig == "flushleft")
2175                 return "raggedright";
2176         if (orig == "flushright")
2177                 return "raggedleft";
2178         if (orig == "center")
2179                 return "centering";
2180         return orig;
2181 }
2182
2183
2184 bool corrected_env(otexstream & os, string const & suffix, string const & env,
2185         InsetCode code, bool const lastpar, int & col)
2186 {
2187         string macro = suffix + "{";
2188         if (noTrivlistCentering(code)) {
2189                 if (lastpar) {
2190                         // the last paragraph in non-trivlist-aligned
2191                         // context is special (to avoid unwanted whitespace)
2192                         if (suffix == "\\begin") {
2193                                 macro = "\\" + correction(env) + "{}";
2194                                 os << from_ascii(macro);
2195                                 col += macro.size();
2196                                 return true;
2197                         }
2198                         return false;
2199                 }
2200                 macro += correction(env);
2201         } else
2202                 macro += env;
2203         macro += "}";
2204         if (suffix == "\\par\\end") {
2205                 os << breakln;
2206                 col = 0;
2207         }
2208         os << from_ascii(macro);
2209         col += macro.size();
2210         if (suffix == "\\begin") {
2211                 os << breakln;
2212                 col = 0;
2213         }
2214         return true;
2215 }
2216
2217 } // namespace
2218
2219
2220 int Paragraph::Private::startTeXParParams(BufferParams const & bparams,
2221                         otexstream & os, OutputParams const & runparams) const
2222 {
2223         int column = 0;
2224
2225         bool canindent =
2226                 (bparams.paragraph_separation == BufferParams::ParagraphIndentSeparation) ?
2227                         (layout_->toggle_indent != ITOGGLE_NEVER) :
2228                         (layout_->toggle_indent == ITOGGLE_ALWAYS);
2229
2230         if (canindent && params_.noindent() && !layout_->pass_thru) {
2231                 os << "\\noindent ";
2232                 column += 10;
2233         }
2234
2235         LyXAlignment const curAlign = params_.align();
2236
2237         if (curAlign == layout_->align)
2238                 return column;
2239
2240         switch (curAlign) {
2241         case LYX_ALIGN_NONE:
2242         case LYX_ALIGN_BLOCK:
2243         case LYX_ALIGN_LAYOUT:
2244         case LYX_ALIGN_SPECIAL:
2245         case LYX_ALIGN_DECIMAL:
2246                 break;
2247         case LYX_ALIGN_LEFT:
2248         case LYX_ALIGN_RIGHT:
2249         case LYX_ALIGN_CENTER:
2250                 if (runparams.moving_arg) {
2251                         os << "\\protect";
2252                         column += 8;
2253                 }
2254                 break;
2255         }
2256
2257         string const begin_tag = "\\begin";
2258         InsetCode code = ownerCode();
2259         bool const lastpar = runparams.isLastPar;
2260         // RTL in classic (PDF)LaTeX (without the Bidi package)
2261         // Luabibdi (used by LuaTeX) behaves like classic
2262         bool const rtl_classic = owner_->getParLanguage(bparams)->rightToLeft()
2263                 && !runparams.useBidiPackage();
2264
2265         switch (curAlign) {
2266         case LYX_ALIGN_NONE:
2267         case LYX_ALIGN_BLOCK:
2268         case LYX_ALIGN_LAYOUT:
2269         case LYX_ALIGN_SPECIAL:
2270         case LYX_ALIGN_DECIMAL:
2271                 break;
2272         case LYX_ALIGN_LEFT: {
2273                 if (rtl_classic)
2274                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2275                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2276                 else
2277                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2278                 break;
2279         } case LYX_ALIGN_RIGHT: {
2280                 if (rtl_classic)
2281                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2282                         corrected_env(os, begin_tag, "flushleft", code, lastpar, column);
2283                 else
2284                         corrected_env(os, begin_tag, "flushright", code, lastpar, column);
2285                 break;
2286         } case LYX_ALIGN_CENTER: {
2287                 corrected_env(os, begin_tag, "center", code, lastpar, column);
2288                 break;
2289         }
2290         }
2291
2292         return column;
2293 }
2294
2295
2296 bool Paragraph::Private::endTeXParParams(BufferParams const & bparams,
2297                         otexstream & os, OutputParams const & runparams) const
2298 {
2299         LyXAlignment const curAlign = params_.align();
2300
2301         if (curAlign == layout_->align)
2302                 return false;
2303
2304         switch (curAlign) {
2305         case LYX_ALIGN_NONE:
2306         case LYX_ALIGN_BLOCK:
2307         case LYX_ALIGN_LAYOUT:
2308         case LYX_ALIGN_SPECIAL:
2309         case LYX_ALIGN_DECIMAL:
2310                 break;
2311         case LYX_ALIGN_LEFT:
2312         case LYX_ALIGN_RIGHT:
2313         case LYX_ALIGN_CENTER:
2314                 if (runparams.moving_arg)
2315                         os << "\\protect";
2316                 break;
2317         }
2318
2319         bool output = false;
2320         int col = 0;
2321         string const end_tag = "\\par\\end";
2322         InsetCode code = ownerCode();
2323         bool const lastpar = runparams.isLastPar;
2324         // RTL in classic (PDF)LaTeX (without the Bidi package)
2325         // Luabibdi (used by LuaTeX) behaves like classic
2326         bool const rtl_classic = owner_->getParLanguage(bparams)->rightToLeft()
2327                 && !runparams.useBidiPackage();
2328
2329         switch (curAlign) {
2330         case LYX_ALIGN_NONE:
2331         case LYX_ALIGN_BLOCK:
2332         case LYX_ALIGN_LAYOUT:
2333         case LYX_ALIGN_SPECIAL:
2334         case LYX_ALIGN_DECIMAL:
2335                 break;
2336         case LYX_ALIGN_LEFT: {
2337                 if (rtl_classic)
2338                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2339                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2340                 else
2341                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2342                 break;
2343         } case LYX_ALIGN_RIGHT: {
2344                 if (rtl_classic)
2345                         // Classic (PDF)LaTeX switches the left/right logic in RTL mode
2346                         output = corrected_env(os, end_tag, "flushleft", code, lastpar, col);
2347                 else
2348                         output = corrected_env(os, end_tag, "flushright", code, lastpar, col);
2349                 break;
2350         } case LYX_ALIGN_CENTER: {
2351                 corrected_env(os, end_tag, "center", code, lastpar, col);
2352                 break;
2353         }
2354         }
2355
2356         return output || lastpar;
2357 }
2358
2359
2360 // This one spits out the text of the paragraph
2361 void Paragraph::latex(BufferParams const & bparams,
2362         Font const & outerfont,
2363         otexstream & os,
2364         OutputParams const & runparams,
2365         int start_pos, int end_pos, bool force) const
2366 {
2367         LYXERR(Debug::LATEX, "Paragraph::latex...     " << this);
2368
2369         // FIXME This check should not be needed. Perhaps issue an
2370         // error if it triggers.
2371         Layout const & style = inInset().forcePlainLayout() ?
2372                 bparams.documentClass().plainLayout() : *d->layout_;
2373
2374         if (!force && style.inpreamble)
2375                 return;
2376
2377         bool const allowcust = allowParagraphCustomization();
2378
2379         // Current base font for all inherited font changes, without any
2380         // change caused by an individual character, except for the language:
2381         // It is set to the language of the first character.
2382         // As long as we are in the label, this font is the base font of the
2383         // label. Before the first body character it is set to the base font
2384         // of the body.
2385         Font basefont;
2386
2387         // If there is an open font-encoding changing command (script wrapper),
2388         // alien_script is set to its name
2389         string alien_script;
2390         string script;
2391
2392         // Maybe we have to create a optional argument.
2393         pos_type body_pos = beginOfBody();
2394         unsigned int column = 0;
2395
2396         // If we are inside an non inheritFont() inset, the real outerfont is local_font
2397         Font const real_outerfont = (!inInset().inheritFont()
2398                                      && runparams.local_font != nullptr)
2399                         ? Font(runparams.local_font->fontInfo()) : outerfont;
2400
2401         if (body_pos > 0) {
2402                 // the optional argument is kept in curly brackets in
2403                 // case it contains a ']'
2404                 // This is not strictly needed, but if this is changed it
2405                 // would be a file format change, and tex2lyx would need
2406                 // to be adjusted, since it unconditionally removes the
2407                 // braces when it parses \item.
2408                 os << "[{";
2409                 column += 2;
2410                 basefont = getLabelFont(bparams, real_outerfont);
2411         } else {
2412                 basefont = getLayoutFont(bparams, real_outerfont);
2413         }
2414
2415         // Which font is currently active?
2416         Font running_font(basefont);
2417         // Do we have an open font change?
2418         bool open_font = false;
2419
2420         Change runningChange = Change(Change::UNCHANGED);
2421
2422         Encoding const * const prev_encoding = runparams.encoding;
2423
2424         os.texrow().start(id(), 0);
2425
2426         // if the paragraph is empty, the loop will not be entered at all
2427         if (empty()) {
2428                 // For InTitle commands, we have already opened a group
2429                 // in output_latex::TeXOnePar.
2430                 if (style.isCommand() && !style.intitle) {
2431                         os << '{';
2432                         ++column;
2433                 }
2434                 if (!style.leftdelim().empty()) {
2435                         os << style.leftdelim();
2436                         column += style.leftdelim().size();
2437                 }
2438                 if (allowcust)
2439                         column += d->startTeXParParams(bparams, os, runparams);
2440         }
2441
2442         // Whether a \par can be issued for insets typeset inline with text.
2443         // Yes if greater than 0. This has to be static.
2444         THREAD_LOCAL_STATIC int parInline = 0;
2445
2446         for (pos_type i = 0; i < size(); ++i) {
2447                 // First char in paragraph or after label?
2448                 if (i == body_pos) {
2449                         if (body_pos > 0) {
2450                                 if (open_font) {
2451                                         bool needPar = false;
2452                                         column += running_font.latexWriteEndChanges(
2453                                                 os, bparams, runparams,
2454                                                 basefont, basefont, needPar);
2455                                         open_font = false;
2456                                 }
2457                                 basefont = getLayoutFont(bparams, real_outerfont);
2458                                 running_font = basefont;
2459
2460                                 column += Changes::latexMarkChange(os, bparams,
2461                                                 runningChange, Change(Change::UNCHANGED),
2462                                                 runparams);
2463                                 runningChange = Change(Change::UNCHANGED);
2464
2465                                 os << "}] ";
2466                                 column +=3;
2467                         }
2468                         // For InTitle commands, we have already opened a group
2469                         // in output_latex::TeXOnePar.
2470                         if (style.isCommand() && !style.intitle) {
2471                                 os << '{';
2472                                 ++column;
2473                         }
2474
2475                         if (!style.leftdelim().empty()) {
2476                                 os << style.leftdelim();
2477                                 column += style.leftdelim().size();
2478                         }
2479
2480                         if (allowcust)
2481                                 column += d->startTeXParParams(bparams, os,
2482                                                             runparams);
2483                 }
2484
2485                 runparams.wasDisplayMath = runparams.inDisplayMath;
2486                 runparams.inDisplayMath = false;
2487                 bool deleted_display_math = false;
2488                 Change const & change = runparams.inDeletedInset
2489                         ? runparams.changeOfDeletedInset : lookupChange(i);
2490
2491                 char_type const c = d->text_[i];
2492
2493                 // Check whether a display math inset follows
2494                 if (c == META_INSET
2495                     && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2496                         if (isDeleted(i))
2497                                 runparams.ctObject = getInset(i)->CtObject(runparams);
2498
2499                         InsetMath const * im = getInset(i)->asInsetMath();
2500                         if (im && im->asHullInset()
2501                             && im->asHullInset()->outerDisplay()) {
2502                                 runparams.inDisplayMath = true;
2503                                 // runparams.inDeletedInset will be set by
2504                                 // latexInset later, but we need this info
2505                                 // before it is called. On the other hand, we
2506                                 // cannot set it here because it is a counter.
2507                                 deleted_display_math = isDeleted(i);
2508                         }
2509                         if (bparams.output_changes && deleted_display_math
2510                             && runningChange == change
2511                             && change.type == Change::DELETED
2512                             && !os.afterParbreak()) {
2513                                 // A display math in the same paragraph follows.
2514                                 // We have to close and then reopen \lyxdeleted,
2515                                 // otherwise the math will be shifted up.
2516                                 OutputParams rp = runparams;
2517                                 if (open_font) {
2518                                         bool needPar = false;
2519                                         column += running_font.latexWriteEndChanges(
2520                                                 os, bparams, rp, basefont,
2521                                                 basefont, needPar);
2522                                         open_font = false;
2523                                 }
2524                                 basefont = (body_pos > i) ? getLabelFont(bparams, real_outerfont)
2525                                                           : getLayoutFont(bparams, real_outerfont);
2526                                 running_font = basefont;
2527                                 column += Changes::latexMarkChange(os, bparams,
2528                                         Change(Change::INSERTED), change, rp);
2529                         }
2530                 }
2531
2532                 if (bparams.output_changes && runningChange != change) {
2533                         if (!alien_script.empty()) {
2534                                 column += 1;
2535                                 os << "}";
2536                                 alien_script.clear();
2537                         }
2538                         if (open_font) {
2539                                 bool needPar = false;
2540                                 column += running_font.latexWriteEndChanges(
2541                                                 os, bparams, runparams,
2542                                                 basefont, basefont, needPar);
2543                                 open_font = false;
2544                         }
2545                         basefont = (body_pos > i) ? getLabelFont(bparams, real_outerfont)
2546                                                   : getLayoutFont(bparams, real_outerfont);
2547                         running_font = basefont;
2548                         column += Changes::latexMarkChange(os, bparams, runningChange,
2549                                                            change, runparams);
2550                         runningChange = change;
2551                 }
2552
2553                 // do not output text which is marked deleted
2554                 // if change tracking output is disabled
2555                 if (!bparams.output_changes && change.deleted()) {
2556                         continue;
2557                 }
2558
2559                 ++column;
2560
2561                 // Fully instantiated font
2562                 Font const current_font = getFont(bparams, i, outerfont);
2563
2564                 Font const last_font = running_font;
2565                 bool const in_ct_deletion = (bparams.output_changes
2566                                              && runningChange == change
2567                                              && change.type == Change::DELETED
2568                                              && !os.afterParbreak());
2569
2570                 // Do we need to close the previous font?
2571                 if (open_font &&
2572                     (current_font != running_font ||
2573                      current_font.language() != running_font.language()))
2574                 {
2575                         // ensure there is no open script-wrapper
2576                         if (!alien_script.empty()) {
2577                                 column += 1;
2578                                 os << "}";
2579                                 alien_script.clear();
2580                         }
2581                         bool needPar = false;
2582                         if (in_ct_deletion) {
2583                                 // We have to close and then reopen \lyxdeleted,
2584                                 // as strikeout needs to be on lowest level.
2585                                 os << '}';
2586                                 column += 1;
2587                         }
2588                         column += running_font.latexWriteEndChanges(
2589                                     os, bparams, runparams, basefont,
2590                                     (i == body_pos-1) ? basefont : current_font,
2591                                     needPar);
2592                         if (in_ct_deletion) {
2593                                 // We have to close and then reopen \lyxdeleted,
2594                                 // as strikeout needs to be on lowest level.
2595                                 OutputParams rp = runparams;
2596                                 column += Changes::latexMarkChange(os, bparams,
2597                                         Change(Change::UNCHANGED), Change(Change::DELETED), rp);
2598                         }
2599                         running_font = basefont;
2600                         open_font = false;
2601                 }
2602
2603                 // if necessary, close language environment before opening CJK
2604                 string const running_lang = running_font.language()->babel();
2605                 string const lang_end_command = lyxrc.language_command_end;
2606                 if (!lang_end_command.empty() && !bparams.useNonTeXFonts
2607                         && !running_lang.empty()
2608                         && running_lang == openLanguageName()
2609                         && current_font.language()->encoding()->package() == Encoding::CJK) {
2610                         string end_tag = subst(lang_end_command, "$$lang", running_lang);
2611                         os << from_ascii(end_tag);
2612                         column += end_tag.length();
2613                         popLanguageName();
2614                 }
2615
2616                 // Switch file encoding if necessary (and allowed)
2617                 if (!runparams.pass_thru && !style.pass_thru &&
2618                     runparams.encoding->package() != Encoding::none &&
2619                     current_font.language()->encoding()->package() != Encoding::none) {
2620                         pair<bool, int> const enc_switch =
2621                                 switchEncoding(os.os(), bparams, runparams,
2622                                         *(current_font.language()->encoding()));
2623                         if (enc_switch.first) {
2624                                 column += enc_switch.second;
2625                                 runparams.encoding = current_font.language()->encoding();
2626                         }
2627                 }
2628
2629                 // A display math inset inside an ulem command will be output
2630                 // as a box of width \linewidth, so we have to either disable
2631                 // indentation if the inset starts a paragraph, or start a new
2632                 // line to accommodate such box. This has to be done before
2633                 // writing any font changing commands.
2634                 if (runparams.inDisplayMath && !deleted_display_math
2635                     && runparams.inulemcmd) {
2636                         if (os.afterParbreak())
2637                                 os << "\\noindent";
2638                         else
2639                                 os << "\\\\\n";
2640                 }
2641
2642                 // Do we need to change font?
2643                 if ((current_font != running_font ||
2644                      current_font.language() != running_font.language())
2645                     && i != body_pos - 1)
2646                 {
2647                         if (in_ct_deletion) {
2648                                 // We have to close and then reopen \lyxdeleted,
2649                                 // as strikeout needs to be on lowest level.
2650                                 bool needPar = false;
2651                                 OutputParams rp = runparams;
2652                                 column += running_font.latexWriteEndChanges(
2653                                         os, bparams, rp, basefont,
2654                                         basefont, needPar);
2655                                 os << '}';
2656                                 column += 1;
2657                         }
2658                         otexstringstream ots;
2659                         bool const non_inherit_inset = (c == META_INSET && getInset(i) && !getInset(i)->inheritFont());
2660                         column += current_font.latexWriteStartChanges(ots, bparams,
2661                                                               runparams, basefont,
2662                                                               last_font, non_inherit_inset);
2663                         // Check again for display math in ulem commands as a
2664                         // font change may also occur just before a math inset.
2665                         if (runparams.inDisplayMath && !deleted_display_math
2666                             && runparams.inulemcmd) {
2667                                 if (os.afterParbreak())
2668                                         os << "\\noindent";
2669                                 else
2670                                         os << "\\\\\n";
2671                         }
2672                         running_font = current_font;
2673                         open_font = true;
2674                         docstring fontchange = ots.str();
2675                         os << fontchange;
2676                         // check whether the fontchange ends with a \\textcolor
2677                         // modifier and the text starts with a space. If so we
2678                         // need to add } in order to prevent \\textcolor from gobbling
2679                         // the space (bug 4473).
2680                         docstring const last_modifier = rsplit(fontchange, '\\');
2681                         if (prefixIs(last_modifier, from_ascii("textcolor")) && c == ' ')
2682                                 os << from_ascii("{}");
2683                         else if (ots.terminateCommand())
2684                                 os << termcmd;
2685                         if (in_ct_deletion) {
2686                                 // We have to close and then reopen \lyxdeleted,
2687                                 // as strikeout needs to be on lowest level.
2688                                 OutputParams rp = runparams;
2689                                 column += Changes::latexMarkChange(os, bparams,
2690                                         Change(Change::UNCHANGED), change, rp);
2691                         }
2692                 }
2693
2694                 // FIXME: think about end_pos implementation...
2695                 if (c == ' ' && i >= start_pos && (end_pos == -1 || i < end_pos)) {
2696                         // FIXME: integrate this case in latexSpecialChar
2697                         // Do not print the separation of the optional argument
2698                         // if style.pass_thru is false. This works because
2699                         // latexSpecialChar ignores spaces if
2700                         // style.pass_thru is false.
2701                         if (i != body_pos - 1) {
2702                                 if (d->simpleTeXBlanks(bparams, runparams, os,
2703                                                 i, column, current_font, style)) {
2704                                         // A surrogate pair was output. We
2705                                         // must not call latexSpecialChar
2706                                         // in this iteration, since it would output
2707                                         // the combining character again.
2708                                         ++i;
2709                                         continue;
2710                                 }
2711                         }
2712                 }
2713
2714                 OutputParams rp = runparams;
2715                 rp.free_spacing = style.free_spacing;
2716                 rp.local_font = &current_font;
2717                 rp.intitle = style.intitle;
2718
2719                 // Two major modes:  LaTeX or plain
2720                 // Handle here those cases common to both modes
2721                 // and then split to handle the two modes separately.
2722                 if (c == META_INSET) {
2723                         if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2724                                 // Greyedout notes and, in general, all insets
2725                                 // with InsetLayout::isDisplay() == false,
2726                                 // are typeset inline with the text. So, we
2727                                 // can add a \par to the last paragraph of
2728                                 // such insets only if nothing else follows.
2729                                 bool incremented = false;
2730                                 Inset const * inset = getInset(i);
2731                                 InsetText const * textinset = inset
2732                                                         ? inset->asInsetText()
2733                                                         : nullptr;
2734                                 if (i + 1 == size() && textinset
2735                                     && !inset->getLayout().isDisplay()) {
2736                                         ParagraphList const & pars =
2737                                                 textinset->text().paragraphs();
2738                                         pit_type const pit = pars.size() - 1;
2739                                         Font const lastfont =
2740                                                 pit < 0 || pars[pit].empty()
2741                                                 ? pars[pit].getLayoutFont(
2742                                                                 bparams,
2743                                                                 real_outerfont)
2744                                                 : pars[pit].getFont(bparams,
2745                                                         pars[pit].size() - 1,
2746                                                         real_outerfont);
2747                                         if (lastfont.fontInfo().size() !=
2748                                             basefont.fontInfo().size()) {
2749                                                 ++parInline;
2750                                                 incremented = true;
2751                                         }
2752                                 }
2753                                 d->latexInset(bparams, os, rp, running_font,
2754                                                 basefont, real_outerfont, open_font,
2755                                                 runningChange, style, i, column);
2756                                 if (incremented)
2757                                         --parInline;
2758
2759                                 if (runparams.ctObject == OutputParams::CT_DISPLAYOBJECT
2760                                     || runparams.ctObject == OutputParams::CT_UDISPLAYOBJECT) {
2761                                         // Close \lyx*deleted and force its
2762                                         // reopening (if needed)
2763                                         os << '}';
2764                                         column++;
2765                                         runningChange = Change(Change::UNCHANGED);
2766                                         runparams.ctObject = OutputParams::CT_NORMAL;
2767                                 }
2768                         }
2769                 } else if (i >= start_pos && (end_pos == -1 || i < end_pos)) {
2770                         if (!bparams.useNonTeXFonts)
2771                           script = Encodings::isKnownScriptChar(c);
2772                         if (script != alien_script) {
2773                                 if (!alien_script.empty()) {
2774                                         os << "}";
2775                                         alien_script.clear();
2776                                 }
2777                                 string fontenc = running_font.language()->fontenc(bparams);
2778                                 if (!script.empty()
2779                                         && !Encodings::fontencSupportsScript(fontenc, script)) {
2780                                         column += script.length() + 2;
2781                                         os << "\\" << script << "{";
2782                                         alien_script = script;
2783                                 }
2784                         }
2785                         try {
2786                                 d->latexSpecialChar(os, bparams, rp, running_font,
2787                                                                         alien_script, style, i, end_pos, column);
2788                         } catch (EncodingException & e) {
2789                                 if (runparams.dryrun) {
2790                                         os << "<" << _("LyX Warning: ")
2791                                            << _("uncodable character") << " '";
2792                                         os.put(c);
2793                                         os << "'>";
2794                                 } else {
2795                                         // add location information and throw again.
2796                                         e.par_id = id();
2797                                         e.pos = i;
2798                                         throw(e);
2799                                 }
2800                         }
2801                 }
2802
2803                 // Set the encoding to that returned from latexSpecialChar (see
2804                 // comment for encoding member in OutputParams.h)
2805                 runparams.encoding = rp.encoding;
2806
2807                 // Also carry on the info on a closed ulem command for insets
2808                 // such as Note that do not produce any output, so that no
2809                 // command is ever executed but its opening was recorded.
2810                 runparams.inulemcmd = rp.inulemcmd;
2811
2812                 // These need to be passed upstream as well
2813                 runparams.need_maketitle = rp.need_maketitle;
2814                 runparams.have_maketitle = rp.have_maketitle;
2815
2816                 // And finally, pass the post_macros upstream
2817                 runparams.post_macro = rp.post_macro;
2818         }
2819
2820         // Close wrapper for alien script
2821         if (!alien_script.empty()) {
2822                 os << "}";
2823                 alien_script.clear();
2824         }
2825
2826         // If we have an open font definition, we have to close it
2827         if (open_font) {
2828                 // Make sure that \\par is done with the font of the last
2829                 // character if this has another size as the default.
2830                 // This is necessary because LaTeX (and LyX on the screen)
2831                 // calculates the space between the baselines according
2832                 // to this font. (Matthias)
2833                 //
2834                 // We must not change the font for the last paragraph
2835                 // of non-multipar insets, tabular cells or commands,
2836                 // since this produces unwanted whitespace.
2837
2838                 Font const font = empty()
2839                         ? getLayoutFont(bparams, real_outerfont)
2840                         : getFont(bparams, size() - 1, real_outerfont);
2841
2842                 InsetText const * textinset = inInset().asInsetText();
2843
2844                 bool const maintext = textinset
2845                         ? textinset->text().isMainText()
2846                         : false;
2847
2848                 size_t const numpars = textinset
2849                         ? textinset->text().paragraphs().size()
2850                         : 0;
2851
2852                 bool needPar = false;
2853
2854                 if (style.resfont.size() != font.fontInfo().size()
2855                     && (!runparams.isLastPar || maintext
2856                         || (numpars > 1 && d->ownerCode() != CELL_CODE
2857                             && (inInset().getLayout().isDisplay()
2858                                 || parInline)))
2859                     && !style.isCommand()) {
2860                         needPar = true;
2861                 }
2862 #ifdef FIXED_LANGUAGE_END_DETECTION
2863                 if (next_) {
2864                         running_font.latexWriteEndChanges(os, bparams,
2865                                         runparams, basefont,
2866                                         next_->getFont(bparams, 0, outerfont),
2867                                                        needPar);
2868                 } else {
2869                         running_font.latexWriteEndChanges(os, bparams,
2870                                         runparams, basefont, basefont, needPar);
2871                 }
2872 #else
2873 //FIXME: For now we ALWAYS have to close the foreign font settings if they are
2874 //FIXME: there as we start another \selectlanguage with the next paragraph if
2875 //FIXME: we are in need of this. This should be fixed sometime (Jug)
2876                 running_font.latexWriteEndChanges(os, bparams, runparams,
2877                                 basefont, basefont, needPar);
2878 #endif
2879                 if (needPar) {
2880                         // The \par could not be inserted at the same nesting
2881                         // level of the font size change, so do it now.
2882                         os << "{\\" << font.latexSize() << "\\par}";
2883                 }
2884         }
2885
2886         column += Changes::latexMarkChange(os, bparams, runningChange,
2887                                            Change(Change::UNCHANGED), runparams);
2888
2889         // Needed if there is an optional argument but no contents.
2890         if (body_pos > 0 && body_pos == size()) {
2891                 os << "}]~";
2892         }
2893
2894         if (!style.rightdelim().empty()) {
2895                 os << style.rightdelim();
2896                 column += style.rightdelim().size();
2897         }
2898
2899         if (allowcust && d->endTeXParParams(bparams, os, runparams)
2900             && runparams.encoding != prev_encoding) {
2901                 runparams.encoding = prev_encoding;
2902                 os << setEncoding(prev_encoding->iconvName());
2903         }
2904
2905         LYXERR(Debug::LATEX, "Paragraph::latex... done " << this);
2906 }
2907
2908
2909 bool Paragraph::emptyTag() const
2910 {
2911         for (pos_type i = 0; i < size(); ++i) {
2912                 if (Inset const * inset = getInset(i)) {
2913                         InsetCode lyx_code = inset->lyxCode();
2914                         // FIXME testing like that is wrong. What is
2915                         // the intent?
2916                         if (lyx_code != TOC_CODE &&
2917                             lyx_code != INCLUDE_CODE &&
2918                             lyx_code != GRAPHICS_CODE &&
2919                             lyx_code != ERT_CODE &&
2920                             lyx_code != LISTINGS_CODE &&
2921                             lyx_code != FLOAT_CODE &&
2922                             lyx_code != TABULAR_CODE) {
2923                                 return false;
2924                         }
2925                 } else {
2926                         char_type c = d->text_[i];
2927                         if (c != ' ' && c != '\t')
2928                                 return false;
2929                 }
2930         }
2931         return true;
2932 }
2933
2934
2935 string Paragraph::getID(Buffer const &, OutputParams const &)
2936         const
2937 {
2938         for (pos_type i = 0; i < size(); ++i) {
2939                 if (Inset const * inset = getInset(i)) {
2940                         InsetCode lyx_code = inset->lyxCode();
2941                         if (lyx_code == LABEL_CODE) {
2942                                 InsetLabel const * const il = static_cast<InsetLabel const *>(inset);
2943                                 docstring const & id = il->getParam("name");
2944                                 return "id='" + to_utf8(xml::cleanID(id)) + "'";
2945                         }
2946                 }
2947         }
2948         return string();
2949 }
2950
2951
2952 pos_type Paragraph::firstWordDocBook(XMLStream & xs, OutputParams const & runparams) const
2953 {
2954         pos_type i;
2955         for (i = 0; i < size(); ++i) {
2956                 if (Inset const * inset = getInset(i)) {
2957                         inset->docbook(xs, runparams);
2958                 } else {
2959                         char_type c = d->text_[i];
2960                         if (c == ' ')
2961                                 break;
2962                         xs << c;
2963                 }
2964         }
2965         return i;
2966 }
2967
2968
2969 pos_type Paragraph::firstWordLyXHTML(XMLStream & xs, OutputParams const & runparams)
2970         const
2971 {
2972         pos_type i;
2973         for (i = 0; i < size(); ++i) {
2974                 if (Inset const * inset = getInset(i)) {
2975                         inset->xhtml(xs, runparams);
2976                 } else {
2977                         char_type c = d->text_[i];
2978                         if (c == ' ')
2979                                 break;
2980                         xs << c;
2981                 }
2982         }
2983         return i;
2984 }
2985
2986
2987 bool Paragraph::Private::onlyText(Buffer const & buf, Font const & outerfont, pos_type initial) const
2988 {
2989         Font font_old;
2990         pos_type size = text_.size();
2991         for (pos_type i = initial; i < size; ++i) {
2992                 Font font = owner_->getFont(buf.params(), i, outerfont);
2993                 if (text_[i] == META_INSET)
2994                         return false;
2995                 if (i != initial && font != font_old)
2996                         return false;
2997                 font_old = font;
2998         }
2999
3000         return true;
3001 }
3002
3003
3004 namespace {
3005
3006 void doFontSwitchDocBook(vector<xml::FontTag> & tagsToOpen,
3007                   vector<xml::EndFontTag> & tagsToClose,
3008                   bool & flag, FontState curstate, xml::FontTypes type)
3009 {
3010         if (curstate == FONT_ON) {
3011                 tagsToOpen.push_back(docbookStartFontTag(type));
3012                 flag = true;
3013         } else if (flag) {
3014                 tagsToClose.push_back(docbookEndFontTag(type));
3015                 flag = false;
3016         }
3017 }
3018
3019 class OptionalFontType {
3020 public:
3021         xml::FontTypes ft;
3022         bool has_value;
3023
3024         OptionalFontType(): ft(xml::FT_EMPH), has_value(false) {} // A possible value at random for ft.
3025         OptionalFontType(xml::FontTypes ft): ft(ft), has_value(true) {}
3026 };
3027
3028 OptionalFontType fontShapeToXml(FontShape fs)
3029 {
3030         switch (fs) {
3031         case ITALIC_SHAPE:
3032                 return {xml::FT_ITALIC};
3033         case SLANTED_SHAPE:
3034                 return {xml::FT_SLANTED};
3035         case SMALLCAPS_SHAPE:
3036                 return {xml::FT_SMALLCAPS};
3037         case UP_SHAPE:
3038         case INHERIT_SHAPE:
3039                 return {};
3040         default:
3041                 // the other tags are for internal use
3042                 LATTEST(false);
3043                 return {};
3044         }
3045 }
3046
3047 OptionalFontType fontFamilyToXml(FontFamily fm)
3048 {
3049         switch (fm) {
3050         case ROMAN_FAMILY:
3051                 return {xml::FT_ROMAN};
3052         case SANS_FAMILY:
3053                 return {xml::FT_SANS};
3054         case TYPEWRITER_FAMILY:
3055                 return {xml::FT_TYPE};
3056         case INHERIT_FAMILY:
3057                 return {};
3058         default:
3059                 // the other tags are for internal use
3060                 LATTEST(false);
3061                 return {};
3062         }
3063 }
3064
3065 OptionalFontType fontSizeToXml(FontSize fs)
3066 {
3067         switch (fs) {
3068         case TINY_SIZE:
3069                 return {xml::FT_SIZE_TINY};
3070         case SCRIPT_SIZE:
3071                 return {xml::FT_SIZE_SCRIPT};
3072         case FOOTNOTE_SIZE:
3073                 return {xml::FT_SIZE_FOOTNOTE};
3074         case SMALL_SIZE:
3075                 return {xml::FT_SIZE_SMALL};
3076         case LARGE_SIZE:
3077                 return {xml::FT_SIZE_LARGE};
3078         case LARGER_SIZE:
3079                 return {xml::FT_SIZE_LARGER};
3080         case LARGEST_SIZE:
3081                 return {xml::FT_SIZE_LARGEST};
3082         case HUGE_SIZE:
3083                 return {xml::FT_SIZE_HUGE};
3084         case HUGER_SIZE:
3085                 return {xml::FT_SIZE_HUGER};
3086         case INCREASE_SIZE:
3087                 return {xml::FT_SIZE_INCREASE};
3088         case DECREASE_SIZE:
3089                 return {xml::FT_SIZE_DECREASE};
3090         case INHERIT_SIZE:
3091         case NORMAL_SIZE:
3092                 return {};
3093         default:
3094                 // the other tags are for internal use
3095                 LATTEST(false);
3096                 return {};
3097         }
3098 }
3099
3100 struct DocBookFontState
3101 {
3102         FontShape  curr_fs   = INHERIT_SHAPE;
3103         FontFamily curr_fam  = INHERIT_FAMILY;
3104         FontSize   curr_size = INHERIT_SIZE;
3105
3106         // track whether we have opened these tags
3107         bool emph_flag = false;
3108         bool bold_flag = false;
3109         bool noun_flag = false;
3110         bool ubar_flag = false;
3111         bool dbar_flag = false;
3112         bool sout_flag = false;
3113         bool xout_flag = false;
3114         bool wave_flag = false;
3115         // shape tags
3116         bool shap_flag = false;
3117         // family tags
3118         bool faml_flag = false;
3119         // size tags
3120         bool size_flag = false;
3121 };
3122
3123 std::tuple<vector<xml::FontTag>, vector<xml::EndFontTag>> computeDocBookFontSwitch(FontInfo const & font_old,
3124                                                                                            Font const & font,
3125                                                                                            std::string const & default_family,
3126                                                                                            DocBookFontState & fs)
3127 {
3128         vector<xml::FontTag> tagsToOpen;
3129         vector<xml::EndFontTag> tagsToClose;
3130
3131         // emphasis
3132         FontState curstate = font.fontInfo().emph();
3133         if (font_old.emph() != curstate)
3134                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.emph_flag, curstate, xml::FT_EMPH);
3135
3136         // noun
3137         curstate = font.fontInfo().noun();
3138         if (font_old.noun() != curstate)
3139                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.noun_flag, curstate, xml::FT_NOUN);
3140
3141         // underbar
3142         curstate = font.fontInfo().underbar();
3143         if (font_old.underbar() != curstate)
3144                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.ubar_flag, curstate, xml::FT_UBAR);
3145
3146         // strikeout
3147         curstate = font.fontInfo().strikeout();
3148         if (font_old.strikeout() != curstate)
3149                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.sout_flag, curstate, xml::FT_SOUT);
3150
3151         // xout
3152         curstate = font.fontInfo().xout();
3153         if (font_old.xout() != curstate)
3154                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.xout_flag, curstate, xml::FT_XOUT);
3155
3156         // double underbar
3157         curstate = font.fontInfo().uuline();
3158         if (font_old.uuline() != curstate)
3159                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.dbar_flag, curstate, xml::FT_DBAR);
3160
3161         // wavy line
3162         curstate = font.fontInfo().uwave();
3163         if (font_old.uwave() != curstate)
3164                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.wave_flag, curstate, xml::FT_WAVE);
3165
3166         // bold
3167         // a little hackish, but allows us to reuse what we have.
3168         curstate = (font.fontInfo().series() == BOLD_SERIES ? FONT_ON : FONT_OFF);
3169         if (font_old.series() != font.fontInfo().series())
3170                 doFontSwitchDocBook(tagsToOpen, tagsToClose, fs.bold_flag, curstate, xml::FT_BOLD);
3171
3172         // Font shape
3173         fs.curr_fs = font.fontInfo().shape();
3174         FontShape old_fs = font_old.shape();
3175         if (old_fs != fs.curr_fs) {
3176                 if (fs.shap_flag) {
3177                         OptionalFontType tag = fontShapeToXml(old_fs);
3178                         if (tag.has_value)
3179                                 tagsToClose.push_back(docbookEndFontTag(tag.ft));
3180                         fs.shap_flag = false;
3181                 }
3182
3183                 OptionalFontType tag = fontShapeToXml(fs.curr_fs);
3184                 if (tag.has_value)
3185                         tagsToOpen.push_back(docbookStartFontTag(tag.ft));
3186         }
3187
3188         // Font family
3189         fs.curr_fam = font.fontInfo().family();
3190         FontFamily old_fam = font_old.family();
3191         if (old_fam != fs.curr_fam) {
3192                 if (fs.faml_flag) {
3193                         OptionalFontType tag = fontFamilyToXml(old_fam);
3194                         if (tag.has_value)
3195                                 tagsToClose.push_back(docbookEndFontTag(tag.ft));
3196                         fs.faml_flag = false;
3197                 }
3198                 switch (fs.curr_fam) {
3199                         case ROMAN_FAMILY:
3200                                 // we will treat a "default" font family as roman, since we have
3201                                 // no other idea what to do.
3202                                 if (default_family != "rmdefault" && default_family != "default") {
3203                                         tagsToOpen.push_back(docbookStartFontTag(xml::FT_ROMAN));
3204                                         fs.faml_flag = true;
3205                                 }
3206                                 break;
3207                         case SANS_FAMILY:
3208                                 if (default_family != "sfdefault") {
3209                                         tagsToOpen.push_back(docbookStartFontTag(xml::FT_SANS));
3210                                         fs.faml_flag = true;
3211                                 }
3212                                 break;
3213                         case TYPEWRITER_FAMILY:
3214                                 if (default_family != "ttdefault") {
3215                                         tagsToOpen.push_back(docbookStartFontTag(xml::FT_TYPE));
3216                                         fs.faml_flag = true;
3217                                 }
3218                                 break;
3219                         case INHERIT_FAMILY:
3220                                 break;
3221                         default:
3222                                 // the other tags are for internal use
3223                                 LATTEST(false);
3224                                 break;
3225                 }
3226         }
3227
3228         // Font size
3229         fs.curr_size = font.fontInfo().size();
3230         FontSize old_size = font_old.size();
3231         if (old_size != fs.curr_size) {
3232                 if (fs.size_flag) {
3233                         OptionalFontType tag = fontSizeToXml(old_size);
3234                         if (tag.has_value)
3235                                 tagsToClose.push_back(docbookEndFontTag(tag.ft));
3236                         fs.size_flag = false;
3237                 }
3238
3239                 OptionalFontType tag = fontSizeToXml(fs.curr_size);
3240                 if (tag.has_value) {
3241                         tagsToOpen.push_back(docbookStartFontTag(tag.ft));
3242                         fs.size_flag = true;
3243                 }
3244         }
3245
3246         return std::tuple<vector<xml::FontTag>, vector<xml::EndFontTag>>(tagsToOpen, tagsToClose);
3247 }
3248
3249 } // anonymous namespace
3250
3251
3252 void Paragraph::simpleDocBookOnePar(Buffer const & buf,
3253                                     XMLStream & xs,
3254                                     OutputParams const & runparams,
3255                                     Font const & outerfont,
3256                                     bool start_paragraph, bool close_paragraph,
3257                                     pos_type initial) const
3258 {
3259         // track whether we have opened these tags
3260         DocBookFontState fs;
3261
3262         if (start_paragraph)
3263                 xs.startDivision(allowEmpty());
3264
3265         Layout const & style = *d->layout_;
3266         FontInfo font_old =
3267                         style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
3268
3269         string const default_family =
3270                         buf.masterBuffer()->params().fonts_default_family;
3271
3272         vector<xml::FontTag> tagsToOpen;
3273         vector<xml::EndFontTag> tagsToClose;
3274
3275         // parsing main loop
3276         for (pos_type i = initial; i < size(); ++i) {
3277                 // let's not show deleted material in the output
3278                 if (isDeleted(i))
3279                         continue;
3280
3281                 Font const font = getFont(buf.masterBuffer()->params(), i, outerfont);
3282
3283                 // Determine which tags should be opened or closed.
3284                 tie(tagsToOpen, tagsToClose) = computeDocBookFontSwitch(font_old, font, default_family, fs);
3285
3286                 // FIXME XHTML
3287                 // Other such tags? What about the other text ranges?
3288
3289                 vector<xml::EndFontTag>::const_iterator cit = tagsToClose.begin();
3290                 vector<xml::EndFontTag>::const_iterator cen = tagsToClose.end();
3291                 for (; cit != cen; ++cit)
3292                         xs << *cit;
3293
3294                 vector<xml::FontTag>::const_iterator sit = tagsToOpen.begin();
3295                 vector<xml::FontTag>::const_iterator sen = tagsToOpen.end();
3296                 for (; sit != sen; ++sit)
3297                         xs << *sit;
3298
3299                 tagsToClose.clear();
3300                 tagsToOpen.clear();
3301
3302                 if (Inset const * inset = getInset(i)) {
3303                         if (!runparams.for_toc || inset->isInToc()) {
3304                                 OutputParams np = runparams;
3305                                 np.local_font = &font;
3306                                 // If the paragraph has size 1, then we are in the "special
3307                                 // case" where we do not output the containing paragraph info.
3308                                 // This "special case" is defined in more details in output_docbook.cpp, makeParagraphs. The results
3309                                 // of that brittle logic is passed to this function through open_par.
3310                                 if (!inset->getLayout().htmlisblock() && size() != 1) // TODO: htmlisblock here too!
3311                                         np.docbook_in_par = true;
3312                                 inset->docbook(xs, np);
3313                         }
3314                 } else {
3315                         char_type c = getUChar(buf.masterBuffer()->params(), runparams, i);
3316                         xs << c;
3317                 }
3318                 font_old = font.fontInfo();
3319         }
3320
3321         // FIXME, this code is just imported from XHTML
3322         // I'm worried about what happens if a branch, say, is itself
3323         // wrapped in some font stuff. I think that will not work.
3324         xs.closeFontTags();
3325         if (runparams.docbook_in_listing)
3326                 xs << xml::CR();
3327         if (close_paragraph)
3328                 xs.endDivision();
3329 }
3330
3331
3332 namespace {
3333
3334 void doFontSwitchXHTML(vector<xml::FontTag> & tagsToOpen,
3335                   vector<xml::EndFontTag> & tagsToClose,
3336                   bool & flag, FontState curstate, xml::FontTypes type)
3337 {
3338         if (curstate == FONT_ON) {
3339                 tagsToOpen.push_back(xhtmlStartFontTag(type));
3340                 flag = true;
3341         } else if (flag) {
3342                 tagsToClose.push_back(xhtmlEndFontTag(type));
3343                 flag = false;
3344         }
3345 }
3346
3347 } // anonymous namespace
3348
3349
3350 docstring Paragraph::simpleLyXHTMLOnePar(Buffer const & buf,
3351                                     XMLStream & xs,
3352                                     OutputParams const & runparams,
3353                                     Font const & outerfont,
3354                                     bool start_paragraph, bool close_paragraph,
3355                                     pos_type initial) const
3356 {
3357         docstring retval;
3358
3359         // track whether we have opened these tags
3360         bool emph_flag = false;
3361         bool bold_flag = false;
3362         bool noun_flag = false;
3363         bool ubar_flag = false;
3364         bool dbar_flag = false;
3365         bool sout_flag = false;
3366         bool xout_flag = false;
3367         bool wave_flag = false;
3368         // shape tags
3369         bool shap_flag = false;
3370         // family tags
3371         bool faml_flag = false;
3372         // size tags
3373         bool size_flag = false;
3374
3375         Layout const & style = *d->layout_;
3376
3377         if (start_paragraph)
3378                 xs.startDivision(allowEmpty());
3379
3380         FontInfo font_old =
3381                 style.labeltype == LABEL_MANUAL ? style.labelfont : style.font;
3382
3383         FontShape  curr_fs   = INHERIT_SHAPE;
3384         FontFamily curr_fam  = INHERIT_FAMILY;
3385         FontSize   curr_size = INHERIT_SIZE;
3386
3387         string const default_family =
3388                 buf.masterBuffer()->params().fonts_default_family;
3389
3390         vector<xml::FontTag> tagsToOpen;
3391         vector<xml::EndFontTag> tagsToClose;
3392
3393         // parsing main loop
3394         for (pos_type i = initial; i < size(); ++i) {
3395                 // let's not show deleted material in the output
3396                 if (isDeleted(i))
3397                         continue;
3398
3399                 Font const font = getFont(buf.masterBuffer()->params(), i, outerfont);
3400
3401                 // emphasis
3402                 FontState curstate = font.fontInfo().emph();
3403                 if (font_old.emph() != curstate)
3404                         doFontSwitchXHTML(tagsToOpen, tagsToClose, emph_flag, curstate, xml::FT_EMPH);
3405
3406                 // noun
3407                 curstate = font.fontInfo().noun();
3408                 if (font_old.noun() != curstate)
3409                         doFontSwitchXHTML(tagsToOpen, tagsToClose, noun_flag, curstate, xml::FT_NOUN);
3410
3411                 // underbar
3412                 curstate = font.fontInfo().underbar();
3413                 if (font_old.underbar() != curstate)
3414                         doFontSwitchXHTML(tagsToOpen, tagsToClose, ubar_flag, curstate, xml::FT_UBAR);
3415
3416                 // strikeout
3417                 curstate = font.fontInfo().strikeout();
3418                 if (font_old.strikeout() != curstate)
3419                         doFontSwitchXHTML(tagsToOpen, tagsToClose, sout_flag, curstate, xml::FT_SOUT);
3420
3421                 // xout
3422                 curstate = font.fontInfo().xout();
3423                 if (font_old.xout() != curstate)
3424                         doFontSwitchXHTML(tagsToOpen, tagsToClose, xout_flag, curstate, xml::FT_XOUT);
3425
3426                 // double underbar
3427                 curstate = font.fontInfo().uuline();
3428                 if (font_old.uuline() != curstate)
3429                         doFontSwitchXHTML(tagsToOpen, tagsToClose, dbar_flag, curstate, xml::FT_DBAR);
3430
3431                 // wavy line
3432                 curstate = font.fontInfo().uwave();
3433                 if (font_old.uwave() != curstate)
3434                         doFontSwitchXHTML(tagsToOpen, tagsToClose, wave_flag, curstate, xml::FT_WAVE);
3435
3436                 // bold
3437                 // a little hackish, but allows us to reuse what we have.
3438                 curstate = (font.fontInfo().series() == BOLD_SERIES ? FONT_ON : FONT_OFF);
3439                 if (font_old.series() != font.fontInfo().series())
3440                         doFontSwitchXHTML(tagsToOpen, tagsToClose, bold_flag, curstate, xml::FT_BOLD);
3441
3442                 // Font shape
3443                 curr_fs = font.fontInfo().shape();
3444                 FontShape old_fs = font_old.shape();
3445                 if (old_fs != curr_fs) {
3446                         if (shap_flag) {
3447                                 switch (old_fs) {
3448                                 case ITALIC_SHAPE:
3449                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_ITALIC));
3450                                         break;
3451                                 case SLANTED_SHAPE:
3452                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SLANTED));
3453                                         break;
3454                                 case SMALLCAPS_SHAPE:
3455                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SMALLCAPS));
3456                                         break;
3457                                 case UP_SHAPE:
3458                                 case INHERIT_SHAPE:
3459                                         break;
3460                                 default:
3461                                         // the other tags are for internal use
3462                                         LATTEST(false);
3463                                         break;
3464                                 }
3465                                 shap_flag = false;
3466                         }
3467                         switch (curr_fs) {
3468                         case ITALIC_SHAPE:
3469                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_ITALIC));
3470                                 shap_flag = true;
3471                                 break;
3472                         case SLANTED_SHAPE:
3473                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SLANTED));
3474                                 shap_flag = true;
3475                                 break;
3476                         case SMALLCAPS_SHAPE:
3477                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SMALLCAPS));
3478                                 shap_flag = true;
3479                                 break;
3480                         case UP_SHAPE:
3481                         case INHERIT_SHAPE:
3482                                 break;
3483                         default:
3484                                 // the other tags are for internal use
3485                                 LATTEST(false);
3486                                 break;
3487                         }
3488                 }
3489
3490                 // Font family
3491                 curr_fam = font.fontInfo().family();
3492                 FontFamily old_fam = font_old.family();
3493                 if (old_fam != curr_fam) {
3494                         if (faml_flag) {
3495                                 switch (old_fam) {
3496                                 case ROMAN_FAMILY:
3497                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_ROMAN));
3498                                         break;
3499                                 case SANS_FAMILY:
3500                                     tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SANS));
3501                                     break;
3502                                 case TYPEWRITER_FAMILY:
3503                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_TYPE));
3504                                         break;
3505                                 case INHERIT_FAMILY:
3506                                         break;
3507                                 default:
3508                                         // the other tags are for internal use
3509                                         LATTEST(false);
3510                                         break;
3511                                 }
3512                                 faml_flag = false;
3513                         }
3514                         switch (curr_fam) {
3515                         case ROMAN_FAMILY:
3516                                 // we will treat a "default" font family as roman, since we have
3517                                 // no other idea what to do.
3518                                 if (default_family != "rmdefault" && default_family != "default") {
3519                                         tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_ROMAN));
3520                                         faml_flag = true;
3521                                 }
3522                                 break;
3523                         case SANS_FAMILY:
3524                                 if (default_family != "sfdefault") {
3525                                         tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SANS));
3526                                         faml_flag = true;
3527                                 }
3528                                 break;
3529                         case TYPEWRITER_FAMILY:
3530                                 if (default_family != "ttdefault") {
3531                                         tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_TYPE));
3532                                         faml_flag = true;
3533                                 }
3534                                 break;
3535                         case INHERIT_FAMILY:
3536                                 break;
3537                         default:
3538                                 // the other tags are for internal use
3539                                 LATTEST(false);
3540                                 break;
3541                         }
3542                 }
3543
3544                 // Font size
3545                 curr_size = font.fontInfo().size();
3546                 FontSize old_size = font_old.size();
3547                 if (old_size != curr_size) {
3548                         if (size_flag) {
3549                                 switch (old_size) {
3550                                 case TINY_SIZE:
3551                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_TINY));
3552                                         break;
3553                                 case SCRIPT_SIZE:
3554                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_SCRIPT));
3555                                         break;
3556                                 case FOOTNOTE_SIZE:
3557                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_FOOTNOTE));
3558                                         break;
3559                                 case SMALL_SIZE:
3560                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_SMALL));
3561                                         break;
3562                                 case LARGE_SIZE:
3563                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_LARGE));
3564                                         break;
3565                                 case LARGER_SIZE:
3566                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_LARGER));
3567                                         break;
3568                                 case LARGEST_SIZE:
3569                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_LARGEST));
3570                                         break;
3571                                 case HUGE_SIZE:
3572                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_HUGE));
3573                                         break;
3574                                 case HUGER_SIZE:
3575                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_HUGER));
3576                                         break;
3577                                 case INCREASE_SIZE:
3578                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_INCREASE));
3579                                         break;
3580                                 case DECREASE_SIZE:
3581                                         tagsToClose.emplace_back(xhtmlEndFontTag(xml::FT_SIZE_DECREASE));
3582                                         break;
3583                                 case INHERIT_SIZE:
3584                                 case NORMAL_SIZE:
3585                                         break;
3586                                 default:
3587                                         // the other tags are for internal use
3588                                         LATTEST(false);
3589                                         break;
3590                                 }
3591                                 size_flag = false;
3592                         }
3593                         switch (curr_size) {
3594                         case TINY_SIZE:
3595                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_TINY));
3596                                 size_flag = true;
3597                                 break;
3598                         case SCRIPT_SIZE:
3599                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_SCRIPT));
3600                                 size_flag = true;
3601                                 break;
3602                         case FOOTNOTE_SIZE:
3603                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_FOOTNOTE));
3604                                 size_flag = true;
3605                                 break;
3606                         case SMALL_SIZE:
3607                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_SMALL));
3608                                 size_flag = true;
3609                                 break;
3610                         case LARGE_SIZE:
3611                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_LARGE));
3612                                 size_flag = true;
3613                                 break;
3614                         case LARGER_SIZE:
3615                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_LARGER));
3616                                 size_flag = true;
3617                                 break;
3618                         case LARGEST_SIZE:
3619                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_LARGEST));
3620                                 size_flag = true;
3621                                 break;
3622                         case HUGE_SIZE:
3623                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_HUGE));
3624                                 size_flag = true;
3625                                 break;
3626                         case HUGER_SIZE:
3627                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_HUGER));
3628                                 size_flag = true;
3629                                 break;
3630                         case INCREASE_SIZE:
3631                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_INCREASE));
3632                                 size_flag = true;
3633                                 break;
3634                         case DECREASE_SIZE:
3635                                 tagsToOpen.emplace_back(xhtmlStartFontTag(xml::FT_SIZE_DECREASE));
3636                                 size_flag = true;
3637                                 break;
3638                         case INHERIT_SIZE:
3639                         case NORMAL_SIZE:
3640                                 break;
3641                         default:
3642                                 // the other tags are for internal use
3643                                 LATTEST(false);
3644                                 break;
3645                         }
3646                 }
3647
3648                 // FIXME XHTML
3649                 // Other such tags? What about the other text ranges?
3650
3651                 vector<xml::EndFontTag>::const_iterator cit = tagsToClose.begin();
3652                 vector<xml::EndFontTag>::const_iterator cen = tagsToClose.end();
3653                 for (; cit != cen; ++cit)
3654                         xs << *cit;
3655
3656                 vector<xml::FontTag>::const_iterator sit = tagsToOpen.begin();
3657                 vector<xml::FontTag>::const_iterator sen = tagsToOpen.end();
3658                 for (; sit != sen; ++sit)
3659                         xs << *sit;
3660
3661                 tagsToClose.clear();
3662                 tagsToOpen.clear();
3663
3664                 Inset const * inset = getInset(i);
3665                 if (inset) {
3666                         if (!runparams.for_toc || inset->isInToc()) {
3667                                 OutputParams np = runparams;
3668                                 np.local_font = &font;
3669                                 // If the paragraph has size 1, then we are in the "special
3670                                 // case" where we do not output the containing paragraph info
3671                                 if (!inset->getLayout().htmlisblock() && size() != 1)
3672                                         np.html_in_par = true;
3673                                 retval += inset->xhtml(xs, np);
3674                         }
3675                 } else {
3676                         char_type c = getUChar(buf.masterBuffer()->params(),
3677                                                runparams, i);
3678                         if (c == ' ' && (style.free_spacing || runparams.free_spacing))
3679                                 xs << XMLStream::ESCAPE_NONE << "&nbsp;";
3680                         else
3681                                 xs << c;
3682                 }
3683                 font_old = font.fontInfo();
3684         }
3685
3686         // FIXME XHTML
3687         // I'm worried about what happens if a branch, say, is itself
3688         // wrapped in some font stuff. I think that will not work.
3689         xs.closeFontTags();
3690         if (close_paragraph)
3691                 xs.endDivision();
3692
3693         return retval;
3694 }
3695
3696
3697 bool Paragraph::isHfill(pos_type pos) const
3698 {
3699         Inset const * inset = getInset(pos);
3700         return inset && inset->isHfill();
3701 }
3702
3703
3704 bool Paragraph::isNewline(pos_type pos) const
3705 {
3706         // U+2028 LINE SEPARATOR
3707         // U+2029 PARAGRAPH SEPARATOR
3708         char_type const c = d->text_[pos];
3709         if (c == 0x2028 || c == 0x2029)
3710                 return true;
3711         Inset const * inset = getInset(pos);
3712         return inset && inset->lyxCode() == NEWLINE_CODE;
3713 }
3714
3715
3716 bool Paragraph::isEnvSeparator(pos_type pos) const
3717 {
3718         Inset const * inset = getInset(pos);
3719         return inset && inset->lyxCode() == SEPARATOR_CODE;
3720 }
3721
3722
3723 bool Paragraph::isLineSeparator(pos_type pos) const
3724 {
3725         char_type const c = d->text_[pos];
3726         if (isLineSeparatorChar(c))
3727                 return true;
3728         Inset const * inset = getInset(pos);
3729         return inset && inset->isLineSeparator();
3730 }
3731
3732
3733 bool Paragraph::isWordSeparator(pos_type pos, bool const ignore_deleted) const
3734 {
3735         if (pos == size())
3736                 return true;
3737         if (ignore_deleted && isDeleted(pos))
3738                 return false;
3739         if (Inset const * inset = getInset(pos))
3740                 return !inset->isLetter();
3741         // if we have a hard hyphen (no en- or emdash) or apostrophe
3742         // we pass this to the spell checker
3743         // FIXME: this method is subject to change, visit
3744         // https://bugzilla.mozilla.org/show_bug.cgi?id=355178
3745         // to get an impression how complex this is.
3746         if (isHardHyphenOrApostrophe(pos))
3747                 return false;
3748         char_type const c = d->text_[pos];
3749         // We want to pass the escape chars to the spellchecker
3750         docstring const escape_chars = from_utf8(lyxrc.spellchecker_esc_chars);
3751         return !isLetterChar(c) && !isDigitASCII(c) && !contains(escape_chars, c);
3752 }
3753
3754
3755 bool Paragraph::isHardHyphenOrApostrophe(pos_type pos) const
3756 {
3757         pos_type const psize = size();
3758         if (pos >= psize)
3759                 return false;
3760         char_type const c = d->text_[pos];
3761         if (c != '-' && c != '\'')
3762                 return false;
3763         int nextpos = pos + 1;
3764         int prevpos = pos > 0 ? pos - 1 : 0;
3765         if ((nextpos == psize || isSpace(nextpos))
3766                 && (pos == 0 || isSpace(prevpos)))
3767                 return false;
3768         return true;
3769 }
3770
3771
3772 bool Paragraph::needsCProtection(bool const fragile) const
3773 {
3774         // first check the layout of the paragraph, but only in insets
3775         InsetText const * textinset = inInset().asInsetText();
3776         bool const maintext = textinset
3777                 ? textinset->text().isMainText()
3778                 : false;
3779
3780         if (!maintext && layout().needcprotect) {
3781                 // Environments need cprotection regardless the content
3782                 if (layout().latextype == LATEX_ENVIRONMENT)
3783                         return true;
3784
3785                 // Commands need cprotection if they contain specific chars
3786                 int const nchars_escape = 9;
3787                 static char_type const chars_escape[nchars_escape] = {
3788                         '&', '_', '$', '%', '#', '^', '{', '}', '\\'};
3789
3790                 docstring const pars = asString();
3791                 for (int k = 0; k < nchars_escape; k++) {
3792                         if (contains(pars, chars_escape[k]))
3793                                 return true;
3794                 }
3795         }
3796
3797         // now check whether we have insets that need cprotection
3798         pos_type size = pos_type(d->text_.size());
3799         for (pos_type i = 0; i < size; ++i) {
3800                 if (!isInset(i))
3801                         continue;
3802                 Inset const * ins = getInset(i);
3803                 if (ins->needsCProtection(maintext, fragile))
3804                         return true;
3805                 if (ins->getLayout().latextype() == InsetLayout::ENVIRONMENT)
3806                         // Environments need cprotection regardless the content
3807                         return true;
3808                 // Now check math environments
3809                 InsetMath const * im = getInset(i)->asInsetMath();
3810                 if (!im || im->cell(0).empty())
3811                         continue;
3812                 switch(im->cell(0)[0]->lyxCode()) {
3813                 case MATH_AMSARRAY_CODE:
3814                 case MATH_SUBSTACK_CODE:
3815                 case MATH_ENV_CODE:
3816                 case MATH_XYMATRIX_CODE:
3817                         // these need cprotection
3818                         return true;
3819                 default:
3820                         break;
3821                 }
3822         }
3823
3824         return false;
3825 }
3826
3827
3828 FontSpan const & Paragraph::getSpellRange(pos_type pos) const
3829 {
3830         return d->speller_state_.getRange(pos);
3831 }
3832
3833
3834 bool Paragraph::isChar(pos_type pos) const
3835 {
3836         if (Inset const * inset = getInset(pos))
3837                 return inset->isChar();
3838         char_type const c = d->text_[pos];
3839         return !isLetterChar(c) && !isDigitASCII(c) && !lyx::isSpace(c);
3840 }
3841
3842
3843 bool Paragraph::isSpace(pos_type pos) const
3844 {
3845         if (Inset const * inset = getInset(pos))
3846                 return inset->isSpace();
3847         char_type const c = d->text_[pos];
3848         return lyx::isSpace(c);
3849 }
3850
3851
3852 Language const *
3853 Paragraph::getParLanguage(BufferParams const & bparams) const
3854 {
3855         if (!empty())
3856                 return getFirstFontSettings(bparams).language();
3857         // FIXME: we should check the prev par as well (Lgb)
3858         return bparams.language;
3859 }
3860
3861
3862 bool Paragraph::isRTL(BufferParams const & bparams) const
3863 {
3864         return getParLanguage(bparams)->rightToLeft()
3865                 && !inInset().getLayout().forceLTR();
3866 }
3867
3868
3869 void Paragraph::changeLanguage(BufferParams const & bparams,
3870                                Language const * from, Language const * to)
3871 {
3872         // change language including dummy font change at the end
3873         for (pos_type i = 0; i <= size(); ++i) {
3874                 Font font = getFontSettings(bparams, i);
3875                 if (font.language() == from) {
3876                         font.setLanguage(to);
3877                         setFont(i, font);
3878                         d->requestSpellCheck(i);
3879                 }
3880         }
3881 }
3882
3883
3884 bool Paragraph::isMultiLingual(BufferParams const & bparams) const
3885 {
3886         Language const * doc_language = bparams.language;
3887         for (auto const & f : d->fontlist_)
3888                 if (f.font().language() != ignore_language &&
3889                     f.font().language() != latex_language &&
3890                     f.font().language() != doc_language)
3891                         return true;
3892         return false;
3893 }
3894
3895
3896 void Paragraph::getLanguages(std::set<Language const *> & langs) const
3897 {
3898         for (auto const & f : d->fontlist_) {
3899                 Language const * lang = f.font().language();
3900                 if (lang != ignore_language &&
3901                     lang != latex_language)
3902                         langs.insert(lang);
3903         }
3904 }
3905
3906
3907 docstring Paragraph::asString(int options) const
3908 {
3909         return asString(0, size(), options);
3910 }
3911
3912
3913 docstring Paragraph::asString(pos_type beg, pos_type end, int options, const OutputParams *runparams) const
3914 {
3915         odocstringstream os;
3916
3917         if (beg == 0
3918             && options & AS_STR_LABEL
3919             && !d->params_.labelString().empty())
3920                 os << d->params_.labelString() << ' ';
3921
3922         for (pos_type i = beg; i < end; ++i) {
3923                 if ((options & AS_STR_SKIPDELETE) && isDeleted(i))
3924                         continue;
3925                 char_type const c = d->text_[i];
3926                 if (isPrintable(c) || c == '\t'
3927                     || (c == '\n' && (options & AS_STR_NEWLINES)))
3928                         os.put(c);
3929                 else if (c == META_INSET && (options & AS_STR_INSETS)) {
3930                         if (c == META_INSET && (options & AS_STR_PLAINTEXT)) {
3931                                 LASSERT(runparams != nullptr, return docstring());
3932                                 getInset(i)->plaintext(os, *runparams);
3933                         } else {
3934                                 getInset(i)->toString(os);
3935                         }
3936                 }
3937         }
3938
3939         return os.str();
3940 }
3941
3942
3943 void Paragraph::forOutliner(docstring & os, size_t const maxlen,
3944                             bool const shorten, bool const label) const
3945 {
3946         size_t tmplen = shorten ? maxlen + 1 : maxlen;
3947         if (label && !labelString().empty())
3948                 os += labelString() + ' ';
3949         if (!layout().isTocCaption())
3950                 return;
3951         for (pos_type i = 0; i < size() && os.length() < tmplen; ++i) {
3952                 if (isDeleted(i))
3953                         continue;
3954                 char_type const c = d->text_[i];
3955                 if (isPrintable(c))
3956                         os += c;
3957                 else if (c == META_INSET)
3958                         getInset(i)->forOutliner(os, tmplen, false);
3959         }
3960         if (shorten)
3961                 Text::shortenForOutliner(os, maxlen);
3962 }
3963
3964
3965 void Paragraph::setInsetOwner(Inset const * inset)
3966 {
3967         d->inset_owner_ = inset;
3968 }
3969
3970
3971 int Paragraph::id() const
3972 {
3973         return d->id_;
3974 }
3975
3976
3977 void Paragraph::setId(int id)
3978 {
3979         d->id_ = id;
3980 }
3981
3982
3983 Layout const & Paragraph::layout() const
3984 {
3985         return *d->layout_;
3986 }
3987
3988
3989 void Paragraph::setLayout(Layout const & layout)
3990 {
3991         d->layout_ = &layout;
3992 }
3993
3994
3995 void Paragraph::setDefaultLayout(DocumentClass const & tc)
3996 {
3997         setLayout(tc.defaultLayout());
3998 }
3999
4000
4001 void Paragraph::setPlainLayout(DocumentClass const & tc)
4002 {
4003         setLayout(tc.plainLayout());
4004 }
4005
4006
4007 void Paragraph::setPlainOrDefaultLayout(DocumentClass const & tclass)
4008 {
4009         if (usePlainLayout())
4010                 setPlainLayout(tclass);
4011         else
4012                 setDefaultLayout(tclass);
4013 }
4014
4015
4016 Inset const & Paragraph::inInset() const
4017 {
4018         LBUFERR(d->inset_owner_);
4019         return *d->inset_owner_;
4020 }
4021
4022
4023 ParagraphParameters & Paragraph::params()
4024 {
4025         return d->params_;
4026 }
4027
4028
4029 ParagraphParameters const & Paragraph::params() const
4030 {
4031         return d->params_;
4032 }
4033
4034
4035 bool Paragraph::isFreeSpacing() const
4036 {
4037         if (d->layout_->free_spacing)
4038                 return true;
4039         return d->inset_owner_ && d->inset_owner_->isFreeSpacing();
4040 }
4041
4042
4043 bool Paragraph::allowEmpty() const
4044 {
4045         if (d->layout_->keepempty)
4046                 return true;
4047         return d->inset_owner_ && d->inset_owner_->allowEmpty();
4048 }
4049
4050
4051 bool Paragraph::brokenBiblio() const
4052 {
4053         // There is a problem if there is no bibitem at position 0 in
4054         // paragraphs that need one, if there is another bibitem in the
4055         // paragraph or if this paragraph is not supposed to have
4056         // a bibitem inset at all.
4057         return ((d->layout_->labeltype == LABEL_BIBLIO
4058                 && (d->insetlist_.find(BIBITEM_CODE) != 0
4059                     || d->insetlist_.find(BIBITEM_CODE, 1) > 0))
4060                 || (d->layout_->labeltype != LABEL_BIBLIO
4061                     && d->insetlist_.find(BIBITEM_CODE) != -1));
4062 }
4063
4064
4065 int Paragraph::fixBiblio(Buffer const & buffer)
4066 {
4067         // FIXME: when there was already an inset at 0, the return value is 1,
4068         // which does not tell whether another inset has been remove; the
4069         // cursor cannot be correctly updated.
4070
4071         bool const track_changes = buffer.params().track_changes;
4072         int bibitem_pos = d->insetlist_.find(BIBITEM_CODE);
4073
4074         // The case where paragraph is not BIBLIO
4075         if (d->layout_->labeltype != LABEL_BIBLIO) {
4076                 if (bibitem_pos == -1)
4077                         // No InsetBibitem => OK
4078                         return 0;
4079                 // There is an InsetBibitem: remove it!
4080                 d->insetlist_.release(bibitem_pos);
4081                 eraseChar(bibitem_pos, track_changes);
4082                 return (bibitem_pos == 0) ? -1 : -bibitem_pos;
4083         }
4084
4085         bool const hasbibitem0 = bibitem_pos == 0;
4086         if (hasbibitem0) {
4087                 bibitem_pos = d->insetlist_.find(BIBITEM_CODE, 1);
4088                 // There was an InsetBibitem at pos 0,
4089                 // and no other one => OK
4090                 if (bibitem_pos == -1)
4091                         return 0;
4092                 // there is a bibitem at the 0 position, but since
4093                 // there is a second one, we copy the second on the
4094                 // first. We're assuming there are at most two of
4095                 // these, which there should be.
4096                 // FIXME: why does it make sense to do that rather
4097                 // than keep the first? (JMarc)
4098                 Inset * inset = releaseInset(bibitem_pos);
4099                 d->insetlist_.begin()->inset = inset;
4100                 return -bibitem_pos;
4101         }
4102
4103         // We need to create an inset at the beginning
4104         Inset * inset = nullptr;
4105         if (bibitem_pos > 0) {
4106                 // there was one somewhere in the paragraph, let's move it
4107                 inset = d->insetlist_.release(bibitem_pos);
4108                 eraseChar(bibitem_pos, track_changes);
4109         } else
4110                 // make a fresh one
4111                 inset = new InsetBibitem(const_cast<Buffer *>(&buffer),
4112                                          InsetCommandParams(BIBITEM_CODE));
4113
4114         Font font(inherit_font, buffer.params().language);
4115         insertInset(0, inset, font, Change(track_changes ? Change::INSERTED
4116                                                    : Change::UNCHANGED));
4117
4118         // This is needed to get the counters right
4119         buffer.updateBuffer();
4120         return 1;
4121 }
4122
4123
4124 void Paragraph::checkAuthors(AuthorList const & authorList)
4125 {
4126         d->changes_.checkAuthors(authorList);
4127 }
4128
4129
4130 bool Paragraph::isChanged(pos_type pos) const
4131 {
4132         return lookupChange(pos).changed();
4133 }
4134
4135
4136 bool Paragraph::isInserted(pos_type pos) const
4137 {
4138         return lookupChange(pos).inserted();
4139 }
4140
4141
4142 bool Paragraph::isDeleted(pos_type pos) const
4143 {
4144         return lookupChange(pos).deleted();
4145 }
4146
4147
4148 InsetList const & Paragraph::insetList() const
4149 {
4150         return d->insetlist_;
4151 }
4152
4153
4154 void Paragraph::setInsetBuffers(Buffer & b)
4155 {
4156         d->insetlist_.setBuffer(b);
4157 }
4158
4159
4160 void Paragraph::resetBuffer()
4161 {
4162         d->insetlist_.resetBuffer();
4163 }
4164
4165
4166 Inset * Paragraph::releaseInset(pos_type pos)
4167 {
4168         Inset * inset = d->insetlist_.release(pos);
4169         /// does not honour change tracking!
4170         eraseChar(pos, false);
4171         return inset;
4172 }
4173
4174
4175 Inset * Paragraph::getInset(pos_type pos)
4176 {
4177         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
4178                  ? d->insetlist_.get(pos) : nullptr;
4179 }
4180
4181
4182 Inset const * Paragraph::getInset(pos_type pos) const
4183 {
4184         return (pos < pos_type(d->text_.size()) && d->text_[pos] == META_INSET)
4185                  ? d->insetlist_.get(pos) : nullptr;
4186 }
4187
4188
4189 void Paragraph::changeCase(BufferParams const & bparams, pos_type pos,
4190                 pos_type & right, TextCase action)
4191 {
4192         // process sequences of modified characters; in change
4193         // tracking mode, this approach results in much better
4194         // usability than changing case on a char-by-char basis
4195         // We also need to track the current font, since font
4196         // changes within sequences can occur.
4197         vector<pair<char_type, Font> > changes;
4198
4199         bool const trackChanges = bparams.track_changes;
4200
4201         bool capitalize = true;
4202
4203         for (; pos < right; ++pos) {
4204                 char_type oldChar = d->text_[pos];
4205                 char_type newChar = oldChar;
4206
4207                 // ignore insets and don't play with deleted text!
4208                 if (oldChar != META_INSET && !isDeleted(pos)) {
4209                         switch (action) {
4210                                 case text_lowercase:
4211                                         newChar = lowercase(oldChar);
4212                                         break;
4213                                 case text_capitalization:
4214                                         if (capitalize) {
4215                                                 newChar = uppercase(oldChar);
4216                                                 capitalize = false;
4217                                         }
4218                                         break;
4219                                 case text_uppercase:
4220                                         newChar = uppercase(oldChar);
4221                                         break;
4222                         }
4223                 }
4224
4225                 if (isWordSeparator(pos) || isDeleted(pos)) {
4226                         // permit capitalization again
4227                         capitalize = true;
4228                 }
4229
4230                 if (oldChar != newChar) {
4231                         changes.push_back(make_pair(newChar, getFontSettings(bparams, pos)));
4232                         if (pos != right - 1)
4233                                 continue;
4234                         // step behind the changing area
4235                         pos++;
4236                 }
4237
4238                 int erasePos = pos - changes.size();
4239                 for (size_t i = 0; i < changes.size(); i++) {
4240                         insertChar(pos, changes[i].first,
4241                                    changes[i].second,
4242                                    trackChanges);
4243                         if (!eraseChar(erasePos, trackChanges)) {
4244                                 ++erasePos;
4245                                 ++pos; // advance
4246                                 ++right; // expand selection
4247                         }
4248                 }
4249                 changes.clear();
4250         }
4251 }
4252
4253
4254 int Paragraph::find(docstring const & str, bool cs, bool mw,
4255                 pos_type start_pos, bool del) const
4256 {
4257         pos_type pos = start_pos;
4258         int const strsize = str.length();
4259         int i = 0;
4260         pos_type const parsize = d->text_.size();
4261         for (i = 0; i < strsize && pos < parsize; ++i, ++pos) {
4262                 // Ignore "invisible" letters such as ligature breaks
4263                 // and hyphenation chars while searching
4264                 while (pos < parsize - 1 && isInset(pos)) {
4265                         odocstringstream os;
4266                         getInset(pos)->toString(os);
4267                         if (!getInset(pos)->isLetter() || !os.str().empty())
4268                                 break;
4269                         pos++;
4270                 }
4271                 if (cs && str[i] != d->text_[pos])
4272                         break;
4273                 if (!cs && uppercase(str[i]) != uppercase(d->text_[pos]))
4274                         break;
4275                 if (!del && isDeleted(pos))
4276                         break;
4277         }
4278
4279         if (i != strsize)
4280                 return 0;
4281
4282         // if necessary, check whether string matches word
4283         if (mw) {
4284                 if (start_pos > 0 && !isWordSeparator(start_pos - 1))
4285                         return 0;
4286                 if (pos < parsize
4287                         && !isWordSeparator(pos))
4288                         return 0;
4289         }
4290
4291         return pos - start_pos;
4292 }
4293
4294
4295 char_type Paragraph::getChar(pos_type pos) const
4296 {
4297         return d->text_[pos];
4298 }
4299
4300
4301 pos_type Paragraph::size() const
4302 {
4303         return d->text_.size();
4304 }
4305
4306
4307 bool Paragraph::empty() const
4308 {
4309         return d->text_.empty();
4310 }
4311
4312
4313 bool Paragraph::isInset(pos_type pos) const
4314 {
4315         return d->text_[pos] == META_INSET;
4316 }
4317
4318
4319 bool Paragraph::isSeparator(pos_type pos) const
4320 {
4321         //FIXME: Are we sure this can be the only separator?
4322         return d->text_[pos] == ' ';
4323 }
4324
4325
4326 void Paragraph::deregisterWords()
4327 {
4328         Private::LangWordsMap::const_iterator itl = d->words_.begin();
4329         Private::LangWordsMap::const_iterator ite = d->words_.end();
4330         for (; itl != ite; ++itl) {
4331                 WordList & wl = theWordList(itl->first);
4332                 Private::Words::const_iterator it = (itl->second).begin();
4333                 Private::Words::const_iterator et = (itl->second).end();
4334                 for (; it != et; ++it)
4335                         wl.remove(*it);
4336         }
4337         d->words_.clear();
4338 }
4339
4340
4341 void Paragraph::locateWord(pos_type & from, pos_type & to,
4342         word_location const loc, bool const ignore_deleted) const
4343 {
4344         switch (loc) {
4345         case WHOLE_WORD_STRICT:
4346                 if (from == 0 || from == size()
4347                     || isWordSeparator(from, ignore_deleted)
4348                     || isWordSeparator(from - 1, ignore_deleted)) {
4349                         to = from;
4350                         return;
4351                 }
4352                 // fall through
4353
4354         case WHOLE_WORD:
4355                 // If we are already at the beginning of a word, do nothing
4356                 if (!from || isWordSeparator(from - 1, ignore_deleted))
4357                         break;
4358                 // fall through
4359
4360         case PREVIOUS_WORD:
4361                 // always move the cursor to the beginning of previous word
4362                 while (from && !isWordSeparator(from - 1, ignore_deleted))
4363                         --from;
4364                 break;
4365         case NEXT_WORD:
4366                 LYXERR0("Paragraph::locateWord: NEXT_WORD not implemented yet");
4367                 break;
4368         case PARTIAL_WORD:
4369                 // no need to move the 'from' cursor
4370                 break;
4371         }
4372         to = from;
4373         while (to < size() && !isWordSeparator(to, ignore_deleted))
4374                 ++to;
4375 }
4376
4377
4378 void Paragraph::collectWords()
4379 {
4380         for (pos_type pos = 0; pos < size(); ++pos) {
4381                 if (isWordSeparator(pos))
4382                         continue;
4383                 pos_type from = pos;
4384                 locateWord(from, pos, WHOLE_WORD);
4385                 // Work around MSVC warning: The statement
4386                 // if (pos < from + lyxrc.completion_minlength)
4387                 // triggers a signed vs. unsigned warning.
4388                 // I don't know why this happens, it could be a MSVC bug, or
4389                 // related to LLP64 (windows) vs. LP64 (unix) programming
4390                 // model, or the C++ standard might be ambigous in the section
4391                 // defining the "usual arithmetic conversions". However, using
4392                 // a temporary variable is safe and works on all compilers.
4393                 pos_type const endpos = from + lyxrc.completion_minlength;
4394                 if (pos < endpos)
4395                         continue;
4396                 FontList::const_iterator cit = d->fontlist_.fontIterator(from);
4397                 if (cit == d->fontlist_.end())
4398                         return;
4399                 Language const * lang = cit->font().language();
4400                 docstring const word = asString(from, pos, AS_STR_NONE);
4401                 d->words_[lang->lang()].insert(word);
4402         }
4403 }
4404
4405
4406 void Paragraph::registerWords()
4407 {
4408         Private::LangWordsMap::const_iterator itl = d->words_.begin();
4409         Private::LangWordsMap::const_iterator ite = d->words_.end();
4410         for (; itl != ite; ++itl) {
4411                 WordList & wl = theWordList(itl->first);
4412                 Private::Words::const_iterator it = (itl->second).begin();
4413                 Private::Words::const_iterator et = (itl->second).end();
4414                 for (; it != et; ++it)
4415                         wl.insert(*it);
4416         }
4417 }
4418
4419
4420 void Paragraph::updateWords()
4421 {
4422         deregisterWords();
4423         collectWords();
4424         registerWords();
4425 }
4426
4427
4428 void Paragraph::Private::appendSkipPosition(SkipPositions & skips, pos_type const pos) const
4429 {
4430         SkipPositionsIterator begin = skips.begin();
4431         SkipPositions::iterator end = skips.end();
4432         if (pos > 0 && begin < end) {
4433                 --end;
4434                 if (end->last == pos - 1) {
4435                         end->last = pos;
4436                         return;
4437                 }
4438         }
4439         skips.insert(end, FontSpan(pos, pos));
4440 }
4441
4442
4443 Language * Paragraph::Private::locateSpellRange(
4444         pos_type & from, pos_type & to,
4445         SkipPositions & skips) const
4446 {
4447         // skip leading white space
4448         while (from < to && owner_->isWordSeparator(from))
4449                 ++from;
4450         // don't check empty range
4451         if (from >= to)
4452                 return nullptr;
4453         // get current language
4454         Language * lang = getSpellLanguage(from);
4455         pos_type last = from;
4456         bool samelang = true;
4457         bool sameinset = true;
4458         while (last < to && samelang && sameinset) {
4459                 // hop to end of word
4460                 while (last < to && !owner_->isWordSeparator(last)) {
4461                         if (owner_->getInset(last)) {
4462                                 appendSkipPosition(skips, last);
4463                         } else if (owner_->isDeleted(last)) {
4464                                 appendSkipPosition(skips, last);
4465                         }
4466                         ++last;
4467                 }
4468                 // hop to next word while checking for insets
4469                 while (sameinset && last < to && owner_->isWordSeparator(last)) {
4470                         if (Inset const * inset = owner_->getInset(last))
4471                                 sameinset = inset->isChar() && inset->isLetter();
4472                         if (sameinset && owner_->isDeleted(last)) {
4473                                 appendSkipPosition(skips, last);
4474                         }
4475                         if (sameinset)
4476                                 last++;
4477                 }
4478                 if (sameinset && last < to) {
4479                         // now check for language change
4480                         samelang = lang == getSpellLanguage(last);
4481                 }
4482         }
4483         // if language change detected backstep is needed
4484         if (!samelang)
4485                 --last;
4486         to = last;
4487         return lang;
4488 }
4489
4490
4491 Language * Paragraph::Private::getSpellLanguage(pos_type const from) const
4492 {
4493         Language * lang =
4494                 const_cast<Language *>(owner_->getFontSettings(
4495                         inset_owner_->buffer().params(), from).language());
4496         if (lang == inset_owner_->buffer().params().language
4497                 && !lyxrc.spellchecker_alt_lang.empty()) {
4498                 string lang_code;
4499                 string const lang_variety =
4500                         split(lyxrc.spellchecker_alt_lang, lang_code, '-');
4501                 lang->setCode(lang_code);
4502                 lang->setVariety(lang_variety);
4503         }
4504         return lang;
4505 }
4506
4507
4508 void Paragraph::requestSpellCheck(pos_type pos)
4509 {
4510         d->requestSpellCheck(pos);
4511 }
4512
4513
4514 bool Paragraph::needsSpellCheck() const
4515 {
4516         SpellChecker::ChangeNumber speller_change_number = 0;
4517         if (theSpellChecker())
4518                 speller_change_number = theSpellChecker()->changeNumber();
4519         if (speller_change_number > d->speller_state_.currentChangeNumber()) {
4520                 d->speller_state_.needsCompleteRefresh(speller_change_number);
4521         }
4522         return d->needsSpellCheck();
4523 }
4524
4525
4526 bool Paragraph::Private::ignoreWord(docstring const & word) const
4527 {
4528         // Ignore words with digits
4529         // FIXME: make this customizable
4530         // (note that some checkers ignore words with digits by default)
4531         docstring::const_iterator cit = word.begin();
4532         docstring::const_iterator const end = word.end();
4533         for (; cit != end; ++cit) {
4534                 if (isNumber((*cit)))
4535                         return true;
4536         }
4537         return false;
4538 }
4539
4540
4541 SpellChecker::Result Paragraph::spellCheck(pos_type & from, pos_type & to,
4542         WordLangTuple & wl, docstring_list & suggestions,
4543         bool do_suggestion, bool check_learned) const
4544 {
4545         SpellChecker::Result result = SpellChecker::WORD_OK;
4546         SpellChecker * speller = theSpellChecker();
4547         if (!speller)
4548                 return result;
4549
4550         if (!d->layout_->spellcheck || !inInset().allowSpellCheck())
4551                 return result;
4552
4553         locateWord(from, to, WHOLE_WORD, true);
4554         if (from == to || from >= size())
4555                 return result;
4556
4557         docstring word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
4558         Language * lang = d->getSpellLanguage(from);
4559
4560         if (getFontSettings(d->inset_owner_->buffer().params(), from).fontInfo().nospellcheck() == FONT_ON)
4561                 return result;
4562
4563         wl = WordLangTuple(word, lang);
4564
4565         if (word.empty())
4566                 return result;
4567
4568         if (needsSpellCheck() || check_learned) {
4569                 pos_type end = to;
4570                 if (!d->ignoreWord(word)) {
4571                         bool const trailing_dot = to < size() && d->text_[to] == '.';
4572                         result = speller->check(wl);
4573                         if (SpellChecker::misspelled(result) && trailing_dot) {
4574                                 wl = WordLangTuple(word.append(from_ascii(".")), lang);
4575                                 result = speller->check(wl);
4576                                 if (!SpellChecker::misspelled(result)) {
4577                                         LYXERR(Debug::GUI, "misspelled word is correct with dot: \"" <<
4578                                            word << "\" [" <<
4579                                            from << ".." << to << "]");
4580                                 } else {
4581                                         // spell check with dot appended failed too
4582                                         // restore original word/lang value
4583                                         word = asString(from, to, AS_STR_INSETS | AS_STR_SKIPDELETE);
4584                                         wl = WordLangTuple(word, lang);
4585                                 }
4586                         }
4587                 }
4588                 if (!SpellChecker::misspelled(result)) {
4589                         // area up to the begin of the next word is not misspelled
4590                         while (end < size() && isWordSeparator(end))
4591                                 ++end;
4592                 }
4593                 d->setMisspelled(from, end, result);
4594         } else {
4595                 result = d->speller_state_.getState(from);
4596         }
4597
4598         if (do_suggestion)
4599                 suggestions.clear();
4600
4601         if (SpellChecker::misspelled(result)) {
4602                 LYXERR(Debug::GUI, "misspelled word: \"" <<
4603                            word << "\" [" <<
4604                            from << ".." << to << "]");
4605                 if (do_suggestion)
4606                         speller->suggest(wl, suggestions);
4607         }
4608         return result;
4609 }
4610
4611
4612 void Paragraph::anonymize()
4613 {
4614         // This is a very crude anonymization for now
4615         for (char_type & c : d->text_)
4616                 if (isLetterChar(c) || isNumber(c))
4617                         c = 'a';
4618 }
4619
4620
4621 void Paragraph::Private::markMisspelledWords(
4622         pos_type const & first, pos_type const & last,
4623         SpellChecker::Result result,
4624         docstring const & word,
4625         SkipPositions const & skips)
4626 {
4627         if (!SpellChecker::misspelled(result)) {
4628                 setMisspelled(first, last, SpellChecker::WORD_OK);
4629                 return;
4630         }
4631         int snext = first;
4632         SpellChecker * speller = theSpellChecker();
4633         // locate and enumerate the error positions
4634         int nerrors = speller->numMisspelledWords();
4635         int numskipped = 0;
4636         SkipPositionsIterator it = skips.begin();
4637         SkipPositionsIterator et = skips.end();
4638         for (int index = 0; index < nerrors; ++index) {
4639                 int wstart;
4640                 int wlen = 0;
4641                 speller->misspelledWord(index, wstart, wlen);
4642                 /// should not happen if speller supports range checks
4643                 if (!wlen) continue;
4644                 docstring const misspelled = word.substr(wstart, wlen);
4645                 wstart += first + numskipped;
4646                 if (snext < wstart) {
4647                         /// mark the range of correct spelling
4648                         numskipped += countSkips(it, et, wstart);
4649                         setMisspelled(snext,
4650                                 wstart - 1, SpellChecker::WORD_OK);
4651                 }
4652                 snext = wstart + wlen;
4653                 numskipped += countSkips(it, et, snext);
4654                 /// mark the range of misspelling
4655                 setMisspelled(wstart, snext, result);
4656                 LYXERR(Debug::GUI, "misspelled word: \"" <<
4657                            misspelled << "\" [" <<
4658                            wstart << ".." << (snext-1) << "]");
4659                 ++snext;
4660         }
4661         if (snext <= last) {
4662                 /// mark the range of correct spelling at end
4663                 setMisspelled(snext, last, SpellChecker::WORD_OK);
4664         }
4665 }
4666
4667
4668 void Paragraph::spellCheck() const
4669 {
4670         SpellChecker * speller = theSpellChecker();
4671         if (!speller || empty() ||!needsSpellCheck())
4672                 return;
4673         pos_type start;
4674         pos_type endpos;
4675         d->rangeOfSpellCheck(start, endpos);
4676         if (speller->canCheckParagraph()) {
4677                 // loop until we leave the range
4678                 for (pos_type first = start; first < endpos; ) {
4679                         pos_type last = endpos;
4680                         Private::SkipPositions skips;
4681                         Language * lang = d->locateSpellRange(first, last, skips);
4682                         if (first >= endpos)
4683                                 break;
4684                         // start the spell checker on the unit of meaning
4685                         docstring word = asString(first, last, AS_STR_INSETS + AS_STR_SKIPDELETE);
4686                         WordLangTuple wl = WordLangTuple(word, lang);
4687                         SpellChecker::Result result = word.size() ?
4688                                 speller->check(wl) : SpellChecker::WORD_OK;
4689                         d->markMisspelledWords(first, last, result, word, skips);
4690                         first = ++last;
4691                 }
4692         } else {
4693                 static docstring_list suggestions;
4694                 pos_type to = endpos;
4695                 while (start < endpos) {
4696                         WordLangTuple wl;
4697                         spellCheck(start, to, wl, suggestions, false);
4698                         start = to + 1;
4699                 }
4700         }
4701         d->readySpellCheck();
4702 }
4703
4704
4705 bool Paragraph::isMisspelled(pos_type pos, bool check_boundary) const
4706 {
4707         bool result = SpellChecker::misspelled(d->speller_state_.getState(pos));
4708         if (result || pos <= 0 || pos > size())
4709                 return result;
4710         if (check_boundary && (pos == size() || isWordSeparator(pos)))
4711                 result = SpellChecker::misspelled(d->speller_state_.getState(pos - 1));
4712         return result;
4713 }
4714
4715
4716 string Paragraph::magicLabel() const
4717 {
4718         stringstream ss;
4719         ss << "magicparlabel-" << id();
4720         return ss.str();
4721 }
4722
4723
4724 } // namespace lyx